Stability of Materials for Use in Space-Based Interferometric Missions

Permanent Link: http://ufdc.ufl.edu/UFE0041495/00001

Material Information

Title: Stability of Materials for Use in Space-Based Interferometric Missions
Physical Description: 1 online resource (237 p.)
Language: english
Creator: Preston, Alix
Publisher: University of Florida
Place of Publication: Gainesville, Fla.
Publication Date: 2010


Subjects / Keywords: breaking, cfrp, dimensional, hydroxide, interferometry, lisa, pzt, sic, telescope
Physics -- Dissertations, Academic -- UF
Genre: Physics thesis, Ph.D.
bibliography   ( marcgt )
theses   ( marcgt )
government publication (state, provincial, terriorial, dependent)   ( marcgt )
born-digital   ( sobekcm )
Electronic Thesis or Dissertation


Abstract: Space-based interferometric missions such as SPIRIT, SPECS, TPF, and LISA will take measurements of the universe with unprecedented results. To do this, these missions will require ultra-stable materials and bonding techniques to be used for critical optical components such as optical benches or support structures. As an example, the telescope support structure for the LISA mission must be made of a material that is stable to better than 1 pm/rtHz at 3 mHz and whose length between the primary and secondary mirrors cannot change by more than 1.2 micrometers over the lifetime of the mission. Although materials such as silicon carbide (SiC) and carbon fiber reinforced polymers (CFRP) have been suggested for this use, they have not been tested to these strict requirements. The author is part of a group at the University of Florida that is developing a materials research facility capable of testing the stability and strength of materials and bonding techniques for use in space-based interferometric missions. This work describes the techniques used to test the stability of materials at the femtometer level. Results are presented for SiC, Zerodur, a carbon fiber reinforced polymer, Super Invar, and k-180 piezoelectric material. In addition, several bonding techniques are described along with potential uses for space missions. Specific attention is paid to a promising new jointing method known as hydroxide bonding. The bonding method, bonding mechanism, and shear test strengths are presented for several material combinations. During this work, it was found that hydroxide bonding could be used to bond SiC to SiC with significant strength in a simple manner and was chosen to bond a prototype telescope support structure for the LISA mission. Initial results and discussion about further improvements to the design are presented. Finally, investigations were done to determine the differential phase noise in counter-propagating beams through an optical macronber. Results using a polarizing Sagnac interferometer showed a diregistered trademarkerential phase noise of 5*10^{-5} cycles/rtHz at 1 Hz. The experimental setup, noise sources, and results are presented.
General Note: In the series University of Florida Digital Collections.
General Note: Includes vita.
Bibliography: Includes bibliographical references.
Source of Description: Description based on online resource; title from PDF title page.
Source of Description: This bibliographic record is available under the Creative Commons CC0 public domain dedication. The University of Florida Libraries, as creator of this bibliographic record, has waived all rights to it worldwide under copyright law, including all related and neighboring rights, to the extent allowed by law.
Statement of Responsibility: by Alix Preston.
Thesis: Thesis (Ph.D.)--University of Florida, 2010.
Local: Adviser: Muller, Guido.

Record Information

Source Institution: UFRGP
Rights Management: Applicable rights reserved.
Classification: lcc - LD1780 2010
System ID: UFE0041495:00001

Permanent Link: http://ufdc.ufl.edu/UFE0041495/00001

Material Information

Title: Stability of Materials for Use in Space-Based Interferometric Missions
Physical Description: 1 online resource (237 p.)
Language: english
Creator: Preston, Alix
Publisher: University of Florida
Place of Publication: Gainesville, Fla.
Publication Date: 2010


Subjects / Keywords: breaking, cfrp, dimensional, hydroxide, interferometry, lisa, pzt, sic, telescope
Physics -- Dissertations, Academic -- UF
Genre: Physics thesis, Ph.D.
bibliography   ( marcgt )
theses   ( marcgt )
government publication (state, provincial, terriorial, dependent)   ( marcgt )
born-digital   ( sobekcm )
Electronic Thesis or Dissertation


Abstract: Space-based interferometric missions such as SPIRIT, SPECS, TPF, and LISA will take measurements of the universe with unprecedented results. To do this, these missions will require ultra-stable materials and bonding techniques to be used for critical optical components such as optical benches or support structures. As an example, the telescope support structure for the LISA mission must be made of a material that is stable to better than 1 pm/rtHz at 3 mHz and whose length between the primary and secondary mirrors cannot change by more than 1.2 micrometers over the lifetime of the mission. Although materials such as silicon carbide (SiC) and carbon fiber reinforced polymers (CFRP) have been suggested for this use, they have not been tested to these strict requirements. The author is part of a group at the University of Florida that is developing a materials research facility capable of testing the stability and strength of materials and bonding techniques for use in space-based interferometric missions. This work describes the techniques used to test the stability of materials at the femtometer level. Results are presented for SiC, Zerodur, a carbon fiber reinforced polymer, Super Invar, and k-180 piezoelectric material. In addition, several bonding techniques are described along with potential uses for space missions. Specific attention is paid to a promising new jointing method known as hydroxide bonding. The bonding method, bonding mechanism, and shear test strengths are presented for several material combinations. During this work, it was found that hydroxide bonding could be used to bond SiC to SiC with significant strength in a simple manner and was chosen to bond a prototype telescope support structure for the LISA mission. Initial results and discussion about further improvements to the design are presented. Finally, investigations were done to determine the differential phase noise in counter-propagating beams through an optical macronber. Results using a polarizing Sagnac interferometer showed a diregistered trademarkerential phase noise of 5*10^{-5} cycles/rtHz at 1 Hz. The experimental setup, noise sources, and results are presented.
General Note: In the series University of Florida Digital Collections.
General Note: Includes vita.
Bibliography: Includes bibliographical references.
Source of Description: Description based on online resource; title from PDF title page.
Source of Description: This bibliographic record is available under the Creative Commons CC0 public domain dedication. The University of Florida Libraries, as creator of this bibliographic record, has waived all rights to it worldwide under copyright law, including all related and neighboring rights, to the extent allowed by law.
Statement of Responsibility: by Alix Preston.
Thesis: Thesis (Ph.D.)--University of Florida, 2010.
Local: Adviser: Muller, Guido.

Record Information

Source Institution: UFRGP
Rights Management: Applicable rights reserved.
Classification: lcc - LD1780 2010
System ID: UFE0041495:00001

This item has the following downloads:

Full Text







0 2010 Alix Preston

This is dedicated to all who were told they would fail, only to prove them wrong


Much of this work would not have been made possible if it were not for the help of

many graduate and undergraduate students, faculty, and staff. I would like to thank Ira

Thorpe, Rachel Cruz, Vinzenz Vand, and Josep Sanjuan for their help and thoughtful

discussions that were instrumental in understanding the nuances of my research. I would

also like to thank Gabriel Boothe, Aaron Spector, Be-_i iiiii:i Balaban, Darsa Donelon,

Kendall Ackley, and Scott Rager for their dedication and persistence to getting the job

done. A special thanks is due for the physics machine shop, especially Marc Link and Bill

Malphurs, who spent many hours on the countless projects I needed. Lastly, I would like

to thank my advisor, Dr. Guido Mueller, who put up with me, guided me, and supported

me in my research.



ACKNOW LEDGMENTS ................................. 4

LIST OF TABLES ....................... ............. 9

LIST OF FIGURES ....................... ........... 10

KEY TO ABBREVIATIONS ............................... 17

KEY TO SYMBOLS .. .. ........... ................ 19

A BSTR A CT . . 20


1 INTRODUCTION ........... ... .. .............. ... 22

1.1 Space-Based Missions ......... .......... ........ 23
1.2 GRACE .......... ... ................ ....... 23
1.3 GRACE Follow-On ......... ..................... 25
1.4 LISA .............. .............. .... 26
1.4.1 Introduction ............................ 26
1.4.2 Sources ...................... .......... 27 Cosmological background sources ...... ..... ... 28 Binary stars ......... .......... .... 28 Chirping sources ........... ............ 29 Extreme mass ratio inspirals (EMRIs) ... 29
1.4.3 General Measurement Overview ............. .. .. 30
1.4.4 The Disturbance Reduction System (DRS) ..... 31
1.4.5 The Interferometric Measurement System (IMS) ... 32 Optical bench ............... ... .. .. 33 Telescope system ..... .......... .... 34
1.5 LATOR ...... ............. ................ .. 34
1.5.1 Mission Design. .................. .......... .. 35
1.5.2 Spacecraft orbits. .................. ......... .. 35
1.5.3 Optical design .................. ........... .. 36
1.5.4 Interferometry .................. ........... .. 37

2 M ATERIALS .................. ................ .. .. 41

2.1 Zerodur .................. .................. .. 41
2.1.1 Production Process .................. ........ .. 42
2.1.2 CTE ................... ... ........ 43
2.1.3 Thermal Cycling .................. .......... .. 44
2.1.4 Long Term Stability .................. ...... .. .. 44
2.1.5 Cryogenic Behavior .................. ........ .. 44

2.2 Invar ... ......... ................ 45
2.2.1 Invar 36 ............... .............. 45
2.2.2 Invar 42 ............... .............. 47
2.2.3 Invar 32-5 ............... ............ 47
2.3 Silicon Carbide ............... .............. 48
2.3.1 Structure of SiC ............... .......... 49
2.3.2 Fabrication Process ............... ........ 49 Acheson process ............... .. 49 Sintered SiC ... ............ ..... 50 Reaction-bonded SiC ...... ........ ... 51 ('C! iii vapor deposition of SiC 52
2.3.3 Properties .................. ............. 52 Density and porosity .................. .. 53 Oxidation resistance .................. .. 54 Flexural and tensile strength .. ...... 54 Thermal conductivity and heat capacity ... 56
2.4 Carbon Fiber Reinforced Polymers .................. .... 56
2.4.1 Fabrication Process .................. ........ 57 Carbon fibers .................. .... 57 Matrix material .................. ... 58
2.4.2 Properties .................. ............. 60
2.5 Lead Zirconate Titanate .................. ......... 60
2.5.1 The Piezoelectric Effect .................. .... 61
2.5.2 Piezoelectric M materials .................. .... 61

3 BONDING TECHNIQUES .................. .......... 65

3.1 Optical Contacting .................. .......... .65
3.1.1 M ethod 1 .. 66
3.1.2 Method 2 (Solution Assisted Optical Contacting) ... 66
3.1.3 Method 3 .................. .......... .. 67
3.2 Anodic Bonding ............... ............. 67
3.3 Brazing . .. .71
3.4 Epoxies . .. .74

4 HYDROXIDE-CATALYSIS BONDING .................. ... 78

4.1 Introduction ............... ............ .78
4.2 Bonding Mechanisms ............... ........... 79
4.2.1 Glass to Glass . 79
4.2.2 Glass to Metals and Metal to Metal ..... ... 80
4.2.3 SiC to SiC ............... ............ 80 X-ray photoelectron spectroscopy 81 SiC-SiC bonding mechanism .. ..... 85
4.3 Bonding Method .................. ............. 87
4.4 Strength Measurements .................. .......... 88

4.4.1 BK7-BK7 Bonding ................ ...... 89
4.4.2 SiC-BK7 Bonding .................. ......... .. 94 Dilution factor .................. .... .. 95 Surface profile .................. ..... .. 96
4.4.3 Super Invar-BK7 .................. ......... .. 97
4.4.4 SiC-SiC .................. .............. .. 98 Hexoloy SA .................. ..... .. 98 POCO SuperSiC ................ .... .. 99 CoorsTek UltraSiC ..... ......... .. 99
4.5 Other Bonding Results ............... ......... 102


5.1 Optical Cavities .................. .............. .. 103
5.1.1 Stability of Optical Resonators ............... 105
5.1.2 Misalignment Analysis ................ ... 107
5.1.3 Cavity Construction .................. ...... .. 108 Zerodur cavities .................. .. .. .. 108 Super Invar cavity ................... 110 SiC cavity .................. ........ .. 110 CFRP cavity .................. ..... .. 112 PZT-actuated cavity .................. .. 115
5.2 Pound-Drever-Hall Laser Frequency Stabilization ..... 116
5.2.1 Conceptual Model .................. ......... .. 117
5.2.2 Quantitative Model .................. ........ .. 118
5.3 Relative Stability Measurements .................. .... 121
5.3.1 Electro-Optical Setup ................... .. .... .. 122
5.3.2 Zerodur-Zerodur. .................. ......... .. 123 Optically contacted mirrors ................ 124 Hydroxide bonded mirrors ..... .... 127
5.3.3 Zerodur-SiC ............... .......... 129
5.3.4 Zerodur-Super Invar... ....... ........ ....... 132
5.3.5 Zerodur-PZT Cavity .................. .. ..... 136
5.3.6 Zerodur-CFRP .................. .......... 146

6 ABSOLUTE STABILITY MEASUREMENTS .................. .. 151

6.1 Saturation Spectroscopy .................. ....... .. .. 151
6.2 Experimental Setup ..... ............. 153
6.2.1 Laser Stabilization Using Modulation Transfer Spectroscopy 154
6.2.2 Laser Stabilization Using Frequency Modulation Spectroscopy 160
6.3 R results .. . .. .. 160
6.3.1 RFAM Noise .................. ............ .. 161
6.3.2 AOM Noise .................. ............ .. 163
6.3.3 C!i i, i, ig C'! 1'1.. i Frequency ................... ... 163
6.3.4 Cesium Cell Temperature C!_i ,i_,, -; .................. .. 163

7 LISA TELESCOPE PROTOTYPE DESIGN ...... ........... 166

7.1 Introduction ...................... ........... 167
7.2 Telescope Design ................... ......... 168
7.3 Telescope Fabrication .................. .......... 172
7.4 Vacuum Tank Design .................. ........... 182
7.5 Experimental Setup .................. ............ 185
7.6 Results ................... ............... 187


8.1 Sagnac Effect ................. ...... 193
8.2 Polarization-Maintaining Optical Fibers .................. 195
8.3 Experimental Setup .................. ............ 197
8.3.1 Qualitative Description .................. .... 197
8.3.2 Quantitative Description .................. ... 200
8.3.3 Voltage to Phase Calibration ................. .. 203
8.4 Stability Results .................. ............. .. 205

9 CONCLUSION .................. ................. .. 212

9.1 Hydroxide Bonding .................. ............ .. 212
9.2 M material Stability .................. ............. .. 213
9.3 Cesium Locking .................. .............. .. 216
9.4 Telescope .................. ............... .. .. 216
9.5 Back-Link Fiber .................. ............ .. .. 217


A THERMAL SHIELD DESIGN .................. ........ .. 219



REFERENCES .................. ................ .. .. 229

BIOGRAPHICAL SKETCH .................. ............ 237




















Selected material properties and their values for Zerodur .

Zerodur class value and corresponding CTE......... .....

Compositional values of Invar 36, 42, and 32-5....... .....

Comparison of selected properties for 3 Invar alloys .

Comparison of density and porosities for selected SiC materials .

Comparison of flexural and tensile strength for selected SiC materials. .

Comparison of thermal conductivity for selected SiC materials .

Selected properties and their values for the k-180 piezoelectric material..

Comparison of properties for select two-part epoxies ......

Parameters of the three Zerodur cavities constructed by either optically c
or hydroxide bonding the mirrors .............. .....

Parameters of the Super Invar cavity with hydroxide bonded mirrors. .

Parameters of the SiC cavity with optically contacted mirrors .

Parameters for the CFRP cavity .............. .....

Parameters for the PZT cavity ............... .....

Diameters and heights of the thermal shield cylindrical shells .

Material properties of Macor, aluminized PET, and aluminum .

A-3 Resistance, capacitance, time constant, and cutoff frequency for each stage of
the therm al shields . ...... .











. 109

. 110

. 112

. 114

. 116

. .. 220

. .. 220


Figure page

1-1 Representation of the GRACE satellites in orbit above the Earth. Image courtesy
of NASA ...................................... 24

1-2 Representation of how the GRACE spacecraft move as they pass over a mass
anomaly. In (a), spacecraft 1 feels an acceleration due to the anomaly. After
spacecraft 1 passes the anomaly, it feels an acceleration in the opposite direction
of spacecraft 2 as shown in (b). After spacecraft 2 passes the anomaly, it feels
an acceleration in the opposite direction as depicted in (c). .......... ..24

1-3 Representation of the LISA orbit. Image courtesy of NASA. ......... ..27

1-4 Simplified view of the assembly (a) and the optical bench (b). ... 33

1-5 Configuration of the LATOR orbits, spacecraft, and interferometer on the ISS.
Image courtesy of NASA. .................. .. ...... 35

1-6 Heterodyne interferometry on one spacecraft with phase-locked local oscillator. 37

1-7 Fiber linked heterodyne interferometry and fiber metrology system (left) and
fiber linked heterodyne interferometry on two spacecraft (right). ... 38

2-1 Zerodur production process. ............... ......... 42

2-2 Zerodur annealing process ............... .......... 43

2-3 Process used to produce sintered SiC. ............... .... 50

2-4 Process used to produce reaction-bonded SiC. .................... .. 52

2-5 Diagram of how 3- and 4-point loading tests are done. ............. .55

2-6 a) Cubic lattice above the Curie point b) Tetragonal lattice below the Curie
point after the poling process has been applied. ................ 62

2-7 Electric dipole moments for the Weiss domains (a) before polarization (b) during
polarization and (c) after polarization. .................. .... 62

2-8 The piezoelectric effect of a cylinder for (a) no external action (b) an applied
compressive force (c) an applied tensile force (d) an applied voltage (e) an applied
voltage of opposite polarity. For clarity only 1 dipole is shown in (a). 63

2-9 Axes designation and directions of deformation. Directions 4, 5, and 6 are the
shear about axes 1, 2, and 3, respectively. ............. .. 64

3-1 Representation of how anodic bonding works. .................. 68

3-2 Proper spacing for brazing materials of differing CTE. ........... .73

3-3 Conceptual design of the GAIA SiC torus. Picture courtesy of ESA. ...... ..75

4-1 Spectrum of an unbonded CoorsTek SiC sample. The peaks from left to right
are the Auger carbon peak, Auger oxygen peak, oxygen Is peak, carbon Is peak,
and the silicon 2s and 2p peaks. .................. ..... 83

4-2 Kinetic energy distribution obtained due to the inelastic scattering of electrons.
The peak corresponds to the photoelectrons emitted near the surface, while the
signal at lower kinetic energies results from the scattering of the electrons deeper
within the sample surface. .................. ........... .. 84

4-3 Ols orbital both before (blue curve) and after (red curve) bonding. The dashed
and dotted lines represent the peak fit values while the solid lines are the raw
data taken from the XPS. The unbonded peak is at 532.61 eV and the bonded
peak is at 532.96 ev. .................. .............. ..86

4-4 Si2p3 orbital both before (blue curve) and after (red curve) bonding. The dashed
and dotted lines represent the peak fit values while the solid lines are the raw
data taken from the XPS. The unbonded peaks are at 102.87 eV and 100.60
eV. The bonded peaks are at 100.42 eV, 103.82 eV, and 102.21 eV. ...... ..87

4-5 Cis orbital both before (blue curve) and after (red curve) bonding. The dashed
and dotted lines represent the peak fit values while the solid lines are the raw
data taken from the XPS. The unbonded peaks are at 282.79 eV, and 285.12
eV. The bonded peaks are at 285.48 eV, 2 : -2' eV, and 282.11 eV. ...... ..88

4-6 Lever arm apparatus used to initially test the hydroxide bonded shear strengths
(left) and MCST apparatus used for later tests (right). 90

4-7 Breaking strengths of BK7-BK7 bonds after being kept in an oven at 330 K.
The zero on the x-axis represents the time after the samples were in the clean
room for approximately 18 hours. All data points were taken using the lever
arm apparatus .................. ................. .. 92

4-8 Averages of the data obtained using the lever arm and modified compression
shear test (\!CST) devices for both a 1:4 and 1:6 dilution factor .... 95

4-9 Surface profiles of areas on the POCO SuperSiC using a Tencor surface profiler.
Image courtesy of Petar Arsenovic at Goddard Space Flight Center. ...... ..99

4-10 Surface profile of one of the large CoorsTek SiC tile using a Tencor surface profiler.
Image courtesy of Petar Arsenovic at Goddard Space Flight Center. ...... ..100

4-11 Surface profile of one of the small SiC pieces using a Tencor surface profiler.
The large spike around 750 pm is most likely a speck of dirt. Image courtesy of
Petar Arsenovic at Goddard Space Flight Center. ..... 100

5-1 Electric fields inside and outside of an optical cavity. ............. ..104

5-2 Plot of the reflected transfer function for a lossless cavity 30 cm long with both
mirrors having reflectivities of r 0.90. The FSR is 500 MHz. .... 105

5-3 The resonator g parameters. .................. .......... 106

5-4 Geometry for analyzing the misalignment of a cavity. .............. ..107

5-5 Completed SiC cavity made of Hexoloy SA. ................ 111

5-6 WYKO images of the polished ends of the SiC spacer. Images produced by Surface
Finishes Co., Inc .................. ................ .. 112

5-7 Fabrication flowchart of the CFRP spacer at the University of Birmingham. 113

5-8 CFRP spacer, Zerodur/mirror plugs, and the alignment jig (left). Finished CFRP
cavity after epoxying the plugs (right). .................. .... 114

5-9 Representation of the PZT-actuated cavity. Note the lengths are not to scale. 116

5-10 Diagram of the PDH locking technique. .................. .... 117

5-11 Two error signals produced from cavities of different finesse. The green curve is
for a cavity of length 200 mm and finesse 500. The red curve is for a cavity of
the same length but finesse of 5000. Both have EOM modulation frequencies of
50 M Hz ..................... ........ ... ..... 121

5-12 Error signals for only the carrier frequency from Figure 5-11. The green curve
is for a cavity with a lower finesse. .................. .... 122

5-13 Electro-optical setup used to test the dimensional stability of select materials. .123

5-14 Time series of the beatnote taken over 28 d iv-. ................ 124

5-15 Noise spectrum of the Zerodur cavities with optically contacted mirrors. Measurements
were taken using the frequency counter, phase meter sampling at 15 Hz and 98
kHz. .................. .............. ... .. 125

5-16 Noise spectrum of the Zerodur cavities with optically contacted mirrors and the
LISA pre-stabilization requirement. The results are from a time series taken
over seven d ,v- and sampled at 1 Hz. .................. ..... 126

5-17 Plots of the Zerodur beatnote and tank temperature (left) and plots of the beatnote
and temperature data after being high pass filtered (right). .... 128

5-18 Comparison of the noise spectrums of the Zerodur cavities with optically contacted
and hydroxide bonded mirrors. .................. ........ 128

5-19 Plots of the SiC beatnote taken over a long weekend (left) and plots of the beatnote
taken over 11 d .,i-(right). ............... .......... 130

5-20 Noise spectrum of the Hexoloy SA SiC cavity with optically contacted mirrors
along with the LISA pre-stabilization and telescope requirements. ... 130

5-21 Inferred temperature stability at the center of Tank 1. ............. 131

5-22 Plots of the Super Invar beatnote and tank temperature (left) and plots of the
beatnote and temperature data after being high pass filtered (right). ..... ..133

5-23 Noise spectrum of the Super Invar cavity with hydroxide bonded mirrors along
with the LISA telescope and pre-stabilization requirement. .... 133

5-24 Inferred temperature stability inside Tank 2. ................ 134

5-25 Voltage amplifier used to provide the applied voltage to the PZT on the PZT
cavity. ..... .............. .................. .. 137

5-26 Noise spectrum of the PZT cavity with no applied voltage. Measurements were
taken using the frequency counter, phase meter sampling at 15 Hz and 98 kHz.
The spike around 20 kHz is most likely the resonance of the PZT ring. .... 139

5-27 Noise spectrum of the PZT cavity with 0 V, and 200 V applied to the PZT along
with the LISA pre-stabilization requirement. ................. 140

5-28 The slope of the beatnote over time with 200 V applied to the PZT. The time
series was taken approximately 10 minutes after the voltage was applied to the
PZT ....................................... .... 141

5-29 Arm-locking setup used to arm-lock the PZT cavity to a Zerodur cavity. The
feedback controls used to initially lock the lasers to the cavities are not shown.
PD: photodiode, PM: phasemeter, FM: frequency meter ............ ..142

5-30 Modeled open-loop transfer function of the arm-locking controller used to suppress
the laser noise ............... ............... .. 143

5-31 Noise spectrum of the PZT cavity without an applied voltage, the noise spectrum
of the PZT cavity using arm-locking, and the LISA pre-stabilization requirement. 143

5-32 Example of a step glitch (top) and spike glitch (bottom). .. ..... 144

5-33 Beatnote of the CFRP cavity in Tank 2 taken over 4 d-.- .......... 147

5-34 Stability of the CFRP cavity along with the LISA telescope and pre-stabilization
requirements ............... ............. .. 148

5-35 Dimensional stability of the CFRP cavity for differing incident intensities along
with the LISA telescope requirement. ............... .... 149

6-1 Intensity as a function of frequency for both an optical cavity and gas cell. .. 152

6-2 Experimental setup used for modulation transfer spectroscopy. ... 155

6-3 Diagram of the oven used to heat the cesium cell. ............... 156

6-4 Intensity profile as the laser frequency is scanned over multiple absorption lines. 157

6-5 Detailed intensity profile of three absorption lines. ............... 158

6-6 Temperature dependance on absorption of molecular cesium. .... 158

6-7 Error signals produced for a range of cesium lines. ............... 159

6-8 Experimental setup used for frequency modulated spectroscopy. ... 161

6-9 Stability results of the modulation transfer and frequency modulated spectroscopy
techniques compared to a free-running laser. .................. 162

6-10 Beatnotes of both the cesium-Zerodur and Zerodur-Zerodur locked systems. 165

7-1 Representation of (a) an on-axis Cassegrain telescope and (b) an off-axis Cassegrain
telescope . . .. 167

7-2 Isometric view of the telescope design. Figure courtesy of Joseph Generie at
Goddard Space Flight Center ............... ... 170

7-3 Payload assembly for one of the optical benches, on-axis telescope, and thermal
shield for one "tube" of the LISA spacecraft. Each spacecraft will consist of two
tubes. Figure courtesy of NASA .. ............ .... .. 172

7-4 Expected operating temperatures on the LISA telescope. Figure courtesy of Angelique
Davis at Goddard Space Flight Center, ................... ...... 173

7-5 Diagram of how the telescope parts fit using the alignment jig. For clarity, not
all of the strut clamps are shown. Courtesy of Jeff Livas at Goddard Space Flight
Center .. ....... .. .. ... .. .. .. .. ... .. .. ..... 175

7-6 Picture of workstation to determine the gap sizes between the struts and either
primary or secondary mirrors. ............... ......... 176

7-7 Gap size when not a strut is not properly aligned (left) and gap size when a
strut is aligned properly (right). In some cases a downward force was needed in
order for the gap to close. Gaps shown on the left side of the figure were typically
10 to 20 pm. .................. .................. .. 177

7-8 Picture of the sodium silicate solution filling in gaps between contact points of
the struts .................. ............. .. .. 178

7-9 Representation of how the struts bonded to the primary and secondary at a tilt.
The view of the secondary is as if one was looking up at the secondary from the
primary. The view of the primary is as if one was looking at it from the secondary.
The shaded areas are where the hydroxide bonds are thicker. .... 179

7-10 Picture of the secondary where the strut had bonded to it (left) and a picture
of part of a strut that had been bonded (right). The bonds appear to be much
thicker on one side, which leads to evidence that the structure twisted during
bonding. .................. .................. .. .. 180

7-11 Diagram of how the sister blocks were used to bond the strut and tile. .... 182

7-12 Representation of the system used to cool the telescope (left) and a picture of
the finished system (right). Not shown on the left figure are the two 'l- i~ of
aluminized PET. The aluminized PET shields are partially shown in the figure
on the right along with the SiC primary after the structure collapsed. ..... .184

7-13 Representation of how the in-band and absolute length noise for the telescope
will be found .................. .................. .. 186

7-14 Representation of how the Michelson interferometer was setup to measure the
distance change between the primary and secondary.. ...... 188

7-15 Photodiode output of the Michelson interferometer as the telescope was cooled
down (left) and a detailed view of the same output over a shorter period of time
(right) ...................... ........ ... .... ... 189

7-16 Detailed view of the oscillations in the Michelson interferometer data. ..... .190

7-17 CTE as a function of temperature for CoorsTek UltraSiC SiC. Data obtained
from David Bath at CoorTek. .................. ....... 191

8-1 Representation of the three LISA spacecraft, the optical benches, and the back-link
optical fiber .................. ............... .. .. 193

8-2 Diagram of a rotating loop interferometer to illustrate the Sagnac effect. Doppler
shifted frequencies are shown with respect to the inertial space. ... 194

8-3 Both the PANDA and bow-tie polarization-maintaining fiber -1, -; are shown
along with the core and stress elements. ................ ..... 196

8-4 Sagnac interferometer used to measure the accumulated differential phase noise
in an optical fiber. HWP1-3 are half wave plates, PBS1-3 are polarizing beam
splitters, QWP is a quarter wave plate, and the balanced photodetector is represented
by BPD...... ................... 198

8-5 Noise spectrum of the detector components of the interferometer. This was found
by placing a mirror in front of PBS1 and aligning the back reflected light. 206

8-6 Plot of the time series before and after modulating the frequency of the laser.
It is clear that modulating the laser frequency produces significantly less noise. 208

8-7 Time series of the input power to the interferometer (left) and the time series
of the powers during the same time in both the clockwise and counterclockwise
beams (right) .................. .................. .. 209

8-8 Time series of the balanced photodetector at the same time the powers of the
counterpropagating beams are measured. ................ ...... 209

8-9 Noise spectrum of the optical fiber with counter-propagating beams. ...... ..210

9-1 Comparison of the noise spectrums of the Zerodur cavities with both optically
contacted and hydroxide bonded mirrors, SiC, CFRP, and Super Invar along
with the LISA telescope requirement. .................. ..... 214

A-i Noise spectrum of the outside tank temperature. .... 219

A-2 Picture of the built thermal shields (left) and a model representation of them
(right). Note the model is not drawn to any scale. ............... 221

A-3 Lumped thermal model of the thermal shields. ................. 221

A-4 Transfer function of the PET thermal shields. ................. 223

B-l Typical surface profile of a SiC surface before cleaning with cerium oxide. .... 226

B-2 Typical surface profile of a SiC surface after cleaning with cerium oxide. .... 227


AOM Acousto-Optic Modulator

BE Binding Energy

CFRP Carbon Fiber Reinforced Polymer

CTE Coefficient of Thermal Expansion

CVD C('! i,, iI Vapor Deposition

DI De-ionized

DRS Disturbance Reduction System

EMRI Extreme Mass Ratio Inspiral

EOM Electro-optic Modulator

ESA European Space Agency

F Finesse

FFT Fast Fourier Transform

FSR Free Spectral Range

FWHM Full-Width Half-Maximum

GRACE Gravity Recovery and C('li ii,. Experiment

GRS Gravitational Reference Sensor

GSFC Goddard Space Flight Center

GW Gravitational Waves

HPF High Pass Filter

I.D Inner Diameter

IMS Interferometeric Measurement System

ISS International Space Station

KE Kinetic Energy

LATOR Laser Astrometric Test of Relativity

LIGO Laser Interferometer Ground Observatory

LISA Laser Interferometer Space Antenna

LPF Low Pass Filter

MCST Modified Compression Shear Test

ND Neural Density

O.D Outer Diameter

PD Photodiode

PDH Pound-Drever-Hall

PZT Piezo-electric transducer

ROC Radius of Curvature

S-D Scratch-Dig

SiC Silicon Carbide

SIM Space Interferometry Mission

SMBH Super-Massive Black Hole

XPS X-Ray Photoelectron Spectroscopy


c speed of light in vacuum

ID flux emerging without being scattered

Io flux of electrons originating at depth D
a Boltzman's constant (1.38065x 10-23 m2kg s-2K-1)

Tr(f) reflected transfer function as a function of laser frequency

Tt(f) transmitted transfer function as a function of laser frequency
Q modulation frequency of the electro-optic modulator

w laser frequency

Abstract of Dissertation Presented to the Graduate School
of the University of Florida in Partial Fulfillment of the
Requirements for the Degree of Doctor of Philosophy



Alix Preston

August 2010

('!C ir: Guido Mueller
Major: Physics

Space-based interferometric missions such as SPIRIT, SPECS, TPF, and LISA will

take measurements of the universe with unprecedented results. To do this, these missions

will require ultra-stable materials and bonding techniques to be used for critical optical

components such as optical benches or support structures. As an example, the telescope

support structure for the LISA mission must be made of a material that is stable to better

than 1 pm/ Hz at 3 mHz and whose length between the primary and secondary mirrors

cannot change by more than 1.2 pm over the lifetime of the mission. Although materials

such as silicon carbide (SiC) and carbon fiber reinforced polymers (CFRP) have been

-l...- -i. ,1 for this use, they have not been tested to these strict requirements.

The author is part of a group at the University of Florida that is developing a

materials research facility capable of testing the stability and strength of materials and

bonding techniques for use in space-based interferometric missions. This work describes

the techniques used to test the stability of materials at the femtometer level. Results are

presented for SiC, Zerodur, a carbon fiber reinforced polymer, Super Invar, and k-180

piezoelectric material. In addition, several bonding techniques are described along with

potential uses for space missions. Specific attention is paid to a promising new jointing

method known as hydroxide bonding. The bonding method, bonding mechanism, and

shear test strengths are presented for several material combinations. During this work,

it was found that hydroxide bonding could be used to bond SiC to SiC with significant

strength in a simple manner and was chosen to bond a prototype telescope support

structure for the LISA mission. Initial results and discussion about further improvements

to the design are presented.

Finally, investigations were done to determine the differential phase noise in

counter-propagating beams through an optical fiber. Results using a polarizing Sagnac

interferometer showed a differential phase noise of ~5x 10- cycles/vHz at 1 Hz. The

experimental setup, noise sources, and results are presented.


In an attempt to learn more about why we are here, where we came from, and

what the future holds for humankind, countless time and resources have been spent

on projects designed to answer these and other questions. Ground-based dark matter

and gravitational wave detectors search for signals coming from our universe to learn

how it has evolved over time. Giant telescopes map the sky for planets that may be

similar to our own. Although ground-based detectors can provide a wide range of

information about both the Earth and the universe we live in, noise sources such as

seismic activity or atmospheric turbulence make it difficult, or near impossible, to make

certain measurements from the ground. For this reason, space-based missions have become

increasingly popular in order to make measurements the Earth-bound detectors cannot.

Often space-based detectors work complimentary to their ground-based counterparts

to provide new and exciting information. While moving detectors into low-Earth orbits

and beyond can overcome many challenges faced on the ground, other difficulties such as

drag-free orbits or the inability to service components provide new challenges to overcome.

As space-based missions are becoming ever more complex, the sensitivity of necessary

measurements has increased as well. This requires the use of ultra-stable materials,

oscillators, and lasers that push their achievable limits. Several materials such as silicon

carbide (SiC) and Super Invar have been used for critical optical components on several

space-based missions, but their dimensional stability has not been studied to determine

if they are suitable for the next generation of detectors. Although this work primarily

focuses on determining the stability of materials and bonding techniques for their use

on the Laser Interferometer Space Antenna (LISA), it can be used as a reference when

designing other critical optical components such as optical benches or telescope support

structures for future space-based missions

1.1 Space-Based Missions

There are a variety of proposed space-based missions that will study everything

ranging from the Earth's climate to gravitational wave sources. Each mission consists of

multiple subsystems that work together for a common goal. Discussion of each mission

and its components would be unfeasible in this work, so only a select few will be discussed

that could potentially benefit form work presented in this dissertation.


The Gravity Recovery and C('liii I, Experiment (GRACE) maps various.-. ,!i--1i I1

processes that produce gravity anomalies [1]. Certain -, .!r,-i. 1 processes generate

gravity anomalies over the surface of the Earth. Measuring these anomalies provides a

better understanding of the structure of the Earth. Short-term mass fluctuations such

as a variation in the water content of the Earth's crust, global sea level changes, or

polar ice sheet balance can be measured with GRACE to determine their impact on the

global climate. Data collected from GRACE has provided the scientific community with

information that has led to a substantial improvement in our understanding of how the

Earth changes over time [2].

GRACE was launched March 17, 2002 and consists of two identical satellites in a near

circular orbit ~500 km above the Earth's surface (see Figure 1-1). The distance between

the satellites is ~220 km along-track and linked by a K-band microwave ranging system.

In addition, each satellite carries Global Positioning System (GPS) receivers, attitude

sensors, and high precision accelerometers. The satellite orbits decay at ~30 m/d iv such

that the satellites do not have a fixed repeat pattern which reduces its measurement

sensitivity. The satellites are 3-axis stabilized such that the K-band antennas are pointed

at each other [1].

In order to produce the necessary precision, several post-processing techniques are

used. To reduce the effects caused by the ionosphere, the dual-frequency one-way K-band

Figure 1-1. Representation of the GRACE satellites in orbit above the Earth. Image
courtesy of NASA.

......... Ellipsoid l.-

a b

Figure 1-2. Representation of how the GRACE spacecraft move as they pass over a mass
anomaly. In (a), spacecraft 1 feels an acceleration due to the anomaly. After
spacecraft 1 passes the anomaly, it feels an acceleration in the opposite
direction of spacecraft 2 as shown in (b). After spacecraft 2 passes the
anomaly, it feels an acceleration in the opposite direction as depicted in (c).

phase measurements transmitted and received by each satellite are combined during

ground processing which removes many of the effects caused by oscillator instability.

Data gathered from the precision accelerometers is used to remove non-gravitational

forces on the satellites. The GPS receivers allow precision time- .--ii:-; of when the data

was taken and is used to determine the inter-satellite distance. Precision estimate of

the inertial orientation is provided by the attitude sensors. The combination of these

measurements along with ancillary data is used to produce a representative gravity field of

the Earth.

The data gathered from the GRACE mission has led to a better understanding of

variations of mass in the Earth, exchanges between ice sheets and the oceans, ground

water storage on land masses, changes due to deep ocean currents, and a detailed map of

the Earth's gravity field [3].

1.3 GRACE Follow-On

Although GRACE continues to provide a wealth of information about continental

water storage, the Earth's geoid, global sea level changes, fluctuations of the polar ice

sheets, and other .1 |1,!1 -ii processes, a follow-on mission has been proposed that

would increase the sensitivity and provide even more information about the Earth's

dynamics[4]. These measurements include the time varying ocean bottom pressure,

monitoring groundwater and polar ice caps, and measuring the lithosphere thickness. To

make these measurements, the proposed follow-on mission would consist of two coplanar

spacecraft in circular, polar, orbits at an altitude of 250-600 km. The spacecraft would

be small satellites approximately 150-200 kg and would require about 100 W of total

power and would have a baseline lifetime of 5 years, with a goal of 10 years. The distance

between the spacecraft would be loosely maintained to 50-100 km and measured using

laser interferometry. MicroNewton FEEP-type thrusters will be used to provide a drag-free

control of the spacecraft in addition to ground track repeating within a few kilometers [4].

For the GRACE follow-on mission to work, the laser interferometry must be capable

of precision phase extraction with lasers having a frequency stability of better than 1 part

in 1016 rms over 1000 seconds. In addition, the spacecraft will need an internal sensor or

.... I. i.11 i i capable of measuring non-gravitational accelerations on the spacecraft

to better than 1 part in 10-13 m/s2 rms over 1000 seconds. It is important that the

inertial sensor be fully integrated with the laser interferometer in order to reduce errors

relating the interferometer reference point to the inertial reference point. While there are

several designs that could be implemented, a method using freely-falling proof masses on

each spacecraft along with heterodyne interferometry containing polarizing optics as a

metrology system appears to be promising [4]. First performance laboratory measurements

of a spacecraft-to-spacecraft interferometry design using a metal breadboard have shown

a sensitivity of 2.5 nm//Hz [4]. Significant improvements in the stability results could

be achieved using an optical bench made of a glass-ceramic such as Zerodur and using

advanced bonding techniques, such as hydroxide bonding, for the optical components.

1.4 LISA

1.4.1 Introduction

To paraphrase John Wheeler, spacetime tells sources how to move, and moving

sources tell spacetime how to ripple. In 1975 Russel Hulse and Joseph Taylor discovered a

binary pulsar whose orbit decay over time was in accordance with what general relativity

predicts from the emission of gravitational radiation [5]. Recently, the outburst timing

of two black holes in the OJ 287 system has provided more indirect evidence that

gravitational waves (GW) exist [6], but have yet to be directly detected despite efforts

from ground-based gravitational wave detectors such as LIGO and VIRGO. Although

ground-based GW detectors have yet to find a signal, the evidence for them is compelling

enough to prepare for space-based detectors that will search for complimentary GW

sources [7],[8]. One such mission is known as the Laser Interferometer Space Antenna

(LISA). LISA is a joint NASA/ESA mission that consists of 3 spacecraft that will form

a triangle whose arms are separated by 5x 09 m. LISA will trail the Earth by 20,

and is inclined to the ecliptic plane by 600. This orientation will allow LISA to provide

angular resolution of sources to about one degree, with sub-degree resolution for many

strong sources when averaged over time periods of 6 to 8 months. LISA will detect

gravitational radiation from super-massive black hole (SMBH) mergers, galactic binaries,

and extreme mass-ratio inspirals (EMRI's) in the 30 pHz to 1 Hz frequency band with

a strain sensitivity of 10-21/vHz at its most sensitive frequency [9]. Since it is thought

that the GW sources detectable by LISA ph-i, an important role in determining how the

Figure 1-3. Representation of the LISA orbit. Image courtesy of NASA.

early universe formed, LISA will effectively provide us with a glimpse into the past, and

subsequently study the structure and evolution of the universe.

1.4.2 Sources

The LISA sensitivity band is expected to be an extremely rich portion of the

gravitational wave spectrum. Observing these gravitational waves will provide an

insight into the dynamic processes that occur in .,-I i -.1 dr, -i since our view of the

universe has been built almost entirely on measurements made using the electromagnetic

spectrum. Detection of gravitational wave sources has been a difficult task because the

field generated by gravitational radiation, given by hij ~ quad V where Mqpad is the

portion of the source's mass that participates in quadrupolar motions, is so small [10].

For any appreciable signal that has hope of being detected, the source must be extremely

massive and compact enough to move at relativistic speeds. Although this does somewhat

limit the sources available for detection, there are still plenty of sources that will provide a

plethora of information about the universe we live in. Cosmological background sources

Of significant interest to the scientific community are the stochastic sources that

are cosmological in origin. Of these sources, the gravitational radiation produced from

the first moments of the Big Bang are the most eagerly sought out. Detection of this

background radiation would provide insight into how the universe developed much

further back in time than measurements of the microwave background can produce since

gravitational waves travel undisturbed to us and carry all of their information from the

past. Unfortunately these waves will not be detectable by the LISA mission due to their

extremely small amplitude. Although this source may be out of reach for LISA, there

are other potential cosmological backgrounds that may be within LISA's grasp [10]. For

example, when the universe was roughly 1 TeV, the electroweak interaction separated

into electromagnetic and weak interactions. Since this separation was not uniform, some

regions transitioned before others, producing a background [11],[12]. It has also been

-,-.;.; -. .1 through superstring theory that a network of cosmic strings produced in the

early universe that have expanded to cosmic size could produce backgrounds detectable by

LISA [13]. Although there is an extremely small chance of LISA finding GW background

signals, the insight provided into out universe will be tremendous and worth looking for. Binary stars

Within the Milky Way alone it is expected that there are tens of millions of compact

binaries that will generate GW's in the LISA band [10]. Most of these are white dwarf

binaries which will slowly spiral towards each other as they radiate gravitational waves.

These objects are extremely important to the LISA mission because they are guaranteed

sources. There are so many of these sources that cannot be distinguished from one

another that there is an expected confused background that will be detected in the LISA

band. The amplitude of the noise will be of significant importance in determining the

distribution of stars within our galaxy [14].

In addition, several binary systems that have already been observed electromagnetically

should produce GW's in the LISA band large enough to be distinguished from the

background. These sources will provide instant verification of the LISA instrumentation[15]. Chirping sources

As objects in a binary system emit gravitational radiation and spiral towards each

other, their orbiting periods will decrease and slowly "chirp" upwards. Since the LISA

band is at such a low frequency, the required mass for this to happen must be extremely

large [10]. The most important chirping sources will consist of systems where both masses

are black holes. Black hole binaries with masses ranging from 104 to 107 solar masses and

mass ratios ranging from 1/20 to around 1 will generate gravitational waves directly in

the most sensitive spot of the LISA band. The signals will sweep across the LISA band

in times ranging from several months to several years. Finding these signals early will be

invaluable. With enough warning, telescopes can be positioned to look at that patch of

the sky where a gravitational wave is thought to have come from. In this manner both an

electromagnetic and gravitational spectrum can be found and provide a more complete

understanding to the constituents and interactions of black holes. Even with pessimistic

formation rates, at least a few super massive black hole coalescence should be observed

over LISA's lifetime, while more optimistic views put this number in the hundreds [16]. Extreme mass ratio inspirals (EMRIs)

EMRIs are binary systems in which the smaller member (such as a stellar mass

compact object) orbits a much larger object (such as a galactic core black hole). Such

systems are created through scattering process in the core of galaxies in which the

larger body captures the smaller body into a highly-eccentric strong-field orbit. Current

estimates -i--.- -1 hundreds of these sources could be detected by LISA each year [17].

The manner in which the smaller body spirals into the larger body will tell a great deal

of information about the binary's spacetime and essentially map out the gravitational

potential of the larger body.

1.4.3 General Measurement Overview

The basic measurement to detect GW sources will be by monitoring the distance

between proof masses on .ildi i:ent spacecraft. Within each spacecraft are two cubical

free-flying proof masses. The spacecraft use capacitive sensing and micro-Newton

thrusters to maneuver around the proof masses as to not disturb them. There are three

measurement arms between the spacecraft, each of which consist of two links. The links

represent a one way travel of the laser from one spacecraft to another. Picometer level

changes between the 5x 09 m proof mass separation will provide the necessary strain

sensitivity of ~10-21 [18].

The distance changes between the proof masses will be measured using laser

interferometry. Light at a wavelength of 1064 nm will be sent from one spacecraft to a

distant spacecraft where it will be interfered with a small amount of light from the local

spacecraft. The precision of this measurement relies on measuring the phase of the radio

frequency (RF) beatnote as recorded by a digital phasemeter with microcycle accuracy.

As a gravitational wave passes through the LISA constellation, the arm lengths will be

modulated. The resulting changes in the beatnote will be recorded and sent back to Earth

where it will be post-processed. The ability to reach the required strain sensitivity will

depend on the arm length, the accuracy to which the relative displacement between the

two proof masses can be measured, and the ability to which the proof masses can be

shielded from disturbances.

Taking these factors into account, the LISA spacecraft design has been driven by

two systems known as the Disturbance Reduction System (DRS) and the Interferometry

Measurement System (IMS). The purpose of the DRS is to keep the proof mass as

undisturbed as possible, while the purpose of the IMS is to measure the distances between

the proof masses using optical interferometry. Both systems work together in a manner to

reach the required sensitivity.

1.4.4 The Disturbance Reduction System (DRS)

The DRS consists of a set of local sensors, actuators, and controls that act together

to meet the disturbance and pointing requirements [18]. The heart of the DRS is the

Gravitational Reference Sensor (GRS) which houses the proof mass. The spacecraft act

to shield the proof masses from disturbance forces such as solar winds or cosmic dust. If

left unattended, these forces would cause the spacecraft to orient itself in a way that would

make measurements impossible. To avoid this, the GRS -, i-, -" where the proof mass is

relative to the spacecraft. MicroNewton thrusters provide the actuation needed in order

to position and orientate the spacecraft in order to provide drag-free operation as well as

provide the necessary pointing of the outgoing laser beams towards the distant spacecraft.

The low frequency end of the LISA band will be limited by the acceleration noise of

the proof masses and spacecraft which are directly dependant on the performance of the

DRS. The noise allocation for the DRS given as an amplitude spectral density for a single

proof mass is given by [18]

<3X10-15 (0.1m ) (11)

To convert Eq. 1-1 into a displacement noise, it is integrated twice. Taking into account

that there are four proof masses in each Michelson-type interferometer and that all of the

noise is uncorrelated will add a factor of two to the total noise. This results in [18]

S 2 < 6 x 10 r 1 +1 t ( mHz.
R (27 f)2 Hz (27 f)2 1 8mH+ f

This error budget is made up of subsystems, each with their own noise sub-allocation.

Such noise sources include electrostatics, Brownian motion of the residual gas around the

proof mass, and thermal effects among others. A detailed acceleration noise budget using

all known disturbance effects is given in reference [19].

Significant research has been done to characterize the noise sources to provide a

better understanding of how the proof masses will behave on the LISA mission [20]. Much

of this work typically involves suspension of the proof mass from a torsion pendulum to

model and measure expected noise sources. In addition, the performance of the GRS will

be directly tested when it is launched on-board the LISA Pathfinder mission [21].

1.4.5 The Interferometric Measurement System (IMS)

The IMS is the part of the measurement system that measures the distances between

proof masses. Since the proof mass orbits vary by about 1 over the course of a year,

the interferometer arms are neither equal nor constant. Because of this, the raw data

measurements taken on board must be post-processed in order to extract the gravitational

wave signals. The measurement performance of the IMS are derived directly from the

single link sensitivity [22]. In order to determine this sensitivity, knowledge of the system

response to a gravitational wave must be known. This response can be calculated using

the sums and differences of the phases measured in each of the six links as they are

combined in post-processing with the appropriate d-1i-' [18]. From the single link strain

sensitivity, the IMS measurement performance can be derived and is given by [18]

SsM(f/) < 12 x 10o127 i(2 j (1-3)

Within this total error budget are subsystems with their own noise sub-allocation. For

instance, the telescope pathlength stability has an allocation of 1 pm/vHz, while the

optical bench dimensional stability has an allocation of 4.5 pm/v/Hz [18]. All IMS

sub-allocations have the same frequency dependance as in Eq. 1-3. AM i.: of the imposed

requirements are currently pushing the limits of what is possible and it is imperative that

all subsystems and their components be tested in order to determine their stability and

 Optical bench

The optical bench consists primarily of three interferometers needed to make the

proof mass to proof mass distance measurements. In addition, it also contains other

components to perform ancillary functions such as monitoring the outgoing laser power or

CCDs to allow establishment of the laser links.

The light-weight baseplate is made of Zerodur and provides the dimensional stability

of all other path lengths. Hydroxide bonding is used to mount the fused silica optical

components on the polished Zerodur baseplate [23]. The bench will be isostatically

mounted to a large interface ring that the GRS is also mounted to.

Light To/From
Other Spacecraft
Optical Bench Interface Ring Other Spacecr
Optical Bench

Light From Light From
Telescope System Support Struts Laser System Other Bench

a b

Figure 1-4. Simplified view of the assembly (a) and the optical bench (b).

Light from the laser system is delivered to the optical bench through the use of a fiber

optic cable. The ini i Pi i ly of this light is then sent to the telescope, with the remainder of

the light used as local measurements. The separation of the incoming and outgoing beams

from .,l.i ient spacecraft is done through the use of a polarizing beam splitter. In order to

make the necessary measurement beatnotes, light from one optical bench needs to be sent

to the other, and vice versa. To do this, an optical fiber will be used that contains both

beams in counter-propagating directions. The transmitted power entering the telescope

from the optical bench is about 2 W, while the total power received from the .1li i'.ent

spacecraft is on the order of 200 pW [23]. Telescope system

Of particular importance for the telescope is to minimize and control the stray light

sources. The most convenient setup to minimize stray light is to use an off-axis two-mirror

telescope. Both the primary and secondary would be made of Zerodur and separated by

a carbon fiber reinforced polymer (CFRP) spacer. The primary and secondary apertures

are ~400 mm and ~30 mm, respectively, and are connected through a ~590 mm CFRP

spacer [23]. The telescope design also includes a re-focusing mechanism to correct for the

differences in alignment on ground and in flight due to the outgassing of the CFRP spacer.

Another possible configuration would be to use an on-axis design made of SiC. The

secondary would be connected to the primary through the use of SiC struts. Although

the struts will produce additional stray light in the on-axis design, the effects need to

be studied more before a decision on the baseline design can be made since the primary

source of stray light will be due to back-reflections from the secondary. Further discussion

of this subsystem is presented in C(i lpter 7.


The Laser Astrometric Test Of Relativity (LATOR) is a proposed mission to measure

the non-Euclidean features of a light triangle with sides passing close by the Sun. The

goal is to probe General Relativity and other alternate theories of gravity by measuring

the deflection of light as it passes close to the Sun [24]. To do this, two spacecraft will be

placed in orbit around the Sun in addition to using an interferometer on the International

Space Station (ISS). Light from both spacecraft are sent to the other spacecraft and to the

ISS. A phase measurement of the interfered light will determine the internal angles of the

spacecraft-spacecraft-ISS triangle.

The angle measured by the spacecraft can then be compared to the angle computed

using Euclidean geometry. A deviation in this angle will provide a wealth of information

LATOR Mission: J
Laser Arometric Test of Rativty
spacecrbut t seon ~- 2 AU f t0 mili n km

D,-~ 5 mileonkm Measured
S3lengths [tI, t, & I angle 19
r Acouralegs needed:
Target Distance. 1
spaoeoraff A. n :0.1 pirad

Figure 1-5. Configuration of the LATOR orbits, spacecraft, and interferometer on the ISS.
Image courtesy of NASA.

about the second post-Newtonian features of gravity, the mass parameter of the Sun

(the spatial metric's I 1,ii, I,'--' ), as well as measuring the solar gravitational field to

unprecedented levels [25].

1.5.1 Mission Design

There are numerous subsystems that will all work together in order for the LATOR

mission to be successful. Only a select few of these subsystems pertaining to this work will

be discussed. For details on these and other subsystems, see references [24], [25].

1.5.2 Spacecraft orbits

To measure the deviation from Euclidean geometry as the light passes through the

Sun's gravitational field, the LATOR mission will place two spacecraft in a heliocentric

orbit on the opposite side of the ISS. The spacecraft are separated by about 1 when

viewed from the ISS [26], and orbit with a 3:2 resonance with the Earth as depicted in

Figure 1-5. The 3:2 resonance occurs when the Earth does three revolutions around the

Sun, while the spacecraft does exactly two. The exact period of the orbit will vary slightly

from a 3:2 orbit by less than one percent depending on the time of launch. The advantage

of choosing this orbit are that it imposes no restriction on the time of launch, it will only

require small propulsion maneuvers, it provides a very slow change in the Sun-Earth-Probe

angle, and an additional maneuver will allow for a small orbital inclination that will enable

measurements at different latitudes.

1.5.3 Optical design

In order to make the necessary angle measurements, the ISS will need to be equipped

with three 20 cm diameter telescopes. One telescope is designed with a narrow bandwidth

laser line filter in front of it and an InGAs camera at its focal plane that is sensitive to

the 1064 nm light that will be used and serves as the acquisition telescope to locate the

spacecraft near the Sun.

The second telescope emits the directing beacon to the other spacecraft. A pair of

piezo controlled mirrors placed on the focal plane directs the light into and out of the

telescope. The collimated light with a power of ~10 W is injected into the telescope focal

plane and properly directed by the piezo mirrors [25].

The third telescope is used for the laser light tracking interferometry and will be used

to measure the relative length changes of the spacecraft to an accuracy of less than 10 pm

using heterodyne interferometry [24]. Two piezoelectric X-Y-Z stages will be used to move

the tips of two single-mode fibers while maintaining focus and beam position on the fibers

and other optics. Dithering at a few Hertz will provide the means to make the necessary

alignment of the fibers and subsequent tracking of the spacecraft completely automatic.

Both spacecraft are identical and contain a 1 W cavity fiber-amplified laser whose

wavelength is 1064 nm and drifts at ~2 MHz per hour. Most of the power is pointed to

the ISS through a 20 cm diameter telescope whose phase is tracked by an interferometer.

The remaining power is directed to the other spacecraft through the use of another

telescope. In addition to the two transmitting telescopes, each spacecraft is equipped

with two receiving telescopes. The telescope that faces the ISS has a laser line filter and

knife-edge coronograph to suppress the light from the Sun to 1 part in 104 of the light

level of the light received from the ISS [25]. The receiving telescope that points towards

the spacecraft does not require a filter and coronograph.

In addition to the four telescopes, the spacecraft will also carry a small telescope with

a CCD camera that will be used to provide the initial alignment of the spacecraft such

that they point at the ISS. Another similar telescope could be installed at right angles to

the first to determine the spacecraft attitude using known, bright stars. Star trackers of

this design have been used for several years and are readily available.

1.5.4 Interferometry

Interferometry will be used to make the precision angle measurements needed. A

simplified schematic of how this can be done is shown in Figures 1-6 and 1-7. In Figure

1-6, two sidereostats are pointed at a target and two fiducials (represented by the corner

cubes) determine the end points of the interferometer baseline.



--- --. Siderostat

LO1 p1 2 L02

Phase-locked LOs

Figure 1-6. Heterodyne interferometry on one spacecraft with phase-locked local oscillator.

Stable local oscillators are interfered with the light from the two arms, and the phase

difference is recorded. If the local oscillators were phase locked, then the angles of the

-,./ ..... / ~_ / ... /

Ba/ ine B / hne

LO1 Fiber Metrology
S211 w12 '21 022

Figure 1-7. Fiber linked heterodyne interferometry and fiber metrology system (left) and
fiber linked heterodyne interferometry on two spacecraft (right).

target with respect to the baseline normal is given by [25]

0 = arcsin (2Wn bl 2a) (1-4)

where A is the wavelength of the laser, n is an unknown integer arising from the fringe

ambiguity, and b is the baseline length [24]. Unfortunately, is extremely difficult to phase

lock the two local oscillators of such a long baseline with the precision needed, so another

method was proposed that uses a single mode fiber. The signal from the local oscillator

can be transmitted to both sidereostats through a single mode fiber. In this configuration,

additional metrology is needed in order to monitor the changes in the path length of the

fiber as the local oscillator propagates through it as shown in the left side of Figure 1-7.

The metrology system measures the distance from one beam splitter to the other and the

resulting angle is given by [25]

0 = arcsin ((2n + ,1 02 + m) (5)
O arcsin )(15)
\ 272b

where mi is the phase variations introduced by the fiber. When an additional arm is

added to this configuration the phase variations due to the fiber are common to both

arms. The differential angle then becomes [25]

Sarcsin ((27(ni n2) + (mi m2))A + ((11 12) (21 22))A6)
Sarcsin (1 6)

This is only a general overview of how the angle would be measured between the two

spacecraft, and in practice the interferometer will have optical paths that are different

lengths and will need to be measured with additional metrology systems.

In order to make the angle measurements, a long baseline will be introduced on the

ISS. Optical packages will be integrated on the ISS to the S6 and P6 truss segments which

are located at opposite ends of the ISS. Both packages will be separated by ~100 m and

will have a clear line of sight path between the two transceivers during the observational

periods in addition to having a clear line of sight to both spacecraft. Location on the ISS

is ideal since a limited amount of Sun tracking that is needed to point the telescopes on

the ISS to the spacecraft can be done using the a-gimbals on the ISS [25].

Two metrology systems will be used on the ISS to make the angular measurements.

The first system measures the optical path difference between the two laser signals while

the other system measures the changes in optical path through the fiber. Michelson

interferometers will be used in order to make the necessary metrology measurements.

Both spacecraft signals will be measured simultaneously using an electro-optic modulator

and modulating at a different frequency for each beam. The metrology signals of the

fiber will be used in the post processing to determine the angular measurements. The

metrology signals sent to the spacecraft from the ISS will be stabilized to better than

one part in 1101 using a temperature controlled Fabry-Perot etalon that is locked using a

Pound-Drever scheme [25].

The current design for LATOR will provide a wealth of knowledge about the

limits of general relativity. Moving from a A = 10 cm (the current standard) to a A

= 1 pm wavelength will gain a factor of 1010 reduction in the solar plasma optical

path fluctuations [25]. Taking advantage of current drag-free technology for use on the

spacecraft will provide ultra-precise orbit determination since the spacecraft will be

insensitive to effects from solar wind and radiation pressure. Making use of existing laser

and optical technologies will keep costs low. The LATOR mission will also benefit from

lasers and optical telecom that have been developed for the Space Interferometry Mission

(SIM) [25]. The technology needed for the LATOR mission has already been demonstrated

and provides a solid conceptual foundation. The LATOR mission is unique in concept and

design, and will lead to robust advancements in the understanding of fundamental physics.


Satellites for space-based missions are made up of multiple subsystems that all

work together to make complex measurements. Often there are strict requirements on

subsystems, such as the optical bench or telescope mirrors, that must be met in order for

the design sensitivity to be achieved. The choice of materials then becomes an important

consideration. In some cases, a material with an extremely low CTE must be chosen

in order to reduce thermal effects. Although there are materials available with CTE's

on the order of 10-s/K, they may exhibit other mechanical or thermal characteristics

which are unwanted, such as a low thermal conductivity, ferromagnetic behavior, are

brittle, or have a low Young's modulus. To fully understand a material and its possible

applications, the thermal and mechanical properties, along with the production process

should be known. For this reason, an introduction to a select, few materials that are

either commonly used, or have great potential for space-based interferometric missions

are presented. The production process, mechanical and thermal properties, and current or

potential applications are discussed.

2.1 Zerodur

Zerodur is a glass ceramic made by SCHOTT whose CTE is extremely low around

room temperature. It is characterized by a phase of evenly distributed nano-crystals

within a residual glass phase. Zerodur contains about 70-78 weight percent of a ( i i ii.1!11

phase with a negative linear thermal expansion. The remaining glass phase consists of

a positive linear thermal expansion. Together, these phases act against each other to

produce an overall CTE that is very close to zero within a certain temperature range,

usually centered around 25C. General properties of Zerodur are presented in Table 2-1

(it should be noted that these values are for room temperature and are typical for a broad

range of Zerodur classes and are not absolute) [27].

Table 2-1. Selected material properties and their values for Zerodur
Property Value
Density 2.53 g/cm3
Modulus of Elasticity 13x 106 psi
Thermal Conductivity 1.43 W/mK
Specific Heat 0.80 cal/g C
Index of Refraction 1.54

2.1.1 Production Process

The production process of Zerodur is outlined in Figures 2-1 and 2-2. The Zerodur

is cast into an annealing oven where the temperature is cooled in a controlled manner

for several weeks. The outer 1. r-i, of the material are removed in order to prevent

uncontrolled ceramization during the next steps. During the ceramization process, the

Raw Materials Batch Preparation Melting Casting

Quality Control (QC) Quality Control (QC)C ~ .*.rII-. _

t all .111 CTE QC Final QC

Figure 2-1. Zerodur production process.

material is heated up again to achieve growth of the (i ~ i 1 phase in a controlled manner.

The heat treatment process has a critical effect on the CTE homogeneity of the blank and

can last several months depending on the Zerodur material needed. Once this step has

been completed, the Zerodur can be processed into its final shape. CTE samples are taken

directly from the final material to ensure the final CTE value and homogeneity. Depending

on the size and shape of the processed parts, the homogeneity of the Zerodur CTE can be

kept to less than 0.03x10-6/K per 18 tons of material [28].

a 1500-
S1000- Devitrification
S\ Nucleation
500- Glassy Annealing Heat
State Treatment


Figure 2-2. Zerodur annealing process.

2.1.2 CTE

Zerodur is optimized for application use from 0-50C, with a CTE zero crossing that

can be tailored within a few degrees around 25C [28]. The CTE of the Zerodur is derived

using the length change measurements from the use of a dilatometer as the material is

heated up or cooled down. Depending on the Zerodur class specified, the variation in

the mean CTE can be as little as 0.02x10-6K-1 for temperatures ranging from 0-50C.

In certain fabrication processes, this variation can be even smaller [28]. The excellent

Table 2-2. Zerodur class value and corresponding CTE.
Zerodur expansion class CTE
0 0 0.02x 10-6K-1
1 0 0.05x 10-6K-1
2 0 0.10x10-6K-1

homogeneity and extremely low CTE have made Zerodur an optimal material for many

telescope mirrors. Telescope blanks for Keck I and II, Grantecan, and HET were produced

using a specifically fabricated Zerodur [28].

2.1.3 Thermal Cycling

Zerodur is designed to operate in a temperature range of 0-50C. Outside this range,

the CTE values can change dramatically. In addition, thermal cycling outside its operating

temperature can produce a range of effects. For temperatures greater than 700C, the

material properties are irreversibly changed and will be determined by the quenching

process. For a temperature cycling in a range between 130-320C, the total Zerodur length

will be changed, in addition to having its CTE zero crossing changed. For most mirror

applications, elevated temperatures are encountered when coating the surface. These

processes are short in time and have shown to have only negligible effect on the surface

figures. Thermal cycles for temperatures between -70 and 40C show a hysteresis in the

CTE and will shift the CTE zero crossing. The only way to reverse the effects of thermal

cycling on the Zerodur is to heat the material above 320 C and cool down in the same

manner as the original ceramization process [28].

2.1.4 Long Term Stability

Over the course of 10 years, a 400 mm long Zerodur bar was examined at a constant

temperature of 20C at the PTB Braunschweig. During this time, the Zerodur showed

a monotone decrease in length, often referred to as aging [28]. The slope of the decrease

in length depends on the thermal treatment and the age after the thermal treatment.

A slower cooling treatment of the Zerodur will decrease the slope, making the Zerodur

appear older. Any thermal treatment will start the aging process anew.

2.1.5 Cryogenic Behavior

The stability of Zerodur at cryogenic temperatures is an important topic due to

its use for mirrors in astronomical space-based missions. The CTE of Zerodur has been

measured from 10-300K to show that Zerodur normally expands upon cooling. In

addition, the observed mirror shape deviations of a 300 mm diameter mirror under a

temperature change from 300 K to 130 K resulted in a figure distortion of less than 0.035

waves RMS over the entire cool down [29].

2.2 Invar

Invar alloys are widely known and used. Their low CTE close to room temperature

have made them an attractive metal for use in many precision instruments [30]. Invar

alloys are commonly used in clocks, scientific instruments, bimetal strips in thermostats,

and many other applications where constant monitoring is necessary. The relative ease in

fabricating and machining have also made Invar alloys popular. While there are several

Invar alloys available that only vary slightly in elemental composition, only the three

most common will be discussed. The production process, chemical compositions, and

applications of each alloy are presented.

2.2.1 Invar 36

Invar 36 is an iron-nickel alloy that possesses a CTE roughly one-tenth that of carbon

steel at temperatures up to 2000C. This alloy is commonly used for applications where

dimensional changes due to temperature fluctuations need to be minimized. This has

made Invar 36 an attractive material for use in radio and electronic devices, aircraft

controls, and optical and laser systems.

Although the heat treatment process will vary slightly depending on manufacturer,

three annealing methods are presented. Each alloy will have its own specific heat

treatment, but in general, the process is similar.

Method 1:

1. Heat parts to 830C and hold at that temperature for one-half hour per inch of

2. Furnace cool at a rate not to exceed 100C per hour to 315C. No additional
machining should be done on these parts.

3. Let parts air cool.

Method 2:

1. Rough machine parts.

2. Heat part to 830C and hold at this temperature for one half hour per inch of

3. Furnace cool at a rate not to exceed 100C per hour to 315C. Air cooling below
315C is acceptable.

4. Heat parts at 315C for 1 hour followed by air cooling.

5. Heat parts at 96C for 48 hours followed by air cooling.

6. Finish machining.

Method 3:

1. Rough machine parts.

2. Heat parts to 830C and hold at this temperature for one half hour per inch of
thickness, then water quench.

3. Semi-finish machining parts.

4. Heat parts for 1 hour at 315C, followed by air cooling.

5. Heat parts at 96C for 48 hours followed by air cooling.

6. Finish machining parts.

The variation in heat treatments and elemental purity and concentration from

manufacturer to manufacturer can produce alloys that are not as reliably made as other

low expansion materials such as Zerodur. This can produce a variation in thermal

properties. Although CTE values can vary, typically Invar 36 has a CTE of approximately

1.6x 10-6/K [31].

A specific Invar alloy was developed by NASA/JPL for the Cassini spacecraft camera.

Dubbed High Purity Invar 36 (HP Invar 36), it has a much improved time varying CTE

and dimensional stability [31]. Very pure elements were weighed, mixed, and pressed into

a mold and sintered in a controlled atmosphere. Analysis showed this alloy to have a

CTE and dimensional stability similar to that of Invar 32-5 [31]. Although this specific

alloy is not readily available, it attests to the extent to which the CTE of Invar 36 can be

pushed. Reference [31] provide an in depth discussion of the heat treatment and stability

of this high purity Invar. It should also be noted that in reference [31], the particular heat

treatment did show some effects on the materials properties, but the differing elemental

compositions had a much greater effect on the material properties, specifically the CTE.

All Invar alloys should be handled carefully. Dropping them may disturb the grain

structure and cause it to become magnetized to a small extent. The addition of impurities

such as sulfur and chromium make Invar 36 easier to machine. This allows for complex

structures to be machined, while still keeping its low CTE.

2.2.2 Invar 42

Invar 42 is an alloy whose chemical composition is presented in Table 2-3. The

primary use of Invar 42 is for glass-to-metal seals in electronic tubes, automotive and

industrial lamps, transformer and capacitor bushings, and other ceramic-to-metal

applications. These applications are due to the similar expansion properties of Invar

42 and other glasses, particularly 1075 glass. The heat treatment for Invar 42 is similar

that of Invar 36.

2.2.3 Invar 32-5

Invar 32-5, also known as Super Invar, is a magnetic alloy containing iron, nickel, and

cobalt. The higher percentage of cobalt in Super Invar provides a lower CTE around room

temperature when compared with Invar 36. Typically, Super Invar has a CTE 3-6 times

less than that of Invar 36, and is used in many applications including structural supports

and optical and laser systems that require precision measurements.

For the lowest thermal expansion and optimum stability, the heat treatment

recommended is to initially heat the materials to 8400C for 1 hour and then water quench.

This is followed by a stress relieving operation at 315C for 1 hour, air cool and age at

96C for 24 hours, then air cool. Due to the high oxidation rate of this alloy above 5380C,

it is recommended the parts be heat treated in a protective environment such as a vacuum

or inert gas.

Although Super Invar's initial temporal stability is much better than Invar 36, it

has been shown that it degrades over time to rates comparable to that of Invar 36. This

process occurred much faster at 600C than at 27C and seems to be evidence that the

temporal stability of Super Invar is highly temperature dependant [32].

Due to the rarity of literature on the chemical composition as well as the mechanical

and thermal characteristics of the Invar family, the following tables are a compiled list of

selected properties for Invar 36, 42, and 32-5.

Table 2-3. Compositional values of Invar 36, 42, and 32-5.
Element 36 42 32-5
Carbon 0.15 .05 max .05
Nickel 36 nominal 41 31.75
Phosphorus .006
Iron balance balance balance
Silicon 0.40 0.20 0.09
Manganese 0.60 0.40 0.39
Sulfur 0.004 -0.01
C('in, 111111i 0.25 -0.03
Cobalt 0.50 -5.36
Aluminum 0.07
Copper 0.08

Table 2-4. Comparison of selected properties for 3 Invar alloys
Property (at room temperature) Invar 36 Invar 42 Invar 32-5
Density (kg/m3) 8055 8110 8150
Thermal Conductivity (W/m K) 10.5 10.7
CTE (x10-6/K) 1.3 4.3 0.63
Modulus of Elasticity (l\ Pa) 148 144
Tensile Strength (\! Pa) 517 144
Specific Heat (J/kg K) 515 500

2.3 Silicon Carbide

Carborundum, more commonly known as silicon carbide (SiC), was first reported in

1810 by Berzelius [33]. In 1890 SiC was accidentally synthesized by Edward G. Acheson

while running an experiment to synthesize diamonds. SiC occurs naturally in meteorites,

although it is found rarely and often only in small amounts. Acheson was able to produce

SiC by passing an electric current through a mixture of sand and petroleum coke at very

high temperatures in an electric furnace. Although this process has been modified since its

inception, it is still used to produce certain types of SiC todc'i. It wasn't until the 1960's

that SiC could be produced in molded form. Five decades later, SiC is produced at almost

one million tonnes per year [33].

Although SiC was discovered 2 centuries ago, not until recently has it become a

popular material to replace the use of metals and alloys for structural purposes, especially

in extreme environments.

2.3.1 Structure of SiC

There are ~200 polytypes of SiC, although only a few are common. Each polytype

can be visualized as being made up of a single tetrahedranol unit in which each silicon

atom is tetrahedrally bonded to 4 carbon atoms. These carbon atoms are then tetrahedrally

bonded to 4 silicon atoms. The difference among the polytypes are the varying orientational

sequences by which these tetrahedral 1l- -is, are stacked. Each 1i-. r can only be stacked in

one of two v-- i- but the varying sequential combinations of these I1-.-is, represent different

crystal polytypes. These polytypes are then grouped into two specific classifications of

SiC known as a-SiC and /-SiC. Although these classifications of SiC differ in < i-i 111iiw

structure, their theoretical densities vary by less than 0.002 g/cm3. The 3-SiC structure

forms at lower temperatures than that of the a-SiC structure [33].

2.3.2 Fabrication Process

SiC is produced in a v i,. I, i of fashions, most of which are proprietary in nature. In

addition, the material characteristics can be varied depending on the specific process and

materials used. For this reason, only a general outline of the production process of SiC will

be presented. Acheson process

The Acheson process uses a mixture of high purity silicon dioxide and petroleum

coke to produce a-SiC. Large quantities of this mixture are then covered in a furnace with

electrodes on opposing ends and a rod in the center to connect the two. A current of ~25

kA is then passed through the electrodes causing the material to heat to 25000C. As the

rod heats up, the mixture heats up form the center outward. This will cause high-purity

a-SiC to form close to the rod and lesser purity SiC to form outward. Depending on the

process used, some /-SiC is formed closer to the outside. The current is left on for 10-20

hours, resulting in a large chunk of SiC [33]. This chunk is then broken, sorted, crushed,

milled, and classified into different sizes and commercial grades of SiC powder. Further

processing of the powder can be done in order to produce powders with crystalline sizes

on the order of tens of nanometers. The crystalline size pl ,i an important role in the

densification process of SiC. The SiC parts made from these powders will have varying

properties (specifically porosity) depending on the crystal size used. Sintered SiC

Sintered SiC is produced by using the a-SiC powder formed from the Acheson

process. In general, sintered SiC parts are made by mixing a-SiC powder and organic

binders such as Boron and Carbon. Carbon is added to remove surface oxygen on the

SiC particles, while boron is added to prevent the growth of grains at lower temperatures

before reaching the sintering point. Other sintering aids include oxides such as alumina,

zirconia, and yttria. The mixed powder is then compacted using cold isostatic pressing.

This is done by placing the mixture into a bag and submerging the bag into a fluid. The

fluid is then exerted with pressure upwards of 200 MPa, resulting in a ,i. I solid block.

This block is then machined into the desired shape, known as the green shape.

Green Block
SiC Powder Cold Processing Gre

Green Block
Polishing Sintenn Gren

Figure 2-3. Process used to produce sintered SiC.

The green shape is then pressureless sintered. The sintering is done at ~2100C in

a non-oxidizing environment. This step may last for dv or weeks, depending on the

thickness of the green shape. In order to reduce the effects of micro-cracks within the

material, the sintered shape is cooled at a rate determined by the thickness of the shape.

The greater the maximum thickness of the shape, the longer the cool down time will be.

This reduces the temperature gradient within the material which may cause micro-cracks

to form. During the sintering process, the product typically shrinks ~t17-. [33].

The addition of sintering aids can produce impurities within the final SiC product,

but much care is taken to remove almost all impurities. High-temperature chlorination can

be used to remove a significant amount of impurities, but may cause depletion of silicon in

the SiC. Each manufacturer has their own techniques used to keep the level of impurities

to a minimum, and in some cases, not even trace amounts can be detected. Reaction-bonded SiC

Reaction-bonded SiC (RBSiC), also called S/SiC or carbon felt SiC in some literature,

uses a different process to create SiC products. Instead of using a pressed SiC shape

made from SiC powder, RBSiC infuses silica into a carbon pre-form. A 3-SiC structure is

formed since the infusion process occurs at lower temperatures than a-SiC forms [33].

The process begins by using a graphite that has already been processed and purified.

This graphite is then machined into the desired shape, known as the green part. The

conversion of the graphite into SiC takes place when they are exposed to silicon-carrying

species, typically SiO gas, at high temperatures. Although a gas is typically used, using

liquid silicon will also produce similar results. The chemical formula for this step is shown


SiO + 2C S SiC + CO

In order for the green part to convert to SiC properly, it is essential for the graphite

to have a reasonable porosity for the SiO to infiltrate the graphite. This reaction is

rate-limited by the pore-diffusion resistance and keeping a large SiO concentration

gradient between the SiO gas and SiC/C interface is essential for this to occur [34]. The

reaction rate at the SiC/C interface is controlled by surface kinetics which are spontaneous

at high temperatures. This means the conversion rate from graphite to SiC is limited by

the inward diffusion of SiO and outward diffusion of the CO gas through the SiC shell.

For this reason, fabricated parts can only be less than a certain thickness if a complete

Graphite Shape Purification -Machining

Polishing SiC Coi cisio.

Figure 2-4. Process used to produce reaction-bonded SiC.

infusion is to take place. Using this infiltration process, the resultant part typically shrinks

by less than 1 of its original shape. Chemical vapor deposition of SiC

Due to its high reflectance in the visible spectrum, SiC parts are often coated with

a dense, thick i- -rV of SiC by use of chemical vapor deposition (CVD). A chemical

reaction of compounds consisting of silicon and carbon are heated to a temperature

between 1000-1800C in the presence of hydrogen. The end result is a 1l- -r of high purity

(typically 99.9' 1' or greater) /-SiC consisting of very fine grains. This process is often

used for large mirrors made of SiC. CVD SiC is a very flexible process and can be tailored

to meet specific needs such as coating thickness and microstructure [33]. Typically the

CVD process produces a SiC with a lower percentage of porosity.

2.3.3 Properties

The variations in the SiC fabrication process from different manufacturers will

produce SiC with differing thermal and mechanical properties. These differences tend to

be small and the effects due to temperature tend to follow the same trend. For this reason,

typical values will be given for material properties. Specific values from venders will be

cited when applicable. Density and porosity

The density of an object is given by the mass per unit volume. This term is general

and it will be necessary to specify which density is being used. The crystallographic

density is calculated from the chemical composition and interatomic distances given by

pC N (2-1)

where M is the gram formula weight of the material (\! 10.09715 for SiC), N is

Avogadros constant (6.0221367x 1023/mole), and V the volume of the unit cell.

The theoretical density is the density of a material as if there were no microstructural

porosity. Also known as the apparent density, it is given by

Pa (2-2)

where m is the mass of the material, and V is the volume occupied by the solids.

Depending on the polytype, the theoretical density lies between 3.166 to 3.249 g/cm3

These measurements have been verified using X-ray spectroscopy and X-ray diffraction[33].

The bulk density is the measured mass in the total bulk volume of the material. The

bulk density is calculated by

Pb (23)
v + Vp
where Vp is the pore volume.

Often reported with the density, porosity is another important physical property that

indicates the amount of free space that is not occupied by solid material. The porosity

is detrimental to the strength of the material. The strength is inversely exponentially

proportional to the total porosity. Open porosity reduces the oxidation resistance and

can present outgassing problems under high vacuum conditions. For RBSiC the open

porosity can range from 0.50' to 21n' depending on the grade of SiC and process

used[34]. Typical porosities for sintered SiC are less than 1 with most having close to

zero porosity [35] [36] [37].

Table 2-5. Comparison of density and porosities for selected SiC materials.
Material Apparent Density Bulk Density Total Porosity Open Porosity
SuperSiC 3.13 g/cm3 2.53 g/cm3 20 % 19 %
SuperSiC-SiC 3.07 g/cm3 3.05 g/cm3 4 % 0.5 %
Hexoloy SA 3.10 g/cm3 < 2 %
CoorsTek SiC 3.10 g/cm3 Oxidation resistance

In general, SiC has excellent oxidation resistance, although this depends largely on

the amount of open porosity and particle size. A larger porosity and particle size increases

the effective surface area exposed to oxygen which will increase the oxidation rate. SiC

is kinetically stable in air up to ~10000C. In the 1000-11500C range, oxidation begins to

occur and a thin film of SiO2 will rapidly form. As the temperature is increased above

this range, the SiO2 l- r begins to thicken and densify which results in a slower rate of

oxygen being able to penetrate into the SiC. This slows down the oxidation rate. If the

temperature is increased above 16500C, a reaction between the SiO2 film and the SiC will

occur and produce SiO and gaseous CO. The CO gas can cause the SiO2 l,- r to rupture,

allowing more oxygen to diffuse into the SiC, hence increasing the oxidation rate. Flexural and tensile strength

Defined as a measure of the ultimate strength of a specimen in bending, the flexural

strength of a material is an important material property [38]. The flexural strength is

most commonly tested using the ASTM\i C-1161 standard with either a 3- or 4-point test

being employ, ,1 The equations for the 3- and 4-point bending test are given by

93pt 22 (2 4)
4pt bc (2 5)
bd 2

IF a F F

3-point 4-point

Figure 2-5. Diagram of how 3- and 4-point loading tests are done.

where 1 is the length of the specimen, F is the total force applied, b is the specimen width,

d is the specimen thickness, and a is the distance between supporting beads. If a material
is ductile, like most metals and alloys, then the material bends before failure. If a material
is brittle, like many ceramics, then there will only be a slight bending of the material
before failure.

Defined as the maximum tensile stress sustained by a material when a pulling force

is applied along the specimen, the tensile strength determines a material's capability of
load bearing in structural applications. The tensile strength is calculated by dividing the
maximum load needed to rupture the material divided by the cross-sectional area. When

determining the tensile strength of many ceramics, the sample preparation is critical.
Micro-cracks in or on the material will propagate rapidly as a load is applied. For this

reason, a statistically significant number of tests need to be performed. The most common
test method used to determine the tensile strength is the ASTM C-1273 standard method
as the elastic modulus can be obtained as well.

Table 2-6. Comparison of flexural and tensile strength for selected SiC materials.
Material Flexural Strength (\! Pa) Tensile Strength (\! Pa)
SuperSiC 147 129
SuperSiC-SiC 269 138
Hexoloy SA 410 210
CoorsTek SiC 410 307 Thermal conductivity and heat capacity

The rate at which heat flows through a material is known as the thermal conductivity

(K). The high thermal conductivity of SiC makes it an appealing material for use in

applications where thermal gradients are needed to be kept low. While the thermal

conductivity of single crystal SiC has been reported to be as high as 500 W/m K, most

commercial grades have a thermal conductivity closer to 50-120 W/mK. Impurities

from the production process will also lower the thermal conductivity. In general, as the

temperature decreases, the thermal conductivity increases.

For many ceramics, the heat capacity is a strong function of temperature. The heat

capacity of a material can be measured through the use of the calorimetry technique, or

for high-purity materials it can be taken from published data [39].

Table 2-7. Comparison of thermal conductivity for selected SiC materials.
Material Thermal Conductivity (at RT in W/mK)
SuperSiC 151
SuperSiC-SiC 218
Hexoloy SA 110
CoorsTek SiC 100

2.4 Carbon Fiber Reinforced Polymers

Carbon fiber reinforced polymers (CFRP), occasionally referred to as "black

aluminum," are a group of materials that have a wide range of thermal and mechanical

properties. CFRP consist of carbon fiber filaments in a matrix material. The purpose of

the matrix is to transfer external loads to the carbon fiber structure. In most cases, the

matrix material is a polymer such as an epoxy resin. Since the orientation of the carbon

fibers and the specific matrix used will determine the overall mechanical and thermal

properties of the CFRP, an almost limitless number of differing CFRP structures can be

made. The fabrication process also allows CFRP to be used in many custom applications.

In this section, the fabrication process and typical mechanical and thermal properties of

CFRP will be discussed.

2.4.1 Fabrication Process

CFRP are composites that consist of two distinctly different components. When

combined, the components remain discrete, but function interactively to make a new

material whose properties can't be predicted by summing those of the components.

Carbon fibers have a tensile strength along the fiber direction many times greater than

that of steel, titanium, or aluminum [40]. The matrix transfers the applied load to the

high-strength fibers in addition to providing crack and environmental resistance damage to

the structure.

The composite part is made from a fiber architecture which is based on a specific

"ply." The orientation, thickness, and fiber type of the ply will determine the properties of

the composite. The composite properties vary from ply to ply, but it is assumed that each

specific ply is homogeneous [40].

Composites are typically made up of several plies with mutually perpendicular planes

of symmetry to create a laminate that is symmetrical and balanced. For stronger and

lighter parts, a high fiber volume should be used. This is ideal for parts such as aircraft

wings, but the price is usually more expensive since the fibers cost more than the resin. A

heavier, more chemically resistant part, such as piping at an oil refinery, can be made from

plies with a lower fiber-to-resin ratio. Carbon fibers

For high-performance applications such as aircraft wings or support structures

in space-based missions, the carbon fibers used are typically made from a variety of

materials. These include polyacrylonitrile (PAN), pitch, and (rarely) ra, in. Fiber made

from these materials are heated and stretched to create a high-strength fiber. The most

versatile of these fibers are PAN-based fibers and typically have high strength and

stiffness[40]. Pitch fibers are made from petroleum or coal tar pitches and typically have

extremely high stiffness and low to negative CTE. Due to these properties, pitch fibers

are often used in applications where thermal contributions need to minimized, such as

instrument housings for spacecraft electronics.

Manufacturers produce carbon fibers in tows consisting of thousands of continuous,

untwisted filaments that are microns thick. The filament count is designated by a number

followed by the letter "K," which designates a multiplication of a thousand. For example,

a 10K tow indicates a bundle containing 10,000 continuous filaments. Tow sizes ranging

from 1K to 12K are typically used for aerospace applications. For application in the

industrial, construction, and automotive industry, filament counts of 48K to 320K

are typically used due to the lower processing cost, and faster processing times. The

production process for higher filament counts tend to result in a lower tensile strength.

When exposed to metals PAN and pitch fibers experience galvanic corrosion. This can

be overcome by using a barrier material or veil ply made of fiberglass or epoxy during the

laminate layup in order to prevent direct contact.

In addition to PAN and pitch fibers, other high-performance fibers include aramid

(commonly referred to by its trademark name Kevlar) and high-modulus polyethylene.

Aramid fibers are composed of aromatic p]" V,! ,iidi,'. and provide exceptional impact

resistance and tensile strength, making them ideal for applications such as ballistic

protection, solid rocket motors, and helicopter blades [40]. Polyethylene fibers are light

weight, have excellent chemical resistance, extremely impact resistant, and have a low

dielectric constant. Many applications for these fibers are found as the same of the aramid

fibers. Quartz, ceramic, basalt, boron, and hybrid fibers are all other options, but not

commonly used. Matrix material

Consisting of polymer chains, the resins used to make CFRP parts fall into two broad

categories known as thermosets and thermoplastics. Thermosets, such as Eccobond 285

or Duroplast, are the most commonly used matrix material to create CFRP composites.

A thermoset permanently cures into a linked network when mixed with a catalyst, is

exposed to heat, or both. Fabricators can control the cure schedule by adding inhibitors

and accelerators to achieve the desired CFRP. For applications in the aerospace industry,

curing often takes place in a vessel that is heated to elevated temperatures and is under

high vacuum. This ensures good consolidation of the laminate to minimize any trapped

air or voids from the initial fabrication of the part that would weaken the structure [40].

Under certain circumstances, pressures can be applied to the part to expedite or improve

the process.

Unsaturated polyester resins are widely used in commercial, mass-produced

applications such as bathware, boats, or car parts due to their ease of 1i ,llii.; relatively

low cost, and good balance of mechanical, thermal, electrical, and chemical properties.

For high-performance applications, polyester resins are typically not used due to their

high shrink rates. Instead, thermosets such as epoxy resins, phenolics, cyanate esters, and

polyimides are used.

For high strength, durability, and chemically resistant composites, epoxy resins are

typically used. Epoxies come in many forms ranging from liquid, to solid, to semisolid

and are typically cured by reacting with amines or anhydrides. Toughening agents such as

thermoplastics or rubber compounds can be added to counteract brittleness. This can be

useful in aerospace applications that use amine-cured epoxies that require the use of an

autoclave at high temperatures that can cause brittleness.

Low cost, flame-resistant, low-smoke products such as aircraft interior panels or

rocket nozzle applications are typically made using phenolic resins due to their excellent

heat absorbing and low char yield qualities. Although phenolic resins have excellent

thermal qualities, their mechanical strengths tend to be lower than when using other

high-performance resins.

Polyimides are a more exotic form of resin and have found their way into high-temperature

products such as jet engine nacelle components. These resins offer low electrical

conductance and high chemical resistance. In some cases, parts made using polyimides

can have service temperatures of 370 C, making them ideal for high-temperature engine

parts [40].

Fabricators of CFRP parts buy the fiber form and resin to best suit the needed

application. Due to the high volume of CFRP parts now made, specially formulated resin

matrixes are reinforced with carbon fibers and partially cured to produce a pI p 1

These prepregs are then frozen and defrosted when they need to be used. The specific

prepreg is chosen based on the application needed and can range from golf clubs to

airplane and jet engine parts.

2.4.2 Properties

As previously stated, the production process, fiber used, and matrix material used will

all determine the mechanical, chemical, electrical, and thermal properties of the CFRP

part. Designers can fabricate CFRP materials with a low CTE, good vibration damping,

excellent fatigue resistance, good resistance to extreme environmental conditions, or a

combination of all of these. For this reason, no specific thermal or mechanical property

values will be given. Consultation with a CFRP manufacturer is ideal to determine

the specific attributes they can provide. As a general rule of thumb, many thermal and

mechanical properties of CFRP are similar to those of SiC.

2.5 Lead Zirconate Titanate

Lead Zirconate Titanate (PZT) is a class of materials whose properties can be tailored

to meet specific needs by changing the zirconate-titanate ratio. Piezoelectric ceramics

are hard, chemically inert, insensitive to atmospheric influences, and have mechanical

properties similar to ceramic insulators. These properties make PZT materials ideal for a

wide range of applications. These include spark igniters, solid-state batteries, sonic and

ultra sonic generators, pressure and acceleration sensors, and actuators. The application of

PZT in this work is limited to its use as an actuator, so only discussion of this use will be


2.5.1 The Piezoelectric Effect

Jacque and Pierre Curie first discovered the piezoelectric effect in 1880 [41]. They

found that if a mechanical strain was applied to certain ( i i --I i- they became electrically

polarized. This polarization was proportional to the applied strain. The Curies also

discovered that when an electric field was applied to these materials, they would deform.

This is known as the inverse piezoelectric effect. This effect is shown to occur in ir inl,-

naturally-occurring (< i-- I including quartz, tourmaline, and sodium potassium tartrate.

These materials have been used for many years as electromechanical transducers. If

a crystal is to exhibit the piezoelectric effect, its structure should have no center of

symmetry. Applying a tensile or compressive stress to such a
separation between the positive and negative charge sites in each elementary cell, causing

a net polarization at the crystal surface. This effect is linear in a large regime and is

direction-dependant such that compressive and tensile stresses will generate voltages of

opposite polarity [41]. The piezoelectric effect also occurs in a reciprocal nature. Hence, if

a voltage is applied to the crystal, then a mechanical deformation will occur. Applying a

voltage of opposite polarity will cause the crystal to deform in the opposite way.

2.5.2 Piezoelectric Materials

PZT are a class of materials that can be considered a mass of tiny crystallites [41].

Above a specific temperature known as the Curie point, these < iv-- i 1ies exhibit simple

cubic symmetry such that there are no dipoles present in the material. Below the Curie

point, the crystalline structures form a tetragonal symmetry such that an electric dipole is

built into the structure (Figure 2-6).

The dipoles are not randomly orientated throughout the material, but rather form

regions of local alignment known as Weiss domains. Within the domain all of the dipoles

are aligned, giving the domain a net dipole moment and polarization, but no overall

polarization is exhibited due to the random orientation of the Weiss domains [41]. The

ceramic may be made piezoelectric in any given direction by use of a poling treatment. In

U Pb, Ba

Ti, Zr + C-

a b
Figure 2-6. a) Cubic lattice above the Curie point b) Tetragonal lattice below the Curie
point after the poling process has been applied.

this process a strong electric field is used at a temperature just below the Curie point to
orientate the Weiss domains in the direction of the electric field (Figure 2-7).
+ ++ +1 +1 + + + li
V+ Pohng I \
++ Voltage + + +

01-0 -----1-+1-+1--- 0 + A

a b c
Figure 2-7. Electric dipole moments for the Weiss domains (a) before polarization (b)
during polarization and (c) after polarization.

When the electric field is removed, the dipoles remain locked into their new
orientation, producing a remanent polarization and permanent deformation. The poling
treatment is usually the final step in creating the piezoelectric material.
The poling process will produce a permanent polarization within the material, but
these properties can be partially or even totally lost if the material is subjected to certain
mechanical, electrical, or thermal conditions [41]. If too large a static electric field or too
large a mechanical stress is applied, or if the material is heated to a temperature near its

Curie point, the ceramic may become depolarized. These values vary depending on the

type of PZT material.

The basic behavior of a PZT cylinder is shown in Figure 2-8. If a tensile or

compressive load is applied to the PZT, the resulting change in dipole moment creates

a voltage to appear between the electrodes. Similarly, if a voltage is applied to the

electrodes, then the PZT material will stretch or compress depending on the polarity of

the voltage.


b c


d e

Figure 2-8. The piezoelectric effect of a cylinder for (a) no external action (b) an applied
compressive force (c) an applied tensile force (d) an applied voltage (e) an
applied voltage of opposite polarity. For clarity only 1 dipole is shown in (a).

The physical constants of PZT materials are given 2 subscripts due to their

anisotropic nature. The direction of positive polarization is chosen to coincide with

the Z-axis of a rectangular coordinate system [42]. The axes X, Y, and Z are designated

by the numbers 1, 2, and 3, respectively, while the shear about these axes are represented

by 4, 5, and 6, respectively (Figure 2-9).

For clarity, 3 examples will be given.

1. d33 is the induced polarization per unit applied stress in direction 3. Alternatively it
is the induced strain per unit electric field in direction 3.




1/ 4
Figure 2-9. Axes designation and directions of deformation. Directions 4, 5, and 6 are the
shear about axes 1, 2, and 3, respectively.

2. da3 is the induced polarization in direction 3 per unit stress applied in direction 1.
Alternatively it is the mechanical strain induced in the material in direction 1 per
unit electric field in direction 3.

3. g15 is the induced electric field in direction 1 per unit shear stress applied about axis
direction 2. Alternatively it is the shear strain induced in the material about axis 2
per unit electric displacement applied in direction 1.

From these numbers the amount of translation and shear can be determined for

a certain applied voltage. A few of these values are given in table 2-8 for the k-180

piezoelectric material [43].

Table 2-8. Selected properties and their values for the k-180 piezoelectric material.
PZT Property Value
Longitudinal Coupling Coefficient (k33) 0.68
Piezoelectric Strain Constant (d33) 165 pm/V
Piezoelectric Strain Constant (d15) 350 pm/V
Piezoelectric Voltage constant (g33) 43 mV m/N
Curie Temperature 350 C
Density 7.7x 103 kg/m3
Coefficient of Thermal Expansion 3.6 x 10-6/C


When designing critical optical components such as optical benches or telescope

support structures for space-based missions, there are two important factors that must

be taken into consideration; the material the component is made of, and the bonding

technique in which the pieces are put together. As the complexity and sensitivity of

space-based missions increases, the jointing technique used needs to be strong in order

to survive launch conditions as well as stable in order to not spoil the measurement


While the previous chapter dealt with a variety of materials that are suitable for

several different structures, the next two chapters will focus on several bonding techniques

that can or have been used to construct optical components.

3.1 Optical Contacting

Optical contacting occurs when two surfaces are polished flat and smooth enough

such that the materials bond without the use of a cement. Typically optical contacting is

done between two glass materials due to the ease of pi.1 i-l Ir,. but it can be done for other

materials, such as SiC and Super Invar, if the surfaces are adequately polished. Once the

bond is formed, the bond line effectively disappears and the 2 materials act as one.

Optical contacting has found its way into many areas such as silicon wafer bonding

in the electronics field, bonding mirrors to optical benches for use in space-based missions,

and bonding optical components for use in lasers. Although optical contacting has been

around for several decades and has gained popularity in various industries, literature

on the subject is scarce and agreement between theory and experiment vary [44]. It is

assumed the primary bonding mechanism is due to the van der Waals interaction at the

surface. The force per unit area derived from the van der Waals attraction is given by [44]

P(D) A (3-1)

where A is the Hamaker constant and D is the distance between the facing plane surfaces.

For most materials, the Hamaker constant is on the order of 10-19 J, with fused quartz

having a value of A?0.6x 10-19 J [44].

There are several optical bonding methods commercially available, but the details

of the process are often difficult to ascertain due to their proprietary nature, although

many of them require moderate heating of the substrate. For this reason, three optical

contacting methods were developed for use at room temperature and have been used on

BK7, SiC, and Super Invar.

3.1.1 Method 1

Clean surfaces as described in Appendix B. Place surfaces on top of each other

and gently press together. If done properly, a bond will have formed within minutes

and strengthens over time. For glass or transparent materials, the quality of the optical

bond can usually be gauged by the number of fringes seen. Fewer fringes implies the

surfaces are closer together, and hence, will produce a stronger bond. The bond can be

non-destructively broken by placing in a sonicator or through thermal stressing.

This method is best used on ultra-polished surfaces. If the surfaces are flat and

smooth enough (typically A/10 or better), an optical contact will form once the 2 surfaces

come in contact and can form appreciable strength with little or no pressure needing to be

applied to the materials.

3.1.2 Method 2 (Solution Assisted Optical Contacting)

Clean surfaces as described in Appendix B. Place a small amount of methanol on

the surface and place the other surface on top. Gently press together. This method

takes approximately 3-4 hours for the methanol to evaporate and for the bond to obtain

appreciable strength. Typical amounts of methanol used are 0.50 to 0.75 pL/mm2.

This method works best when alignment of the two materials is needed. The

methanol works as a lubricant that allows the materials to be moved, but still has some

initial strength. As the methanol evaporates, the bond strengthens.

3.1.3 Method 3

Clean surfaces as described in Appendix B. Place the surfaces on top of each other

and slowly move in a circular motion. After a few seconds, apply a small amount of

pressure and keep applying a circular motion. This should cause the two pieces to

eventually lock together in one or two places. Repeat if necessary until the two pieces

get harder and harder to twist. Once it becomes difficult to move the two pieces, apply

pressure. Let the pieces sit for 24 hours for the bond to settle and gain strength.

This method works best for surfaces that are not polished as well, or when the

surfaces are dirtier than normal. In many cases if the previous two methods do not work,

this method often will succeed, if only to form a partial bond.

These methods have been successfully used on BK7-BK7, SiC-BK7, and Super

Invar-BK7 bonds with breaking strengths typically between 0.1-1.0 MPa. Other optical

contacting methods exist which have reported breaking strengths ranging from several to

tens of megaPascals [45]. Optical contacting was primarily used in this work to attach

optical mirrors to substrate to form an optical cavity whose dimensional stability was then


3.2 Anodic Bonding

Of particular interest in the electronic and electrical engineering fields is forming

a strong bond between a glass and metal surface. These glass-metal bonds have been

extensively used in industrial packaging of microelectromechanical system devices and

hermetic feedthroughs. One method of glass-metal sealing, known as anodic bonding, is

capable of creating a strong bond without the use of an adhesive paste, or heating the

glass to temperatures above its softening point, as is the case with frit bonding.

The basic concept behind anodic bonding is to use the electrostatic attraction

between two materials when a voltage is applied across the bond. Since glasses have an

extremely high electrical resistivity at room temperature, the parts must be heat to a few

hundred degrees before a significant drop in electrical resistivity can be seen. Although the

parts need to be heated, the temperatures used are typically well below the softening point

of the glass. The increased electrical conductivity allows a larger current to pass through

the materials, resulting in a larger electrostatic attraction between the surfaces [46]. The

bond originates at the contact points between the surfaces and spreads out from there. For

this reason, flatter and smoother surfaces will produce a bond that will form faster and

with greater strength.

Vgap applied
Glass -Vlass

Xgap x ggap

gap 2 gap
Metal Fgap^ ^ ogap

Figure 3-1. Representation of how anodic bonding works.

When one places a well polished surface on top of another, the surfaces usually only

contact in a few spots. This can easily be seen when two pieces of well polished glass are

placed on top of one another and interference fringes can be seen radiating outward from

one or two points. With anodic bonding, these surfaces are brought together by the use of

an electrostatic force. The voltage applied across the glass sets generates an electric field

within the gap that, in turn, creates an electrostatic force (see Figure 3-1) that pulls the

surfaces together. If glass behaved as a perfect insulator, then the voltage needed would

be a function of the glass thickness. In addition, bonding would occur irrespective of the

voltage polarity used. With most glasses, this is not observed. Glasses with thicknesses

ranging from 0.05 to 25.4 mm show no voltage dependence and only bond if the correct

polarity is used [46]. This is because at the temperatures used for anodic bonding most

glasses contain mobile positive ions compensated by almost immobile negative ions. This

will leave a net negative charge on the bonding surface of the glass. If a positive potential

is placed on the glass, then negative ions will build up on the metal surface as well,

resulting in a repulsion. If the polarity is switched, then positive ions will be on the metal

surface and negative ions will be on the glass surface. This will result in the necessary

electrostatic attraction needed to bond. This also explains why the bond forms where the

surfaces are in contact and then spreads out. The electrostatic force will be greatest where

the gap between the surfaces is the smallest. This occurs where the surfaces are in contact.

When dealing with anodic bonding, there are six process parameters that must be

taken into account; temperature, voltage, time of applied voltage and cooling rate, surface

finish, CTE, and atmosphere [46].

Most work done with anodic bonding occurs in the 300 to 600 C range, although this

can vary depending on the temperature dependence of the electrical resistivity of the glass.

Zerodur, for instance, has a resistivity of about 1 MQ cm at 300C [46].

Typical working voltages are in the range of 200 to 2000 V. For glasses with a higher

resistivity, more voltage should be used to create a larger surface charge at the bonding

interface. The upper limit on the voltage is set to inhibit sparking which may damage the


The time required in order for the bonding to occur is typically 1-5 minutes (after the

parts have been heated and the applied voltage is turned on). Slow cooling of the parts

may be needed for differences in CTE. The slow cooling will work only if the temperature

used is high enough that the glass begins to soften. In this case, as the glass cools, it

distorts itself such that it absorbs some of the stresses in the bond [47].

For a strong bond to form, it is necessary for the surfaces to be as flat and smooth

as possible. This allows fewer and smaller gaps between the surfaces. Typically a surface

finish of better than 50 nm rms is used.

As with other glass-metal seals, matching the CTE of the two parts is critical. Due to

the heating of the glass and metal components, it is critical for the CTE's of both parts to

match as closely as possible over the temperature range in which it is heated. A difference

in CTE can cause induced stresses that the bond may not be able to compensate for. If

this happens, it may be possible that the bond significantly weakens, or breaks as it is

being cooled.

Anodic bonding typically occurs in room air, although other inert gasses could be

used as well. For metals that oxide rapidly, it may be useful for bonding to take place in

an inert gas as the formation of an oxide may cause the surface profile to change.

Using an aluminum coated piece of ULE, van Elp et al [46] were able to anodically

bond Zerodur with a tensile strength of 9 MPa. The purpose was to develop a complicated

electrostatic clamp for the use in next generation lithography machines.

The investigations into anodic bonding for this work were done to determine potential

bonding methods in order to bond low expansion metals, such as Super Invar, to other

glasses in order to make an optical cavity to test the dimensional stability of the metal.

Using anodic bonding instead of a paste or frit bonding would test only the dimensional

stability of the metal, and not that of the metal/bond system. Several investigations

into bonding Super Invar to BK7 glass were done, but with extremely poor results. The

materials were chosen because of their similar CTE's and that BK7 has a similar chemical

composition to that of Zerodur and should allow for a similar current to be passed through

it. The Super Invar piece was 50 mm x 50 mm x 6.3 mm and had a mirror finish on it

with a global flatness of A/4. A variety of BK7 pieces were used ranging in thickness from

6 mm to 1 mm and in diameter from 25.4 mm to 6 mm. The pieces were heated in an

oven to 300C and allowed to soak for at least an hour. A voltage ranging from 200 to

1000 V was applied and left for times between 5 minutes and several hours. After the

voltage was applied, the oven was turned off and allowed to cool to room temperature.

This usually took several hours. In each case, a bond did not form. At best, one bond

formed that formed that showed a slight resistance to being pulled apart by hand.

The only bond that was able to form with significant strength was between a BK7

disk approximately 6 mm in diameter and 1 mm thick and a piece of aluminum foil. The

pieces were heated to 3000C and a voltage of 1000 V was applied for 1 hour. A weight

was applied on the aluminum foil in order to keep the surfaces in good contact with each

other. The pieces were cooled to room temperature by turning the oven off and allowing

the temperatures to equilibrate. Although a bond did form and was difficult to pull apart

using a force in a shear direction, it could be peeled off easily.

The lack of results is most likely due to the difference in CTE between the Super

Invar and BK7. Although there is a significant difference in the CTE of aluminum and

BK7, a bond was most likely able to form because the aluminum foil is thin and able

to stretch and contract in order to handle the stresses induced from cooling the pieces.

Because of this, anodic bonding was abandoned as a way to bond a low expansion metal

to mirrors in order to make an optical cavity.

3.3 Brazing

Brazing is a method that joins two materials through the use of heat and a filler

metal in a cost-effective method that produces significant strength and has excellent stress

distribution and heat transfer properties. Although brazing is commonly used for metals,

its use has begun to expand into other materials as well. In this section, a general outline

of the brazing procedure will be discussed. In addition, examples of brazing SiC parts for

mirrors and p loadd structures is presented.

Brazing joins two materials through the use of a filler material. It is performed at a

temperature high enough to melt the filler material, but low enough as to not melt the

materials being joined. Because of this, brazing is ideal for jointing dissimilar metals.

Ferrous and non-ferrous metals can easily be joined by brazing, as well as metals of

different melting points. The joints are strong and dependant on the strength of the filler

material. In some cases, the braze-joint will exhibit a tensile strength greater than that

of the filler material [48]. Filler materials are chosen such that the tensile strength of the

joint is similar to that of the materials being brazed.

Brazing creates a metallurgical bond between the surfaces of the two materials being

joined and the filler material. Heat is applied broadly to the jointing materials to raise it

to a temperature greater than the melting point of the filler. The filler is then placed in

contact with the surfaces and is drawn completely through the joint via capillary action.

In contrast, welding uses temperatures higher than the melting point of the two materials

and fuses them together with a filler material. Thus, welding temperatures start at the

melting points of the materials. This intense heat can cause warping or distortions around

the joint, or stresses around the weld area [48]. Often, the extreme temperatures used for

welding do not make it suitable for joining thin pieces together, whereas brazing can be

done easily.

A good fit of the materials with proper clearance is essential for brazing due to the

capillary action needed to pull the filler material through the joint. Typical clearances

are 25 to 125 pm. The clearance will also have a significant impact on the strength of

the joint[48]. If the clearance is too small, then there will not be enough force to pull

the liquified filler through. If the clearance is too large, then the joint strength may be

reduced to that of the filler. Highly polished surfaces can also impede the filler flow. A

typical "mill finish" will provide a surface roughness allowing for the creating of capillary

SI' !i for the filler to flow through. In addition, when brazing materials with differing

CTEs, the clearance must be adjusted for the elevated temperatures accordingly. For

example, consider brazing a brass bushing into a steel sleeve (as shown in Figure 3-2).

Brass has a higher CTE than steel and if it is to be the inner part of the assembly, a

larger clearance than if both were the same materials should be used. Conversely, if the

brass now becomes the outer member and the steel the inner, then less clearance should be


Suitable gap Narrow gap

Steel Brass

Brass Steel

Figure 3-2. Proper spacing for brazing materials of differing CTE.

Of similar importance is the need to clean the surfaces before brazing. Capillary

action will work only when the surfaces are clean. For this reason it is essential that all

surfaces to be brazed are adequately cleaned.

When heating the materials, it is possible that an oxide li.-r will form on the surface.

This is not ideal for brazing, and may produce a weaker bond since the oxide will prevent

the filler from adequately bonding to the surface. For this reason fluxing of the parts

should be done. Flux is a chemical compound that is applied to the joint surface before

brazing that prevents oxide formation. The flux will dissolve and absorb any oxides that

will form during the heating process, allowing the filler to freely flow. Although fluxing

the surfaces can be omitted in some cases, even then it can be advantageous to use a small

amount of flux as this may improve the wetting action of the filler [48].

When 'L i. 1ir. both materials should be heated as uniformly as possible so they will

reach the brazing temperature at the same time. When heating materials of different

thicknesses or thermal conductivities, attention should be paid to apply the heat as evenly

as possible. For this reason, using an oven is often the best method to braze differing

materials. In most cases, observing the color and viscosity of the flux might indicate that

the pieces are being heated appropriately. When the assembly has reached the proper

temperature, the brazing rod is then placed in contact with the joint area. A portion of

the rod will melt off as it is touched to the surface and the capillary forces will draw it

through the joint. Once the assembly has been brazed the excess flux should be removed.

Flux residue is corrosive and if not removed can attack the material, possibly weakening

the material or joint.

Although typically used for metals, brazing has become an important jointing method

for SiC parts as well. It is often difficult, or even impossible, to build some SiC structures

in one piece. An example of this is the Herschel Space Telescope primary mirror. With a

diameter of 3.5 m, this mirror was far to big to be made in one single piece. To construct

this mirror, several smaller sintered segments were brazed together [49]. The filler material

was a doped silicon compound with a CTE that closely matched the CTE of the SiC

pieces. The brazing process used by Boostec to create Herschel primary produced a very

thin, uniform joint. The typical braze thickness is less than 50 pm, and in most cases less

than 10 pm. Producing such a thin joint resulted in a joint strength comparable to that of

the SiC alone.

In addition to producing the Herschel primary, Boostec also constructed the SiC

torus for the GAIA mission using their brazing technique. The torus consisted of 17

individual segments that were brazed into one coherent structure. The final structure is a

quasi-octagonal structure that is three meters in diameter and will support the two GAIA

telescopes and the focal plane assembly.

3.4 Epoxies

Epoxies are generally considered to be structural adhesives and can form excellent

bonds to materials including metals, glass, wood, plastics, rubbers, and masonry products.

Epoxies are commonly used to replace other forms of fasteners such as screws and rivets

because they can provide a lighter weight structure for lower cost and are easier to

assemble. Epoxies are either a one- or two-part system. For most high performance

applications, such as in aerospace applications, typical epoxies used are a two-part system.

For this reason the focus will be on this area of epoxies, although a leap from two- to

Figure 3-3. Conceptual design of the GAIA SiC torus. Picture courtesy of ESA.

one-part epoxies is easy with suitable knowledge of two-part epoxies. Advantages of

two-part epoxies are that they typically have a high peel and shear strength, outstanding

chemical resistance, good thermal properties which make them ideal for cryogenic use, low

cost, and good adhesion [50]. In addition to being used for bonding, epoxies are often used

as coatings because of their excellent chemical and environmental resistance.

A two part epoxy adhesive is a type of thermosetting polymer. As defined by Daly,

a thermoset is a material that has the property of undergoing a chemical reaction by the

action of heat, catalyst, ultraviolet light, or other means leading to a relatively infusible

state [51]. Two-part epoxies consist of a resin and a hardener. Also known as the curing

agent, the hardener is required to convert the resin into a plastic-like solid. Once the

resin and hardener are thoroughly mixed, the epoxy will begin to set. Often when mixing

the components, an extreme amount of heat may be released. For this reason caution

should be taken when mixing large amounts of epoxy. The release of heat may also cause

the epoxy to harden faster since elevated temperatures often make the epoxy set faster,

while lower temperatures tend to do the opposite. It is often recommended that elevated

temperatures are used to set the bond since this typically results in a stronger bond.

Working times can range from a few minutes to several hours depending on the specific

epoxy, and setting times can range from one to several d-i,- Table 3-1 lists a few epoxies

with select properties.

Table 3-1. Comparison of properties for select two-part epoxies.
Product Application Work Lifetime Shear Strength Total Mass Loss
(min) (\!Pa at 20 C) (.)
100 B/A Glass-to-metal 5 10.3 5
Mil-Bond Glass-to-metal 30 1
A-12 Glass-to-metal 180 17.2 0.9
Eccobond 285/11 Wide temp. range 240 14.5 0.3
Hysol 1C Vacuum sealing 30 -0.8
Epo-Tek 375 Fiber optics 300 15.9

Percent total mass loss is a measure of the mass loss after a specified period of time

when exposed to elevated temperature. For space-based applications, less than 1 is

desirable. Some vendors, such as Master Bond Inc., have special epoxies that are NASA

approved for low outgassing and percent total mass loss [52].

M. ii: vendors are capable of taking a specific epoxy and modifying it slightly to

produce the desired mechanical, thermal, or chemical properties. For example, small

aluminum beads can be added to lower the CTE of an epoxy. Although this may lower the

CTE of the epoxy, it may cause the bonding strength to change [50].

There are several v- -i-s to adhere surfaces using epoxy. The strength of the bond will

be determined not only by the epoxy itself, but by the quality of the surfaces as well.

For high strength, all paint, oil, dust, etc.. should be removed. Often, a specific surface

roughness is needed for full strength. Although the vendor should be consulted to ensure

proper application of the epoxy, a few cleaning methods are -i-i .: -I. '1 below for a few

common surfaces as adapted from reference [50]. Steel or aluminum:

1. Wipe the surfaces from all contaminants using a solvent such as acetone or alcohol.

2. Abrade using clean fine grit abrasives (180 grit or finer).

3. Wipe again with solvents to remove any contaminants.

4. Apply a thin coating (typically less than 12 pm) of epoxy and cure according to
specific epoxy. Slightly elevated temperatures tend to produce a stronger bond.


1. Wipe with isopropanol. Using acetone may damage the plastic or rubber.

2. Proceed as with steps 2-4 above.


1. Clean using acetone, or solvent wipe. Allow time for surfaces to completely dry.

2. Apply a thin coating (typically less than 12 pm) of epoxy and cure according to
specific epoxy. Slightly elevated temperatures tend to produce a stronger bond.

To break the epoxy bond, mechanical, chemical, or thermal means may be employ, l1

although this may cause damage or destruction of the pieces. These methods include using

a pick, screwdriver, razor blade, heating the bond, or soaking in strong solvents such as



C'i ipter 3 describes a few select bonding techniques that have been used in a variety

of applications. Unfortunately, these techniques suffer from drawbacks which may not

make them ideal for some space-based interferometric missions. For example, optical

contacting works well over a large temperature range when small areas with similar CTE's

are used. Although optical contacting can be done between materials with differing CTE's,

changing the temperature at which the materials were bonded will induce stresses the

bond may not be able to handle. Brazing essentially makes the bonded components

as one, but the elevated temperatures needed may cause the material properties to

change, or may effect possible coatings applied to the material (as with a telescope for

example). A chemical bonding technique, colloquially known as hydroxide bonding,

provides strong bonds, is capable of using smaller surface areas that do not need to

meet the high polishing tolerances of optical ( ,nl 1 iiri .- and can be used for precision

alignment needed for optical benches [53]. This chapter will detail the hydroxide bonding

mechanism and present strength results for several materials that have been bonded using

the hydroxide bonding technique.

4.1 Introduction

Hydroxide-catalyzed bonding was originally developed at Stanford University by Gwo

to adhere optical components for the Gravity Probe B mission [54]. The characteristics of

hydroxide bonding make it ideal for optical benches due to the high bond strength while

still allowing for precision alignment. As described by Gwo, the hydroxide bonding process

is capable of forming bonds as long as a silicate-like network can be created between

the surfaces. The most obvious of materials for use of hydroxide bonding is Zerodur.

Extensive measurements have been made involving the mechanical and thermal properties

of hydroxide bonded Zerodur, specifically for the LISA mission [53]. Over the past few

years, this method has been expanded significantly and the list of materials suitable for

use with hydroxide bonding has grown.

4.2 Bonding Mechanisms

The hydroxide bonding mechanism takes place in 3 steps:

1. Hydration and etching: In this step, a hydroxide solution (typically NaOH or KOH)
is added to one of the surfaces where the free OH- ions act to liberate silicate ions.

2. Polymerization: The silicate ions dissociate and form siloxane chains.

3. Dehydration and bonding: As the water evaporates, the siloxane chains intertwine
and bond the 2 surfaces together.

Although the overall process for hydroxide bonding is the same, it varies slightly

depending on the materials and bonding solution used. In the case of glass-glass bonds,

the hydration and etching step is easily done since silica materials easily liberate silicate

ions in the presence of OH-. For metals, there are no silicate ions for the siloxane chains

to attach to, so the second step must occur in a different manner [53]. In the case of

SiC, which contains silicon but is extremely chemical resistant, the manner in which the

siloxane chains attach to the SiC surface must be different than that of glass-glass bonds.

In the following subsections a more complete analysis of the hydroxide bonding mechanism

will be discussed for the cases of glass-glass, glass-metal, metal-metal, and SiC-SiC bonds.

4.2.1 Glass to Glass

For glass-glass hydroxide bonding to take place, a small amount of an alkaline

solution such as sodium hydroxide or sodium silicate is placed on one surface and the

other silica material is placed in contact. The OH- in the bonding solution etches the

silica surface and liberates silicate ions. This process can be described by

SiO2 + OH- + 2H20 Si(OH)5. (41)

As the etching takes place, the number of OH- ions are reduced and the pH of the

bonding solution decreases. Once the pH is below 11, the silicate ions dissociate to form

Si(OH)- as described by

Si(OH)5 -- Si(OH)4 + OH-. (4-2)

The Si(OH)4 molecules then combine to form siloxane chains and water as described by

2Si(OH)4 (OH)3SiOSi(OH)3 + H20. (4-3)

The siloxane chains start to form as the water evaporates and provide the overall rigidity

of the bond [54]. As the dehydration continues, a 3D network of siloxane chains form and

eventually join the two surfaces together.

4.2.2 Glass to Metals and Metal to Metal

For other oxide materials which cannot form a silicate-like network on the surface, an

existing network can be introduced into the hydroxide solution for the surface to bond to.

This is typically done by using a sodium silicate solution with ,1' NaOH and ''7'. SiO2

as the bonding solution. This solution is readily available from companies such as Sigma

Aldrich. The SiO2 in the solution forms the necessary network for the surfaces to attach

to. It is critical for the materials to be easily oxidized, as the oxygen will provide the link

for the siloxane chains to bond to. This makes many metals such as aluminum, copper,

and silver possible to bond to silicate materials using the hydroxide bonding technique.

It is a simple step in logic that the silicate material could also be replaced with another

readily oxidized material to form a hydroxide bond between two non-silica containing

materials. This has been achieved with materials such as copper and aluminum [54].

4.2.3 SiC to SiC

Although SiC is a material which largely consists of silicon, its extreme chemical

resistance to oxidation and alkaline solutions should not allow SiC to bond using the

hydroxide bonding technique. It has been shown in reference [55] that hydroxide bonding

can take place with limited results by growing an SiO2 l1,-r on the SiC surface by heating

the SiC to temperatures in excess of 10000C. The oxidized SiC li. r allows the siloxane

chains to bond the two surfaces in the same manner that glass-glass bonding occurs.

Although promising results were obtained using this method, the oxidation of the SiC

makes it unattractive for use in space-based missions, as this would ruin any optical

coating that may be necessary for components such as mirrors on the optical bench. For

this reason other methods to make SiC-SiC bonds using the hydroxide bonding method

were investigated. As it turns out, the oxidation step is not necessary and bonding with

significant strengths can be done by simply using a sodium silicate solution. The reason

as to why this should happen was at first perplexing, since there should be no oxide 1lv,,T

on the SiC surface to bond to, and virtually no silicate ions should be liberated from the

alkaline solution.

In order to determine the bonding mechanism between SiC and SiC, the chemical

compositions of the surfaces were analyzed both before and after bonding. This was

easily done as the SiC-SiC bond almost ahv--, broke at the bond line. To determine

the surface composition of the silicon carbide surfaces during SiC-SiC bonding, an

X-Ray Photoelectron Spectroscopy (XPS) system was used. The XPS system is

based on the photoelectric effect and is the least destructive of all the electron or ion

spectroscopy techniques. The XPS is routinely used in industry and research when

elemental or chemical state analysis of a surface needs to be done. For a solid, XPS probes

approximately 5-500 A deep, depending on the material, the energy of the x-rays and

emitted photoelectrons, and the angle with respect to the surface of the measurement. X-ray photoelectron spectroscopy

The XPS works on the basis of the photoelectric effect. Photons of energy hv

(typically 1486.6 eV in most commercial systems) produced from an x-ray source saturate

the surface. The photo-ionization process begins with the photon absorption which then

causes an electron emission. This electron then either travels from the surface or through

the material into the vacuum where it is detected by a spectrometer and its energy is

measured. Using the principles of conservation of energy, we find that the binding energy

of the electron to first order is simply

KE = h BE (4-4)

where KE is the kinetic energy of the photoelectron, hv is the energy of the x-ray, and

BE is the binding energy of the particular atom concerned. Thus, element identification

can be done by analyzing the photoelectron energy distribution. The measured spectrum

will exhibit photoelectron signals attributed to core-level (Ols, Cis, Fe2p, etc...), valence

level (Fe3s, Fe3p, etc), and Auger emission. Auger emission occurs from the filling of the

core-level vacancy by an electron in an outer l-1-r state. As this electron relaxes, energy is

emitted. If the energy is absorbed by another electron, this electron will be ejected from

the material and detected on the spectrometer. The Auger peaks are much broader than

the photoelectron peaks due to the multiple paths the electron may escape from within the


The tightly bound core-level electrons of an atom have energies that are nearly

independent of the chemical species in which the atom is bound because they are not

involved in the bonding process. Thus, a compound will have core-level energies that

are similar to the elemental energy values. For example, in nickel carbide, the CIs BE is

within a few eV of its value for elemental carbon. In addition, the Ni2p BE is within a

few eV of its natural value for Ni metal. Identifying the core-level BE's of certain elements

will provide signatures of the elements in the compound. Although cross section values are

needed to determine relative atomic concentrations, when comparing relative intensities

of the same atomic core-level to get compositional data, the cross section values do not

need to be taken into account. Thus the relative abundances can be taken directly from

the relative peak heights. For example, figure 4-1 shows the sample constituents as well

as their relative abundances. These numbers are calculated using the cross sections of the

elements and their relative peak intensities. Figure 4-5 on the other hand only represents

the Cis orbital and the relative concentrations can be directly taken from the area under

the peaks. In this case, peak 1 and peak 2 are of similar abundances.









Figure 4-1.

0 I I I I
1000 800 600 400 200 0
Binding Energy (ev)

Spectrum of an unbonded CoorsTek SiC sample. The peaks from left to right
are the Auger carbon peak, Auger oxygen peak, oxygen Is peak, carbon Is
peak, and the silicon 2s and 2p peaks.

The x-ray source of the XPS will produce photoelectrons to the depth that it

penetrates the sample. Photoelectrons that are ejected from the top l. -ir of the sample

will be detected with minimal scattering and show up as peaks. The electrons that are

released deeper within the sample will interact with the sample and multiple scattering

will take place, reducing their chances to make it to the detector. These electrons show up

at lower KE and mostly end up in the background after the XPS peaks (Figure 4-2).

Thus, peaks come primarily from the surface photoelectrons, while the background

is a result of the scattered electrons within the material. The intensity of the flux of


Background Step
(Scattered Electrons)

KE (eV)

Figure 4-2. Kinetic energy distribution obtained due to the inelastic scattering of
electrons. The peak corresponds to the photoelectrons emitted near the
surface, while the signal at lower kinetic energies results from the scattering of
the electrons deeper within the sample surface.

electrons as a function of depth and electron angle emission is given by
I = 0in() (4-5)

where Io is the flux of electrons originating at depth D, ID is the flux emerging without
being scattered, D/sinO is the distance traveled through the solid at that angle, 0 is
the angle of electron emission, and A is the inelastic mean free path. The value of A
depends on the KE of the electron and the material through which it travels. It also
determines what the surface sensitive of the measurement is. The values of A can be
found in standard tables from books [56]. To detect the surface composition, the grazing
emission angle, 0, can be lowered, while detecting deeper l~-. -is requires a larger angle. SiC-SiC bonding mechanism

The individual SiC pieces were first cleaned and then analyzed using the XPS. The

SiC pieces were bonded and then broken. The broken SiC samples were cleaned and

analyzed again. Cleaning the SiC pieces after they were broken removed almost all of the

sodium that was in the bonding solution. This was verified using the XPS and finding

either no sodium peak, or a very small peak which was typically less than a few percent of

the total chemical composition.

Samples of Hexoloy SA from Saint Gobain, POCO Super SiC from POCO graphite,

and UltraSiC from CoorsTek were analyzed in the XPS. An integrated background was

subtracted from the raw data and Gauss-Laguerre peaks were fitted using AugerScan

software. Once the energy of the peaks were found, they were matched to the NIST online

database to determine the chemical composition. Although some of the peaks did not

match up exactly with the NIST database, all peaks were within 0.10-0.15 eV of the

database, with the next suitable candidate being significantly further away. From this

information a bonding mechanism was formulated consisting of the following steps:

1. The hydroxide ions in the sodium silicate solution break the SiO bonds on the
SiC surface. The SiO is present on the surface possibly due to either the polishing
process or from oxidation during the manufacturing process.

2. The broken SiO bonds then attach to the SiO2 in the sodium silicate solution.

3. As the water in the solution evaporates, the aqueous SiO2 forms a solid 1.- r that
binds the two SiC surfaces together.

The end result is that two pieces of SiC are bonded together by a l- -r of SiO2 [57]. This

can be seen from the surface composition of the SiC before and after bonding. Figure

4-3 shows the Ols orbital energies both before and after bonding. Before bonding, only

the SiO peak is present. After bonding, an SiO/SiO2 peak is present [57]. This would

indicate that a bond between the SiO and Si02 has occurred. Due to the different number

of detected counts before and after bonding, the figures show the normalized data. While

this will not show the relative abundance of compounds both before and after bonding,

it shows the relative abundance of the chemical compounds at the time the sample was

tested. Analyzing the Si2p3 orbital, we find that only SiC and SiO are present before

bonding, but after bonding an SiO2 peak is found (Figure 4-4). In a similar manner,

analyzing the Cis orbital, we find that only SiC and graphite are found before bonding,

but an additional SiO2 + SiC peak is present after bonding (Figure 4-5). This supports

the bonding process presented above.

0.9- SiO/SiO2 / SiO
(D /
.N 0.6

0.3 /
0.2 /
0.1 /

536 534 532 530 528

Figure 4-3. Ols orbital both before (blue curve) and after (red curve) bonding. The
dashed and dotted lines represent the peak fit values while the solid lines are
the raw data taken from the XPS. The unbonded peak is at 532.61 eV and the
bonded peak is at 532.96 ev.

Using the XPS to probe deeper into the samples, it was found that as the depth

increased, the effects of the SiO2 diminished, and the chemical compositions became

identical. This mechanism also explains the poor bonding results when an NaOH solution

is used. Since SiC is extremely resistant to both acids and bases as well as being an

extremely poor oxidizer at room temperature, there is no intermediate material that is

either produced or present during bonding when an NaOH solution is used.

.N 0.6
-FU / : \
I5 Si02
S0.4 ,

0.3 -
/ \.
0.2 -

106 104 102 100 98 96

Figure 4-4. Si2p3 orbital both before (blue curve) and after (red curve) bonding. The
dashed and dotted lines represent the peak fit values while the solid lines are
the raw data taken from the XPS. The unbonded peaks are at 102.87 eV and
100.60 eV. The bonded peaks are at 100.42 eV, 103.82 eV, and 102.21 eV.

4.3 Bonding Method

In order to get the best bond, it is important to keep the surfaces as clean as possible.

In addition, the smoother and flatter a surface, the more likely a strong bond will form.

Although there are many different methods to clean the surfaces, and many surfaces

capable of bonding, below describes the bonding method used for the hydroxide bonding


1. Preparation of the surfaces: The surfaces were cleaned as described in Appendix B.

2. Application of solution: After the surfaces have been cleaned and sonicated, they
are then taken out of the sonic bath and cleaned again with an optical wipe and
methanol. This step helps to remove any particulate that may still be on the
surfaces as well as removing the surface of any DI water left from the sonic bath.

3. Jointing of the 2 surfaces: After the surfaces have been cleaned, one piece is placed
on a leveling table and the solution is then applied. The other piece is then placed
on top. A leveling table is used to avoid any slipping of the bonding pieces after the
bonding solution has been used. A small weight (typically a few to several ounces)
can then be placed on top of the pieces, but this is usually only needed for surfaces

0.9h I. -
S 7 SiC
N 0.6

0.2 -

0.1 ,\V..
280 281 282 283 284 285 286 287 288

Figure 4-5. Cis orbital both before (blue curve) and after (red curve) bonding. The
dashed and dotted lines represent the peak fit values while the solid lines are
the raw data taken from the XPS. The unbonded peaks are at 282.79 eV, and
285.12 eV. The bonded peaks are at 285.48 eV, 2 : ; eV, and 282.11 eV.

that are not very flat. If a weight is not ideal, then pressure can be administered by
pressing down on the top piece. The pieces are then not disturbed for at least 18

While all preparation and bonding was done in a clean room, any clean area, such as a

vented hood, will suffice. A clean room reduces the chance of particulates contaminating

the surface which may lower the bonding strength.

4.4 Strength Measurements

For all materials tested, the mode of failure with the least amount of applied force will

be parallel to the bond area. By testing the shear strength of the hydroxide bonded

materials, a lower bound on the breaking strength can be determined. To test the

breaking strength, two different apparatuses were used. The first consisted of a lever

arm with a 3 mm diameter stainless steel wire on one end, and a spring-based force meter

with a maximum force indicator (a spring scale from Cabela's) on the other as shown in

Figure 4-6. The lever arm is approximately 39 cm long and has a 5:1 torque advantage.

The bonded pieces were held in place by a slot machined in a solid piece of aluminum that

was held in place using a clamp that was mounted to a table.

The wire was then wrapped around the bonded piece and a downward force was

applied on the other until the test piece broke. The applied force was measured by reading

the maximum force indicator from the spring scale. It should be noted that similar to

experiments in reference [53], by placing the wire a few millimeters from the bond, a

slight torque or peeling effect was created on the bonding area perpendicular to the

direction of the shear stress. After extensive testing using this apparatus, promising

results were obtained. However, the variation of the measured bond strength was higher

than expected, so a more precise method of measuring breaking strengths was used. This

method uses a modified compression shear test (\!CST) as described in reference [58]

(Figure 4-6). A linear actuator with an 8 mm/s travel speed was used to apply the force

while a LabVIEW program sampled the amplified output of an MLP series load cell from

Transducer Techniques at 1 kHz. Sampling at this rate showed a clearly defined peak

and produced an error in the applied force measurements of less than 1.5 lbs. In every

instance, the range of the measurements was greater than the error produced from the

measuring device. To compare the lever arm apparatus and MCST, eight BK7-BK7 bonds

were made using 0.50 pL of solution and then half were broken in each device. For each

apparatus, the range of breaking strengths were similar, but the average using the MCST

was higher. This is most likely due a slight cutting in the glass from the wire used in the

lever arm apparatus as the force is being applied, which causes the glass to break at lower

strengths. In our results both the data obtained using the lever arm apparatus and MCST

are used.

4.4.1 BK7-BK7 Bonding

If hydroxide bonding is to be used on space-based interferometric missions, the bond

must be able to withstand the significant accelerations endured during launch. A typical

optical component on an optical bench with a bond area of 100 mm2, an approximate

Figure 4-6. Lever arm apparatus used to initially test the hydroxide bonded shear
strengths (left) and MCST apparatus used for later tests (right).

mass 10 mg, and subject to 35g accelerations during launch conditions results in a

requirement that the bonds be able to withstand on the order of tens of kiloPascals.

Results from the University of Glasgow have shown hydroxide bonds to be significantly

stronger than what is required [53], although more research is needed to better understand

the process and its limitations. In this section I elaborate on existing hydroxide bonding

results to gain a more complete understanding of this process.

While Zerodur or ULE are usually used for optical components in space-based

interferometric missions, BK7 glass has a similar molecular structure and composition

to these materials and bonds in the same manner with similar breaking strengths [45].

Consequently, BK7 provides us with a cost-effective alternative to Zerodur or ULE to

test the hydroxide bonds. The BK7 glass rods (CTE of 7 to 8x10-6/C) are 12.7 mm in

diameter and 10 mm thick. One side was polished to a A/8 global flatness with a 20-10

scratch-dig finish, while the other side had a ground polish. This allowed the testing of

bonds with both rough and smooth surfaces.

The dehydration of the water molecules in the hydroxide solution pl 'i- an important

role in the hydroxide bonding process. As the water evaporates or moves into the bulk

of the material, the siloxane chains form a 3D network and join the two surfaces. If it

takes appreciable amount of time for the bonds to reach their maximum strength, then

the bonds must be left to cure for this time before their breaking strength should be

measured. In addition, if the dehydration process occurs over an extended period of time,

then this may cause the bonded pieces to shift slightly and may cause problems when

used for precision alignment for uses such as optical benches. To determine the effects

of dehydration time on the shear strength, multiple samples were prepared using 0.40

pL/cm2 of a volumetric 1:4 sodium silicate to deionized water solution between two BK7

pieces using the unpolished side. Each sample had 1.27 cm2 of bonding area and -I li-,,

in the clean room approximately 18 hours. The shear strength of one bonded piece was

then tested using the lever arm apparatus. The other pieces were kept in an oven at 50C

until broken in the same manner. Two bonded pieces were also left in the clean room for

11 d iv to determine if the elevated temperature within the oven had any effect on the

bonding strength (see Figure 4-7). The temperature the oven was set at was based on

previously bonded samples that had been heated to various temperatures to determine

what temperature range the hydroxide bonds could survive [59]. Most either failed or

weakened significantly above 100C. Using an oven temperature of 50C will expedite the

dehydration process, but shouldn't cause the bond to weaken or fail.

It appears the dehydration time has no appreciable effect on the breaking strength

when using a sodium silicate solution and the bond reaches significant strength after

approximately 18 hours [59]. After breaking the pieces that were left in the clean room

for 11 d i-- it was found that they had a similar breaking strength to the pieces left in

the oven for the same amount of time. It should be noted that when breaking the bonded

glass pieces with the lever arm apparatus, the failure was almost albv-- a few millimeters

above the bond where the wire on the apparatus applied the force on the glass, and not at

the glass to glass interface.

* In oven, failure at glass
* In oven, failure at bond
* In clean room, failure at glass
* In clean room, failure at bond

2 4 6 8 10
Time (days)

Figure 4-7. Breaking strengths of BK7-BK7 bonds after being kept in an oven at 330 K.
The zero on the x-axis represents the time after the samples were in the clean
room for approximately 18 hours. All data points were taken using the lever
arm apparatus.

It was -Ii--.: -I. l in reference [53] that the water in the hydroxide solution may be

weakening the glass material, causing the fracture to occur in the glass and not along

the bond interface. Through extensive testing of multiple samples, it is believed that

this effect is most likely caused by the small surface area of the wire that makes contact

with the glass, which in turn, creates a much higher localized pressure at the wire-glass

interface, and not over the bond area. This results in the wire cutting or scoring the glass,

0 *


* U

which then creates a small crack. As more force is applied, the micro-crack propagates

through the BK7, causing the glass to break at the glass-wire interface.

To test this, multiple samples that had been broken at the glass-glass interface after

having their shear strength tested had their bulk shear strength tested using the MCST.

The values of these data points were compared to the bulk shear strengths of glass pieces

broken using the MCST that had not been bonded. Both bonded and unbonded samples

had similar shear strengths. From the similar breaking strength values, it is assumed that

the water in the hydroxide solution does not weaken the glass in an appreciable manner, if

at all [59].

The primary advantage the MCST has over the lever arm apparatus is that the

MCST applies the force evenly over the glass rod, which reduces localized forces and

scratching. Comparing the breaking strengths of the bonded BK7-BK7 samples using the

MCST and the lever arm, it was found that the MCST breaking strengths were higher

than those using the lever arm. In addition, it was also found that all pieces tested using

the MCST broke at the bond, and did not break the glass. This supports the assumption

that the reason the glass pieces were breaking at the glass-wire interface when using the

lever arm device is because of a cutting effect produced when using the wire, and not due

to a weakening of the glass from the water used in the bonding solution.

It was shown in reference [53] that the molarity of the hydroxide solution was varied,

a maximum breaking strength was found. This value was different each alkali metal used.

Another parameter that could effect the bonding strength is the amount of solution used

per square centimeter. Using a volumetric 1:4 sodium silicate to deionized water solution

and varying the amounts of solution used from 0.20 pL to 1.25 pL while keeping the bond

area constant and using the polished surfaces of the glass rods, multiple samples were

made and kept in a clean room at least 18 hours before being tested. The data consisted

of three bonded pieces using 0.20 pL of solution, seven pieces using 0.50 pL, four pieces

using 0.75 pL, five pieces using 1.00 pL, and two pieces using 1.25 pL.

The majority of the data fell close to each other within approximately .'. of their

mean, but in one case the deviation was as high as 91i'. [59]. In this case, it was a single

data point that was well above the average. From the data it was found that there was

no significant correlation between the shear strength of the bonds and the amount of

solution used. It should be noted that as the amount of solution used was increased, more

solution would spill out of the sides. Unfortunately there was no way of determining how

much fluid was spilling out. The average of the 21 samples was 3.2 MPa with a range

of 1.7 to 7.6 MPa. Comparing the average breaking strength of the hydroxide bonds to

those obtained in reference [45] by optically contacting similar materials, the average

breaking strength of the hydroxide bonds is approximately 2.5 times greater using the

hydroxide bonding technique when compared to optical contacting. In addition, the

greatest breaking strength found in reference [45] was comparable to the lowest breaking

strength when hydroxide bonding is used.

4.4.2 SiC-BK7 Bonding

While glass to glass bonding for multiple glass materials has been studied for both

optical contacting and hydroxide bonding, only limited experimental results are available

for SiC bonds using hydroxide bonding [60] [61]. To determine the shear strength for

SiC-BK7 hydroxide bonds, a volumetric 1:4 dilution of sodium silicate to DI water was

used. The amount of solution was varied while keeping the bonding area constant and

using the polished side of the glass rods. The SiC tiles are made of Hexoloy SA (CTE

of 4x 10-/C) are 50.8mm on each side and 6.4mm thick with a A/8 global flatness and

60-40 scratch-dig on both large sides. The SiC tiles were cleaned using cerium oxide and

sodium bicarbonate as described in Appendix B. Although the surfaces were cleaned

using cerium oxide, similar results should be obtained using only a sonic bath to clean

the SiC. The data consisted of three glass rods bonded using 0.12, 0.35, 0.60, 0.75, 1.00,

and 1.20 pL of solution, five rods at 0.25 pL, and four rods at 0.50 pL of solution. The

typical deviation from the average for each amount of solution used was below 50'(. As

with the BK7-BK7 bonds, there appears to be no significant correlation between the

amount of solution used and the breaking strength of the bonds. The average of the 27

data points is 3.0 MPa with a range of 1.0 to 5.0 MPa [62]. For a comparison between the

breaking strength of SiC-BK7 hydroxide bonds and SiC-BK7 optically contacted bonds,

the breaking strength of one optically contacted SiC-BK7 bond was broken and found to

have a breaking strength of a little under 1 MPa. More data is need to draw a conclusion

on the comparison between optically contacted and hydroxide bonds between SiC and

BK7. It should be noted that optically contacting SiC to BK7 is difficult to do with much



D *
D 0

O BK7-SiC (1:4) MCST
H BK7-SiC (1:4) Lever Arm
BK7-SiC (1:6) MCST
BK7-SiC (1:6) Lever Arm

0.4 0.6 0.8 1
Amount of Solution Used (uL)

0 0.2

1.2 1.4

Figure 4-8. Averages of the data obtained using the lever arm and modified compression
shear test (\ICST) devices for both a 1:4 and 1:6 dilution factor. Dilution factor

It may be possible that the shear strength is effected by the concentration of the

solution used. Using a dilution factor of 1:4 may produce a stronger bond than a 1:6

dilution factor since the 1:4 solution will have a greater bond thickness and may allow the

siloxane chains to bond to the SiC surface more readily. To test this, multiple samples

were prepared and bonded in the same manner as above, although a 1:6 sodium silicate

solution was used instead of a 1:4 solution (Figure 4-8). Two glass rods each bonded using

0.12, 0.40, 0.60, and 1.30 pL of solution, six rods using 0.25 pL, and three rods at 0.50,

0.75, and 1.00 pL. The variability in the measurements -1 ,, .1 consistent as with what was

obtained above and no significant correlation between the amount of solution used and the

shear strength could be observed [59]. The average of the 23 data points is 3.4 MPa, with

a range of 1.0 to 6.3 MPa. Surface profile

The surface profile of the bonding surface pl i an important role in determining

what solution should be used. If the surface is too rough, then a bond cannot form. A

sodium silicate solution can be used as a filler/intermediary to fill the gaps and allow

the siloxane chains to bond. The data taken above used the polished surface of the

glass. Using the rough surface may produce different breaking strengths since there is a

greater probability the bond will not completely form and gaps will be created within the

bond[53]. To test this, three glass rods were bonded using the rough side of the glass with

0.25 pL of 1:4 sodium silicate solution, two rods at 0.40, 0.75, and 1.00 pL, and six rods

were bonded using 0.50 pL of solution. The variation was again comparable to previous

measurements with an overall mean of 3.0 MPa and a range of 1.1 to 7.3 MPa. This is

comparable to what was achieved when using the polished side of the BK7 [59]. The fact

that a ground finish will suffice to bond these materials and still produce a substantial

bond has a significant cost-reducing effect. It is often expensive to polish glass or SiC

flat and smooth enough to obtain a strong optical contact, especially as the surface area

increases. Using hydroxide bonding allows the faces of the pieces to be bonded to have

significantly less demanding polishing tolerances, which, in turn, reduces the cost of the


4.4.3 Super Invar-BK7

Bonding glass to metals with appreciable strength in a non-destructive manner

can be a difficult process. While anodic bonding or frit bonding can be used for some

glass-metal bonds, it often requires the metal to be very thin, or that the pieces be heated

to several hundreds of degrees Celsius [63] [64] [65], or the glass and metal have close

CTE's. This makes it virtually impossible to bond bulk materials with differing CTE's in

a non-destructive manner. Using the hydroxide bonding technique allows metals ranging

from aluminum to copper [66] [67] to Super Invar to be bonded to glass while retaining

their thermal and mechanical characteristics. To test the shear strength of the Super Invar

BK7 hydroxide bond, two bonds using 0.20, 0.40, 0.60, 0.80, and 1.00 pL of 1:4 sodium

silicate solution were made using the polished side of the glass rods. The Super Invar

tiles (CTE of 0.7x10-6/C) are 38.1mm on each side and 6.4mm thick and polished to a

A/4 global flatness and 60-40 scratch-dig on both large sides. The Super Invar tiles were

cleaned using cerium oxide as described in Appendix B. All bonds were tested using the

MCST and no significant correlation between the amount of solution used and the shear

strength could be obtained. The average of the measurements is 5.8 MPa with a range of

2.3 to 8.1 MPa.

In addition to these test, anecdotal experiments were done to determine what thermal

stresses the Super Invar-BK7 bonds could endure. In one instance, a BK7 glass window

25 mm in diameter and 6 mm thick was bonded to a Super Invar tile and then heated to

100C at 0.700C/min, left for 1 hour, and then quenched in ice water. Although the glass

window broke, pieces of the glass were still bonded to the Super Invar. In another case, a

12 mm diameter plano-convex lens was bonded to another Super Invar tile and heated in

the same manner. After being quenched in ice water, no visible degradation of either the

bond or the materials could be detected. This remarkable property needs to be studied in

greater detail since it may be possible the siloxane chains in the hydroxide bond may be

reducing the thermally induced stresses of the materials.

4.4.4 SiC-SiC

A variety of tests were performed to characterize the strength and durability of the

SiC to SiC bonds on three types of SiC. The three SiC materials tested were Hexoloy SA,

POCO SuperSiC, and CoorsTek UltraSiC. To determine the shear strength of the SiC-SiC

hydroxide bonds, the MCST was used on all tests. Hexoloy SA

A 50.8 mm x 50.8 mm x 6.4 mm tile of Hexoloy SA was obtained from Saint-Gobain

Ceramics. It was polished to a A/8 global flatness with a mirror quality surface finish. A

section 9mm x 6mm was cut from this tile and both polished faces were bonded using 1.0

pL of 1:4 sodium silicate solution. The resulting bonded piece was left to cure for several

weeks. It was then cooled to approximately liquid nitrogen temperatures. Although the

bond strength was not tested while it was cooled, the bond could not be broken trying

to pry it apart using hand strength. From previous testing experience, this results in a

breaking strength that is greater than approximately 100 kPa [68]. The SiC was then

allowed to return to room temperature and the bond strength was tested and found to be

6 MPa.

Another section of the same tile 10 mm x 10 mm was cut and the flat faces were

bonded using 1.0 pL of 1:4 sodium silicate solution. The bonded piece was then left to

cure for 10 dv4 and was cut with a diamond tile saw rotating at 3600 rpm. The cut was

made approximately 3mm from the outside and took roughly 45 minutes to perform. No

degradation of the bond could be seen and the resulting cut SiC could not be pried apart

using hand strength. Although the bond retained significant strength after cutting, it may

be possible that prolonged exposure to water or other cutting fluids may cause the bond to

fail. The bonded pieces were then dipped in liquid nitrogen for several seconds and then

taken out for a few seconds. This was done multiple times. The bond could not be broken

using hand strength after this process. POCO SuperSiC

Although it was claimed in reference [55] that the surfaces need a peak-to-valley

flatness of <60 nm, it was found that it is possible to bond with surfaces as rough as

24 pm over a 2 mm length. An unpolished disc 3 mm thick and 50.8 mm in diameter of

SuperSiC was obtained from POCO Graphite. The surface roughness was found to be as

large as 24 pm over 2 mm (figure 4-9). A 10mm x 10mm section was cut from this piece

and bonded using 3.0pL of solution. The resulting shear strength was found to be 2.45

MPa [68]. CoorsTek UltraSiC

Two 50.8 mm x 50.8 mm x 7.6 mm and six 10 mm x 10 mm x 7.6 mm SiC samples

were obtained from CoorsTek. They were polished at the CoorsTek facilities and the larger

tiles were found to have a flatness between 0.95 pm and 0.99 pm, both with a surface

finish of 0.05 pm Ra. The smaller SiC pieces were found to have a flatness ranging from

0.07 pm to 0.30 pm with a surface finish ranging from 0.03 pm Ra to 0.09 pm Ra.

The smaller SiC pieces were bonded to the larger SiC tiles using differing amounts of

sodium silicate solution and the shear strengths were measured. Two pieces were bonded

using 1.00 pL of solution and had breaking strengths of 7.12 MPa and 4.00 MPa. Two

pieces were bonded using 0.50 pL of solution and had breaking strengths of 5.78 MPa and

m mTnir I an

A -y

r MM-- o11 I I I=-i--
-- il i AI -Im

Figure 4-9. Surface profiles of areas on the POCO SuperSiC using a Tencor surface
profiler. Image courtesy of Petar Arsenovic at Goddard Space Flight Center.

100 200 300 400 500
Scan Length (pm)

600 700 800 900 1000

Figure 4-10.

Surface profile of one of the large CoorsTek SiC tile using a Tencor surface
profiler. Image courtesy of Petar Arsenovic at Goddard Space Flight Center.






-80nm -



-l0OOnm ____

-1800nm ____
0 100 200 300 400 500 600 700 800 900 1000
Scan Length (im)

Figure 4-11.

Surface profile of one of the small SiC pieces using a Tencor surface profiler.
The large spike around 750 pm is most likely a speck of dirt. Image courtesy
of Petar Arsenovic at Goddard Space Flight Center.

3.25 MPa. One piece was bonded using 0.25 pL of solution and had a breaking strength of

2.22 MPa [68].

It was previously shown that by using this testing method of the shear strengths,

the typical variance in breaking strengths was approximately .' The variation in our












measured breaking strengths could be from either a lack of samples, or from a possible

roll-off of surface curvature in some of the samples due to the polishing process. Although

the surface finishes and flatness were all measured to be similar, the sampled areas were

only taken at random places and only in local areas. Hence, although the local areas

tested all had similar profiles, the overall profile (taken over the entire surface) was not

measured. In some cases, a noticeable rolling off of the edges could be seen by visual

inspection. When a roll off in surface profile could be noticed, the breaking strength of the

bond was significantly less than those when no roll off could be noticed, holding all other

bonding factors constant.

Another piece of SiC was bonded using 1.50 pL of solution. The bonded pieces were

cured overnight and then cooled to 77 K in a 2-step process. In the first step, the SiC

was placed on an aluminum block that sat in a small reservoir. The reservoir was filled

with liquid nitrogen which slowly cooled the aluminum block. As the temperature of the

aluminum block was lowered, the temperature of the bonded SiC piece fell as well. Once

the boiling rate of the liquid nitrogen appeared to be constant, the bonded SiC piece was

taken off the aluminum block and dipped a few times into the liquid nitrogen for a few

seconds. The bond showed no noticeable signs of degradation. Although the SiC piece was

not tested for its shear strength after it warmed back to room temperature, it could still

not be broken using hand strength, and is assumed to have similar breaking strength to

the pieces tested above.

In addition to cooling the bonded pieces, the sample that was cooled was also heated

to determine the effects of heating on the bond. The sample was heated in an oven from

room temperature to 80C at 5C/min and allowed to soak at 80C for 1 hour. After this

was done, the bond could not be broken using hand strength. The bond was then heated

to 100C at 5C/min and allowed to soak at 100C for 30 minutes. The bond could not

be broken using hand strength. The sample was then further heated to 130C at 5C/min

and allowed to soak for 30 minutes. The bond was then broken using a significant amount

of strength. Although this bond was not broken using the MCST, based on previous

bonding experience, the bond strength at 100C is at least 100 kPa.

4.5 Other Bonding Results

In addition to these breaking strength measurements, hydroxide bonding was

attempted between other materials, although the shear strength of these bonds were

not tested. Successful bonding occurred between copper-copper, aluminum-aluminum,

and a silver coated PZT material and BK7. In many cases, the surfaces of these materials

were not polished and in the case of the aluminum-aluminum bond, the surface polish was

done by a mill in the machine shop. Although the breaking strengths were not tested, they

could not be broken using hand strength, implying a shear strength of greater than 100

kPa. In addition, a bond between a strip of nichrome and aluminum was attempted, but

the bond failed. This supports the theory that metals that are oxidized easily are capable

of hydroxide bonding, since nichrome has an extreme resistance to oxidation. This also

supports the theory that the oxide p1 i' a critical role in the bonding process, and not the

material itself as nichrome has a composition with elements in it that are present in other

materials that have successfully bonded.


Continuous-wave (cw) lasers have 1pl id, d a n i, i". role in advancing precision

measurements in many fields. Frequently, the linewidth of such lasers is not adequate

for irn nir: applications without active stabilization of the laser frequency. For example,

some atomic clocks require extremely stable laser sources to probe sub-Hertz linewidths

available in laser cooled samples [69]. Interferometric gravitational wave detectors, such as

LIGO and LISA, require laser systems with extremely low frequency and amplitude noise.

Through the use of an optical cavity and feedback control system, the stability of the

cavity can be transferred to the laser which can then be measured against another stable

laser source. In this chapter a versatile feedback control known as Pound-Drever-Hall

(PDH) frequency stabilization will be discussed for its application to measure the stability

of select materials on the femtometer level. A brief discussion of optical resonators is

presented as well as how the optical cavities were made in order to test the dimensional

stability of several materials. Lastly, the relative dimensional stability of Zerodur, Super

Invar, SiC, CFRP, and a tunable PZT-actuated cavity are presented.

5.1 Optical Cavities

The core concepts needed to make measurements at the femtometer level revolve

around knowledge of both optical resonators and PDH frequency stabilization. In this

section, a few select concepts of optical cavities is presented. Referring to Figure 5-1, the

laser fields of an optical cavity are described by

El = itlE, + rlE2

E2 r2Elei2L(2f)/c

Et = it2EeiL(2Wf)/c

Er itlE2 + E, (5-1)

where f is the laser frequency, L is the length of the cavity, c is the speed of light, and ri

and ti are the reflection and transmission coefficients of mirror i. The transmitted and

Ei E

E r E 2

rl, t r2 t2

Figure 5-1. Electric fields inside and outside of an optical cavity.

reflected transfer functions, Tt(f) and T,(f), can be found by solving Eq. 5-1 in terms of

Et/Ei, and Er/Ei,, respectively. Doing this results in

-tl_ -7 rLIc
1 rlr2e2i(27f)L/c
2 + tT 2 i2 2 1 7 (')L/c
T,(f) -' (5-2)
1 rr1262i(27f)L/c

For a lossless cavity (r, + t 1) the transfer function of the reflected field simplifies to

1 -- ')L/c
T,(f) 1- i ,L/c. (5-3)

Eq. 5-3 will present resonant behavior when T,(f) is a minimum and occurs when

2(2x f)L
( 27n), neN. (5-4)

This equation can also be written as

fn = n- (5-5)

The frequency difference between neighboring resonant modes is constant and is known as

the free spectral range (FSR). It is given by

n+1 n c
FSR = f, f,= c c (5-6)
2L 2L 2L

L 0.6

t 0.5
H 0.4

| 0.3

( 0.2 J -


0 ---- 1 1 -
0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8
Frequency (Hz) x 109

Figure 5-2. Plot of the reflected transfer function for a lossless cavity 30 cm long with
both mirrors having reflectivities of r 0.90. The FSR is 500 MHz.

In addition to the FSR, another important parameter of optical cavities is the finesse,

F. The finesse is defined as the ratio of the FSR to the linewidth. This can also be

represented in terms of the reflectivities or transmissivities by

FSR w r2
F (5-7)
linewidth 1 rir2 tl + t2

The mirrors used for the optical cavities were made from fused silica and were quoted by

the vendor (AT Films) as having transmissivities near 360 ppm.

5.1.1 Stability of Optical Resonators

Light within the cavity will bounce back and forth between the mirrors and

experience a periodic focusing action. Depending on the radius of curvature of the

two mirrors, the periodic focusing can be classified as either stable or unstable [70]. A

cavity with spherical mirrors is an example of a periodic focusing that can result in either

stable or unstable behavior. Using what are known as the g parameters, the stability of

= 1-L g2 = 1
g 1R, g2 R2

Figure 5-3. The resonator g parameters.

the resonator can be determined. Each mirror is given a g-value defined by

gi = 1 (5-8)

where L is the length of the cavity, and Ri is the radius of curvature of the ith mirror (i=1

is the front mirror and i=2 is the back mirror). For a cavity to be a stable resonator, the


0 < g2 < 1 (5-9)

must be met. The waist spot size, o0, of the resonator can then be written in terms of the

g parameters, namely
SLA gig2(1 glg2)
Uo 2 2. (5-10)
V (91 + g2 2gg2)2
The spot sizes at the ends of the resonator are given by

2 LA / g2
i r gi(1 gig2)

2 LA 15 g
2 (5- 11)
7 92t2 9192)

In the case when the front mirror is flat (as is the case with all cavities presented in this
work), then gi=1 and Eqs. 5-11 reduces to

2 2 L 92

2 LA. (512)
S g2(1 g2)

Knowing only the length of the cavity and the radius of curvature of the two mirrors
will determine the resonant mode of the cavity. From this data the laser beam can be
mode-matched using lenses to fit the mode of the cavity.
5.1.2 Misalignment Analysis

In addition to specifying the RMS polishing tolerances of the spacer ends, the
parallelism needs to be specified as well. The mirrors have a high reflectivity (HR) coating
for the center 5 mm. If the displacement of the optical axis is too large, then the beam
may be misaligned so much that the beam does not hit the coated area on the mirror.
This would produce a cavity with a lower finesse in addition to producing additional noise.
For this reason, parallelism tolerances need to be specified as well. Using Figure 5-4 as

__Ax1 __I Ax

01 02

Figure 5-4. Geometry for analyzing the misalignment of a cavity.

a guide, where L is the length of the cavity, 01 and 02 are the small angular rotations of
the two end mirrors, Axl and Ax2 are the small sid-.i--- translations of the new optical
axis at the point where it intercepts the end mirrors, and C1 and C2 are the centers of

curvature of the two end mirrors, the parallelism tolerance can be found. Based on the HR

coated area of the mirrors and a typical spot size of 300 pm, the displacement should be

kept to within 1 mm. From Figure 5-4 and some geometry [71], it can be found that the

misalignments are given by

Ax 92 x L1 + 1 x L02 (5-13)
1 gig2 1 gi92

Ax2 = x L01 + x L02. (5-14)
1 gig2 1 gi92
The front face can be assumed to be the reference, so 01 can be assumed as zero. This

holds as long as the center hole is large enough for the beam to easily pass through and

the perpendicularity of the front face to the spacer is reasonable. In this case, Eq. 5-14

reduces to

Ax2 91 x L02. (5-15)
1 gig2
Since all the front mirrors are flat for the cavities produced in this work, gi 1, and Eq.

5-15 then reduces to Ax2 = L-02. Most of the cavities in this work are approximately 25

cm long with a flat front mirror, and a back mirror with a radius of curvature of 1 m. For

these parameters, the typical parallelism needs to be better than ~3 mrad. This tolerance

is not hard to obtain as most polishing companies can easily achieve a parallelism of better

than 1 mrad.

5.1.3 Cavity Construction Zerodur cavities

Three Zerodur cavities were made to test their relative stability. Two consisted of

optically contacted mirrors, and one with hydroxide bonded mirrors. The two cavities

with optically contacted mirrors were constructed as part of another group members

dissertation work [72]. The third cavity with hydroxide bonded mirrors was constructed by

the author. Table 5-1 specifies the length of the Zerodur spacer, the radius of curvature for

the mirrors used, the FSR, linewidth, and finesse of all 3 cavities. The 'i--, -1 difference

Table 5-1. Parameters of the three Zerodur cavities constructed by either optically
contacting or hydroxide bonding the mirrors.
Cavity L R1 R2 FSR linewidth Finesse
Z1 260 mm oo 0.5 m 576 MHz 50 kHz 11,500
Z2 225 mm oo 2.0 m 647 MHz 140 kHz 4,600
Z3 210 mm oo 1.0 m 715 MHz 60 kHz 11,700

between the three cavities is that Z2 has a much higher linewidth than the other two

cavities. This is most likely from the mirrors that were used to make the cavities. Both Z1

and Z3 were made using mirrors that had been made at Wave Precision and then coated

at AT Films. The mirrors on Z2 were older mirrors from a different vendor that had

possible contaminations of the coatings. The specifics of the construction of Zi and Z2 can

be found in reference [72].

To construct Z3, the Zerodur spacer was first cleaned as described in Appendix B. It

was allowed to dry overnight in order to make sure the surfaces were completely dry. The

curved mirror was then cleaned in the same manner. After it was cleaned, it was allowed

to dry for approximately five minutes. During this time, the surface of the Zerodur spacer

was cleaned using a cleanroom wipe that had been wetted with methanol. Both surfaces

to be bonded were then inspected for any contaminates on their surfaces. The Zerodur

spacer was then placed such that the surface to be bonded was facing straight up. 1.90

pL of hydroxide solution that had been diluted by a factor of 1:128 KOH:H20 was then

pippetted on the Zerodur surface and the mirror was then placed on top. The mirror was

then aligned by eye to be in the approximate center. The bond was left to cure for two

di,- To bond the flat mirror, it was cleaned in the same manner as the curved mirror.

The Zerodur spacer was flipped around and the surface to be bonded was cleaned with a

cleanroom wipe wetted with methanol. The same solution and amount was used to bond

the flat mirror as the curved mirror. A small amount of the hydroxide solution leaked into

the center but did not spread enough such that it would be in the path of the beam as it

passed through the cavity. Super Invar cavity

There are several manufacturers of Super Invar in the United States. The specific

manufacturer of the material used to create the Super Invar cavity is unknown as the

material was already in possession before work on constructing the cavity began. The

Super Invar bar is 178 mm x 44.5 mm x 44.5 mm. The smaller area surfaces were polished

by Surface Finishes Co., Inc with a surface flatness of A/4, 60-40 S-D finish or better, and

both ends parallel to better than 0.127 mm.

To make the Super Invar cavity, a similar method was used when constructing the

Zerodur cavity. The Super Invar spacer was cleaned as described in Appendix B. It was

then allowed to dry overnight. The curved mirror was cleaned in the same manner as the

Zerodur mirrors were. 1.70 pL of a sodium silicate solution that had been diluted by a

factor of 1:4 volumetrically with DI water was used. The bond cured for two div- The

flat mirror was then cleaned in the same manner and bonded using the same amount of

solution using the solution from the previous bonding.

The cavity parameters for the Super Invar cavity are given in table 5-2. Although

Table 5-2. Parameters of the Super Invar cavity with hydroxide bonded mirrors.
L R1 R2 FSR linewidth Finesse
178 mm o0 1.0 m 843 MHz 167 kHz 5030

the measurement for the linewidth of the Super Invar cavity was done in air, measurements

of the linewidth for Z3 were done both in air and in vacuum and were within 10' of each

other. Similar values should hold for the Super Invar measurements. SiC cavity

Hexoloy SA SiC made at Saint Gobain was used to make the SiC cavity. The specific

type of SiC was chosen because of its excellent mechanical strength and low CTE. SiC

is a possible material for the telescope support structure on the LISA mission. For this

reason, a spacer design similar to what might be expected on the LISA mission was made,

though scaled down in length by a factor of approximately 3. The SiC spacer is shown

in Figure 5-5. It consists of three struts that represent the struts that would hold the

secondary in place. The overall structure of the SiC spacer is that of a tube 250 mm in

length and 28.6 mm in diameter with a 12.7 mm hole through the center. The struts are

6.35 mm thick and spaced 1200 apart. The ends were polished at Surface Finishes Co.,

Figure 5-5. Completed SiC cavity made of Hexoloy SA.

Inc with a surface flatness of A/8, mirror finish, and end parallelism to better than 0.05

mm. Due to the extreme hardness of SiC, polishing to these specifications can be difficult.

In particular, the flatness can be hard to achieve. Although the polishing was still kept

within the stated specifications, a small dips could be seen on the polished ends and made

optical contacting difficult (Figure 5-6). The SiC spacer and mirrors were cleaned in

the same manner as the previous two cavities. In addition to being cleaned using a sonic

bath, the SiC spacer was cleaned thoroughly using cleanroom wipes wetted with methanol

because of the extremely dirty nature of the struts due to the fabrication process. After

this was done, the spacer was sonicated again. The polished faces of the spacer were

significantly cleaner and did not need the extra cleaning using the methanol. Both mirrors

were optically contacted using a combination of methods 2 and 3 as presented in C'! lpter

3.1. A small amount of methanol was put on the surface and the mirror was then placed

on top. The mirror was then rotated in a circular fashion until the mirror stuck. The

+0 500- +1.000-
+0.000- --------------------?-- ------- +0.375-
S-0500- 0 3 510.
S-1000-0 6-75
-1.500_ -1.500-
0 50 100 150 0 5 100 150
Distance (pix) Distance (plx)

Figure 5-6. WYKO images of the polished ends of the SiC spacer. Images produced by
Surface Finishes Co., Inc.

mirror was then pushed and rotated until it was close to centered on the spacer. The

cavity was left for two di,- to allow all of the methanol to evaporate. The linewidth was

then measured in air and found to be 97 kHz.

Table 5-3. Parameters of the SiC cavity with optically contacted mirrors.
L R1 R2 FSR linewidth Finesse
250 mm o0 1.0 m 600 MHz 97 kHz 6200 CFRP cavity

A CFRP spacer was made at the University of Birmingham in the United Kingdom

by the Center for Space and Gravity Research group. This group has experience in making

CFRP structures for space-based missions. They have had their CFRP parts flown on

several missions including Solar-B [73]. The spacer consisted of a thin outer shell that was

used to provide additional support to an inner tube whose thickness is larger than the

outer shell. Struts connected the outer shell to the inner tube (see Figure 5-8). The outer

shell has a diameter of ~200 mm and is a few millimeters thick. The inner tube has an

outer diameter of ~45 mm and an inner diameter of ~30mm. The length of both the tube

and shell are 230 mm. The struts connecting the shell and tube are a few millimeters thick

and span the length of the spacer. The process flow used to create the CFRP cavity is

shown in Figure 5-7.

Fabricate and Stress Relief Clean in Accordance
Machine Tube 3 Cycles (+60 C with S2-UGB-PR-3007
to -20 C)

V\actuum Bake at 100 C Butter Machined/ V\acuium Bake at 100 C
foi 1 Week (Monitor Exp d C P for 1 \eek (cMonitor
Exposed CFRP
\\ith RGA) \\ith RGA)

Figure 5-7. Fabrication flowchart of the CFRP spacer at the University of Birmingham.

In addition to providing the CFRP spacer, the University of Birmingham group also

sent Zerodur tubes that had been polished on one face. One flat mirror and one mirror

with ROC of 1 m were optically contacted to the Zerodur tubes. These Zerodur/mirror

plugs were then epoxied into the center of the CFRP tube. To do this, an alignment jig

was made such that the plugs could be slid in and out of the tube while still keeping the

alignment. A laser was then aligned through the cavity such that the cavity had an I1',

visibility. The plugs were taken out and epoxied using eccobond 285 epoxy that was mixed

with catalyst 9 and a small amount of acetone to make the mixture less viscous. Eccobond

285 was recommended by colleagues at the University of Birmingham since it is the epoxy

used to create the CFRP structure. The amounts of epoxy, catalyst 9, and acetone used

were 100 g, 3.5 g, and 2.0 g, respectively. The contents were mixed thoroughly using a

glass rod in a glass beaker for approximately 5 minutes. After about 15 minutes the paste

becomes less viscous. After 30 minutes the epoxy had a consistency that could be spread

onto the plugs and then inserted.

The CFRP spacer had been left in atmosphere approximately 4 months while working

on the alignment jig and the alignment process. During this time, it is speculated that the

CFRP spacer should have absorbed a small amount of water vapor from the surroundings.

For this reason, the CFRP spacer was baked in an oven at 100 C for 7 di to get rid of

any absorbed water. After this was done, the plugs were epoxied into the CFRP spacer.

Figure 5-8. CFRP spacer, Zerodur/mirror plugs, and the alignment jig (left). Finished
CFRP cavity after epoxying the plugs (right).

The curved mirror was first epoxied into place using the alignment jig and allowed

to cure for 2 d, i. The flat mirror was then epoxied and the laser was aligned such that

the cavity had a visibility close to II'. The cavity was then left for 1 week and when

the visibility was checked again, it was close to 4(0'. The laser was then realigned and a

visibility of ,-' was achieved again. The change in visibility is most likely from the epoxy

settling and causing the mirrors to move slightly.

The CFRP cavity was then left in atmosphere another 3 months. It could not be

moved into vacuum due to another ongoing experiment. The Finesse and linewidth of

the cavity was measured to be 2900 and 224 kHz, respectively, in air. After the CFRP

was moved into high vacuum and allowed to stay there for approximately 5 months, the

Finesse and linewidth were measured to be 4060 and 160 kHz, respectively. The error

in both measurements was ~10' The cause of the change in linewidth and Finesse is


Table 5-4. Parameters for the CFRP cavity.
L R1 R2 FSR linewidth Finesse
230 mm oc 1.0 m 650 MHz 225 kHz 2900 PZT-actuated cavity

In order to suppress the 12 orders of laser frequency fluctuations necessary for

gravitational wave detection on the LISA mission, a combination of pre-stabilization,

arm-locking, and time-delay interferometry (TDI) will be used. Pre-stabilization consists

of locking a free-running laser to an optical cavity to transfer the stability of the cavity

to the laser. The next step, known as arm-locking, will stabilize the laser frequency using

the LISA arms as a reference. The last step, TDI, is a post-processing technique that will

simulate an equal arm interferometer.

To use arm-locking, it will become necessary for the pre-stabilized laser to be tuned

to follow the changing LISA arms. This requires a continuous tuning of the pre-stabilized

lasers over 20 MHz. There are several methods available for this including sideband

locking and phase locked lasers [74]. Another possible method is to use a tunable cavity.

By changing the length of a cavity, the resonant frequency of the cavity will change

as well. This can be done by bonding a PZT actuator to a stable spacer material and

applying a voltage to the PZT. To make this cavity, a PZT ring made of k-180 material

from Piezo Technologies was used as the actuator. The specific PZT material was chosen

due to its low CTE and high d33 coefficient. The PZT ring has a 41 mm O.D, 12.7

mm I.D, is 3.2 mm thick, and the wide faces are covered with a silver coating. The

capacitance of the PZT is 1.3 nF and was found using a multimeter. The spacer material

was chosen as Zerodur for its high stability and has a 25.4 mm O.D, 6.4 mm I.D, and is

approximately 51 mm long. The Zerodur spacer was polished by Mindrum Precision to a

A/10 global surface flatness, 40/20 S-D, and <12.7 pm parallelism. The silver coated faces

allow hydroxide bonding to take place between both the Zerodur spacer and the mirror

(see Figure 5-9).

To construct the cavity, the Zerodur spacer was cleaned as described in Appendix

B. It then dried overnight. The PZT was cleaned in the same manner as the Zerodur.

A cleanroom wipe wetted with methanol was used to wipe the surface of the Zerodur to

PZT Material Zerodur

Flat Mirror Curved Mirror


Hydroxide Bonds Optical
Figure 5-9. Representation of the PZT-actuated cavity. Note the lengths are not to scale.

be bonded. 1.75 pL of a sodium silicate solution (Sigma Aldrich) that had been diluted
volumetrically with deionized water by a ratio of 1:4 was applied to the Zerodur tube and
the PZT material was placed on top and allowed to cure overnight. The next di a flat
mirror was cleaned in the same manner and bonded using 1.75 pL of the same sodium
silicate solution. The bond cured for 2 d-,v. A mirror with a ROC of 1.0 m was then
cleaned and optically contacted to the Zerodur spacer using 0.75 pL of methanol. Kapton
coated wires were soldered to the PZT using a silver solder.
The Finesse and linewidth were measured in air and found to be 6900 and 400 kHz,

Table 5-5. Parameters for the PZT cavity.
L R1 R2 FSR linewidth Finesse
54 mm o0 1.0 m 2780 MHz 400 kHz 6900

5.2 Pound-Drever-Hall Laser Frequency Stabilization
The PDH technique is both powerful and simple. The idea is that a laser's frequency
noise can be measured with a Fabry-Perot cavity, and this measurement is fed back to the
laser to suppress frequency fluctuations. In this manner, the stability of the cavity can

be transferred to the laser. The basis for PDH is a form of nulled lock-in detection that

decouples the frequency measurement from the laser's intensity [75]. A similar technique

used extensively in atomic physics is known as frcqu, -ii' -n-modulation (FM) spectroscopy

and is discussed in further detail in C'! lpter 6. The foundations of both are similar and by

understanding one, the other is often easy to understand.

Optical Spectrum Modulated Spectrum

Isolator PBS Optical Cavity
Laser --- M ----



Servo I XC-

Figure 5-10. Diagram of the PDH locking technique.

5.2.1 Conceptual Model

The basic layout of a PDH stabilization system is shown in Figure 5-10. The laser

beam is passed through a Faraday isolator to reduce back-reflections from coupling

into the laser. The beam then passes through an electro-optic modulator (EOM) which

modulates the carrier by frequency 2 to produce sidebands on both sides of the carrier.

The beam then passes through a polarizing beam splitter and A/4 waveplate. These

components act as an optical diode for the reflected light from the cavity. If the sidebands

of the beam are well outside the cavity resonance, then they will be reflected from the

cavity input mirror with no significant phase shift. When the laser carrier matches the

cavity resonance, the reflected light from the cavity at the carrier frequency will be in

phase with the incident light. The reflected carrier and sidebands produce two heterodyne

beats at the frequency of phase modulation (Q) and are detected by the photodiode. If the

carrier is on resonance, then the two beats will be 1800 out of phase with respect to each

other. Hence, the beats will cancel each other and there will be no photodiode signal at

the frequency of modulation.

When the laser is slightly off resonance, the reflected light carrier will experience a

phase shift. The sign of the phase shift is determined from the deviation off resonance. For

example, if the carrier frequency is higher than resonance and produces a positive phase

shift, then a frequency lower than the resonance will produce a negative phase shift. In

this case, the heterodyne beats will no longer cancel since they are not 1800 out of phase.

Figure 5-10 shows how the amplitude of the beat is extracted from the photodiode signal.

The photodiode signal is mixed with the local oscillator that drives the EOM to produce

an error signal. The phase shifter (typically differing lengths of coaxial cable) is used to

ensure the mixer inputs have the correct phase. The mixed signal can then be fed into

control electronics that produce signals that are fed back to the laser.

5.2.2 Quantitative Model

To provide a more complete understanding of how the error signal is produced for the

PDH technique, a brief mathematical derivation is described. The electric field of the laser

beam as is comes out of the laser can be described as Elser = Eoei"t. After the laser has

passed through the EOM which is driven at frequency 2, the electric field then becomes

EafterEOM = Eoei(t+sin(t)), where m is the modulation depth or modulation index of the

EOM. Expanding EafterEOM in terms of Bessel functions results in

EafterEOM et(Jo(m) + 2iJl(m)sin(Qt))

S Eo(Jo(mn + Ji(m)ei(w+t Ji(m> _l). (5-16)

From Eq. 5-16 it becomes apparent that the EOM adds two sidebands in frequency space,

each one located Q/27 Hz away from the carrier frequency as shown in Figure 5-10. The

powers in the carrier and first-order sidebands then become

Pc = J(m) Po

Ps = J (m) Po,

where Po = |Eo 2 is the total power in the beam before the EOM. When the modulation

depth is small, almost all of the power is in the carrier and first-order sidebands.

After the laser has passed through the EOM, it passes through the polarizing beam

splitter and A/4 waveplate. The beam is then incident on the cavity and either reflected or

transmitted through the cavity. The reflected beam passes through the A/4 waveplate and

beamsplitter again and is reflected onto the photodiode. Since the reflected beam is used

for actuating the laser, the reflected field will be considered, and the transmitted signal

will be ignored. Using the transfer function in reflection as defined by Eq. 5-3, where w =

27f, the reflected electric field is

E,,f = Eo(T (L)JJo (m^ '- + T (L + Q)J(m)ei(w T(w )J1 (mr, T "'). (5-17)

Since power is measured by a photodiode and not the electric field, Pref = |Eref 12 is the

signal of interest. After some algebra, this can be expressed as

Pref P~ I T,() 12 + P (I T,( + ) 2 + I T,( ) 2)

+2 PP(Re[Tr()T,(u; + Q) Tr(w)T,(u Q)]cos(Qt)

+Im[Tr(w)T,(w + Q) T (w)Tr(w Q)]sin(Qt)) + (2Qterms). (5-18)

Eq. 5-18 represents three waves of different frequencies being mixed together. The

2Q terms represent the interference produced from mixing the two sidebands with each

other. There may also be some contribution from higher order terms that were neglected

when Eq. 5-16 was expanded in terms of Bessel functions. These may make a significant

contribution to the 2Q terms, but are not considered here [75]. When using a photodiode

to measure Eq. 5-18, all terms will show up. Near resonance, the first two terms in Eq.

5-18 are negligible since most of the power in the carrier will be transmitted through the

cavity, and the power in the sidebands is typically small. This leaves only the sine and

cosine terms.

Mixing the photodiode signal with the driving signal from the EOM isolates the sine

term. Thus, the error signal becomes

c 2 VPcPIm[T,P(w)Tr(wP + Q) Tr(wP)T, (P Q)1.

Since the laser frequency must be near a cavity resonance in order to lock, we are only

interested in the error signal near this frequency. In this case, the transfer function of the

reflected field for the sidebands will be nearly one. This results in an error signal near

resonance of the form

res 2 VPP Im[T7() Tr(w)] -4 v Im[Tr(w )].

Linearizing Tr(wu) for small u and to first order yields

Tr r2 2ir2(rf 1) 6L
1 rr2 (1 rl)2

S- i- lil ii- and letting rl = r2 = r (since all mirrors used have the same coatings), the

error signal now becomes

_(/2r(r2 1) 6LL r 6wL
(1 r2)2 C T c

Making use of the fact that
rr c/2L
T Af'

where Af is the linewidth of the cavity, and using w= 27f, the error signal close to

resonance becomes

Acres 8P 6-. (5-19)
From this, it is easy to see that the error signal is linear around resonance. Figure

5-11 shows two error signals for the same cavity length and EOM modulation frequencies,



S 0
m 0

Frequency (free spectral ranges)

Figure 5-11. Two error signals produced from cavities of different finesse. The green curve
is for a cavity of length 200 mm and finesse 500. The red curve is for a cavity
of the same length but finesse of 5000. Both have EOM modulation
frequencies of 50 MHz.

but different Finesse. Three error signals can be seen. The center error signal is produced

from the carrier interfering with the sidebands while the signals on either side corresponds

to the error signals produced from the sidebands. Figure 5-12 shows a closer view of the

carrier beatnote from Figure 5-11. The narrower curve corresponds to the cavity with a

higher Finesse as is expected from Eq. 519.

5.3 Relative Stability Measurements

The PDH frequency locking scheme is a powerful tool that can be used for a variety

of applications. Of importance to many space-based interferometric mission will be

the dimensional stability of select materials for their implementation in critical optical

components. This section presents how the relative dimensional stability of select

materials was found in addition to providing results for these materials.
materials was found in addition to providing results for these materials.


co 0.5


m 0

i -0.5

0.98 0.99 1 1.01 1.02
Frequency (free spectral ranges)

Figure 5-12. Error signals for only the carrier frequency from Figure 5-11. The green curve
is for a cavity with a lower finesse.

5.3.1 Electro-Optical Setup

To test the relative dimensional stability between certain materials, the PDH locking

scheme is used to lock the frequency of one laser to one of the cavities described above.

The experimental setup consists of two vacuum tanks which house the cavities (as shown

in Figure 5-13). Tank 1 uses five cylindrical lrz-;- s of gold-coated stainless steel shells

which are separated by Macor spacers to provide passive thermal isolation for two cavities

that are placed in the middle. Tank 2 uses five l z.-,rs of aluminized PET (basically a thick

Mylar) separated by Macor spacers to provide passive thermal isolation to one cavity that

is placed in the center of the thermal shields (see Appendix A for a complete analysis of

the aluminized PET thermal shields). In this manner, the dimensional stability can be

tested under common and uncommon thermal conditions.

Laser 1 is locked to Cavity 1 and Laser 2 is locked to Cavity 2 using the PDH

stabilization method. Pick-offs from each laser are then superimposed on each other to

produce a beatnote. The beatnote is sent into a photodiode where the frequency is read by

Faraday I sc/er P
Isolator Glass PD / Cavity 3 D
PLaser 3
EOM Lens Pol 1/4 WP
Frequency PBS
Counter Tank 2
Pow BS T
Control -
Electronics | y
Cavity 2
Laser2 U C 2

Laser I
-Cavity 1

Electroncs Tank 1

Figure 5-13. Electro-optical setup used to test the dimensional stability of select materials.

a frequency counter. Using the relationship

6v 61
F, 1

where 6v is the linear spectral density of the measured beatnote, v is the laser frequency,

61 is the linear spectral density of the length of the cavity, and 1 is the length of the cavity,

the relative noise spectrum of the cavity can be found.

Laser 3 is locked to Cavity 3 in Tank 2 using the PDH method. A pick-off from

this laser is superimposed on the pick-off from Laser 2 to determine the relative stability

between Cavity 2 and Cavity 3.

5.3.2 Zerodur-Zerodur

As previously stated, Zerodur is an ultra-stable material with a low CTE around

room temperature. This makes it an ideal material for use in laser frequency stabilization

since the stability of the cavity is transferred to the laser by using the PDH locking

technique. Narrow linewidth lasers are critically important for space-based interferometric

missions since the frequency noise of the laser will couple into the measurement sensitivity.

The more stable the frequency of a laser, the more sensitive a measurement can be made.

For this reason the frequency stability of Zerodur with optically contacted and hydroxide

bonded mirrors was studied. Optically contacted mirrors

To test the relative frequency stability of two Zerodur cavities with optically

contacted mirrors, the cavities described above were both placed in Tank 1 and Laser

1 and Laser 2 were locked to Cavity 1 and Cavity 2, respectively, using the PDH

locking technique as depicted in Figure 5-13. The beatnote between the lasers was

continuously monitored for 28 d ,i- using a frequency counter sampled every 20 seconds.

The resultant beatnote is shown in Figure 5-14. In addition to this measurement,

36 ..




LL 34
w 2-

a 33.5-


32.5 -
0 5 10 15 20 25
Time (days)

Figure 5-14. Time series of the beatnote taken over 28 d4 -

frequency measurements were taken using a phase-meter sampling at 98 kHz and 15 Hz.

To determine the frequency stability, the frequency time series is broken into smaller

pieces. These pieces have a linear regression fit to the data sets that is subtracted. A

fourier transform is done on the residuals and the resultant linear spectral density (LSD)

in frequency-space is given. For a discussion on the process that was used, see Appendix

C. The resultant LSD for the frequency counter and phase meter measurements is shown

in Figure 5-15. Although measurements less than 10 Hz/vHz at 1 Hz have been reported

for this experimental setup [72], typical results are between 20-100 Hz//Hz. For the


T 104

(D 2



10-5 100
Frequency (Hz)

Figure 5-15.

Noise spectrum of the Zerodur cavities with optically contacted mirrors.
Measurements were taken using the frequency counter, phase meter sampling
at 15 Hz and 98 kHz.

LISA mission, the prestabilization step could be done by locking the Nd:YAG lasers to

cavities made of Zerodur. The required stability for this step is [76]

v( f = 1 + (3mHz/f (5-20)

The relative stability of the Zerodur cavities is shown in Figure 5-16 along with the LISA

pre-stabilization requirement. The results in Figure 5-16 were taken from a data run that

lasted for 7 d ,v and was sampled at 1 Hz. From Figure 5-16, it can be seen that the

Zerodur cavities easily meet the LISA pre-stabilization requirement in the entire frequency


T 10" \

I 105

I 102


10-4 10-3 10-2 1071
104 103 102 10
Frequency (Hz)

Figure 5-16. Noise spectrum of the Zerodur cavities with optically contacted mirrors and
the LISA pre-stabilization requirement. The results are from a time series
taken over seven d-iv and sampled at 1 Hz.

There are several noise sources within the LISA band that may be affecting the

achieved sensitivity. Among them are the temperature fluctuations of the cavity due

to changes in the lab temperature, the RFAM due to the EOM, and the heating of

the mirrors due to the laser power stored within the cavity. Noise from the feedback

controllers is also a potential source, although measurements of these noise sources are

typically an order of magnitude or more below the measured noise. An additional noise

source due to acceleration from the optical table is also significant, although it is found

outside the LISA band and is typically averaged out by the frequency counter when

sampled at a low enough frequency. The peak between approximately 8-60 Hz shown in

Figure 5-15 is most likely due to this acceleration noise [77].

Models have shown that frequency noise below ~1 mHz tend to be dominated by

temperature changes in the lab propagating through the thermal shields into the cavity

which cause an expansion and contraction of the cavity. The five I v. -is of gold-coated

stainless steel thermal shields provide an excellent low pass filter for temperature

fluctuations, but at low enough frequency, the thermal shields will not provide any

suppression. To test this, the temperature of the outside of Tank 1 was simultaneously

monitored while the 28 di beatnote was taken as shown in Figure 5-17. The temperature

was monitored using an OM-PLTT temperature 1. .-.-, r from Omega Engineering, Inc.

and provides a temperature resolution of 0.10 F using a thermocouple probe. From the

figure, it is clear that as the temperature of the tank increased, a notable decrease in the

beatnote could be observed. The ~3.5 MHz beatnote change over the 28 d4v is almost

the largest change in beatnote frequency observed. The beatnote between the two Zerodur

cavities has changed by less than 5 MHz over 3 years. This is most likely due to seasonal

temperature changes in the lab and electronic offset effects in the feedback controls.

The large temperature swing over a relatively short time (typically these temperature

changes occur over six months, not one) means the temperature of the lab was not in a

steady state. Although it is obvious from Figure 5-17 there is a temperature dependance

on the beatnote, by high pass filtering the data a more steady state transient can be

seen. Both the beatnote and temperature data were high pass filtered using a 2-pole

Butterworth filter whose 3 dB frequency was set at 5x 10-6 Hz. The resulting high

pass filtered (HPF) data is shown in Figure 5-17. Again, it is clear there is a definitive

correlation between the changing temperature of the tank and the change in beatnote

frequency. Although the cavities are in the same thermal environment, this shows there is

some non-common temperature effects due to the changing temperature in the lab. This

may be due to the different lengths and orientation of the cavities. Hydroxide bonded mirrors

The hydroxide bonding technique has shown to be a versatile way to joint a variety

of materials with significant strength. Although there are many bonding techniques that

produce similar, or even greater, bonding strengths, they most likely do not provide the

precision alignment needed for critical optical components such as optical benches. This,

determining the amount of noise the bonds produce will be critical as interferometric

0 5 10 15
Time (days)

?n 95

Time (days)

Figure 5-17. Plots of the Zerodur beatnote and tank temperature (left) and plots of the
beatnote and temperature data after being high pass filtered (right).

missions keep pushing measurement limits. The frequency stability of the hydroxide

bonding technique was tested by hydroxide bonding mirrors to a Zerodur spacer as

described in Sect. This cavity replaced Z2. The resultant noise spectrum is shown

in figure 5-18 [78].

Zerodur Optically Contacted Mirrors
07 Zerodur Hydroxide Bonded Mirrors
0- LISA Pre-stabilization Requirement

106 -

2 104

(U 103

Z 102


10-4 10-3 10-2 101
Frequency (Hz)

Figure 5-18. Comparison of the noise spectrums of the Zerodur cavities with optically
contacted and hydroxide bonded mirrors.

The noise spectrum shown in Figure 5-18 compares the noise spectrum from the

cavities with optically contacted mirrors and that when one of the cavities has hydroxide

bonded mirrors. Above approximately 3 mHz, the noise spectrum is the same. Below this

frequency the noise spectrum is slightly higher. This may be due to different temperature

fluctuations within the lab when the measurements were taken, or it could be noise

attributed to the hydroxide bonds. Unfortunately, these noise sources were not studied at

the time the measurements had been taken. Despite this, it appears that the hydroxide

bonding process adds little or no additional noise and, more importantly, the noise

spectrum is still below the LISA pre-stabilization requirement.

5.3.3 Zerodur-SiC

SiC has become an ever increasingly popular material for use in space-based

interferometric missions due to its high strength, low CTE, non-magnetic nature,

and its ability to be made in complex shapes [62]. In addition, using the hydroxide

bonding technique on SiC open up new possibilities for its use ranging from telescope

support structures to optical benches. To test the stability of Hexoloy SA, a cavity was

constructed as described in Sect. and locked to Laser 1 in Tank 1. A time series of

the beatnote between Laser 1 and Laser 2 taken over a long weekend is shown in Figure

5-19 (left). The temperature stability in the lab is typically much better over weekends,

so there is a lack of an oscillatory nature in the beatnote as is commonly seen. For longer

time periods, or when the beatnote is taken during the week, the beatnote would oscillate

with an amplitude between 80 and 120 MHz over several di,- as shown in Figure 5-19


The resultant noise spectrum for the beatnote shown on the left side of Figure

5-19 is shown in Figure 5-20 and meets both the LISA telescope and pre-stabilization

requirements. The spike in noise around 0.20 Hz is most likely from having too much

gain in the temperature feedback controller to the laser and is not present in other

measurements. The noise above ~0.01 Hz is similar to that of the Zerodur cavities [79].


w 125

N 80




1156 20
0 05 1 15 2 25 3 35 0 2 4 6 8 10
Time (days) Time (days)

Figure 5-19. Plots of the SiC beatnote taken over a long weekend (left) and plots of the
beatnote taken over 11 d ii;ht).

This si-.-- -1 a limitation of the measurement system is being reached. Although the

mirrors were optically contacted to the SiC spacer, it is assumed that hydroxide bonding

the mirrors should not produce any additional noise that would be detectable.

S Hexoloy SA SiC
10 LISA Telescope Requirement
104 LISA Pre-stabilization Requirement






10-5 10-4 10-3 10-2 10-1
Frequency (Hz)

Figure 5-20. Noise spectrum of the Hexoloy SA SiC cavity with optically contacted
mirrors along with the LISA pre-stabilization and telescope requirements.

The CTE of the SiC is around two orders of magnitude larger than of Zerodur.

Because of this, the beatnote measurement between SiC and Zerodur will be much more

sensitive to temperature fluctuations than the Zerodur-Zerodur measurements. If the

limiting noise source is assumed to be from temperature fluctuations in the lab that

propagate into length changes in the SiC cavity, then the temperature stability of Tank

1 can be found. The inferred temperature stability is shown in Figure 5-21. At 1 mHz,

the temperature stability is ~3 pK/vHz which is close to what is expected on the LISA



10-2 ... .


E 10-4

V 10-6

10-4 10-3 10-2 10-1
Frequency (Hz)

Figure 5-21. Inferred temperature stability at the center of Tank 1.

The surprising result is that the SiC cavity performed so well at high frequencies

and had noise levels the same as the Zerodur cavities. Temperature stabilization of the

cavity could bring down the low frequency noise, while using less input power would

lower the noise in the 10-3 to 10-2 frequency band (as will be discussed in the CFRP

cavity section). Thus, it would appear that SiC would be a viable candidate for the LISA

telescope support structure.

5.3.4 Zerodur-Super Invar

Super Invar is a high stability, machinable metal that has been used on previous

space missions [65]. Its high strength, low CTE, and ease of machining make it ideal for

support structures such as the narrow angle camera on the Cassini spacecraft. Although

the dimensional stability of Super Invar has been measured, it has not been measured

at the fm//Hz level. In addition, this test will provide additional information about the

stability of the hydroxide bonding technique when a sodium silicate solution is used.

To test the dimensional stability of Super Invar, a cavity was constructed as described

in Sect. and placed in Tank 2. Laser 3 was locked to the cavity and the beatnote

between Laser 2 and Laser 3 was monitored. A 28 dwi measurement was made at the

same time the Zerodur measurement in Sect. was made. The temperature of the

tank was measured at the same time using another OM-PLTT temperature logger. The

resultant beatnote and corresponding tank temperatures are shown in Figure 5-22 (left).

The corresponding dimensional stability is shown in Figure 5-23. Not only does the Super

Invar meet the LISA telescope requirements, but it meets the pre-stabilization requirement

as well, and is similar to what was found with the SiC cavity. Above ~10 mHz, the noise

level is almost identical to that obtained from the Zerodur cavities. This would also imply

that an experimental noise floor for our system is being met.

The resulting noise spectrum at high frequencies of the Zerodur-Super Invar

measurement again shows that there is no detectable addition of noise from using the

hydroxide bonding technique. Two separate solutions which produced two different bond

thicknesses both showed no additional noise.

From Figure 5-22, it is apparent that as the temperature of the tank changes, so does

the beatnote frequency. To further test this, both the beatnote and temperature data sets

were high pass filtered as described in Sect. to look at a more steady state of the

data. The results are shown in Figure 5-22 (right). It is evident the beatnote is influenced

x 10

Time (days)

Time (days)

Figure 5-22. Plots of the Super Invar beatnote and tank temperature (left) and plots of
the beatnote and temperature data after being high pass filtered (right).

by the tank temperature and the beatnote change lags the tank temperature change by

approximately one day.


Z 102

E 101

O) 100



10-4 10-3 10-2 10-1
Frequency (Hz)

Figure 5-23. Noise spectrum of the Super Invar cavity with hydroxide bonded mirrors
along with the LISA telescope and pre-stabilization requirement.

Knowing the temperature stability inside Tank 2 will be critical in order to determine

temperature effects on the material. The gold-coated stainless steel thermal shields in

Tank 1 provide excellent suppression of temperature fluctuations due to the low emissivity

and high mass. A cost-effective alternative to these thermal shields was designed and built

using an aluminized PET material that is separated by Macor spacer (see Appendix A

for a full description). If it is assumed that the limiting noise source at low frequencies

is from temperature fluctuations in the lab that couple into length changes of the cavity,

then an upper limit on the temperature stability of Tank 2 can be found. Although this

assumption will be valid at lower frequencies, at higher frequencies only an upper bound

can be determined since it is assumed the temperature effects are significantly lower than

other noise sources such as electronic or acceleration noise. The estimated temperature

stability of the inside of Tank 2 is shown in Figure 5-24.




E 10


10-5 10-4 10-3 10-2 10-1 100
Frequency (Hz)

Figure 5-24. Inferred temperature stability inside Tank 2.

It should be noted that to make this calculation, the CTE of the Super Invar cavity

needs to be known. Unfortunately the CTE of the Super Invar is not known since it was

bought by a previous group member and records of where it was purchased can not be

found. After searching several sources, a range of CTE values was found. Although the

stated values for the CTE of Super Invar can vary from 0.2x10-6/K to 2.0x10-6/K, a

typical value of 0.7x10-6/K is used.

An estimation of the CTE of the Super Invar cavity can be done using the outside

tank temperature and the beatnote time series. To estimate the CTE of the Super Invar

cavity, measurements of both the tank temperature and beatnote were simultaneously

taken for spans of several d4iv- Since the thermal shields will have little attenuation of the

temperature changes in the tank at low frequencies, if the tank temperature is constantly

changing in the same direction over a long period of time, then the beatnote will be

continuously changing in a similar manner. By cutting off the first and last 1-2 d iv- of

data, and looking at the change in beatnote compared to the change in tank temperature,

the CTE of the Super Invar can be found. Once the change in beatnote frequency per

degree is found, the CTE can be found via

At A 1
CTEpp A ,t (5-21)
ACTET FSR 2L' (5 21)

where Av is the change in beatnote frequency, AT is the corresponding change in tank

temperature, A is the wavelength of the laser, FSR is the free spectral range of the cavity,

and L is the length of the cavity. Doing this, the CTE of the Super Invar is found to be

approximately 0.2-0.3 ppm, which is within the CTE values previously stated. This is,

in fact, a lower bound on the CTE, since it is expected that there is some attenuation at

frequencies of 10-5 Hz as shown in Figure A-4 and would produce a larger CTE value. For

this reason, a CTE value of 0.7x10-6/K is an appropriate approximation to the CTE of

the Super Invar cavity.

For the inferred temperature stability inside Tank 2, a CTE of 0.7x10-6/K was

assumed. Comparing the thermal stability of Tank 1 and Tank 2, Tank 2 appears to

have approximately an order of magnitude worse temperature noise at low frequencies.

Although the temperature stability of Tank 2 is worse than that of Tank 1, the difference

in cost is substantial, and making a few changes to the thermal shield design could result

in a significantly better temperature stability while still keeping the cost low.

5.3.5 Zerodur-PZT Cavity

Several methods can be used to tune the frequency of a pre-stabilized laser by tens or

hundreds of megaHertz. These include using an acousto-optic modulator, offset sideband

locking, and offset phase-locking, among others. Probably the most straightforward

method is to integrate a PZT into the cavity. By applying a voltage to the PZT, the

length of the cavity will change, hence changing the resonant frequency of the cavity. This

has a particular interest for use in the AURIGA and LISA gravitational wave detectors. In

the case of AURIGA, a PZT cavity has been tested and shown promising results, despite

its more complicated design than the PZT cavity presented above [80]. For LISA, a PZT

cavity could be implemented for use in arm-locking. To utilize arm-locking, it will become

necessary for the pre-stabilized laser to be tuned to follow the changing distances of the

LISA arms. This requires a continuous tuning of the pre-stabilized lasers over 20 MHz.

Due to the length stability necessary for LISA, the PZT cavity must be extremely stable

as well.

To test the stability of the PZT-actuated cavity, it was placed in Tank 2 and Laser 3

was locked to it. The stability of the two Zerodur cavities with optically contacted mirrors

were put in Tank 1 and the relative stability between the two were measured while the

PZT cavity measurements were taken. The stability of the two Zerodur cavities were

found to be similar to what was measured above and typically showed a noise spectrum of

20 to 50 Hz//Hz at 1 Hz.

To determine the tuning coefficient of the PZT, the PZT cavity was locked and

an 8 V peak-to-peak triangle wave at 0.10 Hz was applied to the PZT. The measured

tuning coefficient was found to be ~1.5 MHz/V [81]. The current estimates for the

necessary tuning frequency for arm-locking is 20 MHz and is based on the variations in

the LISA arms. Due to the drifts of the cavity itself, which are on the order of Hz/s, a

tuning capability of 100 MHz would provide enough margin to ensure that the locking

stretches are not interrupted by a saturated high voltage amplifier. The required voltage

for this tuning would only be 70 V. It is also important to note that LISA will not take

data continuously and that the science runs will be interrupted for short periods due to

maintenance such as rotations of the high gain antennas.

Commercially available amplifiers rarely specify their performance in the LISA band.

For this reason, a low cost (less than $100), low noise alternative to a commercially

produced amplifier was built. The design shown in Figure 5-25 is almost identical to the

design in reference [82], except for a 60 mHz low-pass filter that is placed on the output

of the amplifier, and the 1000 V source is provided by a commercial high-voltage DC

source. The input voltage is provided by a 12 V voltage reference and voltage divider in

which one of the resistors is a potentiometer. The input voltage can be tuned through the



Q2 60 mHz
+ low pass filter
Cl 1
R4 D1 100
10k R3 Vout
3x330k R6
Vin + Q1 5
SOP27 Vmonitor
R1i 50k <

Figure 5-25. Voltage amplifier used to provide the applied voltage to the PZT on the PZT

When the amplifier needs to charge the piezo, the charge rate is fixed at the value

of the current source, which is determined by Q2 and R2. Since the Q2-R2 value also

determines the quiescent current of the amplifier, the quiescent current must be low

(around 1 mA) in order to not overheat the device when it is working at 1 kV. For a 10 nF

piezo, the slew rate is 0.1 V/ps for a 1 mA quiescent current. The up and down slew rates

can be trimmed using R2 and R1, respectively.

When the amplifier needs to discharge the piezo, the output of the op amp goes up

and Q1 acts as a sink whose current is limited by the set value of R1. The diode provides

a path for the discharge current through Q1.

The gain is set at 100 through the R3 and R4 resistors. Both R3 and C1 must

support a 1 kV drop and are rated such that the operating voltage and wattage can

support this. R5 is used to isolate the capacitive load in addition to improving the

stability. Resistors R6 and R7 act as a voltage divider to monitor the output voltage.

The beatnote between Laser 2 and Laser 3 was measured using a phasemeter

sampling at 98 kHz and 15 Hz, as well as using a frequency counter when no voltage was

applied to the PZT. The noise spectrum is shown in Figure 5-26. The two peaks in the 10

Hz regime are most likely due to the seismic noise and the different damping mechanisms

of the two vacuum tanks. The peak around 20 kHz is most likely from the resonant

frequency of the PZT [83]. The beatnote frequency was also measured using a frequency

counter for several di -. The resultant noise spectrums for 0 V and 200 V applied to the

PZT are shown in Figure 5-27. A typical noise spectrum for applied voltages up to 200

V shows that the PZT cavity meets or exceeds the LISA pre-stabilization requirement

in all but a small frequency band. For applied voltages ranging from 0-100 V, the noise

spectrums showed similar levels. For applied voltages between approximately 100-200

V, the noise increased slightly. Above 200 V, the noise increased as the applied voltage

increased. The visibility of the cavity and peak-to-trough voltage of the error signal were

found to be similar up to 300 V. A decrease of less than 10' was found for an applied

voltage of 450 V. This might be due to the shearing effects of the PZT.

A typical beatnote between the laser locked to the Zerodur cavity and the laser

locked to the PZT cavity will drift at 50-350 Hz/s and is assumed to be from the changing



(C 103

Z 102


10-4 10-2 100 102 104
Frequency (Hz)

Figure 5-26. Noise spectrum of the PZT cavity with no applied voltage. Measurements
were taken using the frequency counter, phase meter sampling at 15 Hz and
98 kHz. The spike around 20 kHz is most likely the resonance of the PZT

temperature within the lab that cause the cavity to expand or contract [81]. When

a voltage was applied, changed, or removed from the PZT, a temporal change in the

beatnote drift could be observed. The amount of time for the beatnote to return to normal

increased as the voltage increased. For example, when 10 V was applied to the PZT, the

initial drift of the beatnote increased to approximately 2000 Hz/s. After about 30 minutes,

the drift in the beatnote returned to 100-400 Hz/s. When 200 V was applied, the initial

drift increased to approximately 18000 Hz/s, and the time for the beatnote to return to

normal increased to several hours (Figure 5-28). Monitoring the voltage output of the

voltage amplifier showed the change in beatnote frequency could not be attributed to

any drift in the voltage amplifier. The slope of the beatnote decays exponentially when

a voltage was changed on the PZT, and is believed to be attributed to the relaxation

phenomenon of the charged domains slowly returning to their final state in the PZT

material [84]. This could be seen since the beatnote would change in one direction as the


N 105


UO 103


S PZT Cavity No Applied Voltage
S PZT Cavity 200 V Applied
-- LISA Pre-stabilization Requirement

1 1.

10-4 10-3 10-2 10-1
Frequency (Hz)

Figure 5-27. Noise spectrum of the PZT cavity with 0 V, and 200 V applied to the PZT
along with the LISA pre-stabilization requirement.

voltage was applied, and then change in the other direction after applying the voltage.

This implies that the relaxation process occurs in the opposite direction of when a voltage

is applied.

Arm-locking uses the stability of the LISA arms to further suppress the frequency

noise of a pre-stabilized laser. An error signal can be generated from a pre-stabilized laser

and its transponder signal from the far spacecraft. This error signal is then fed into an

arm-locking controller (ALC) which serves as a tuning mechanism for the frequency of the

laser [74].

It has already been experimentally verified that by using an ALC the frequency noise

of a pre-stabilized laser can be further suppressed and that the experimental results are in

close agreement to what is theoretically expected [74].

To test this, the beat signal between the Laser 2 and Laser 3 was measured with a

phasemeter. The signal was d, 1 i, .1 by 0.98 s and Doppler shifted using a delay unit. The

d, 1, i .1 and Doppler shifted signal is then mixed with the prompt signal to form the arm

1111111 1 1 1111111 1 1 1111111 1 1 1111111 1 I I

x 104



| 1.4

0 1


w 0.6



0 2000 4000 6000 8000 10000
Time (sec)

Figure 5-28. The slope of the beatnote over time with 200 V applied to the PZT. The time
series was taken approximately 10 minutes after the voltage was applied to
the PZT.

locking error signal upconverted by the Doppler shift similar to the upconverted signal

in LISA. A second phasemeter is used to demodulate the Doppler-shifted signal with the

Doppler frequency. Any differential laser frequency fluctuations will change the phasemeter

output. This signal is appropriately filtered, amplified, and applied to the PZT to change

the length of the PZT cavity and stabilize the beat frequency of the lasers to the delay

line. Note that Laser 2 is only providing a reference to be able to electronically d. 1iv the

phase noise of the main laser.

The frequency noise of the beat signal is shown in Figure 5-31. The frequency noise

between 10-4 and 10-2 Hz is 4 to 6 orders below the requirements and then rolls up with

~f2 up to 0.5 Hz. The peaks at multiples of the inverse delay time are caused by the fact

that the arm locking sensor is insensitive at these frequencies [81]. Using this method,

the cavity can be stabilized from noise levels that are just at the LISA pre-stabilization

requirement, to levels that are several orders of magnitude below the requirement. Using

this method alone can suppress seven of the twelve orders of magnitude in frequency

suppression from a free running-laser needed for the LISA mission.

The main difference in this setup when compared to what will be experienced on

LISA, is that there is a 0.98 s delay instead of the 32 s delay that the light will experience

between the LISA spacecraft. However, LISA will use a linear combination of the signals

from both arms. This linear combination is insensitive to laser frequency noise at multiples

of the inverse of the light travel time difference. These nulls in the response function are at

or above 1 Hz for most of the time of the mission, which is equal to the response we have

in our system.

Control FM

Laser 2


Laser 1

Figure 5-29. Arm-locking setup used to arm-lock the PZT cavity to a Zerodur cavity. The
feedback controls used to initially lock the lasers to the cavities are not
shown. PD: photodiode, PM: phasemeter, FM: frequency meter

Although the noise spectrum of the PZT cavity almost meets the LISA requirements

in most of LISA band, sudden jumps in the time series of the beatnote do occur.

Depending on the size and frequency of these data glitches, they are often not seen in

the noise spectrum due to the averaging of the time series before the FFT is performed.

These glitches may cause problems in keeping the laser properly locked if arm-locking is to

be used. A sudden jump in frequency may have a large enough slope such that the ALC

controller is not able to track the beatnote, subsequently causing the laser to lose lock.



103 102 101 100
Frequency (Hz)

Figure 5-30. Modeled open-loop transfer function of the arm-locking controller used to
suppress the laser noise.



o 10o



10-3 10-2 101 100
Frequency (Hz)

Figure 5-31. Noise spectrum of the PZT cavity without an applied voltage, the noise
spectrum of the PZT cavity using arm-locking, and the LISA pre-stabilization

The amplitude, frequency, and cause of the glitches are important factors to consider when

designing the ALC as well as determining their propagational effects within the LISA


There are 2 distinct types of data glitches that occur within the data. The first is

the step-glitch (Figure 5-32, top) and is characterized by a sudden jump in frequency

that does not return to the normal beatnote frequency. The second type of glitch is the

spike-glitch (Figure 5-32, bottom) and is characterized by a jump in frequency that returns

to the normal beatnote frequency.

x 108
C 7.1086

4.534 4.536 4.538 4.54 4.542 4.544 4.546 4.548 4.55
Time (sec) x 104
x 108
7.1093 -
N 7.1092
S 7.1091 -
7.109 -
CO 7.1089 -

4.545 4.55 4.555 4.56 4.565
Time (sec) x 104

Figure 5-32. Example of a step glitch (top) and spike glitch (bottom).

Three different causes for the step-glitches were investigated. Sudden changes in the

alignment of the laser beam caused by human activity in the lab is ah--,i-i an issue but

can usually be accounted for. Electronic disturbances such as sudden changes in an offset

voltage within the laser feedback system produces step-glitches. In addition, internal

relaxation processes within the Zerodur or PZT material cannot be ruled out as a cause

for step-glitches. However, once the feedback system was optimized, step glitches occurred

on average only every 3.8 d-i,- Their amplitude ranges from 3000 to 6000 Hz with an

average of 3900 Hz [81].

The overwhelming 1 I i ii ily of the glitches that occur are spike-glitches. The detection

rate of spike-glitches can depend on the sampling rate. Only spike-glitches where the

average frequency variation during a sample time is large compared to the normal rms

noise can be detected. Consequently, shorter spikes require a larger amplitude to be

detectable. However, when we increased the sampling rate from 1 Hz to 8 Hz, the rate

-I li-, d essentially the same. In contrast to this, when the sampling rate was reduced to

0.1 Hz, most of the spike-glitches disappear into the rms noise. The only identifiable

source for these spike-glitches was the PZT voltage. At 50 V, the spike-glitch rate was

about once every 3-4 hours. A typical amplitude of 1000 Hz at a 1 Hz sampling rate was

typically measured, while the amplitude rarely exceeded 4000 Hz when sampled at 8 Hz.

The scaling indicates that the amplitude is probably not low pass filtered by the sampling


Although spike-glitches were shown to increase as the applied voltage increased, they

were not able to be completely suppressed with no voltage applied to the PZT. Both the

beatnote and the voltage applied to the fast actuator of the laser were simultaneously

measured and showed that spikes in the beatnote often showed a spike in the laser

actuator voltage as well. These spike-glitches are most likely caused by electronic noise in

the laser feedback controllers and/or human activity in the lab [81].

To test the mechanical feasibility of using the constructed PZT cavity on the LISA

mission, a -!i I:.. test" was done at GSFC. To measure the vibrational signature of the

PZT cavity and determine if the cavity would survive expected launch conditions, two

accelerometers were used. One sensor was cemented to the PZT to sense motion along the

cavity axis only. The other sensor detects motion in all three axes and was cemented to

the top of the mounting clamp that holds the cavity. Two rounds of testing were done for

a total of six tests. One round of testing consisted of vibrating the PZT cavity transverse

to its optical axis using a 0.25 g sine sweep for frequencies of 20-2000 Hz, a 0.14 g2/Hz

amplitude with random frequency vibrations, and a 0.25 g sine sweep for frequencies of

20-2000 Hz after the random vibrational test was done. These tests were then repeated for

a vibrational direction axial to the optical axis. The cavity showed no appreciable change

based on the vibration signature, and would be expected to survive launch conditions.

Further tests will need to be done to determine if the vibration testing caused the stability

of the cavity to decrease, but it is expected that the stability should be similar to what

was originally found.

5.3.6 Zerodur-CFRP

SiC and CFRP have become widely used in many applications due to their high

mechanical strength and low CTE. CFRP has been used in support structures for space

based missions, and is under consideration for use as an off-axis secondary mirror support

for the LISA mission [23]. To the authors knowledge, there is no information on the

stability of CFRP materials, although it is claimed by manufacturers of space-qualified

CFRP that the outgassing is well understood and can be accounted for through the use

of optics to adjust the beam as the strut outgasses. To test both the dimensional stability

and outgassing properties of CFRP, the CFRP cavity described above was placed in Tank

2 and Laser 3 was locked to the cavity. A time series between Laser 2 and Laser 3 taken

over 4 d iv is shown in Figure 5-33 [68]. Typical beatnote fluctuations range from 250-600

MHz, but have been measured to be over 1 GHz. For this reason it is difficult to get data

runs that last several weeks because the beatnote will drift out of range of either the

photodiode or the frequency counter.

Typical drifts in the beatnote frequency are a few kiloHertz per second. When the

CFRP cavity was initially placed in the vacuum tank, the tank was pumped down using a

roughing pump for approximately one d4iv. The turbo pump was then turned on and left

on for several weeks before the ion pumps were turned on. The first measurements of the

CFRP beatnote were taken approximately one di,- after the turbo pump were turned on


o 400

L 350



0 0.5 1 1.5 2 2.5 3 3.5 4
Time (days)

Figure 5-33. Beatnote of the CFRP cavity in Tank 2 taken over 4 d4-v

and showed a drift in the beatnote of ~12 kHz/s. The next d4i-, the drift was observed

to be ~ 6 kHz/s, and the following di- it was observed to be ~3 kHz/s [68]. It then

continued to drift at this rate for several d4i It is assumed this drift is most likely due to

the outgassing of the CFRP cavity.

Although the CTE of the CFRP cavity is not known, it is assumed to be larger than

that of the Super Invar cavity since the beatnote oscillations are significantly higher than

those of the Super Invar cavity. Comparison of the tank temperature fluctuations when

the Super Invar cavity was measured and when the CFRP cavity was measured show to

be within a factor of two. This can not explain the order of magnitude larger frequency

fluctuation of the CFRP beatnote and it is assumed the large swing is due to a larger

CTE than that of the Super Invar cavity. Based on this, it is expected that the CTE of

the CFRP should be about an order of magnitude larger than that of the Super Invar


To estimate the CTE of the CFRP cavity, the same process was done to find the CTE

of the Super Invar cavity. Using Eq. 5-21 the resultant CTE was found to be 1-3 ppm. As

expected, this CTE is about an order of magnitude higher than that of the Super Invar.

1 LISA Telescope Requirement
10 -- LISA Pre-stabilization Requirement

I 103


E 101

'5 100


10-4 10-3 10-2 101
Frequency (Hz)

Figure 5-34. Stability of the CFRP cavity along with the LISA telescope and
pre-stabilization requirements.

Although the CFRP cavity meets the LISA telescope requirement at higher

frequencies, the causes of the high noise at lower frequencies is of interest in order to

determine what processes are responsible. It is known from the previous cavities that

the low frequency noise is due to the high CTE of the CFRP material. The frequency

these temperature effects typically occur at are below 1x 10-3 Hz. The additional noise

between 10-3 and 10-2 Hz was of interest and it was speculated that the mirror coatings

could be the cause.

The high reflectivities of the cavity mirrors will cause the light to bounce back and

forth within the cavity, causing a power build-up. This will cause a local heating of the

mirrors and mirror coatings that will translate into a length expansion. The effect of the

incident power on the noise spectrum can be tested by changing the input power of the

laser. To change the incident power, a neutral density (ND) filter was placed before the

polarizing beam splitter. The ND filter was then aligned such that the visibility of the

cavity was not changed compared to when there was no ND filter. The laser was then

maximized on the photodiode used to produce the PDH error signal. Without any ND

filter the DC voltage was 3.7 V and the incident power was about 3 mW. One ND filter

caused the PD voltage to drop to 1.8 V, and another ND caused the PD voltage to drop

to 0.18 V. These voltages correspond to 1.6 mW and 150 pW, respectively. A time series

at each of the photodiode voltages were taken over three d i and the resulting noise

spectrums are shown in figure 5-35. It is clear that by decreasing the incident power, the

noise spectrum decreases in the 10-4 to 10-2 frequency range. With a low enough incident

power, the CFRP cavity noise spectrum will meet the LISA telescope requirement.

8 3.7 V
10 1.8 V
--0.18 V
107 LISA Telescope Requirement

N 106

o 104

10-4 10-3 10-2 10-1
Frequency (Hz)

Figure 5-35. Dimensional stability of the CFRP cavity for differing incident intensities
along with the LISA telescope requirement.

From the noise spectrum alone, it would appear that CFRP would be a viable option

for use on the LISA mission. Unfortunately, the outgassing properties and absolute length

stability still need to be studied in order to determine if they meet the LISA requirements.

In addition, the shape of the CFRP cavity is not indicative of what would be designed

for the LISA spacecraft. The outer shell around the CFRP cavity was designed such that

when the CFRP outgassed, the bending of the structure would be kept to a minimum.

This supporting ring would not be viable for the LISA telescope support and a more

realistic design needs to be made and tested before a final decision on material should be

chosen for the LISA telescope support structure.


In some cases, it will be necessary to know the absolute stability of a material in

addition to its relative stability. An example of this is the telescope support structure

for the LISA mission. The material chosen must have a relative stability of better than 1

pm/ /Hz above 3 mHz in addition to expanding or contracting by less than 1.2 pm over

the lifetime of the mission in order to prevent de-focusing of the laser. Although results

for the relative stability of materials were presented in C'! lpter 5, the experimental setup

used is not adequate to determine the absolute stability of a material over month time

scales or longer due to the large frequency dependence on laboratory temperature and the

densification of the optical cavities. The strong temperature dependence can be reduced

using heaters and feedback controllers, but the densification of the material used for the

optical cavity is not very well known. C'! ,ih5iig the frequency reference from an optical

cavity to an atomic transition will provide a much more suitable frequency reference

over long periods of time. For this reason, two methods known as fr c- ii -n. '-modulated

spectroscopy (FMS) and modulation transfer spectroscopy have been used to determine

the absolute dimensional stability of a material.

6.1 Saturation Spectroscopy

Figure 6-1 shows a laser beam transmitted through both a gas cell containing

an absorbing molecular vapor and a laser transmitted through a Fabry-Perot cavity.

The transmitted intensity for the optical cavity will peak around resonance, while the

transmitted light for the gas cell will dip around the molecule's transition. Deriving an

error signal to lock a laser to the resonant frequency of an optical cavity has already been

presented and is similar to using a molecular transition [85]. In the case of a molecular

transition, the linewidth of the transition is significantly larger than that of a high Finesse

optical cavity.

To understand the concepts of saturation spectroscopy, consider the situation where
the laser is being modulated, with the frequency of the modulation being fairly low and
the amplitude of that modulation is small. If the laser frequency is in the vicinity of
an absorption line, then the frequency modulation causes the absorption to modulate
synchronously which causes the laser frequency modulation to map onto the laser's
transmitted frequency [85]. Another way of stating this is that the frequency modulation
on the laser converts to an amplitude modulation by the absorption.

D IT (V)

Gas Cell v


Optical Cavity v

Figure 6-1. Intensity as a function of frequency for both an optical cavity and gas cell.

Near the frequency of maximum absorption, the conversion from frequency modulation
to amplitude modulation will be small, and goes to zero at the transition line's center.
After it passes through the line's center, the amplitude conversion increases, but the
phase relationship has now reversed. If the laser's frequency is well outside the absorption
line, the amplitude modulation goes to zero due to the lack of absorption [85]. The
phenomenon is know as frequency modulation spectroscopy and is extremely useful in
locking a laser to the frequency of an absorption line of a gaseous substance [85].
When the beam passes through the gas cell, the laser frequency does not have to be in
exact resonance with the transition to experience an absorption of light. This is due to the
Doppler effect. The Doppler broadening of the transition line occurs because the atoms or
molecules are in motion [85]. If an atom or molecule is moving such that it has a velocity

component that is parallel to the direction of the beam, then the resonant frequency of the

Doppler shifted atom or molecule will be [85]

v= o(1+ ), (6-1)

where v is the resonant frequency of the moving molecule, v0 is the resonant frequency of

the molecule with no velocity, c is the speed of light, and vj is the parallel component of

the moving particle with respect to the laser beam. It is clear from Eq. 6-1 that as the

gas is heated up, the Doppler frequency will be larger, resulting in a broader transition

line [85].

Let's now consider a weak beam passing through a gas cell that is incident on a

photodiode and a strong beam propagating in the opposite direction as the weak beam

such that the two beams nearly overlap. It will be shown that this results in significantly

narrowing the widths of the absorption lines. As the weak beam passes through the gas

cell, it will experience a Doppler broadening of the transition line. The strong beam will

experience a Doppler shift, although in the opposite direction, and so both beams interact

with a completely different group of atoms as long as the laser frequency is far enough

off resonance. If the laser frequency is near vo, then both beams are now competing for

the same atoms. The strong beam will quickly deplete the number of atoms with v| = 0,

leaving very few atoms to interact with the weak beam [85]. This causes the weak beam

to pass through the cell with little loss in intensity, resulting in a spike in intensity at

resonance on the photodetector. This effect can be seen in Figure 6-6 and can be used to

generate an error signal that the laser can lock to. In this manner, an absolute frequency

reference can be established.

6.2 Experimental Setup

The only known absorbers near 1064 nm are 133Cs2, C02, C2H2, and C2HD [86] [87].

Of these absorbers, only molecular cesium (133Cs2) can be investigated by standard

sub-Doppler techniques. The other absorbers have low cross-sections and require

saturation powers that are very high. Because of this, special techniques, such as using

a high-finesse cavity, have been developed to study their sub-Doppler spectra [86]. A\ iii:

difficulties associated with laser locking to molecular cesium can be circumvented by using

second-harmonic generation and locking to sub-Doppler lines of molecular iodine (12212) as

discussed in ('! lpter 7.

The majority of the development of frequency standards at 1064 nm has been done

using Cs2 [88]. Cesium spectroscopy in the 0.90 to 1.14 pm regime occur in the A1 Z+

-- X1E band and have been reported by Orlov and Ustyugov [86]. It has also been

shown by Mak et al. that laser frequency stabilization is possible using sub-Doppler

lines of a cesium cell which was heated to 2200C and achieved a stability of 6x10-1 at a

measurement time of one second [89]. Investigations into the sub-Doppler spectroscopy of

molecular cesium were done in addition to using two different laser locking methods known

as modulation transfer spectroscopy and frequency modulation spectroscopy to provide a

frequency reference for use in determining the absolute stability of a material.

6.2.1 Laser Stabilization Using Modulation Transfer Spectroscopy

Using modulation transfer spectroscopy allows the sub-Doppler spectrum of molecular

cesium to be studied in great detail. The experimental setup used for cesium spectroscopy

is shown in Figure 6-2. The laser passes through a Faraday isolator, half-wave plate, and

polarizing beam splitter. By rotating the half-wave plate, the power in the pump and

probe beams can be easily changed. The probe beam then passes through the oven and

cesium cell. A detailed drawing of the oven and cesium cell are shown in Figure 6-3.

The cesium cell is 150 mm long, 19 mm in diameter, and has Brewster windows at

each end to prevent back-reflections from interfering with the incident beam. The cesium

cell is held in place by two circular copper supports at each end of the cell. These supports

are then held in place by stainless steel rods. The support structure is able to slide in

and out of a thin copper tube 200 mm long and 50 mm in diameter. The copper tube is

wrapped in a heating mesh whose power is controlled by a temperature controller (model



Figure 6-2. Experimental setup used for modulation transfer spectroscopy.

BriskONE from BH Thermal Corporation). A K-type thermocouple is attached to the

end of the glass tube using high temperature adhesive tape and is fed into the controller

to regulate the temperature. The heating mesh is wrapped with a high temperature

insulator, and this is then surrounded by additional insulation. The tube and insulation

are enclosed in a high temperature glass with a high transmission at 1064 nm. The

temperature stability of the cesium cell is better than 0.10C over 10 s as measured by the

thermocouple and temperature controller.

After the probe beam passes through the cesium cell it passes through a 50/50 power

beam splitter and is incident on a high-bandwidth photodetector. Simply using the probe

beam in this setup will allow spectroscopy of molecular cesium to be done, although

results with a low signal-to-noise ratio were typically achieved. This is due to the fact

that cesium is a poor absorber at 1064 nm, so temperatures in excess of 220C are needed

in order to see significant absorption which also causes additional noise. Typical cesium

lines at 240 C are shown in Figures 6-4 and 6-5. To determine the wavenumber of the

dips, a pick-off from the laser was sent into a wavemeter and the location of the dips were

matched to line numbers in references [90] and [89].

Cs Cell

Copper Support
--------- High Temp. Glass


Heating Mesh

Cesium Cell with
Brewster Windows

Copper Tube


Figure 6-3. Diagram of the oven used to heat the cesium cell.

The pump beam was phase modulated at 12 MHz using an EOM and then shifted in

frequency by ~20 MHz using an acousto-optic modulator (AOM) to prevent interference

between the probe and scattered pump beams. The pump beam had a diameter of ~2.0

mm with a power of 5 mW and the probe beam a diameter of ~1.3 mm and power of 1

mW as they passed through the cesium cell. A four-wave mixing interaction of the pump

and one of its side sidebands and the probe beam induce two new fields in the direction

of the probe beam at frequencies of w Q, where w is the probe carrier frequency and

Q is the modulation frequency of the EOM [90]. In this manner, modulation transfer

spectroscopy has no Doppler-induced baseline offsets, since the four-wave mixing is not

influenced by the Doppler background [90]. For this reason additional modulation, such as

using a chopper, is not needed to remove the Doppler background.

A sub-Doppler error signal was obtained in a similar manner as that of the PDH

signal described in section 5.2 by mixing the photodetector signal with the EOM

modulation frequency. Due to the low signal to noise ratio produced from the lack of

high absorbing cesium lines, a significant amount of signal conditioning had to be done

S0 8 -4
4245 93947

010 10 250 60 70 80 90 100
Time (s)

Figure 6-4. Intensity profile as the laser frequency is scanned over multiple absorption

to produce a suitable error signal. The AC signal from the photodiode was amplified by

approximately two orders of magnitude before it was mixed with the function generator

used to power the EOM. After the photodiode and oscillator signal were mixed, the error

signal was filtered using a 30 Hz low pass filter and amplified again, typically by a factor

of ten to fifty depending on the cesium line used.

Molecular cesium is a poor absorber at 1064 nm at room temperature and requires

significant heating before noticeable absorption occurs. To determine the temperature

dependence on the absorption percentage, the cesium cell was heated to 2000C, 2200C, and

250C while scanning over the same line. The results are shown in Figure 6-6 and clearly

show that as the temperature increases, the absorption percentage increases as well. In

addition to determining the absorption percentage, the linewidths of the transitions can be

determined as well. It was found that most transitions had FWHM linewidths of 300 to

500 MHz.

From figure 6-6 it would seem that to obtain the best error signal, temperatures

in excess of 2500C should be used to increase the absorbtion. Unfortunately at these





7 35 30 35 40 45 50 55 60

Figure 6-5. Detailed intensity profile of three absorption lines.

- 200 C
- 220 C
- 250 C

400 600 800 1000
Frequency (Mhz)

1200 1400 1600

Figure 6-6. Temperature dependance on absorption of molecular cesium.

temperatures, the cesium reacts violently with the glass cell and will often degrade the

cell to a point where micro-cracks are induced in the cell and air enters and reacts with

the molecular cesium and becomes useless [90]. Even if the cesium cells were kept at

temperatures less than 2400C, they would often have lifetimes less than eight months.

In addition to the accelerated degradation of the cesium cells at high temperatures,

significantly higher beam powers are needed to overcome the increased Doppler effects.

Typically above 250C, the signal to noise ratio of the error signal did not improve. For

these reasons, the cesium cell was typically operated at 235C in order to obtain a good

error signal in addition to lengthening the lifetime of the cesium cell.

Using the modulation transfer spectroscopy technique and scanning over a wide

range of frequencies will produce a set of error signals. In addition to providing a set of

suitable locking points, the error signals also provide information about the spectroscopy

of cesium. A larger peak-to-peak voltage implies the line is a better absorber. Using

the error signals also makes it easier to see where the cesium lines are. This is easily

seen by comparing Figures 6-4 and 6-7. The dips in Figure 6-4 are often hard to spot

without having a reference to match them up to. Alternatively, the lines are easily





Figure 6-7. Error signals produced for a range of cesium lines.

determined from Figure 6-7. In this manner the cesium spectra can be obtained with

high precision. The knowledge of the cesium spectra and strength of the error signal is

important when measuring the absolute stability of a material. While cavities have set

frequency shifts defined by their FSR, there is no pattern that can be used for locking

to cesium. This becomes difficult as many cesium lines are gigaHertz apart and most

commercial photodectectors have a range of less than a few gigaHertz. By knowing the

cesium spectra, a suitable line can be chosen and locked to in order to make the necessary

measurement to determine the absolute stability.

6.2.2 Laser Stabilization Using Frequency Modulation Spectroscopy

Although modulation transfer spectroscopy is ideal for cesium spectroscopy, and

presents a good signal-to-noise ratio, it was found in reference [90] that better results

were obtained using FM spectroscopy. For this reason, both modulation transfer and FM

spectroscopy methods were studied for laser locking.

The experimental setup used for FM spectroscopy is shown in Figure 6-8 and is

similar to that of modulation transfer spectroscopy. In the case of FM spectroscopy, the

pump is phase modulated and a chopper is used to extract the signal from the Doppler

background. The modulated probe beam is detected on the photodiode and is mixed

with the signal generator that is fed into the EOM. This signal is then fed into a lock-in

amplifier whose reference is provided by the chopper source. Most of the amplification

and low pass filtering was done by the lock-in amplifier and only a small amount of signal

conditioning was needed to produce an error signal with similar peak-to-peak and noise

levels as those obtained using modulation transfer spectroscopy. A range of chopper

frequencies were studied and it was found that using chopper frequencies on the order

of hundreds to a few thousands Hertz produced better error signals than using slower


6.3 Results

A pick-off from the cesium stabilized laser is beat with a pick-off from Laser 2 in

Figure 5-13. In this manner, the absolute stability of the material Laser 2 is locked to

Laser -
EOM Oven


chopper AOM



Figure 6-8. Experimental setup used for frequency modulated spectroscopy.

can be determined. In this case, the material chosen was a Zerodur cavity with optically

contacted mirrors.

Although it was reported in reference [90] that better results were achieved when FM

spectroscopy was used, the opposite was found to be true for my setup. The results for

using modulation transfer spectroscopy, FM spectroscopy, and a free-running laser are

shown in Figure 6-9. For almost all frequencies, using modulation transfer spectroscopy

was better than using a free-running laser, but locking a laser using FM spectroscopy

was worse than using a free-running laser. The cause of the excessive noise using FM

spectroscopy is unknown and several noise sources were investigated for both experimental

setups. The most plausible noise sources in both setups are RFAM noise produced from

the EOM, noise produced from the AOM, noise produced from a changing chopper

frequency, and temperature changes in the cesium cell.

6.3.1 RFAM Noise

Misalignment of the EOM will cause residual amplitude modulation on the photodiode

that ultimately shows up as noise in the beatnote. This can easily be seen if the output

of the photodiode is monitored by a spectrum analyzer around the frequency the EOM

10-1 100

Figure 6-9. Stability results of the modulation transfer and frequency
spectroscopy techniques compared to a free-running laser.


is modulated at. As the beam alignment is changed, the peak amplitude changes as well.

A change in the input polarization of the beam will also cause the residual amplitude

modulation (RAM) to change. In addition, if the alignment and polarization are left

constant, the RAM will change over time due to the length change of the EOM crystal

which is caused by temperature fluctuations of the crystal. All of these sources will cause

a changing RAM which is then detected on the photodiode.

To determine if the RFAM produced from the EOM was a limiting noise source, the

incident beam was misaligned such that the RFAM detected on a spectrum analyzer was

ten times more than normal. The noise spectrums between the aligned and misaligned

beams were then compared and showed no difference.

Modulation Transfer Technique
S FM Spectroscopy
S Free Running Laser
- Free Running Laser

^1. IA


| 107



10-3 10-2
Frequency (Hz)

6.3.2 AOM Noise

In addition to RFAM noise produced from the EOM, it is possible the instability

of the AOM could be causing a source of noise. If the shifted frequency changes

over time, this will cause a misalignment into the cesium cell, which will cause the

counter-propagating beams to overlap less, which causes the lamb dip to decrease. It is

also possible that by changing the AOM frequency, spurious interference between the

pump and probe beam can occur. The net effect if that by changing the AOM frequency,

the beatnote frequency will change as well.

To test the effects of a changing AOM frequency, the laser was locked and the

beatnote was monitored as the AOM frequency was changed. This effect was noticeable

but small, and it was concluded that this was not a significant enough noise source to be

causing the low frequency drift in the beatnote.

6.3.3 Changing Chopper Frequency

The optical chopper has a rotating blade that periodically blocks the laser beam

passing through it. When a chopper is used on the pump it will periodically affect the

absorption of the probe beam only when the frequency is near a resonance. Consequently,

only then will the chopping of the pump beam result in a modulation of the probe beam.

By measuring the probe beam with a photodiode and sending it and the chopper signal

into a lock-in amplifier, an error signal can be generated.

To test the effects of changing the chopper frequency, the laser was locked and the

beatnote was monitored as the chopper frequency was changed. The effect was negligible

and it was concluded that this was an extremely small contribution to the noise.

6.3.4 Cesium Cell Temperature Changes

The high frequency noise in Figure 6-9 is assumed to be from noise produced from

atom-molecule collisions in the cesium cell. This is because the atomic pressure is almost

three orders of magnitude higher than the molecular pressure at 2200C [90]. Thus,

operating at lower temperatures would be ideal, but this becomes difficult since the

signal-to-noise ratio decreases dramatically as the cell temperature is decreased. While

there are methods that would allow the signal-to-noise ratio to significantly improve at

lower temperatures [90], they are not discussed here.

Although the high frequency noise is assumed to be from operating the cesium cell at

elevated temperatures, it was found that the low frequency noise could not be attributed

to drifts in the temperature of the cesium cell. Figure 7 in reference [90] shows how the

beatnote frequency changes as the temperature of the cell is increased. Similar trends were

obtained from my experimental setup, with a change in frequency per degree Celsius found

to be 250 kHz/C. The assumed temperature stability of the cesium cell is 0.10C which

was determined by monitoring the feedback controller used for temperature stabilization.

From these two numbers, it was determined that the low-frequency noise could not be

attributed to the drifts in the cesium cell temperature. In addition, if the low frequency

noise was due to the changing cell temperature, the beatnote should show an oscillatory

nature that would arise from the cell heating and cooling. This effect was not seen.

After an analysis of the potential noise sources, it was postulated that the low

frequency noise seen in the noise spectrum was due to the expansion and contraction of

the Zerodur cavity, and not from an inherent noise source in the system. To test this,

the cesium-Laser 2 (or cesium-Zerodur) beatnote was recorded at the same time the

Laser 1-Laser 2 (or Zerodur-Zerodur) beatnote was recorded and the results are shown in

Figure 6-10. The kink in the Zerodur-Zerodur beatnote is due to one of the lasers coming

unlocked during the run. The kinks in the cesium-Zerodur beatnote is unknown, but were

commonly present in many data runs. It is possible the kink in data is from the control

electronics locking onto a different point due to the high level of noise in the error signal.

Despite the kinks in both data sets, it is evident that both beatnotes are drifting in the

same direction. A comparison of other, shorter, data runs shows that the cesium-Zerodur

beatnote consistently drifted ~1 MHz for every 10 kHz the Zerodur-Zerodur beatnote





0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6
Time (days)

Figure 6-10. Beatnotes of both the cesium-Zerodur and Zerodur-Zerodur

drifted. The exact cause of this is unknown, but it may possibly be from

drifts in the cavity, or laser intensity drifts over time.


32.86 a



locked systems.



In order for the LISA mission to achieve its goal of detecting low frequency

gravitational waves, the distances between proof masses on .,.i i'.ent spacecraft must

be measured to picometer precision. To do this, three distances must be measured; The

distance from proof mass to optical bench on one spacecraft, the distance from the optical

bench on one spacecraft to the optical bench on the other spacecraft, and the distance

from the proof mass to the optical bench on the other spacecraft. In order to make these

measurements, the light on .,I.i ient spacecraft must be sent to each other. This will be

done via a two-mirror reflecting telescope. The light on the optical bench will be focused

and sent to a secondary mirror where it will be magnified and reflected to the primary

mirror. The end result will be a highly collimated beam that will be sent to the optical

bench on the other spacecraft. Through the use of the same telescope mechanism mounted

on the second spacecraft, the light will be refocused and used on the second optical bench.

In this manner the necessary measurements can be taken to determine the distance

between proof masses on opposing spacecraft. Far-field wavefront distortions caused by

misalignment of the mirrors, a change in mirror separation, clipping from the secondary

support structure, or errors in pointing the telescope will decrease the sensitivity of the

length measurements needed. In some cases, such as the de-focusing of the mirrors or

beam pointing, if the effect is large enough, then the measurements may not be able to

be taken, and the link will essentially be lost. This makes the telescope and its support

structure critical optical components for the LISA mission. Despite the importance of the

telescope to the LISA mission, a design for either an on- or off-axis telescope had yet to be

designed, fabricated, and tested. Through collaboration with several group members at the

Goddard Space Flight Center (GSFC) and the University of Florida, an on-axis telescope

was designed, fabricated, and currently is in the process of determining its feasibility for

the LISA mission.

7.1 Introduction

The choice of telescope design for the LISA mission is restricted to only reflecting

telescopes since the size of the spacecraft ultimately constrains the length of the optical

bench plus telescope (in general reflectors are physically shorter than the corresponding

refractor of the same aperture). While there are many reflecting telescopes available,

the Cassegrain type is ideal due to its long focal length but short physical length. In

addition, either an on- or off-axis design can be employ, ,1 as shown in Figure 7-1. The

On-Axis Off-Axis



Optical Optical
Bench S Bench


(a) (b)

Figure 7-1. Representation of (a) an on-axis Cassegrain telescope and (b) an off-axis
Cassegrain telescope.

choice of material used for the secondary support structure will have direct influence on

the overall telescope design. The in-band length requirement of 1 pm/-/Hz above 3 mHz

and the expected temperature stability of ~10-6 K/ /Hz at 3 mHz at the telescope require

the material chosen to have a CTE of ~10-6/K [18]. Currently, the two most appealing

materials for the task are SiC and CFRP due to their high strength, extreme stiffness, and

low CTE. Both materials can be implemented in both on- and off-axis designs, but CFRP

is typically favored more for the off-axis design, and SiC is favored more for the on-axis

design. The current baseline for the LISA mission is an off-axis design using CFRP for

an f/1 telescope [23]. Although the viability of both on- and off-axis designs using both

CFRP and SiC has been done using computer models, more concrete evidence is needed

before a final design can be chosen. For this reason, a telescope support structure was

designed and built in order to gain more insight into what is needed in order to meet the

LISA telescope requirements.

7.2 Telescope Design

The shot noise limited phase error between the received and transmitted beams at

a far spacecraft is a significant noise source in the displacement sensitivity of LISA [91].

The magnitude of this noise will depend on several factors including the intensity of the

detected laser light, the beam width at the output of the telescope, and how well the beam

is focused. The quality of the telescope will directly effect these factors and must be taken

into account.

As an example, take into account the defocusing of the telescope. The distance

between the primary and secondary mirrors is a critical component of the telescope design.

If the distance between the primary and secondary change by too much, then the f/1

telescope may defocus the laser by too much. The following is an brief analysis that

describes the requirement on the primary-secondary distance change.

The telescope design calls for a flat wavefront of the outgoing beam of size w

immediately after the primary. If the wavefront is not flat and has a sagitta of size h,

the beam appears to have come from a focus with a radius given by [91]

o (7-1)

An analysis has shown that a sagitta of h = A/27 will decrease the apparent beam radius

by v2 which will increase the divergence angle by a factor of /2 and the received power

at the far spacecraft by a factor of two [91]. The f/1 design of the telescope results in a

strongly diverging beam. At the primary, the beam is already in the far field regime and

the radius of curvature is equal to the distance from the apparent focus, which is behind

the secondary. For a perfect adjustment, the primary will add a negative sagitta of


which flattens the wavefront [91].

If we assume that we move the primary by 6z, then this will cause a change in the

radius of curvature of the wavefront by the same amount which introduces a mismatch to

the primary and the resulting sagitta after reflection at the primary becomes [91]

L)2 6Z
h (7-3)
2R R

For the current design, estimates of h < A/20, R = 0.6 m, and w = 17.8 cm are appropriate[91].

A value of w = 17.8 cm is the Gauss beam radius which optimizes the intensity at the far

spacecraft for a 40 cm aperture Gaussian beam. Substituting these numbers into Eq. 7-3

results in an allowable length change of

6z < 1.2pm. (7-4)

This means that in order for the telescope to meet the wavefront distortion requirement

of A/20, the distance between the primary and secondary mirrors cannot change by more

than 1.2 pm over the lifetime of the mission. Other noise sources, such as the tilt of the

received wavefront, beam divergence, and pointing stability, will p1 i, a critical role in the

achieved sensitivity of LISA [92].

Taking these potential noise sources into account and current material fabrication

limitations, a design for the telescope was chosen as shown in Figure 7-2. It consists

of a 400 mm diameter primary, a 135 mm diameter secondary, and four 5 mm x 30

mm struts that connect and separate the primary and the secondary by 600 mm. SiC

was chosen to make the entire structure due to its high stiffness, low CTE, and high

thermal conductivity. The high stiffness and thermal conductivity will reduce temperature

gradients in the telescope which will reduce the effects of tilting and pointing errors, while

Figure 7-2. Isometric view of the telescope design. Figure courtesy of Joseph Generie at
Goddard Space Flight Center.

the low CTE will keep the in-band noise below the required 1 pm/n/Hz. In addition,

results for the relative dimensional stability of SiC were already known during the

design process, while those of the CFRP cavity were yet to be studied. Although it

is claimed that there are manufacturers that can make CFRP with a low CTE, high

thermal conductivity, high stiffness, and are able to describe the outgassing and shrinking

process, SiC was chosen primarily due to the familiarity of the material. SiC also has the

advantage that numerous bonding techniques, specifically hydroxide bonding, can be used

to joint parts together with high precision.

Although a three strut design seems like a more logical choice than four struts from

a mechanical point of view since a four strut system will be an over-constrained system,

the decision to use four struts was made due to the main quadrant photodiode that is used

downstream in the incoming beam optical path. This photodiode not only detects the

interference signal between the received and reference beams, but also acts as a wavefront

sensor by comparing the relative phase of the signal in the four quadrants. Using four

struts will produce four equal shadows that will be imparted on each photodiode quadrant.

By aligning the photodiode such that the shadows are 450 off the photodiode axis, then

the effect of the shadows produced by the struts is equal in each quadrant. Although

it may be possible to align three struts such that the shadows seen on the photodiode

doesn't interfere with the wavefront sensing, further analysis is needed. The cross section

of the struts was chosen such that the shadow produced on the quadrant photodiode was

minimized while still providing enough area such that multiple bonding techniques could

be used.

Finally, the holes in the primary and secondary were designed to allow a large

flexibility in measurements. The use of either a Fabry-Perot cavity or Michelson

interferometer can easily be implemented or changed depending on the specific measurement

that needs to be made. The outer holes are ideal for Michelson interferometers that will

measure the induced tilt of the telescope as it cools down, while the center holes are ideal

for using a Fabry-Perot cavity to determine the in-band noise.

Once the telescope design had been finalized, the next task was to determine what

temperatures would be expected on the telescope. The current design for an on-axis

telescope has the secondary effectively seeing open space temperatures, while the primary

experiences the heat generated from the optical bench. A finite element analysis of

the LISA spacecraft was done at the Goddard Space Flight Institute and it was found

that the expected operating temperature of the telescope was -71C and that there was

an axial gradient of about 1.5C and negligible radial gradient. The low temperature

gradient is due to the high thermal conductivity of the SiC. This is ideal since the

uniform temperature will decrease the tilting effects of the telescope. In order to meet

Optical Bench

ESA Strongback

Goddard Strongback

Secondary Mirror Support truths

Figure 7-3. Payload assembly for one of the optical benches, on-axis telescope, and
thermal shield for one "tube" of the LISA spacecraft. Each spacecraft will
consist of two tubes. Figure courtesy of NASA.

the parallelism tolerances needed in the design of the telescope support structure, a

high-stability, strong bonding technique needed to be chosen that also provided precision

alignment. The hydroxide bonding technique was chosen due to its flexibility and ease of

use, strong bond strength, and it ability to survive thermal cycling from room temperature

to liquid nitrogen temperature.

7.3 Telescope Fabrication

A significant portion of the telescope design process occurred while discussing options

with several SiC vendors. Each vendor has their own brand of SiC and their properties

and design tolerances vary depending on the vendor and type of SiC. Eventually, it was

decided that CoorsTek would supply the SiC parts. This was decided primarily due to

the fact that they could supply the parts as needed in addition to meeting the polishing

,4 I AT=-1 50


AT=- 00

Figure 7-4. Expected operating temperatures on the LISA telescope. Figure courtesy of
Angelique Davis at Goddard Space Flight Center,

tolerances required. Brazing and using an epoxy were considered as options to assemble
the telescope, but were eventually discarded. Brazing and sintering would require elevated
temperatures and could ruin any optical coatings that would be on the final telescope
design, and so it was considered an unrealistic option. Using epoxy was discarded as
an option due to the extreme difficulty in getting the right bond thickness in order to
keep the parallelism requirement between the primary and secondary mirrors. It was
decided that hydroxide bonding would be used because it allowed for precision bonding,
has significant bonding strength, and could survive thermal cycling beyond the operating
temperatures of the telescope. The use of hydroxide bonding also required that significant
dimensional tolerances be placed on the telescope parts since large gaps cannot be filled
using a sodium silicate solution.

The most stringent polishing specifications require that the struts all be to within 2

pm in length of each other and that the global flatness on the strut ends be to better than

2 pm. This was in order to avoid a gap large enough such that hydroxide bonding would

not be able to bridge [68]. Estimates of an upper limit on the gap size hydroxide bonding

could fill were placed as 8 pm and were based on data from the University of Glasgow. An

analysis by the group members at GSFC showed that if the following polishing tolerances

were met, then there would be gaps small enough that the hydroxide bonding process

would be able to bridge:

Primary disk tolerance:

1. Flatness on four strut mount pads 1 pm

2. Flatness on four small mirror mount pads, and center mirror mount surface 5 pm

3. Flatness on primary disc backside 0.5 mm

4. Surface finish on four strut mount pads, four mirror mount pads, and center mirror
mount surface 0.1 pm

Secondary disk tolerances:

1. Flatness on bottom side 2 micron

2. Surface finish on bottom side 0.1 pm

Strut tolerances:

1. Flatness on bottom surface 1 pm

2. Parallelism between top surface and bottom surface 2 pm

3. Perpendicularity between bottom surface and side surfaces and angularity- 1mm
(This means the top of the strut could be 1 mm out of tolerance on all four sides at
top of strut. The 1 mm tolerance is peak-to-valley)

4. Surface Finish on top and bottom ends 0.1 pm

5. Variation in length between all four struts 2 pm

The other polishing requirements (specifically the 0.1 pm requirement) were such that
the surfaces would be flat enough for a strong bond to form when using the hydroxide
bonding technique.
The primary, secondary, and struts were all fabricated and polished at the CoorsTek
facility and shipped to GSFC where the tolerances were checked. It was found that the
requirements had been met. An alignment jig was designed at GSFC and built by an
outside company (Figure 7-5). The alignment jig held the struts in place and allowed them
to slide up and down in a reproducible manner. In three of the four cases the alignment
jig allowed the struts to be lifted and then placed back on the primary with no change
in the surface contact between the strut and primary. In the other case, a small weight


Telescop .Alignment
Sp ac e and
:,, assembly
H jig

Figure 7-5. Diagram of how the telescope parts fit using the alignment jig. For clarity, not
all of the strut clamps are shown. Courtesy of Jeff Livas at Goddard Space
Flight Center.

(approximately 2.5 kg) was needed in order to keep the strut in suitable contact with the

primary. This small weight was needed to counteract a spurious restoring force that would

push the strut upwards. This sometimes resulted in small gaps unless a weight was placed

on top of the secondary. All four struts were placed such that they were flush with the

primary and the secondary was then placed on the struts. After doing this, a noticeable

gap could be seen and the alignment of the struts were modified to produce the best

suitable contact for hydroxide bonding.

Figure 7-6. Picture of workstation to determine the gap sizes between the struts and either
primary or secondary mirrors.

To produce this alignment, a weight of approximately 5 kg needed to be placed on

top of the secondary to make sure the struts il flush with the primary and secondary

as much as possible. From previous bonding experience, the additional weight should not

cause a weakening of the bond. In fact, several SiC-BK7 bonds have been produced with

significant strength by applying a downward force on the BK7 rod when bonding. All

SiC pieces were cleaned thoroughly before bonding. The struts had been cleaned using

methanol and the struts ends were sonicated in DI water for 30 minutes before being

Figure 7-7. Gap size when not a strut is not properly aligned (left) and gap size when a
strut is aligned properly (right). In some cases a downward force was needed
in order for the gap to close. Gaps shown on the left side of the figure were
typically 10 to 20 pm.

cleaned with methanol and clean room wipes. The primary and secondary were too big

to be placed in the sonicator, so they were cleaned using methanol and clean room wipes.

The bonding was done in a clean room that had been used a few weeks before to assemble

the Lunar Reconnaissance Orbiter (LRO).

In order to bond the telescope pieces together, all four struts were lifted and

approximately 3.0 fL of sodium silicate solution that had been diluted volumetrically

by a 1:4 ratio was placed under the strut one at a time. After the solution had been

applied the strut was then lowered and kept in place by the alignment jig. Once all four

struts were lowered onto the primary, the same sodium silicate solution was applied to the

top of the struts and the secondary was placed on top along with the weight. The bonds

cured overnight in the clean room and were analyzed using the same method described for

the alignment of the struts. Analysis of the bonds showed that all gaps had been filled by

the sodium silicate solution, and in some cases a gap of ~5 pm was seen to be filled.

The telescope was then transported by car from GSFC to the University of Florida

where it was kept in a low-traffic, low-dirt room. It 1 ,[ in this room while preparation

Figure 7-8. Picture of the sodium silicate solution filling in gaps between contact points of
the struts.

were made to test the telescope. It was kept in this room for several weeks with no visible

degradation to the structure. Within a month, it was noticed that one of the struts had

come loose and was leaning against the opposite strut but was fully intact. The most

plausible explanation was that it had been bumped by workers that had been in the room

although no concrete evidence could be found to support this. There was also no evidence

that would -i-i- -1 the bond broke due to any other reason. The bonds were visually

inspected and it was decided that the telescope should be cooled to -70C to see if the

damaged structure would survive the thermal cycling. A Michelson interferometer was

set up in the center of the telescope to monitor the distance between the primary and

secondary to determine if the bonds failed as the telescope is cooled. Both the vacuum

chamber and Michelson interferometer are described in detail in section 7.4 and 7.6,

respectively. Temperature sensors on the primary and secondary indicate they were cooled

to -70C and -60C respectively before being brought back up to room temperature. A

visual analysis of the telescope showed that it was still in-tact, but collapsed after applying

a small amount of force to the secondary in the direction perpendicular the optical axis.

One of the struts fell onto the mounting structure for the Michelson interferometer and

broke. The other two struts and secondary were caught before any damage could be done

to them.

To gain an insight as to why the structure collapsed, the hydroxide bonds were

investigated. It was determined that the alignment jig constrained the struts well in

the in the radial plane, but not it a torsional motion, and because of this the telescope

experienced a twisting motion as it was bonded (Figure 7-9). This was verified by a visual

inspection of the bonds using a low-power microscope camera. Using the camera, it

3 4 Secondary

2 4


2 4

Figure 7-9. Representation of how the struts bonded to the primary and secondary at a
tilt. The view of the secondary is as if one was looking up at the secondary
from the primary. The view of the primary is as if one was looking at it from
the secondary. The shaded areas are where the hydroxide bonds are thicker.

was clear the bond was thicker on one side of the struts than it was on the other. When

looking at the primary and secondary, the thicker part of the bond was seen on opposite


After a visual inspection of the bonds it was thought that another reason the

telescope collapsed could have been to the bond not fully forming over the bonding

surface. To test this, a piece of the strut that had broken was cut and analyzed using

an XPS. One piece was cut where it looked like the bonding occurred, and one piece

Figure 7-10. Picture of the secondary where the strut had bonded to it (left) and a picture
of part of a strut that had been bonded (right). The bonds appear to be
much thicker on one side, which leads to evidence that the structure twisted
during bonding.

was cut where it appeared little or no bonding occurred. If no bonding took place, then

there should be little or no SiO2 peaks in the spectrum. Analysis of the XPS data proved

inconclusive evidence for this theory. While the sample that appeared to bond showed a

higher relative abundance of SiO2 than that of the piece that appeared not to bond, the

piece that appeared not to bond showed a spectrum similar to that of previously pieces

that had bonded. The most likely theory for this is that the bond thickness of the sample

that appeared not to bond is very thin, and so appears to not have bonded since the

bonding agent is glass. In the sample that appeared to have bonded, the bond is most

likely significantly thicker and is easier to see. Thus, it would appear that the ini i ly of

the surface did bond, but the strut bonded at a tilt which produced a thicker bond on one

side of the strut than it did on the other [68].

After the analysis of the bonds were done, it was decided that the telescope parts

would be repolished and rebounded. The primary and secondary were sent back to be

repolished at the CoorsTek facility. When the original telescope pieces were made, copies

were made by CoorsTek in case something happens to the original. Since one of the struts

was broken, it was decided that the back-up struts would be used and polished instead

of the original struts. In order to prevent tilting of the struts on the next assembly,

the alignment jig was redesigned to prevent motion in that direction. In addition, an

aluminum mock-up of the telescope was fabricated to not only test out the alignment

jig modifications, but to also determine how feasible hydroxide bonding will be for the


Since no conclusive evidence could be found as to why one of the struts came loose,

a combination of hydroxide bonding and "sister-block" bonding will be used. The idea

behind sister-block bonding is that a small rectangular piece of SiC (roughly 3 mm

x 4 mm x 20 mm) will be epoxied to both sides of the strut where the strut meets

the primary and secondary. Using the sister-block bonding will provide extra strength

in addition to the hydroxide bonds. After extensive research was done into different

types of epoxies, it was decided that EP21TCHT-1 from Master Bond would be used

for the sister-block bonding. This choice was primarily due to the fast cure time at

room temperature, low oul ._ i-: low CTE, high strength, and ability to be used at

cryogenic temperatures. As discussed in C'! lpter 3, for an epoxy to form appreciable

strength, a rough surface and elevated temperatures are usually required to form the best

bond. To assemble the telescope neither of these conditions would be ideal. The surface

recommended would be too rough to keep the required parallelism of the telescope, and

heating the telescope too much may cause the hydroxide bonds to fail. To determine what

bond strength would be obtained using the sister-block bonding, a strut was bonded to a

tile with the same polish as the telescope using the sister blocks and epoxy (Figure 7-11).

The strut used for this test was cut using a tile saw from the strut that had broken after

thermally cycling the telescope. The use of the tile saw made the strut such that it didn't

sit flush with the tile, and a small gap could be seen.

The sister blocks were bonded and allowed to cure for two di at room temperature.

The shear strength of the sister-block bonding was tested using the MCST and found

Broken Strut

Epoxy Bonds
S/ Sister Blocks

Gap SiC Tile

Figure 7-11. Diagram of how the sister blocks were used to bond the strut and tile.

to be 8 MPa. This is half of what is quoted by Master Bond and is most likely due

to the smoother surface finish and lack of using an elevated temperature to cure the

epoxy. Although the EP21TCHT-1 epoxy is capable of curing at room temperature, it

is recommended that an elevated temperature be used to form the strongest bond. It is

most likely these two factors that contributed to the lower shear strength. Despite the

lower than expected shear strength, the sister-block bonding proved to be a method that

produced significant strength (larger than using hydroxide bonding) and would be feasible

for use when assembling the telescope again [68]. With this result, it was decided that

hydroxide bonding will be used to allow for the precision alignment needed to keep the

primary and secondary within parallelism tolerances, and the sister-block bonding will be

used to provide additional strength to the telescope structure.

7.4 Vacuum Tank Design

After the expected operating temperature of the telescope was known, the next

1n i', r task was to design a vacuum chamber that was capable of housing the telescope

and all of the optics needed to make the necessary measurements while keeping the

telescope at the correct temperature (-70C) with the expected temperature stability.

A vacuum chamber 1.4 m in diameter and 1.8 m high with several ports was already in

possession, but not being used at the time. The expected temperature of the telescope

made it difficult to design thermal shields and a cooling mechanism. Although solid

carbon dioxide has a temperature close to what is needed, the heat flow from the outside

would require frequent refilling, and the source would have to be close to the telescope,

which was nearly unfeasible. The next logical substance for cooling the telescope then

became liquid nitrogen. Placing a reservoir beneath the telescope was considered, but

ultimately ruled out due to the potential vibrational noise from filling the reservoir and

having the liquid nitrogen evaporate that would couple into the sensitivity measurements.

Using a lumped thermal model, a passive thermal shielding design with a liquid nitrogen

reservoir close to the top of the tank was decided on. The initial design did not get to the

desired temperature, but a few changes were made and the right temperature was met

(Figure 7-12). To impede the flow of heat from the tank base to the telescope, a series

of stainless steel tubes and plates were used. The plates are 1.2 m in diameter, 3.1 mm

thick and the bottom face is covered in a 1-v.-r of aluminized PET. The tubes are 25.4

mm long, 25.4 mm in diameter, and have a wall thickness of 1 mm. Ten tubes are placed

on the tank base followed by a stainless steel plate. This 1 -, li-ii, is then repeated. Four

aluminum rods 38 mm in diameter and 1 m long are screwed into the top stainless steel

plate and a support structure that holds the reservoir in place. The top of the reservoir

is equipped with a flange that allows a filling tube to be removed so the tank lid can be

placed over the entire structure. In this manner, the reservoir and thermal shielding are

enclosed in vacuum, but filling the reservoir can be done outside the tank. Four Macor

rods 50.8 mm in diameter and 25.4 mm long are used to support an aluminum baseplate.

This baseplate has holes drilled and tapped in a 25.4 mm square grid to allow optics to

be placed on it. The aluminum baseplate is connected to the liquid nitrogen reservoir

through the use of four copper rods 25.4 mm in diameter and 38 cm long. In addition, two

1-v.. -i of aluminized PET that are directly connected to the liquid nitrogen reservoir are

used to shield the telescope from thermal radiation. The first 1liv.r surrounds the reservoir

Fill Tube
Fill Tubei LN2 Reservoir

Copper Rod

Telescope Aluminum Rod
Aluminum Plate-- aco
Layer 2 (SS)
Tank Top
Tank Top Layer 1 (SS)
Tank Base

Figure 7-12. Representation of the system used to cool the telescope (left) and a picture of
the finished system (right). Not shown on the left figure are the two 1'l ri of
aluminized PET. The aluminized PET shields are partially shown in the
figure on the right along with the SiC primary after the structure collapsed.

and second stainless steel plate. The second lI--r surrounds the telescope and aluminum

baseplate. Cooling an aluminum mock-up of the telescope using this configuration

produces a temperature of approximately -95 C on the primary. To raise the temperature

to the necessary -70C, a feedback controller using heaters and temperature sensors are

used to produce the desired temperature. In this manner the temperature of the telescope

can be changed over a wide range of temperatures. Although the computers models

showed the nominal operating temperature of the LISA telescope was -70C, this number

varied depending on the model used, material properties, and were based on estimates

of heat generated from the optical bench. Testing over a wide temperature range will

give a greater insight into how well the telescope design will work on the LISA mission.

Using the combination of thermal shields, temperature sensors, and heaters, a temperature

range of -90 to -500C can be obtained with a temperature stability of better than 100

pK/vHz above 1 mHz. The temperature measurement is limited by the sensitivity of the

temperature sensors and the actual temperature stability of the telescope is expected to be

better above 1 mHz. In addition, the secondary faces one of the thermal shields that is in

direct contact with the liquid nitrogen reservoir. In this manner, the secondary is seeing

an environment that is similar to what would be expected on the the LISA spacecraft and

the gradient across the optical axis of the telescope should form naturally.

7.5 Experimental Setup

The most significant aspects of the telescope are the in-band noise, the absolute

length stability (or de-focusing), and the tilt between the primary and secondary. To

test the in-band noise, a Fabry-Perot cavity will be made using mirrors placed on the

primary and secondary. Initially a low Finesse (~600) cavity will be used to get a general

characterization of the telescope and the optical setup. The lower Finesse cavity will

allow a better tracking of how the optical system changes due to the temperature change.

Although the noise levels are expected to be higher using the low Finesse cavity, once an

understanding of how the optical system works, a higher Finesse cavity can be made using

mirrors with a higher reflectance. Using the same method as described in ('! Ilpter 5, by

beating a laser locked to a cavity on the telescope and a laser locked to a Zerodur cavity,

the relative in-band noise can be found (Figure 7-13). By varying the temperature of the

telescope, the in-band noise can be characterized as a function of temperature.

To test the absolute length stability of the telescope, a laser locked to a molecular

iodine transition will be used. Iodine will be used instead of cesium due to the difficulties

associated with operating the cesium cell in an oven. The process is similar to that

described in ('! i pter 6, although an iodine cell and frequency doubling < -l -I 1 are used

instead of an oven and cesium cell. The frequency doubling ( i i--I I1 is needed in order to

convert the 1064 nm light into 532 nm light since iodine has strong absorption lines at 532

nm. The details of the experimental setup will not be discussed due to the similarity of

the cesium locking experiment presented in ('!i lpter 6. By locking a laser to a molecular

transition of iodine and beating it with a laser locked to a cavity on the telescope, the

absolute stability of the telescope can be measured. This work is currently underway by

other members of the group

Beam Splitter

Zerodur Cavity

Laser /

Tank 1

Figure 7-13. Representation of how the in-band and absolute length noise for the telescope
will be found.

Although it would be ideal to determine the absolute stability of the telescope at

-70C, it may be more feasible to determine the absolute stability at room temperature.

Determining the absolute stability at higher temperatures should provide an upper limit

on the compactification of the material since dislocations are less likely to propagate

through a material at lower temperatures, and are thought to be the leading cause of

the shrinking of a material. Once the compactification has been characterized at higher

temperatures, the temperature of the telescope could then be lowered for several weeks to

determine the effects of lower temperatures on the densification process.

The telescope was designed such that the tilt between the primary and secondary

could be measured in several v-wi. One way would be to use Fabry-Perot cavities placed

around the central cavity used to measure the in-band noise. Locking a laser to each of

the cavities and monitoring the beatnotes between them would provide a precise way to

determine the induced tilt in the telescope, but would require the addition of four extra

lasers. A more feasible way would be to set up Michelson interferometers around the

Iodine Cell

central cavity and monitor the outputs over time. Placing a power beamsplitter with

a highly-reflective coating on one face of the beam splitter and positioning it on the

primary and securing a mirror to the secondary will allow the length change between

the primary and secondary to be monitored with sub-micron precision. Using Michelson

interferometers will provide the necessary measurements while not requiring the addition

of several additional lasers since a pick-off from the primary laser beam can be used to

form several Michelson interferometers.

7.6 Results

The failure analysis of the bonds, determining the proper actions to re-bond the

structure, and repolishing the struts set the project back close to a year. The only

stability data that was gathered from the telescope is from the Michelson interferometer

that was running while the telescope was cooled down (Figure 7-15). To set up the

interferometer, a non-polarizing beam splitter was mounted on the primary and a mirror

was mounted on the secondary (see Figure 7-14). A mirror was placed on the primary a

few inches from the beam splitter. Since the distance between the primary and secondary

is about a factor of ten greater than the distance between the mirror on the primary

and the beam splitter, the interferometer will essentially measure the change in distance

between the primary and secondary.

From the data there are four noticeable characteristics in the data and each one

will be discussed. The first characteristic is the fact that the time between peaks slowly

increases over time. This is expected and is due to the temperature of the telescope

reaching its equilibrium value. Initially the telescope cools down rapidly, causing the peaks

to be closer together. As time increases, the telescope temperature reaches its equilibrium

temperature slowly, causing the peaks to spread out in time. Analysis of the telescope

temperature data shows that as the slope of the temperature decreases, the peaks spread

out more. Towards the end of the data, the peaks begin to get closer again, which is due

to the telescope heating up.


Stand -

Figure 7-14. Representation of how the Michelson interferometer was setup to measure the
distance change between the primary and secondary..

The second trend in the interferometer data is that the peak-to-peak values slowly

decrease over time. This is due to the misalignment of the interferometer as the rods that

hold the aluminum baseplate shrink. Realignment of the interferometer was done as the

telescope was cooling down and this can be seen from the offset of the peaks increasing in

addition to the peak-to-peak values increasing.

Another characteristic of the data are the spikes that can be seen. The spikes where

the voltages drops to zero can be accounted for from occasionally blocking the laser

when realignment of the interferometer was done. These are almost ahv-,i seen before

a change in amplitude of the voltage can be seen. The other spikes were seen when the

liquid nitrogen reservoir was refilled. The exact cause of this is unknown, but it is most

likely due to a settling or shifting of the thermal shields that slightly misaligned the


The last characteristic of the data is the high frequency noise of the data as shown

in Figure 7-16. This is believed to be due to a spurious interferometer that was induced

25 22

Time (hours) Time (hours)
Figure 7-15. Photodiode output of the Michelson interferometer as the telescope was

cooled down (left) and a detailed view of the same output over a shorter
0 5 10 15 20 25 30 35 40 10 1 16 17 18 19 20

cooled down (left) and a detailed view of the same output over a shorter
period of time (right).

from using an ND filter before the detection photodiode to avoid saturation. Subsequent

interferometers on the final telescope design will use substantially less power, and so this

problem should not be present. Noise associated from the drift in the laser was ruled out

due to the periodic nature of the signal, and the fact that the laser drifts by less than 50

MHz/di-, and is significantly less than the FSR of the long arm of the interferometer.

By counting the number of fringe peaks and using the equation Ad = nA/2, where

Ad is the change in distance, n is the number of fringe peaks, and A is the wavelength of

the laser, the total distance the telescope shrunk can be found. From the interferometer

data, it was calculated that the telescope shrunk by ~60 pm. The distance the telescope

shrunk can also be found using the telescope temperature and the CTE as a function of

temperature provided by CoorsTek. The telescope initially started at 200C and cooled

down such that the final temperature of the secondary was -600C and the primary was

-700C. Taking the average of these values, the assumed final temperature of the telescope

is -650C. The plot of the CTE versus temperature is shown in Figure 7-17 and is roughly

linear between 200 and 300 K. Fitting a straight line between this temperature range

results in a CTE given by CTEppm = 0.0lxT 1.11, where T is the temperature of the



a, 1.2
2 1.15


320 340 360 380 400 420
Time (sec)

Figure 7-16. Detailed view of the oscillations in the Michelson interferometer data.

telescope in Kelvin. Using the average of the CTE at 200C and -650C, and multiplying by

the change in temperature (85C) and length between the primary and secondary (0.6 m),

results in a length change of ~85 pm.

Although the amount the structure shrank using the Michelson interferometer should

be less than the value found using the CTE values since the interferometer arm on the

primary will shrink as the telescope cools down, the discrepancy between the two should

not be as large as it is. The cause of the discrepancy is not known and further testing will

need to be done on the aluminum mock-up before being used on the SiC telescope.

Currently, work is underway to rebuild the telescope spacer. The alignment jig

was redesigned and used to bond an aluminum mock-up using both hydroxide bonding

and sister-block bonding. The mock-up was cooled to -800C, warmed back to room

temperature, and no degradation of the bonds could be found. Thus, the combination of

using hydroxide bonding and epoxying the sister blocks should provide more than enough





a 1.6





50 100 150 200 250 300
Temperature (K)

Figure 7-17. CTE as a function of temperature for CoorsTek UltraSiC SiC. Data obtained
from David Bath at CoorTek.

strength to the structure. It remains to be seen how the epoxy will behave during the

relative stability measurements.


The LISA mission consists of three spacecraft each separated by five million

kilometers. Each of these spacecraft contain two optical benches that contain lasers

which point to the .,.i i,'ent spacecraft. Each optical bench contains a proof mass that

the spacecraft orientates itself around to provide a drag-free environment. Due to the

geodesics of the proof masses, the lengths between the spacecraft will change by as much

as 1 over the course of a year. To track the breathing of the spacecraft, the angle

between the optical benches on a spacecraft will have to change by as much as 1.

If the LISA constellation is thought of as a Michelson interferometer, then one

spacecraft acts as a beamsplitter, while the other two spacecraft act like mirrors. Because

two separate lasers are used at the beamsplitter spacecraft, they will need to be interfered

in order to make the appropriate phase measurements. This then becomes a problem since

the angle between optical benches will change over time.

The proposed solution is to use an optical fiber to transfer the beam from one bench

to the other, and vice-versa. The fiber will contain both counter-propagating beams and

will induce additional phase noise as it travels through the fiber but can be subtracted out

through the use of appropriate metrology. If the sensitivity of LISA is not to be spoiled

from induced fiber noise, then the differential phase noise of the counter-propagating

beams must be less than 10-6 cycles//Hz.

The current project plan is to use a single-mode polarization-maintaining fiber

approximately 0.5 to 1.0 m to accommodate the counter-propagating beams. Experiments

at both the Jet Propulsion Laboratory (JPL) and the Albert Einstein Institute (AEI) in

Hannover have failed to confirm that a bi-directional fiber meets the LISA requirement

and both appear to be limited by systematic noise [93], [94].

In order to test the differential phase noise induced in counter-propagating beams

through an optical fiber, a Sagnac interferometer was constructed and an optical fiber was

SC 2 SC 3

0 0

Optical Bench Laser

Back-Link Fiber

SC 1
Figure 8-1. Representation of the three LISA spacecraft, the optical benches, and the
back-link optical fiber.

placed in its path. The Sagnac configuration takes advantage of the fact that all noise will
be common except that of the noise induced from the fiber since the optical paths are the
same, but with opposite directions. This chapter contains an introduction to the Sagnac
effect and interferometer, the experimental setup used to test the differential noise in an
optical fiber, and the results obtained from this setup.
8.1 Sagnac Effect
Detection of rotation with the use of light was first demonstrated by Sagnac in
1913 by showing that two waves acquired a phase difference by propagating in opposite
directions around a loop interferometer that was rotating. The kinematic description

of the Sagnac effect is simple and is demonstrated by considering a hypothetical

interferometer with a circular path in vacuum as shown in Figure 8-2

co-1/2 Ac)

w0+1/2 A)D

/ 'R

P i

Figure 8-2. Diagram of a rotating loop interferometer to illustrate the Sagnac effect.
Doppler shifted frequencies are shown with respect to the inertial space.

When the interferometer is not rol ,li in then the counterpropagating beams will

have equal path lengths since the light travels at the same velocity in both directions.

This will result in both beams returning to the same point Pi of injection after a time of

- = 27R/c, where R is the path radius and c is the speed of light in vacuum. When the

interferometer is rotating at a rate f and the observer is stationary in the inertial frame,

then the point Pi moves through an angle Qfr during the propagation time r. Hence, the

difference in propagation times to lowest order in RQ/c is given by [95]

(27 + 2r)R (27 r)R 47R2
T (8 1)
c c c

For a continuous wave of frequency w, the corresponding phase shift becomes

AQ = wAT- R2 (8-2)

This result will remain unchanged when the interferometer is filled with a medium of

refractive index n since the Fresnel-Fizeau drag effect due to the movement of the medium

will compensate for the increased optical path lengths [96].

The Sagnac phase shift may be calculated using the Doppler frequency shift, but will

lead to the same result as in Eq. 8-2 since the Doppler shift will cause the wavelength

from the wave propagating in one direction to be different from that of the wave traveling

in the other direction. Thus, the two interpretations of the Sagnac effect are equivalent (to

lowest order in RQ/c).

A more rigorous approach to the Sagnac phase difference is to solve the electrodynamic

propagation equations in the rotating medium [97], but will not be considered here. What

can be derived from the electrodynamic formulism is that the phase shift depends only on

the light frequency u and on the dot product of the equivalent area vector of the mean

optical path and the rotation rate vector. This results in a phase difference of

A = A A (8-3)

Thus, the Sagnac effect is due to the difference in propagation times of the waves as they

travel around the rotating loop. The waves frequency acts as a clock to measure this time

difference and is independent of the velocity of the physical substrate used. Using a higher

frequency light will result in a larger phase difference.

8.2 Polarization-Maintaining Optical Fibers

Even if a fiber has a circularly symmetric design, it will exhibit some degree of

birefringence because there is ahv--, some amount of stress which will break the

symmetry. As this happens, the polarization state of the light traveling in the fiber

will change in an uncontrolled manner which is typically wavelength and temperature


To circumvent this problem, a fiber with a strong built-in birefringence will

preserve the polarization state even if the fiber is bent as long as the input polarization

state is aligned with one of the birefringent axes. This type of fiber is known as a

polarization-maintaining optical fiber. The propagation constants of the different

polarization states are different since the strong birefringence will cause the relative phase

of copropagating modes to drift away rapidly. The most common method to introduce

a strong birefringence is to include two stress rods of a modified glass composition in

the preform on opposite sides of the core as shown in Figure 8-3. The stress rods do not

necessarily need to be cylindrical, but the modified glass composition must have a different

coefficient of thermal expansion.

Glass Cladding Polymer Coating
Slow \

SStress \

PANDA style Bow-tie style

Figure 8-3. Both the PANDA and bow-tie polarization-maintaining fiber -I,! -; are shown
along with the core and stress elements.

The two most common styles of polarization-maintaining fibers are the PANDA

and bow-tie, although others are available. The two -I -1. are based on the stress rods

used that run parallel to the fiber's core. PANDA stress rods are cylindrical, while the

bow-tie design uses trapezodial prism stress rods. For most applications other than the

telecommunications industry, either style can be used, although PANDA style fibers are

typically used more often since the uniformity of the fiber is much easier to maintain.

In the case of the PANDA design, the optical core of the fiber is manufactured

independently of the stress rods, allowing it to be made with high chemical purity. Core

compositions are either a germanium-doped or pure silica core. The stress rods are then

inserted into circular holes drilled into the pristine preform allowing for a higher degree of

splicing and coupling to be done. This design also allows for lower splice loss and higher

extinction ratios when compared to other designs. The PANDA construction allows for

independent control over the optical core, stress rod size, location, and composition.

The control of these four factors will determine the performance of the fiber, and the

ability to control each one of them makes the PANDA design extremely versatile, all while

maintaining the polarization state in the optical fiber.

Polarization-maintaining fibers should not be confused with single-polarization fibers,

which can only guide light of a certain polarization. Polarization-maintaining fibers are

used in applications where the polarization state cannot be allowed to drift as a result of

temperature changes, such as for use in fiber optic gyroscopes, fiber interferometers, and

certain fiber lasers [98]. These fibers usually require a much greater degree of alignment

than for a standard fiber, and the polarization direction is also required for optimal

performance. Propagation losses are also higher than for standard fibers, and not all types

of fibers are easily obtained in polarization-preserving form.

8.3 Experimental Setup

In order to determine the amount of differential phase noise induced by a fiber optic

cable with counter-propagating beams, a polarizing Sagnac interferometer was constructed

and a fiber was placed in the path of the beam as shown in Figure 8-4. Both a qualitative

and quantitative description of the setup are given.

8.3.1 Qualitative Description

The interferometer consists of three polarizing beam splitters (PBS1-PBS3), three

half wave plates (HWP1-HWP3), one quarter waveplate (QWP), two collimating lenses

to focus the light into the fiber, one fiber, one calcite wedge, one balanced photodetector

(BPD), and four mirrors that are used for alignment into the fiber. The collimating lenses

are free-spaces lenses and are not directly attached to the fiber. In order to get the

differential phase signal, the incident beam is split before the fiber, then recombined after



Figure 8-4. Sagnac interferometer used to measure the accumulated differential phase
noise in an optical fiber. HWP1-3 are half wave plates, PBS1-3 are polarizing
beam splitters, QWP is a quarter wave plate, and the balanced photodetector
is represented by BPD.

the fiber in the following 15 steps (for a reference, the numbers in figure 8-4 match up

with the numbers below):

1. The beam from the laser passes through a lens that produces a beam with a
Rayleigh range of ~1 m. This allows the beam to be well collimated throughout
the interferometer. The beam then passes through a polarizing beam splitter (PBS)
to produce a beam with almost all p-polarization.

2. The beam then passes through another polarizing beam splitter (PBS1) that acts as
both an additional polarization cleaner for the incident beam and the pick-off to the
detector for the outgoing beam from the interferometer.

3. The beam passes through a half wave plate (HWP1) such that there is now an equal
amount of s- and p-polarization in the beam.

4. The beam is incident on a polarizing beam splitter (PBS2) that splits the beam into
its polarization components. The transmitted beam (green) contains p-polarization
only, while the reflected beam (red) contains primarily s-polarization, with a small
amount of p-polarization due to the beam splitters.

5. The transmitted beam (green) continues towards the fiber, while the reflected
beam (red) passes through a half wave plate (HWP2) orientated such that the
polarizations are rotated by 90.

6. The transmitted beam (green) passes through a collimating lens and is incident on
the fiber. The reflected beam (red) passes through another polarizing beam splitter
(PBS3) to isolate the spurious s-polarization that is produced from PBS2. In this
manner both beams incident on the fiber have the same polarization as they go
through the fiber.

7. Both beams emerge from the fiber with 1' efficiency.

8. The transmitted beam (green) passes through PBS3 cleanly, although ~5'. of the
power is lost. The reflected beam (red) continues towards PBS2.

9. The transmitted beam (green) passes through HWP2 and is rotated by 900, while
the reflected beam (red) continues towards PBS2.

10. The beams then recombine at PBS2. If there was no differential phase noise
induced from the system, then the resulting beam would emerge from the Sagnac
interferometer with linear polarized light in equal amounts. If any differential phase
noise is picked up from the interferometer, then the beams will recombine to produce
elliptically polarized light. The minor axis of the elliptically polarized light serves as
a measure of the differential phase accumulated in the interferometer.

11. The elliptically polarized light passes through HWP1 and changes the !i i,,.' .li."
of the beam. This means that if the beam was rotating clockwise, it would now be
rotating counter-clockwise, and vice-versa.

12. Once the elliptically polarized light is incident on PBS1 the majority of the
s-polarized light will be reflected, while only a fraction of the p-polarization will
be reflected. The result is such that an elliptically polarized beam is reflected,
but with a much lower power. If PBS1 was ideal and only the s-polarization was
reflected, then the output of this port would serve as a measure of the accumulated
differential phase.

13. The quarter wave plate (QWP) is orientated such that the beam is linearly polarized
after the elliptically polarized beam passes through.

14. The linearly polarized light then passes through a half wave plate (HWP3) with
an orientation such that the emerging beam has equal amplitudes of s- and

15. The beam then passes through a calcite wedge which separates the polarizations that
are then separately incident on a balanced photodetector (BPD). A calcite wedge is
used to separate the polarization more cleanly than through the use of a polarizing
beam splitter. The output on the BPD will read zero (or possibly a small offset) if
the orientations of the waveplates are correct, and the beam is properly aligned into
the fiber. Any accumulated differential phase from the interferometer will cause the
s-polarization at step 12 to be larger and will propagate through the waveplates such
that the amount of power on the BPD is not the same and a non-zero voltage will be

The advantage of this setup is that both the reflected and transmitted beams from PBS2

see the same optical components and both experience two transmissions and one reflection

through a polarizing beam splitter. These noise sources are common to each beam and will

cancel on the balanced photodetector, resulting in only the differential noise induced from

the optical fiber.

8.3.2 Quantitative Description

Now that a qualitative description of the Sagnac interferometer has been presented, a

more rigorous approach is taken to understand how to extract the differential phase noise.

The following analysis assumes an ideal case of the interferometer to determine the phase

shift from the balanced photodetector.

After both beams have traveled in opposite directions around the interferometer, the

resulting electric field after they recombine (step 10) is given by

Eo.t = EGeiOGp + EReiRs,

where EG and QG are the electric field and phase of the beam after traveling the green

path in Figure 8-4, and ER and QR are the electric field and phase of the beam after

traveling the red path. If we use the following definitions

E = (EG + ER)

2 ( + OcR)


A (QG O),

then our electric field after recombining becomes

E,,t = eil[(E + AE)ei&Pp + (E AE)e-i s]. (8-6)

The common phase shift (ei) can be ignored since it is a factor that will be carried

around and eventually canceled when finding the photodiode signals. Sorting Eq. 8-6 into

a more manageable form results in

Eot = (p + s)(EcosAQ + iAEsinMA) + (p s)(iEsinMA + AEcosAQ).


As the combined electric field in step 10 passes through HWP1, the half waveplate will

rotate both polarizations by 45. This will transform the electric fields in the following


(s+p) -+p


(8 8)

which transforms the electric field at step 11 into

En = (EcosAQ + iAEsinAf)p + (iEsinMA + AEcosA)Os.




As the laser passes through PBS1, it will reflect all of the s-polarized light, and a fraction,

r, of the p-polarized light. This results in an electric field at step 12 of

E12 = r(EcosAQ + iAEsinAf)p + (iEsinMA + AEcosA)^s. (8-10)

Aligning QWP such that its fast axis is aligned with p and its slow axis with s is the ideal

case since this will only produce an addition of a 900 phase shift to the s-polarized light.

This results in an electric field at step 13 of

E13 = r(EcosAQ + iAEsinA)p (iEsinA AEcosA)^s. (8-11)

In the ideal case, HWP3 would be aligned such that it rotates both polarizations by 45.

If there were no differential phase shifts accumulated in the interferometer, then there

would be no s-polarized light, and HWP3 would rotate the p-polarization such that an

equal amount of light fell on both photodiodes on the balanced photodetector. Rotating

Eq. 8-11 by 450 results in

14 = r(EcosAQ + iAEsinMA)(p + s) (iEsinA AEcosAQ)(s p). (8-12)

Rearranging Eq. 8-12 into its s- and p-polarizations results in

E14 = [(rEcosA( + EsinA) + i(rAEsinA AEcosAQ)]p

+[(rEcosA EsinMA) + i(rAEsinMA + AEcosA)]^s.

Using a calcite crystal instead of a polarizing beam splitter will split the polarizations

more efficiently, so the polarization dependance on the reflection and transmission

coefficients can be ignored and the polarizations are assumed to be completely separate

from each other. This results in only one polarization being incident on one photodiode

of the balanced photodetector, and the other polarization being incident on the other


Thus, at step 15 the powers in the s- and p-polarization are given by

Pp = E(rcosAO + sinA 2)2 + AE2(rsinA cosA)2 (8 13a)

P, E2(rcosA sinA)2 + AE2(rsinA + cosA)2. (8-13b)

To first order in AQ and second order in AE, cosAQ t 1 and sinAQ w AQ, and the power

levels are given by

P = E2(r2 2rA) + AE2 (814a)

P, E2(r2 + 2rA) + AE2 (8-14b)

and the difference in the photodiode signals is given by

P, Pp = -4rE2 (815)

Eq. 8-15 is the resulting balanced photodetector signal under ideal conditions for the

interferometer. The p-polarization leakage from PBS1 acts as a local oscillator and if no

differential phase is introduced, then the resulting signal from the balanced photodetector

will be zero.

8.3.3 Voltage to Phase Calibration

Although a relation to find the differential phase was found in Eq. 8-15 assuming

ideal operations, an experimental verification was done in order to determine the exact

relation between the output voltage of the balanced photodiode and the differential change

in phase. To do this, a 10 m fiber-EOM (model PM1060HF from Jenoptik) was put

in series with the fiber already in the interferometer and a voltage was applied to the

EOM to produce a tone at 10 MHz. This tone was then detected after HWP3 by a high

bandwidth photodetector made by the EOTech company since the balanced photodetector

does not have a bandwidth high enough to detect a 10 MHz signal. Rotating HWP3

will cause the amplitude of the tone to change. From this signal, the overall calibration

factor was determined. The conversion factor can then be applied to the time series of the

balanced photodetector signal to get the correct differential phase induced.

To determine the calibration factor, a 10 MHz, 3 V peak-to-peak signal was applied

to the EOM. The output on the EOTech photodetector was measured with a spectrum

analyzer and found to have a peak height of 70 mV. Using the impedance of the spectrum

analyzer (50 Q) and the efficiency of the photodetector (0.75 A/W), the RMS power

incident on the photodetector was found to be

0.070V 1
x = 1.87mW.
502 0.75A/W

To convert the RMS power to the actual power, the RMS power is multiplied by v/2 to get

an incident power of 2.64 mW.

The phase induced from applying the voltage to the EOM was found by multiplying

a conversion factor by the applied voltage. The conversion factor was taken from

experiments done at the AEI in Hannover and was found to be 7.2 rad/V for the

particular EOM used [99]. Using this factor and the applied voltage, results in a total

phase change of 22 rad, or ~3.5 cycles. Using the incident power, the amount of phase

change from the EOM, and the responsivity of the balanced photodetector (1.28 V/mW),

the overall conversion factor can be found using

r Pactual
cal =

where r] is the responsivity of the balanced photodetector and Pactual = 2.64 mW as

calculated above. Substituting these numbers, the calibration factor is found to be ~1

V/cycle. Although this was found to be the calibration signal, there was substantial

variability in the peak height of the tone as was measured by the spectrum analyzer.

The peak height varied due to the high frequency noise of the experimental setup and

an accurate value was difficult to determine. As a result, the peak height of 70 mV was

chosen because it was on the lower end on the peak height variability. If a larger peak

height was used, then the calibration factor would become larger, and the differential

phase noise would appear to be lower than it actually is.

8.4 Stability Results

In order to determine the sensitivity of the phase detector (in this case, this means

the combination of PBS1, HWP1, HWP3, QWP, and the calcite wedge), a mirror was

placed after HWP1 and the beam was aligned such that the reflected beam overlapped

the incoming beam as much as possible. Placing the mirror before the interferometer

acts as if the interferometer were ideal, and so the noise levels of the output fringe of the

interferometer can be found. It was found using this configuration that the amount of p-

to s-polarization after the calcite wedge was 7:1. HWP3 was rotated such that the powers

were then equal on the balanced photodetector. The output of the balanced photodetector

was sent into an amplifier where it was amplified by a factor of 20 and low pass filtered

at 1 Hz. The gain on the output of the photodetector was needed so the measurement

would be above the noise floor of the data acquisition card, while the low pass filter was

used to filter some of the noise at higher frequencies. The setup was covered in cardboard

to provide some thermal stability. The results are shown in Figure 8-5 and appears to be

very flat.

Once the noise floor of the interferometer was known, a fiber was placed in the optical

path, and the differential phase noise was measured. Two different optical fibers were

tested and both produced similar results. The first fiber is 10 m in length with FC/APC

polished ends and was bought from the Nufern company. The second fiber was bought

from the Schafter and Kirchhoff company and is 10 m long with FC/APC polished ends.

The Nufern fiber has a significantly thicker outer cladding than the other fiber, but this

didn't appear to effect the measurements. The ends of the fiber were polished at an

angle to help avoid back-scattering of the incident light. Both fibers are single-mode

polarization-maintaining fibers and have a numerical aperture (NA) of 0.12. Light was

focused into the fibers through the use of free-space collimating lenses with an NA of 0.15.

(D 10-6



10-4 10-3 10-2 101 100
Frequency (Hz)

Figure 8-5. Noise spectrum of the detector components of the interferometer. This was
found by placing a mirror in front of PBS1 and aligning the back reflected

With proper alignment, most of the time 70- i' of the light would go through the fiber

each way using the collimating lenses. All optics pertaining to the interferometer were

placed on a Super Invar breadboard bought from N. '. port (represented by the square in

Figure 8-4). Typical input powers into the interferometer were between 10 to 30 mW and

produced less than a few milliWatts of power on the detector photodiodes.

The interferometer was placed on a vacuum tank flange (represented by the circle in

Figure 8-4). The tank lid could then be lowered on top of the flange through the use of

a pulley. Several ports were available for light to go in and out of the tank. The initial

alignment of the interferometer was done with the lid off and the top was then lowered

after the alignment was done. This sometimes caused a small change in alignment. When

the tank was on and not under vacuum, one of the ports was left open so the light didn't

pass through a glass plate. Aluminum foil was placed over the port to help restrict air

currents and provide a small amount of temperature stability inside the tank. A small hole

was punched in the aluminum foil to allow the beam to pass through. A noticeable effect

could be seen in the noise spectrums when comparing runs with the tank top on and with

it off. When the tank top was on, the noise levels were typically a factor of two to five

lower than when the tank was off.

The rest of the optics outside the vacuum tank were shielded using cardboard. The

noise levels didn't noticeably change because of it. The reason for the lower noise levels

with the tank lid on is unknown, but it is assumed that the temperature fluctuations are

less when the tank lid is on.

To try and improve the stability, the port on the tank that had been covered with

aluminum foil was replaced with a glass flange and the tank was pumped down using

a roughing pump. Unfortunately, as the pressure dropped, the alignment into the fiber

became worse and eventually no light was going through the fiber. The misalignment

as the pressure drops most likely occurs due to stresses that are induced in the optical

components that change the beam trajectory. Attempts to realign the beam into the fiber

through mirrors outside of the tank were unsuccessful.

After several tests with the fiber in the interferometer and the tank placed on top,

it was found that a parasitic interferometer was causing the photodiode signal to be

significantly more noise [100]. The solution was to modulate the laser frequency by

applying a triangle wave with amplitude of 3 V peak-to-peak and frequency of 5 kHz.

The noise dropped by almost an order of magnitude as shown in Figure 8-6. Modulating

at this frequency changed the laser frequency by 75 GHz/s. By modulating the laser

frequency at this rate, the fields reaching the detector have a frequency different from the

main input beam [101]. In this manner, the noise from the parasitic interferometer is now

well outside of the measurement band.

After this problem had been solved, it was then noticed that over the course of a

few d -iv the amount of light going through the fiber dropped significantly. Often the

fiber would be aligned such that s I' of the light would be transmitted and a few di-,

later this would drop to 50' or lower. Typically the light transmitted from PBS2 was




10 50 1I.0 150 200 250 300
Time (sec)

Figure 8-6. Plot of the time series before and after modulating the frequency of the laser.
It is clear that modulating the laser frequency produces significantly less noise.

affected more than the reflected light. To determine the cause of this, glass plates were put

in at each end of the fiber and were aligned such that the angle the surface of the glass

plate made with the output of the fiber was close to the Brewster angle. The powers of

both glass plate reflections were measured using a photodiode. In addition, another glass

plate was placed after HWP1 in the same manner to determine if the input power was

changing as well. The results of all three measurements taken at the same time are shown

in Figure 8-7. It is easy to see that the input power is -I ,vii.-; almost constant while the

output powers from the fiber are changing over time. The exact cause of this was never

determined, but it is speculated that it could be from pointing issues into the fibers.

While the powers measurements were being taken, the output of the balanced

photodetector was monitored as well as shown in Figure 8-8. The exact relationship

between the changing fiber output powers and the change in voltage of the balanced

photodiode is difficult to determine. As the power through the fiber changed, it was

also determined that the mode of the fiber after HWP3 changed as well. When the

0.5 1 1.5
Time (sec)

2 2.6
x 10

Figure 8-7. Time series of the input power to the interferometer (left) and the time series
of the powers during the same time in both the clockwise and counterclockwise
beams (right).

S-- Balanced PD signal





0 0.5 1 1.5 2 2.5
Time (sec) x 105

Figure 8-8. Time series of the balanced photodetector at
counterpropagating beams are measured.

the same time the powers of the

Po,:er Lhrough C: W
Power through CW

interferometer was properly aligned, the resulting beam shape after HWP3 appeared to

be that of a TE AFi., mode. After 8-12 hours, the mode would become oddly shaped and

was most representative of a TF ii,, or TF i,,-. mode. Realignment of the interferometer

would cause the beam to revert back to its original shape. The exact cause of both the

power loss through the beam and the change in made is difficult to determine from the

data taken, but it may be that they are linked somehow.

The best results for the differential phase noise are presented in Figure 8-9. At higher

frequencies, the noise is on the 10-5 level, but at lower frequencies, the noise is about two

orders of magnitude higher. If the differential phase noise scales with the length of the

fiber, then the levels would be 10 to 20 times lower since the LISA back-link fiber will

only be 0.5 to 1.0 m in length. This would almost meet the LISA requirement at high

frequencies, but a significant amount of work will need to be done in order to determine

the cause of the low frequency noise.


i I
(D 10-4


10-3 1C02 101 100 10
Frequency (Hz)

Figure 8-9. Noise spectrum of the optical fiber with counter-propagating beams.

Although the LISA requirement was not met using the Sagnac interferometer, a

significant amount of information was learned about the stability of the optical fiber and

noise sources that are associated with it. The stability results are within a factor of three

to five of what other groups at the AEI in Hannover [102] and JPL were able to achieve

using a more LISA-like experiment. In both these experiments, the noise appears to be

limited by the back-scatter of the beam into the fiber. Information about pointing and

back-scatter from the fiber will be used to design another experiment that will hopefully

be sensitive enough to reach the necessary requirements.


9.1 Hydroxide Bonding

There are a variety of techniques available to joint two pieces together. The bond

thickness, bonding material, and operating temperature are all factors that need to be

considered when choosing the bonding technique. Hydroxide bonding is a new bonding

technique that has already shown promising results. Although the this technique had

been previously studied in some detail, results in this dissertation expand further on the


It was found that the reason the glass was breaking in the first tests were due to the

wire used in the lever-arm apparatus, and not from the water in the solution weakening

the glass. From this, a better way to test the materials was found and the MCST was

developed. Using this device, existing glass to glass results were expanded on by varying

the amount of solution used, varying the concentration of the solution, and testing

bonds made with rough surfaces. It was found that there is no significant dependance in

shear strength of the bond based on the amount of solution used, whether a 1:4 or 1:6

concentration is used, or if a rough surface is used instead of a polished surface. Typical

shear breaking strengths were 2 to 5 MPa.

It was also found that hydroxide bonding could be used on Super Invar, which is due

to its ease of oxidizing. Super Invar to BK7 bonds proved to be successful, and varying the

amount of solution used showed no significant dependance on the shear breaking strength.

The average breaking strength was found to be ~5 MPa.

Although bonding SiC to SiC had already been done, the bonding process required

the materials to be heated to over 1000C and the bonds were not reliably made with

significant strength. A more reliable and simpler way to bond SiC was found, and

produced typical breaking strengths of 3 to 5 MPa. The surface profile of the materials,

the amount of solution used, and the dilution factor of the sodium silicate solution were

all varied and showed no dependance on the shear strength of the bonds. In addition, the

bonding mechanism was found for the SiC to SiC bonds using a sodium silicate solution.

The SiO on the SiC acts to bond the Si02 in the solution and produces a bond with

significant strength.

The versatility of the hydroxide bonding process demonstrates that it is a promising

bonding technique that can be used for many space-based missions. Although promising

results have been reported, further investigations are needed to determine how to achieve

the optimum bond strength in a repeatable manner, and under what conditions the bond

will survive. By further understanding this bonding technique, the limits to which it can

be used will be found.

9.2 Material Stability

Zerodur, SiC, Super Invar, PZT, and CFRP are all materials that have either been

used on space-based missions, or are promising materials to be used for future missions.

While some of the materials presented in this work have been studied already, all but

Zerodur have yet to be studied at the fm//Hz level.

Zerodur with both optically contacted and hydroxide bonded mirrors showed

almost identical noise levels and were well below the LISA pre-stabilization requirement.

Although the hydroxide bonds may produce a higher noise spectrum, it was not detectable

with our system.

Both SiC and CFRP were found to meet the LISA telescope requirement. While this

would appear to indicate that both materials could be used for the LISA telescope support

structure, CFRP needs to be studied in greater detail before a definitive conclusion can

be reached. This is due to the fact that the CFRP cavity outgassed and the overall shape

is not similar to what would appear on the LISA mission. Although it was seen that

the CFRP cavity initially outgassed, both the overall length contraction and outgassing

constituents need to be studied further. It would also be useful to study a cavity that has

a shape to what would be expected on the LISA mission, and not a cavity that was made

for maximum stability. While the outer cylinder on the CFRP cavity helps to provide the

inner cylinder from twisting and contracting unevenly, its large radius is not indicative of

what would be used on the LISA mission. From the results presented, it appears that SiC

would be a more viable material for use as a telescope support structure than CFRP.

-- Zerodur Optically Contacted Mirrors
14 Zerodur Hydroxide Bonded Mirrors
Super Invar
103 -- CFRP
I..... \ LISA Telescope Requirement
T 10
Z. 101

100 \ "",
/) 100

Z 10 1 '


10-5 10-4 10-3 10-2 10-1
Frequency (Hz)

Figure 9-1. Comparison of the noise spectrums of the Zerodur cavities with both optically
contacted and hydroxide bonded mirrors, SiC, CFRP, and Super Invar along
with the LISA telescope requirement.

Although Super Invar was shown to meet the LISA telescope requirement, it will

not be used on the LISA mission due to it magnetic properties. Despite this, its ease of

machinability, high strength, high stability, and low CTE make it a material that may

be suited for other missions that require ultra stable materials. In addition to finding

the stability of Super Invar, it was also shown that using a sodium silicate solution to

bond the mirrors appeared to produce no additional noise. This is significant since the

bonds produced using the potassium hydroxide solution when bonding the mirrors for the

Zerodur cavity are about a factor of four thinner than those when sodium silicate solution

are used. Thus, the hydroxide bonding technique is more than suitable for use on the

LISA mission and could be used to form complex optical structures from other materials

such as SiC, and not just Zerodur.

Despite the PZT cavity producing noise levels slightly above the LISA pre-stabilization

requirement, it was shown that through the use of an additional feedback control that is

fed to the PZT, the noise spectrum of the cavity can be pushed orders of magnitude below

the pre-stabilization requirement. It was also shown that a voltage of up to ~200 V could

be applied to the PZT without a noticeable increase in the noise spectrum. This is over an

order of magnitude larger than what would be needed for arm-locking. The combination

of its low noise levels with an applied voltage, its ability to be used for arm-locking, and

the fact that it survived vibrational conditions similar to what would be expected during

launch make the PZT cavity viable for use on the LISA mission.

Even though the PZT cavity is viable for the LISA mission, by using the same

PZT material and thickness and increasing the length of the Zerodur spacer, the noise

of the cavity should decrease. It should also be noted that the expected temperature

stability on the LISA optical bench is approximately an order of magnitude better than

the temperature stability of the tank the PZT cavity was tested in. Although a beatnote

drift is noticeable when a voltage is applied to the PZT, a slowly changing applied voltage

should not produce any significant drifts.

From the presented results, all materials except the PZT and CFRP cavity reached

a common noise floor above ~0.01 Hz (see Figure 9-1), which is well below the LISA

pre-stabilization requirement. This would imply a common noise limit of the measurement

system has been reached and further improvements will be needed in order to push the

noise floor down. At frequencies below ~1 mHz, the noise is limited by temperature

fluctuations in the lab that propagate through the thermal shields. Although the thermal

shields provide significant attenuation at high frequencies, at low enough frequencies the

thermal shields have almost no attenuation as can be seen in the beatnote frequency

oscillations. For materials with large CTE's, the beatnote oscillations will be larger,

which produces a higher noise spectrum at low frequencies. To determine the amount

of compactification or outgassing that occurs in materials such as SiC or CFRP over

year-long time scales, active temperature stabilization will be needed since the beatnote

fluctuations are most likely larger than the expected beatnote fluctuations due to the

compactification process. Although further improvements are needed to fully determine

the noise spectrum of these materials, promising results have been achieved already.

9.3 Cesium Locking

Both the modulation transfer spectroscopy and frequency modulation spectroscopy

experimental setups were tested, and the modulation transfer technique produced a lower

noise spectrum. In addition, the noise spectrum using the modulation transfer technique

were about an order of magnitude lower than using a free-running laser at low frequencies.

Several systematic noise sources in this setup were investigated and ruled out as the

cause of the drift in the beatnote. While the exact cause is unknown, it is suspected that

the beatnote drift is due to an expansion or contraction in the cavity from temperature

changes in the lab, or from laser intensity fluctuations that change the thickness of the

mirror coating which appear as a length change. Although promising initial results

were achieved, cesium stabilization thus far has not proven to be a viable use for long

term measurements due to the cesium cells consistently breaking after several months of

operation, and the high frequency noise associated from heating the cesium cell to over

2200C. Significant improvements in the cesium cell design and an implementation of a

fabry-perot cavity would allow the operating point of the cesium cell to be lowered and the

absorption increased, which would provide lower noise.

9.4 Telescope

A design for an on-axis telescope support structure using SiC was developed and

made. The design, known as the "quad-pod" for its four struts that separate the primary

and secondary, was built using an alignment jig that held the struts in place while they

could be hydroxide bonded to both the primary and secondary. The first prototype was

successfully built and transported to the University of Florida, where one of the struts

ini ,-I i. usly came undone. Despite this, the structure was cooled to approximately -65C

using a specially designed vacuum chamber. After warming up and a visual inspection,

it appeared the telescope was still in tact, but applying a small force to the secondary

caused the entire structure to collapse. An inspection of the bonds on the primary and

secondary showed that the telescope had twisted in the alignment jig during the bonding

process. While this may or may not have caused the strut to become un-bonded, a

redesign of the alignment jig was done to account for this. It was also decided that using

SiC "sister blocks" epoxied to the strut and either primary or secondary would provide

additional strength to the overall structure. In this manner, hydroxide bonding will

be used to provide the initial alignment, while the sister blocks will provide additional

strength. Bonding of the next prototype is currently underway along with setting up

the optics needed to test both the relative and absolute stability of the structure. At

least one optical cavity will be made in the center of the structure to determine the

relative stability, and an iodine stabilized laser is in the process of being set up in order to

determine the absolute length change of the telescope. These future experiments will be

used to determine the viability of the telescope design for the LISA mission.

9.5 Back-Link Fiber

A Sagnac interferometer was constructed in order to determine the differential phase

induced in an optical fiber with counter-propagating beams. It was found that a spurious

interferometer was in induced in the setup and by modulating the laser frequency, the

noise could be reduced by almost an order of magnitude. It was also found that over time

the power coming out of the fiber decreased, despite the input power being constant. This

is most likely due to pointing issues and could be remedied through active stabilization

of the interferometer mirrors using PZT mirror mounts. A calibration tone was injected

into the system using a fiber EOM to determine the relationship between the output

voltage on the balanced photodetector and the differential phase noise. From this, the

differential phase could be measured. At higher frequencies, it was found that the noise

was a little over an order of magnitude higher than the 10-6 cycles/V/Hz requirement.

This result is within a factor of three to five of what other groups at the JPL and the AEI

in Hannover were able to achieve using a significantly more complex experimental design.

Additionally, if the noise scales with the length of the fiber, then within some of the LISA

band, the fiber meets the requirements. It is expected that the results using the Sagnac

interferometer would improve if placed inside a vacuum chamber, but this was not able to

be done because once the pressure in the tank began to decrease, the alignment through

the fiber became misaligned so bad that no output signal could be detected. Using PZT

mirror mounts would prove useful in this case since active alignment could be done as

the pressure drops inside the tank. The information learned from the noise sources will

be used to design another experiment that will hopefully show that the differential phase

using counter-propagating beams in a fiber can be kept below 10-6 cycles/v/Hz.


Temperature fluctuation inside the vacuum tank will cause the cavity to expand and

contract based on its CTE. The purpose of the thermal shields will be to act as a passive

low pass filter to attenuate the temperature fluctuations in the lab. A typical temperature

noise spectrum for the outer wall of Tank 2 is shown in Figure A-i. Using the equations




(- 10-3

Z 10-4


10-4 10-3 10-2 10
Frequency (Hz)

Figure A-i. Noise spectrum of the outside tank temperature.

v L

6L = aL6T.,a


where 6v is the LSD of the frequency, v is the frequency of the laser, 6L is the LSD of the

length of the cavity, L is the length of the cavity, a is the CTE of the cavity, and JTcfa is

the LSD of the temperature of the cavity, the corresponding frequency fluctuations can be

found via

6v = vaTcav. T


To keep the frequency noise of the cavity due to thermal fluctuations low, either the

CTE of the cavity needs to be extremely low, the temperature stability in the tank needs

to be low, or a combination of both should be used. Although using feedback controls

and a heater or cooler could be used to actively control the temperature of the cavity, a

focus on passive thermal shielding will be presented here. Based on Figure A-i, several

orders of magnitude of temperature suppression will be needed in order for the frequency

fluctuations to be below 1000 Hz/vHz. The adopted set up to provide the thermal

insulations is shown in Figure A-2.

The thermal shields consist of 5 cylindrical lvr-is of an aluminized PET whose base

is a sheet of aluminum. Each successively smaller cylinder is thermally insulated from the

next using Macor spacers. In the center of the thermal shields is a long stand that elevates

the cavity to a height such that the laser can pass through the cavity. The aluminized

Table A-i. Diameters and heights of the thermal shield cylindrical shells.
Shell 1 Shell 2 Shell 3 Shell 4 Shell 5
diameter (m) 1.08 1.00 0.93 0.85 0.77
height (m) 0.91 0.86 0.81 0.76 0.71

PET has a low emissivity and acts to reflect the radiation from its surrounding. The

large heat capacity of the aluminum allows the thermal shield to absorb large amounts

of energy while changing its temperature very little. Although the aluminized PET has

a higher heat capacity, its thickness is only 0.12 mm thick and has a negligible mass

when compared to the mass of the aluminum. The low thermal conductivity of the Macor

spacers impedes the heat flow from one cylinder to the next. A summary of properties

for all three materials is listed is Table A-2. To determine the attenuation the thermal

Table A-2. Material properties of Macor, aluminized PET, and aluminum.
Material Thermal Conductivity Emissivity Density Heat Capacity
(W/mK) (kg/m3) (J/kg C)
Aluminized PET 0.03 1380 1190
Aluminum 250 ~0.15 2700 902
Macor 1.46 2520 790

~ -

Figure A-2. Picture of the built thermal shields (left) and a model representation of them
(right). Note the model is not drawn to any scale.

shields will provide, a thermal lumped model will be used. Only conduction and radiative

heat transfer will be considered since the convective process will be negligible due to the

high vacuum within the tank. A simplified lumped thermal model of the thermal shields

is shown in Figure A-3. The thermal resistance for both conduction and radiation can be

Stage 1 Stage 2 Stage 3 Stage 4 Stage 5 Stage 6
Rcond Rcond Rcond Rcond Rcond Rcond
Ttank A A Tcav

Rrad Rrad Rrad Rrad Rrad Rrad

Tank to S1 S1 to S2 S2 to S3 S3 to S4 S4 to S5 S5 to cav

Figure A-3. Lumped thermal model of the thermal shields.

found using







where Rcod is the conductive thermal resistance, I is the length of the element between

the two bodies at different temperature, K is the thermal conductivity of the element,

Acond is the cross section of the element, Rrad is the radiative resistance, a is Boltzmann's

constant (5.67x10-s W m-2K-4), Arad is the area of the body which is heated due to

radiation, T is the initial or mean temperature of the bodies, and 3 is a factor used when

determining the radiative resistance of cylindrical shells defined by

S 1 2 r (A5)
3 r2" (A 5)
C1 C2 r2

The total thermal resistance and capacitance of each stage is given by

R RradRcond
Rrad + Rcond
C = cpV, (A-6)

where c is the heat capacitance, p is the density, and V is the volume of the aluminum

base plates. The time constant of the low pass filter, 7, for each cylindrical shell is

7 = RC. (A-7)

The first order cutoff frequency for the filter is then

fcntoff = 2 (A-8)

These values for each stage of the thermal shields are given in Table A-3. The expected

Table A-3. Resistance, capacitance, time constant, and cutoff frequency for each stage of
the thermal shields.
Stage Rrad Reond R C 7 fc
(KW-1) (KW-1) (KW-1) J kg-1 (s) (Hz)
1 68.6 0.08 0.08 40325 3085 5x10-5
2 33.3 0.075 0.075 1626 122 1.30 x 10-3
3 33.3 0.075 0.075 1626 122 1.30 x 10-
4 33.3 0.075 0.075 1626 122 1.30 x 10-
5 33.3 0.075 0.075 1626 122 1.30 x 10-
6 1.91 0.40 0.33 3262 1083 1.47x10-4

global low pass filter can be found using the values in Table A-3 from

H(f) 1 (A-9)
H(f) f 27arif + t

where ,i are the values of 'r at stage i. Plotting IH(f)| as a function of frequency will result

in the amount of suppression the thermal shields will provide at that frequency. This

is shown in Figure A-4. The attenuation is not shown below 10-5 in order to gain an


rU 0.01

4 0.001

W 0.0001
Z 0.00001

1. x1060.00001 0.0001 0.001 0.01 0.1
Frequency (Hz)

Figure A-4. Transfer function of the PET thermal shields.

understanding of the transfer function at low frequencies.

To compare this effectiveness of this model, the transfer function of the thermal

shields can be used with the tank temperature data shown in Figure A-i to determine

what the expected beatnote fluctuations should be due to the temperature fluctuations

of the tank. Unfortunately, the expected frequency fluctuations do not match up well

with the measured beatnote fluctuations and are approximately an order of magnitude

larger than what is actually measured at low frequencies. This is most likely due to

the limitations of using a lumped thermal model and simplifications of the tank design

that were used. In the lumped thermal mode, it was assumed that there is only a large

aluminum block that sits on the tank bottom that supports the thermal shields. In fact,

there are three large posts that support the aluminum block which would provide a

significant amount of thermal noise suppression at low frequency due to their large mass.

A COMSOL model would provide a more accurate description of the transfer function and

would provide more accurate estimates of beatnote fluctuations.


When cleaning optical components for bonding, great care must be taken to ensure

the surfaces are as clean as possible. Using a clean room is optimal as this allows a smaller

chance of particles to get on the surfaces. Although a clean room is preferred, optical

and hydroxide bonding have been done using areas that have been cleaned thoroughly. A

general guideline is presented below to clean the optical surfaces before bonding.

1. Inspect surfaces to be bonded. If there are any large particulates on the surface,
wash off with methanol and clean room wipe.

2. Fold a microfiber cloth in half and place at the bottom of the sonic bath. Place
pieces to be bonded on top of the cloth so that the surfaces to be bonded are facing

3. Fill sonic bath with deionized (DI) water.

4. Run sonic bath for 15-30 minutes. The dirtier the optical components, the longer the
sonic bath should run.

5. After the sonic bath has stopped running, drain the DI water.

6. Wipe the components to be bonded using a clean room wipe wetted with methanol.
If possible, use compressed nitrogen to blow off any excess water, methanol, or dirt
still on the components. If any streaks or particulates can be seen on the surfaces to
be bonded, repeat this step until they have been removed.

If a sonic bath is not available, an alternate cleaning method can be used to clean the

surfaces to be bonded. This method is described below.

1. Wet a microfiber cloth with DI water.

2. Dab the damp part of the cloth into some powdered cerium oxide.

3. Rub on the surface to be bonded in a circular motion. Do this for approximately 5
minutes. The dirtier the surfaces, the longer this step should be done. You may need
to dab the cloth into the cerium oxide again (or many times) during this time to
ensure ample amount of cerium oxide is on the surface.

4. Rinse off excess cerium oxide using methanol.

5. Wet another part of the microfiber cloth with DI water again and dab with sodium
carbonate. You may need to dab the cloth into the sodium carbonate many times
during this step. Make sure to use plenty of sodium carbonate to ensure any cerium
oxide residue is cleaned off.

6. Rub in a circular motion for approximately 5 minutes.

7. Rinse components thoroughly using methanol. If a film can be seen on the surface,
then an optical wipe wetted with methanol can be used to removed the residue.

Another possible method to clean the bonding surfaces is to initially clean the surfaces

using cerium oxide and then place the pieces in a sonic bath as described above.

Although using the cerium oxide will help to clean and polish the surfaces, a "pitting

effect has been noticed when this is done. To determine the effects of using cerium oxide,

a WYKO profilometer was used to determine the surface profile on several spots of a

SiC tile. A typical surface profile before using cerium oxide is shown in Figure B-1. The

SiC tile was then polished and cleaned as described above and the surface profile was

determined using the WYKO once again. The resulting surface profile is shown in Figure

B-2. Although the rms roughness of the tile seems to have become smoother, there are

noticeable spots where the surface has deep pits. The cause of this effect is unknown.

g reB-. Tal n IDf pC sa- 7o
Figur B-f. Typical surfae p lrium

Iz 1,4 lum 0
Rt 17!1 .u 200


Processed Options:


,0'u I I O--tS ,r0 4y-I

1 ig t5r 2BO- 2T5p sW

Figure B-i. Typical surface profile of a SiC surface before cleaning with cerium oxide.

Surface Statistics:
Ra 11 80nm
Rq- 2270nm
Rz 1.02um
Rt 201 um

Set-up Parameters:
Size 736 X480
Sampling 40977nm

Processed Options:
Terms Removed



0 50 100 150 200 250 3i0

Figure B-2. Typical surface profile of a SiC surface after cleaning with cerium oxide.



The linear spectral densities (LSD) computed for the spectrums in (C! lpters 5-8 were

done using a Matlab script. The script computes the power spectral density (PSD) using

Welch's method of overlapping segmented averages of modified periodograms and is based

on techniques described in "Spectrum and spectral density estimation by the Discrete

Fourier Ti ,-I,-!. ii G. Heinzel, et. al. [103].

A time series with a known sampling frequency (fs) is loaded into Matlab. The time

series is then divided into sections of N data points as defined by the user. It should be

noted that the bin spacing will be equal to the lowest frequency and is equal to fs/N.

A linear regression is fit to the time segments and subtracted out. This residual is then

multiplied by a Hanning window (not to be confused with a Hamming window) and a

discrete Fourier transform is performed on the windowed residual. The power spectral

density for the ith segment is then calculated and an average is taken over all segments

to get the power spectral density for the entire set. The linear spectral density is then

computed by taking the square root of the power spectral density.

The time series measurements of the frequency counter can be converted to a length

spectrum using
6V 61

where 6v is the linear spectral density of the measured beatnote, v is the laser frequency,

61 is the linear spectral density of the length of the cavity, and 1 is the length of the cavity.


[1] B. Tapley, S. Bettadpur, M. Watkins, and Ch. Reigber. "The Gravity Recovery and
('Cin ii.. Experiment: Mission Overview and Early Results". American G(.e i/' ;,'. l
Union, (2004).

[2] J. Davis, P. Elosequi, J. Mitrovica, and M. Tamisiea. "C('lii il.-driven deformation of
the solid Earth from GRACE and GPS". G(e .// Res. Let., 31(L24605).

[3] G. Ramilliena, F. Frapparta, A. Cazenavea, and A Gountner. "Time variations
of land water storage from an inversion of 2 years of GRACE ..i 1- Earth and
Planetary Science Letters, 2 :-. 2 ;-301, (2005).

[4] M. Dehne, F. Guzmin Cervantes, B. Sheard, G. Heinzel, and K. Danzmann. "Laser
interferometer for spaceborne mapping of the Earth's gravity field". Journal of
Ph,-,. Conference Series, 154, (2009).

[5] J. Taylor and J. Weisberg. "Further experimental tests of relatavistic gravity using
the binary pulsar PSR 1913 + 16". A;/',../,,;/, J., 345:434-450, (1989).

[6] M. Valtonen et al. "A massive binary black-hole system in OJ 287 and a test of
general 1i .111,v Nature letters, 452:851-853, (2008).

[7] NASA. "LISA Laser Interferometer Space Ai.l, ii I for more information, go to
http://www.lisa.nasa.gov, accessed 5 May 2010, Copyright 2010.

[8] M. Ando et al. "DECIGO Pathfinder". Journal of Ph;-. Conference Series, 120,

[9] LISA Study Team. "Laser Interferometer Space Antenna for the detection and
observation of gravitational v- i., Pre-phase A report, July 1998.

[10] S. Hughes. "LISA sources and science". arXiv:0711.0188vl [gr-qc], 1 Nov 2007.

[11] B. Allen. relativisticc gravitation and gravitational radiation". In edited by
J. A. Marck and J. P. Lasota, editors, Proceeding of the Les Houches School of
Ph;,-. page 373. Cambridge University Press, 1997.

[12] A. Kosowsky, A. Mack, and T. Kahniashvili. Phys. Rev. D, 55:7368, 1997.

[13] C. Hogan. "Gravitational Wave Sources from New Phy- In edited by
S. M. Merkowitz and J. C. Livas, editors, Laser Interferometer Space Antenna,
page 397. AIP Conference Proceedings 823, Mellville, New York, 2006.

[14] M. Benacquista and Holley-Bockelmann. A,hI-li,;- J., 645:589, 2006.

[15] A. Stroeer and A. Vecchio. "the LISA verification binaries". Class. Quant. Gray.,
23:S809-S818, 2006.

[16] M. Haehnelt. page 45. AIP Conference Proceedings 456, 1998.

[17] C. Hopman. "A-l ii. ,1 -iics of extreme mass ratio inspiral sources". In edited by
S. M. Merkowitz and J. C. Livas, editors, Laser Interferometer Space Antenna, page
397. AIP Conference Proceedings 823, Mellville, New York, 2006.

[18] Laser Interferometer Space Antenna (LISA) Requirements Flowdown Guide,
February 2007.

[19] DRS ITAT. "LISA DRS Acceleration Noise B- 1i I LISA Project internal report,

[20] A. Cavalleri, G. Ciani, R. Dolesi, A. Heptonstall, M. Hueller, D. Nicolodi, S. Rowan,
D. Tombolato, S. Vitale, P. Wass, and W. Weber. "A new torsion pendulum for
testing the limits of free-fall for LISA test 1i .--. Class. Quant. Gray., 26, (2009).

[21] S. Anza. "the LTP experiment on the lisa pathfinder iii--,i"o Class. Quant. Gray.,
22:S125-S138, 2005.

[22] J. Armstrong, F. Estabrook, and M. Tinto. "Time d., i iii. I, I1. .1 Ii. l Class.
Quant. Gray., 20:S283-S289, (2003).

[23] M. Sallusti, P. Gath, D. Weise, M. Berger, and H Schulte. "LISA system design
hi;,1,!i!,i- Class. Quant. Grav., 26, (2009).

[24] S. Turyshev, M. Shao, and K. Nordtvedt Jr. "Science, Technology, and Mission
Design for the Laser Astrometric Test of R. !1 A,.1iiy A,-'. /*'i;:.. and Space Science
I1.:I,,; i 349:473-543.

[25] S. Turyshev, M. Shao, and K. Nordtvedt Jr. "Experimental Design for the LATOR
I !:--i~.! arXiv:gr-qc/0410044vl, 8 Oct 2004.

[26] J. Yu. "Laser Astrometric Test of relativity (LATOR) 1iii--. i- In Edited
by M. Shao, editor, JPL Technical Memorandum, volume 2200. SPIE, 1994.

[27] TIE-31: Mechanical and thermal properties of optical glass. July 2004, Version 2.

[28] TIE-37: Thermal expansion of ZERODUR, August 2006.

[29] J. Baer and W. Lotz. "Figure testing of 300 mm ZERODUR mirrors at cryogenic
temp. i ii ii. In C,,'i .i. ,..:. Optical Sii-, m,- and Instruments IX, SPIE Proc.,
volume 4822, (2002).

[30] D. de C'i inl>ure, R. Laine, K. van Katwijk, and P. Kletzkine. X\I\!'s X-Ray
Telescopes". XMM project bulletin 100, December 1999.

[31] W. Sokolowski, S. Jacobs, M. Lartc, I. O'Donnell, and C. Hsieh. "Dimensional
Stability of High Purity Invar 36". Mechanical S,-/. i'- Engineering and Research
Division, Jet Propulsions Laboratory.

[32] S. Jacobs, S. Johnston, and D. Schwab. "Dimensional instability of Ii. ii- Applied
Optics, letter to the editor, (1984).

[33] M. van V-.--1. The Basic Angle Monitoring system: picometer -/,l.:/i:; with Silicon
Carbide optics. PhD thesis, Technische Universiteit Eindhoven, 2007.

APPLICATIONS. Department of Mechanical Engineering, The University of Texas
at Austin, Austin, TX 78712.

[35] POCO Grahpite. "SUPERSiC material proper- i see http://www.poco. com/
tabid/124/Default. aspx for details, accessed 5 May 2010, Copyright 2010.

[36] Saint Gobain. "Hexoloy SA Silicon Carbide Technical D ,I i see http://www.
hexoloy. cor/data-sheets for details, accessed 5 May 2010, Copyright 2003.

[37] CoorsTek. "Technical C( i 1,i!!1 see http://www. coorstek. com/materials/
ceramics.asp for details, accessed 5 MA i- 2010, Copyright 2009.

[38] S. Suyama, T. Kameda, and Y. Itoh. "Development of high-strength
reaction-sintered silicon carbide". Diamond and Related Materials, 12:1201-1204,

[39] O Kubaschewski and C Alcock. M. l.lli,,l:. Thermochemistry. Pergamon Press,
New York, 5th edition edition, 1979.

[40] S. Black. Getting to Know "Black Aluminum". Technical report, 2008.

[41] Electroceramics. Technical report, Morgan electroceramics.

[42] C. Poizat and A. Benjeddou. "On analytical and finite element modelling
of piezoelectric extension and shear bimorphs". Computers and Structures,
84:1426-1437, (2006).

[43] Piezo Technologies. "K-180 Lead Zirconate Titanate Piezoelectric Ceramic with High
Shear Coupling Coefficient, Low Permittivity, and High Temperature Stability". for
more information, go to http://www.piezotechnologies .com/kl80.htm,accessed 5
M i-v 2010, Copyright 2010.

[44] V. Greco, F. Marchesini, and G. Molesini. "Optical contact and van der Waals
interactions: the role of the surface topography in determining the bonding strength
of thick glass pl! ii Journal of Optics A: Pure and Aplied Optics, 3:85-88, (2000).

[45] J.-J. Ferme. "Optical Contactiing In D. Rimmer G. Roland, R. Geyl and L. W\ ii
editors, SPIE Proc., volume 5252 of Optical Fabrication, Testing, and Metr '/. 'i;,

[46] J. van Elp, P. Giesen, and J. van der Velde. "Anodic bonding using the low
expansion glass ceramic Zerodur". American Vacuum S .. ;/, (2005).

[47] M. Powers. "Hermetic Glass-to-Aluminum Seals for RF Packaging Applications: A
University-Industry Design Project Collaboration". In editors: John Stephens and
K. Scott Weil, editors, Brazing and Soldering, Proceedings of the 3rd International
Brazing and Soldering Conference, (2006).

[48] A. Belohlav. Understanding Brazing Fundamentals. Technical report, The American
Welder, 2000.

[49] D. Bath, D. Spain, E. Ness, S. Williams, and M. Bougoin. "Evaluation of segmented
and brazed mirror assemblies". Technical report, collaboration between CoorsTek
and Boostek companies, technical note.

[50] S. Gibb. Two Part Epoxy Adhesives. Technical report, 2006. A Technical Memo
Prepared for Opti 521 Introduction to Opto-Mechanical Engineering.

[51] J. Daly. "Structural Adhesives for Optical Bonding; SPIE short course notes,

[52] Master Bond Inc. \! Ilter Bond Polymer System EP21TCHT-1 Product
Description". For more information see http://www.masterbond. com/tds/
ep21tcht-1.html, accessed 5 M ,i-, 2010 Copyright 2009.

[53] E. Elliffe, J. Bogenstahl, A. Deshpande, J. Hough, C. Killow, S. Reid, D. Robertson,
S. Rowan, H. Ward, and G. Cagnoli. "Hydroxide-Catalysis Bonding for Stable
Optical Systems for Sp ... Class. Quant. Gray., 22:S257-S267, (2005).

[54] D. Gwo. "Ultraprecision Bonding for Cryogenic Fused-Silica Optics". In SPIE Proc.,
volume 136 of 3435, (1998).

[55] A. van V -.- 1, D. van den Endeb, J. Bogenstahl, S. Rowanc, W. Cunninghamc,
G. Gubbels, and H. Nijmeijer. "Hydroxide catalysis bonding of silicon carbide".
Journal of the European Ceramic S .. .. I/; 28:303-310, (2008).

[56] G. S.-ii i i "C'i, i,.:-ry in Two Dimensions: Sn i,.i. `". Cornell University Press,
Ithaca, N.Y., 1981.

[57] A. Preston and G. Mueller. "Bonding SiC to SiC using a sodium silicate solution".
In preparation, (2010).

[58] B. Lauke K. Schneider and W. Beckert. "Compression Shear Test (CST) A
Convenient Apparatus for the Estimation of Apparent Shear Strength of Composite
Mat i I.- Applied Composite Materials, 8:43-62, (2001).

[59] A. Preston, B. Balaban, and G. Mueller. "Hydroxide-Bonding Strength
Measurements for Space-Based I--., ,i- Int. J. Appl. Ceram. Technol.,
5(4):365-372, (2008).

[60] A. van V.- 1, D. van den Ende, J. Bogenstahl, S. Rowan, W. Cunningham,
G. Gubbels, and H. Nijmeijer. "Hydroxide Catalysis Bonding of Silicon Carbide".
Journal of the European Ceramic S .. .. ;I, 28:303-310, (2007).

[61] S. Rowan, J. Hough, and E. Elliffe. "Silicon Carbide Bondiin Please contact Mr.
D. Whiteford for further information: D.Whiteford@admin.gla.ac.uk.

[62] A. Preston, B. Balaban, G. Boothe, and G. Mueller. "Stable Materials and Bonding
Techniques for Space-Based Optical Systems". In NASA Science and T. I,,,./.iI;
Conference, page Session C1 P3, (2007).

[63] J. Berthold III. Dimensional Still'i of Low E/..';;,... ;; Materials Time Dependent
Cli.i,., in Optical I,.., i f.i and Phase S/.li- on R.fl7. i/.. from M.ll.7.n ,
Dielectrics. PhD thesis, University of Arizona, 1976.

[64] G. Wallis and D. Pomerantz. "Field Assisted Glass-Metal Sealiin Journal of
Applied Ph.,-:.- 40:3946-3949, (1969).

[65] J. Berthold III, S. Jacobs, and M. Norton. "Dimensional Stability of Fused Silica,
Invar, and Several Ultra-Low Thermal Expansion Mat( 1 i !- Metrologia, 13:9-16,

[66] D. Gwo. "Hydroxide-Catalyzed Bondiin United States Patent No. U.S. 6 548 176
Bl, 2003.

[67] Personal communications with Suvrath Mahadeven at the University of Florida
department of Astronomy.

[68] A. Preston. "Stable Materials for LISA". Talk presented at the Frontiers in Optics
conference, October 2009.

[69] B. Young, F. Cruz, W. Itano, and J. Bergquist. "Visible Lasers with Subhertz
Linewidths". Ph,;,-.. .il Review Letters, 82(3799), (1999).

[70] H. Kogelnik and T. Li. "Laser Beams and Resonators". Applied Optics, 5(10),

[71] A. Seigman. Lasers. Stanford Science Books, 1986.

TIME DELAY INTERFEROMETRY. PhD thesis, University of Florida, 2006.

[73] Personal communications with the Space and Gravity Research group at the
University of Birmingham.

[74] J. Livas, J. Thorpe, K. Numata, S. Mitryk, G. Mueller, and V. Wand.
"Fre -ii i-,-tunable pre-stabilized lasers for LISA via sideband locking; Class.
Quant. Gray., 26, (2009).

[75] E. Black. "An introduction to PoundDreverHall laser frequency stabilization".
American Journal of Ph;,.- 69(1):79-87, (2001).

[76] D. Shaddock et al. "LISA frequency control white paper". Technical report, LISA
project technical note LISA-JPL-TN-823, 2009.

[77] S. Webster, M. Oxborrow, and P. Gill. "Vibration insensitive optical cavity". Phys.
Rev. A, 75:011801(R), (2007).

[78] A. Preston, R. Cruz, J. Thorpe, G. Mueller, G. Boothe, R. Delgadillo, and
S. Guntaka. "Dimensional Stability of Hexoloy SA Silicon Carbide and Zerodur
Materials for the LISA Mission.

[79] A. Preston, R. Cruz, I. Thorpe, G. Mueller, and R. Delgadillo. "Dimensional
Stability of Hexoloy SA Silicon Carbide and Zerodur Glass Using
Hydroxide-Catalysis Bonding for Optical Systems in Sp .. In Joseph Antebi
Edited by Eli Atad-Ettedgui and Dietrich Lemke, editors, Proceedings of the SPIE:
Optomechanical Technologies for Astronr'ii., volume 6273, (2006).

[80] L. Conti, M. De Rosa, and F. Marin. "High-spectral-purity laser system for the
AURIGA detector optical readout". Journal of the Optical S... : Ii of America B,
20(3):462-468, (2003).

[81] A. Preston and G. Mueller. "PZT-actuated tunable cavity for use in the LISA
1nii--' i In preparation, (2010).

[82] A. Stumner. "1-kV Piezo Amplifier Keeps Cost, Noise L( .
published on http://electronicdesign. com/article/components/
l-kv-piezo-amplifier-keeps-cost-noise-low3789.aspx, August 2001. accessed
25 April, 2010.

[83] S. Lin. "Study on the equivalent circuit and coupled vibration for the longitudinally
polarized piezoelectric ceramic hollow cylinder". Journal of Sound and Vibration,
275:859-875, (2004).

[84] B. C'!. ing and M. Reese. "Stress relaxation and estimation of activation volume in a
commercial hard pzt piezoelectric c(i ,ii:. Bull. Mater. Sci., 24(2):165-167, April

[85] New Focus Corporation. 1 I Spectroscopy With Tunable Diode Lasers, Application
note 7". see http://www.newfocus.com/products/documents/literature/
apnote7.pdf for more details, accessed 5 May 2010, Copyright 2001.

[86] O. Orlov and V. Ustyugov. "Molecular cesium frequency reference for frequency
stabilization of a 1.06 pm nd:yag 1 i Sov. Tech. Phys. Lett., 12:120-121, 1986.

[87] P. Fritschel and R. Weiss. "Frequency match of the nd:yag laser at 1.064 pm with a
line in co2". Appl. Opt., 31:1910-1912, 1992.

[88] R. Benedict, D. Drummond, and L. Schlie. "Absorption spectra of the Cs2
molecule". J. Ch, ii Phys., 66:4600-4607, 1977.

[89] A. Mak, O. Muravitsky, O. Orlov, and V. U- i gov. \, v-- laser for interferometry
with long-term frequency stabilization at 1.06 pm onto molecular cesium st i ,l i Il .
In Edited by Z. Jaroszewicz and M Pluta, editors, in Interferometry '89, volume
1121, pages 478-484. SPIE, 1989.

[90] E. Inbar, V. Mahal, and A. Arie. "Frequency stabilization of nd:yag lasers to 133Css
sub-doppler lines near 1064 nm". J. Opt. Soc. Am. B, 13(7):4600-4607, July 1996.

[91] R. Schilling, P. Bender, J. Livas, T. Pedersen, A. Martino, P. McNamara,
O. Jennrich, A. Preston, and G. Mueller. Notes taken from the LISA telescope
subgroup teleconferences. internal document, February 2006.

[92] D. Robertson and J. Hough. Interferometry for LISA. Class. Quantum Grav.,
13:A271-A277, 1996.

[93] et. al. R Fleddermann. \!. i-n cement of the non-reciprocal phase noise of a
polarization maintaining single-mode optical fiber". Journal of Ph; -. Conference
Series, 154, (2009).

[94] K. McKenzie. "Fiber noise measurements and photoreciever d, v, 1'. pi, ill Pre-
sented at the 7th International LISA Symposium, (2009).

[95] R. Bergh, H. Lefevre, and H. Shaw. "An Overview of Fiber-Optic Gyroscopes".
Journal of Lightwave T. ,,.,....,i; LT-2(2), (1984).

[96] E. Post. "Sagnac effect". Rev. Mod. Phys., 39:475-494, (1967).

[97] H. Lefevre and H. Arditty. "Theoretical basis of Sagnac effect in fiber gyroscopes".
Fiber-Optic Rotation Sensors and Related Technologies, 32(8), (1982).

[98] J. Noda and et. al. "Polarization-maintaining fibers and their applications". J.
Lightwave Technol., 4(8):1071, (1986).

[99] S. Barke. "Phase C('I, i. 'terization of EOM Sidebands well within the Mission
Requirements of the Laser Interferometer Space Ai., 'I i i Technical Report
Presentation given at the 7th annual LISA symposium, AEI in Hannover, 2009.

[100] A. Preston. "Back-link Fiber Stability". Talk presented at the 7th Inernational LISA
Symposium, 2008.

[101] P. Beyersdorf, R. Byer, and M. F, i r. "Results from the Stanford 10 m Sagnac
ini, 1 ... 1, 1 Claasical and Quantum Gi,; .:/I, 19(7):1585-1589, (2002).

[102] R. Fleddermann et al. \ i -nrement of the non-reciprocal phase noise of a
polarization maintaining single-mode optical fiber". In Journal of Ph;i-. Confer-
ence Series, volume 154, (2009).

[103] G. Heinzel, A. Rudiger, and R. Schilling. "Spectrum and spectral density estimation
by the Discrete Fourier Transform (DFT), including a comprehensive list of window
functions and some new flat-top wind .--- Technical report, Max Plank Institute
for Graviational Physics, Hannover Germany, February 15 2002.


Alix Preston was born in Bartlesville, Oklahoma, but moved to Fort Collins, Colorado

when he was five. He spent his childhood and adolescence fascinated with math and

science. He took several Advanced Placement courses in High School and graduated as

class Valedictorian. Alix attended the Colorado School of Mines in Golden, Colorado and

graduated in the top 2 1' of his class with a Bachelor of Science in engineering physics

in addition to taking several graduate level courses in both math and physics. In the Fall

of 2004 Alix began his studies at the University of Florida and soon began working with

Dr. Guido Mueller on determining the dimensional stability of materials and bonding

techniques for the Laser Interferometer Space Antenna (LISA). This work carried into

researching several other aspects of LISA, and is the bulk of his dissertation.








Muchofthisworkwouldnothavebeenmadepossibleifitwerenotforthehelpofmanygraduateandundergraduatestudents,faculty,andsta.IwouldliketothankIraThorpe,RachelCruz,VinzenzVand,andJosepSanjuanfortheirhelpandthoughtfuldiscussionsthatwereinstrumentalinunderstandingthenuancesofmyresearch.IwouldalsoliketothankGabrielBoothe,AaronSpector,BenjaminBalaban,DarsaDonelon,KendallAckley,andScottRagerfortheirdedicationandpersistencetogettingthejobdone.Aspecialthanksisdueforthephysicsmachineshop,especiallyMarcLinkandBillMalphurs,whospentmanyhoursonthecountlessprojectsIneeded.Lastly,Iwouldliketothankmyadvisor,Dr.GuidoMueller,whoputupwithme,guidedme,andsupportedmeinmyresearch. 4


page ACKNOWLEDGMENTS ................................. 4 LISTOFTABLES ..................................... 9 LISTOFFIGURES .................................... 10 KEYTOABBREVIATIONS ............................... 17 KEYTOSYMBOLS .................................... 19 ABSTRACT ........................................ 20 CHAPTER 1INTRODUCTION .................................. 22 1.1Space-BasedMissions .............................. 23 1.2GRACE ..................................... 23 1.3GRACEFollow-On ............................... 25 1.4LISA ....................................... 26 1.4.1Introduction ............................... 26 1.4.2Sources .................................. 27 .............. 28 .......................... 28 ....................... 29 ........... 29 1.4.3GeneralMeasurementOverview .................... 30 1.4.4TheDisturbanceReductionSystem(DRS) .............. 31 1.4.5TheInterferometricMeasurementSystem(IMS) ........... 32 ......................... 33 ....................... 34 1.5LATOR ..................................... 34 1.5.1MissionDesign .............................. 35 1.5.2Spacecraftorbits ............................. 35 1.5.3Opticaldesign .............................. 36 1.5.4Interferometry .............................. 37 2MATERIALS ..................................... 41 2.1Zerodur ..................................... 41 2.1.1ProductionProcess ........................... 42 2.1.2CTE ................................... 43 2.1.3ThermalCycling ............................. 44 2.1.4LongTermStability ........................... 44 2.1.5CryogenicBehavior ........................... 44 5


....................................... 45 2.2.1Invar36 ................................. 45 2.2.2Invar42 ................................. 47 2.2.3Invar32-5 ................................ 47 2.3SiliconCarbide ................................. 48 2.3.1StructureofSiC ............................. 49 2.3.2FabricationProcess ........................... 49 ........................ 49 .......................... 50 ..................... 51 .............. 52 2.3.3Properties ................................ 52 ..................... 53 ..................... 54 ................. 54 .......... 56 2.4CarbonFiberReinforcedPolymers ...................... 56 2.4.1FabricationProcess ........................... 57 ......................... 57 ........................ 58 2.4.2Properties ................................ 60 2.5LeadZirconateTitanate ............................ 60 2.5.1ThePiezoelectricEect ......................... 61 2.5.2PiezoelectricMaterials ......................... 61 3BONDINGTECHNIQUES ............................. 65 3.1OpticalContacting ............................... 65 3.1.1Method1 ................................. 66 3.1.2Method2(SolutionAssistedOpticalContacting) .......... 66 3.1.3Method3 ................................. 67 3.2AnodicBonding ................................. 67 3.3Brazing ..................................... 71 3.4Epoxies ..................................... 74 4HYDROXIDE-CATALYSISBONDING ....................... 78 4.1Introduction ................................... 78 4.2BondingMechanisms .............................. 79 4.2.1GlasstoGlass .............................. 79 4.2.2GlasstoMetalsandMetaltoMetal .................. 80 4.2.3SiCtoSiC ................................ 80 .............. 81 ................. 85 4.3BondingMethod ................................ 87 4.4StrengthMeasurements ............................. 88 6


........................... 89 4.4.2SiC-BK7Bonding ............................ 94 ........................ 95 ......................... 96 4.4.3SuperInvar-BK7 ............................ 97 4.4.4SiC-SiC ................................. 98 .......................... 98 ....................... 99 ...................... 99 4.5OtherBondingResults ............................. 102 5RELATIVESTABILITYMEASUREMENTS ................... 103 5.1OpticalCavities ................................. 103 5.1.1StabilityofOpticalResonators ..................... 105 5.1.2MisalignmentAnalysis ......................... 107 5.1.3CavityConstruction ........................... 108 ........................ 108 ...................... 110 ........................... 110 ......................... 112 ..................... 115 5.2Pound-Drever-HallLaserFrequencyStabilization .............. 116 5.2.1ConceptualModel ............................ 117 5.2.2QuantitativeModel ........................... 118 5.3RelativeStabilityMeasurements ........................ 121 5.3.1Electro-OpticalSetup .......................... 122 5.3.2Zerodur-Zerodur ............................. 123 ................. 124 .................. 127 5.3.3Zerodur-SiC ............................... 129 5.3.4Zerodur-SuperInvar ........................... 132 5.3.5Zerodur-PZTCavity .......................... 136 5.3.6Zerodur-CFRP ............................. 146 6ABSOLUTESTABILITYMEASUREMENTS ................... 151 6.1SaturationSpectroscopy ............................ 151 6.2ExperimentalSetup ............................... 153 6.2.1LaserStabilizationUsingModulationTransferSpectroscopy .... 154 6.2.2LaserStabilizationUsingFrequencyModulationSpectroscopy ... 160 6.3Results ...................................... 160 6.3.1RFAMNoise ............................... 161 6.3.2AOMNoise ............................... 163 6.3.3ChangingChopperFrequency ..................... 163 6.3.4CesiumCellTemperatureChanges ................... 163 7


................... 166 7.1Introduction ................................... 167 7.2TelescopeDesign ................................ 168 7.3TelescopeFabrication .............................. 172 7.4VacuumTankDesign .............................. 182 7.5ExperimentalSetup ............................... 185 7.6Results ...................................... 187 8LISABACK-LINKFIBERPHASESTABILITY ................. 192 8.1SagnacEect .................................. 193 8.2Polarization-MaintainingOpticalFibers ................... 195 8.3ExperimentalSetup ............................... 197 8.3.1QualitativeDescription ......................... 197 8.3.2QuantitativeDescription ........................ 200 8.3.3VoltagetoPhaseCalibration ...................... 203 8.4StabilityResults ................................ 205 9CONCLUSION .................................... 212 9.1HydroxideBonding ............................... 212 9.2MaterialStability ................................ 213 9.3CesiumLocking ................................. 216 9.4Telescope .................................... 216 9.5Back-LinkFiber ................................. 217 APPENDIX ATHERMALSHIELDDESIGN ........................... 219 BCLEANINGCOMPONENTSFORBONDING .................. 225 CCALCULATIONOFTHELINEARSPECTRALDENSITY ........... 228 REFERENCES ....................................... 229 BIOGRAPHICALSKETCH ................................ 237 8


Table page 2-1SelectedmaterialpropertiesandtheirvaluesforZerodur ............. 42 2-2ZerodurclassvalueandcorrespondingCTE. .................... 43 2-3CompositionalvaluesofInvar36,42,and32-5. .................. 48 2-4Comparisonofselectedpropertiesfor3Invaralloys ................ 48 2-5ComparisonofdensityandporositiesforselectedSiCmaterials. ......... 54 2-6ComparisonofexuralandtensilestrengthforselectedSiCmaterials. ...... 55 2-7ComparisonofthermalconductivityforselectedSiCmaterials. .......... 56 2-8Selectedpropertiesandtheirvaluesforthek-180piezoelectricmaterial. ..... 64 3-1Comparisonofpropertiesforselecttwo-partepoxies. ............... 76 5-1ParametersofthethreeZerodurcavitiesconstructedbyeitheropticallycontactingorhydroxidebondingthemirrors. .......................... 109 5-2ParametersoftheSuperInvarcavitywithhydroxidebondedmirrors. ...... 110 5-3ParametersoftheSiCcavitywithopticallycontactedmirrors. .......... 112 5-4ParametersfortheCFRPcavity. .......................... 114 5-5ParametersforthePZTcavity. ........................... 116 A-1Diametersandheightsofthethermalshieldcylindricalshells. .......... 220 A-2MaterialpropertiesofMacor,aluminizedPET,andaluminum. .......... 220 A-3Resistance,capacitance,timeconstant,andcutofrequencyforeachstageofthethermalshields. .................................. 222 9


Figure page 1-1RepresentationoftheGRACEsatellitesinorbitabovetheEarth.ImagecourtesyofNASA. ....................................... 24 1-2RepresentationofhowtheGRACEspacecraftmoveastheypassoveramassanomaly.In(a),spacecraft1feelsanaccelerationduetotheanomaly.Afterspacecraft1passestheanomaly,itfeelsanaccelerationintheoppositedirectionofspacecraft2asshownin(b).Afterspacecraft2passestheanomaly,itfeelsanaccelerationintheoppositedirectionasdepictedin(c). ............ 24 1-3RepresentationoftheLISAorbit.ImagecourtesyofNASA. ........... 27 1-4Simpliedviewoftheassembly(a)andtheopticalbench(b). .......... 33 1-5CongurationoftheLATORorbits,spacecraft,andinterferometerontheISS.ImagecourtesyofNASA. .............................. 35 1-6Heterodyneinterferometryononespacecraftwithphase-lockedlocaloscillator. 37 1-7Fiberlinkedheterodyneinterferometryandbermetrologysystem(left)andberlinkedheterodyneinterferometryontwospacecraft(right). ......... 38 2-1Zerodurproductionprocess. ............................. 42 2-2Zerodurannealingprocess. .............................. 43 2-3ProcessusedtoproducesinteredSiC. ........................ 50 2-4Processusedtoproducereaction-bondedSiC. ................... 52 2-5Diagramofhow3-and4-pointloadingtestsaredone. .............. 55 2-6a)CubiclatticeabovetheCuriepointb)TetragonallatticebelowtheCuriepointafterthepolingprocesshasbeenapplied. .................. 62 2-7ElectricdipolemomentsfortheWeissdomains(a)beforepolarization(b)duringpolarizationand(c)afterpolarization. ....................... 62 2-8Thepiezoelectriceectofacylinderfor(a)noexternalaction(b)anappliedcompressiveforce(c)anappliedtensileforce(d)anappliedvoltage(e)anappliedvoltageofoppositepolarity.Forclarityonly1dipoleisshownin(a). ...... 63 2-9Axesdesignationanddirectionsofdeformation.Directions4,5,and6aretheshearaboutaxes1,2,and3,respectively. ..................... 64 3-1Representationofhowanodicbondingworks. ................... 68 3-2ProperspacingforbrazingmaterialsofdieringCTE. .............. 73 10


....... 75 4-1SpectrumofanunbondedCoorsTekSiCsample.ThepeaksfromlefttorightaretheAugercarbonpeak,Augeroxygenpeak,oxygen1speak,carbon1speak,andthesilicon2sand2ppeaks. ........................... 83 4-2Kineticenergydistributionobtainedduetotheinelasticscatteringofelectrons.Thepeakcorrespondstothephotoelectronsemittednearthesurface,whilethesignalatlowerkineticenergiesresultsfromthescatteringoftheelectronsdeeperwithinthesamplesurface. .............................. 84 4-3O1sorbitalbothbefore(bluecurve)andafter(redcurve)bonding.ThedashedanddottedlinesrepresentthepeaktvalueswhilethesolidlinesaretherawdatatakenfromtheXPS.Theunbondedpeakisat532.61eVandthebondedpeakisat532.96ev. ................................. 86 4-4Si2p3orbitalbothbefore(bluecurve)andafter(redcurve)bonding.ThedashedanddottedlinesrepresentthepeaktvalueswhilethesolidlinesaretherawdatatakenfromtheXPS.Theunbondedpeaksareat102.87eVand100.60eV.Thebondedpeaksareat100.42eV,103.82eV,and102.21eV. ....... 87 4-5C1sorbitalbothbefore(bluecurve)andafter(redcurve)bonding.ThedashedanddottedlinesrepresentthepeaktvalueswhilethesolidlinesaretherawdatatakenfromtheXPS.Theunbondedpeaksareat282.79eV,and285.12eV.Thebondedpeaksareat285.48eV,283.22eV,and282.11eV. ....... 88 4-6Leverarmapparatususedtoinitiallytestthehydroxidebondedshearstrengths(left)andMCSTapparatususedforlatertests(right). .............. 90 4-7BreakingstrengthsofBK7-BK7bondsafterbeingkeptinanovenat330K.Thezeroonthex-axisrepresentsthetimeafterthesampleswereinthecleanroomforapproximately18hours.Alldatapointsweretakenusingtheleverarmapparatus. .................................... 92 4-8Averagesofthedataobtainedusingtheleverarmandmodiedcompressionsheartest(MCST)devicesforbotha1:4and1:6dilutionfactor. ......... 95 4-9SurfaceprolesofareasonthePOCOSuperSiCusingaTencorsurfaceproler.ImagecourtesyofPetarArsenovicatGoddardSpaceFlightCenter. ....... 99 4-10SurfaceproleofoneofthelargeCoorsTekSiCtileusingaTencorsurfaceproler.ImagecourtesyofPetarArsenovicatGoddardSpaceFlightCenter. ....... 100 4-11SurfaceproleofoneofthesmallSiCpiecesusingaTencorsurfaceproler.Thelargespikearound750mismostlikelyaspeckofdirt.ImagecourtesyofPetarArsenovicatGoddardSpaceFlightCenter. ................. 100 5-1Electriceldsinsideandoutsideofanopticalcavity. ............... 104 11


........... 105 5-3Theresonatorgparameters. ............................. 106 5-4Geometryforanalyzingthemisalignmentofacavity. ............... 107 5-5CompletedSiCcavitymadeofHexoloySA. .................... 111 5-6WYKOimagesofthepolishedendsoftheSiCspacer.ImagesproducedbySurfaceFinishesCo.,Inc. ................................... 112 5-7FabricationowchartoftheCFRPspacerattheUniversityofBirmingham. .. 113 5-8CFRPspacer,Zerodur/mirrorplugs,andthealignmentjig(left).FinishedCFRPcavityafterepoxyingtheplugs(right). ....................... 114 5-9RepresentationofthePZT-actuatedcavity.Notethelengthsarenottoscale. 116 5-10DiagramofthePDHlockingtechnique. ....................... 117 5-11Twoerrorsignalsproducedfromcavitiesofdierentnesse.Thegreencurveisforacavityoflength200mmandnesse500.Theredcurveisforacavityofthesamelengthbutnesseof5000.BothhaveEOMmodulationfrequenciesof50MHz. ........................................ 121 5-12ErrorsignalsforonlythecarrierfrequencyfromFigure 5-11 .Thegreencurveisforacavitywithalowernesse. ......................... 122 5-13Electro-opticalsetupusedtotestthedimensionalstabilityofselectmaterials. 123 5-14Timeseriesofthebeatnotetakenover28days. .................. 124 5-15NoisespectrumoftheZerodurcavitieswithopticallycontactedmirrors.Measurementsweretakenusingthefrequencycounter,phasemetersamplingat15Hzand98kHz. ......................................... 125 5-16NoisespectrumoftheZerodurcavitieswithopticallycontactedmirrorsandtheLISApre-stabilizationrequirement.Theresultsarefromatimeseriestakenoversevendaysandsampledat1Hz. ........................ 126 5-17PlotsoftheZerodurbeatnoteandtanktemperature(left)andplotsofthebeatnoteandtemperaturedataafterbeinghighpassltered(right). ............ 128 5-18ComparisonofthenoisespectrumsoftheZerodurcavitieswithopticallycontactedandhydroxidebondedmirrors. ........................... 128 5-19PlotsoftheSiCbeatnotetakenoveralongweekend(left)andplotsofthebeatnotetakenover11days(right). .............................. 130 12


........ 130 5-21InferredtemperaturestabilityatthecenterofTank1. .............. 131 5-22PlotsoftheSuperInvarbeatnoteandtanktemperature(left)andplotsofthebeatnoteandtemperaturedataafterbeinghighpassltered(right). ...... 133 5-23NoisespectrumoftheSuperInvarcavitywithhydroxidebondedmirrorsalongwiththeLISAtelescopeandpre-stabilizationrequirement. ............ 133 5-24InferredtemperaturestabilityinsideTank2. .................... 134 5-25VoltageamplierusedtoprovidetheappliedvoltagetothePZTonthePZTcavity. ......................................... 137 5-26NoisespectrumofthePZTcavitywithnoappliedvoltage.Measurementsweretakenusingthefrequencycounter,phasemetersamplingat15Hzand98kHz.Thespikearound20kHzismostlikelytheresonanceofthePZTring. ..... 139 5-27NoisespectrumofthePZTcavitywith0V,and200VappliedtothePZTalongwiththeLISApre-stabilizationrequirement. .................... 140 5-28Theslopeofthebeatnoteovertimewith200VappliedtothePZT.Thetimeserieswastakenapproximately10minutesafterthevoltagewasappliedtothePZT. .......................................... 141 5-29Arm-lockingsetupusedtoarm-lockthePZTcavitytoaZerodurcavity.Thefeedbackcontrolsusedtoinitiallylockthelaserstothecavitiesarenotshown.PD:photodiode,PM:phasemeter,FM:frequencymeter ............. 142 5-30Modeledopen-looptransferfunctionofthearm-lockingcontrollerusedtosuppressthelasernoise. .................................... 143 5-31NoisespectrumofthePZTcavitywithoutanappliedvoltage,thenoisespectrumofthePZTcavityusingarm-locking,andtheLISApre-stabilizationrequirement. 143 5-32Exampleofastepglitch(top)andspikeglitch(bottom). ............. 144 5-33BeatnoteoftheCFRPcavityinTank2takenover4days. ............ 147 5-34StabilityoftheCFRPcavityalongwiththeLISAtelescopeandpre-stabilizationrequirements. ..................................... 148 5-35DimensionalstabilityoftheCFRPcavityfordieringincidentintensitiesalongwiththeLISAtelescoperequirement. ........................ 149 6-1Intensityasafunctionoffrequencyforbothanopticalcavityandgascell. ... 152 6-2Experimentalsetupusedformodulationtransferspectroscopy. .......... 155 13


................. 156 6-4Intensityproleasthelaserfrequencyisscannedovermultipleabsorptionlines. 157 6-5Detailedintensityproleofthreeabsorptionlines. ................. 158 6-6Temperaturedependanceonabsorptionofmolecularcesium. ........... 158 6-7Errorsignalsproducedforarangeofcesiumlines. ................. 159 6-8Experimentalsetupusedforfrequencymodulatedspectroscopy. ......... 161 6-9Stabilityresultsofthemodulationtransferandfrequencymodulatedspectroscopytechniquescomparedtoafree-runninglaser. .................... 162 6-10Beatnotesofboththecesium-ZerodurandZerodur-Zerodurlockedsystems. .. 165 7-1Representationof(a)anon-axisCassegraintelescopeand(b)ano-axisCassegraintelescope. ....................................... 167 7-2Isometricviewofthetelescopedesign.FigurecourtesyofJosephGenerieatGoddardSpaceFlightCenter. ............................ 170 7-3Payloadassemblyforoneoftheopticalbenches,on-axistelescope,andthermalshieldforone\tube"oftheLISAspacecraft.Eachspacecraftwillconsistoftwotubes.FigurecourtesyofNASA. .......................... 172 7-4ExpectedoperatingtemperaturesontheLISAtelescope.FigurecourtesyofAngeliqueDavisatGoddardSpaceFlightCenter, ....................... 173 7-5Diagramofhowthetelescopepartstusingthealignmentjig.Forclarity,notallofthestrutclampsareshown.CourtesyofJeLivasatGoddardSpaceFlightCenter. ......................................... 175 7-6Pictureofworkstationtodeterminethegapsizesbetweenthestrutsandeitherprimaryorsecondarymirrors. ............................ 176 7-7Gapsizewhennotastrutisnotproperlyaligned(left)andgapsizewhenastrutisalignedproperly(right).Insomecasesadownwardforcewasneededinorderforthegaptoclose.Gapsshownontheleftsideofthegureweretypically10to20m. ..................................... 177 7-8Pictureofthesodiumsilicatesolutionllingingapsbetweencontactpointsofthestruts. ....................................... 178 7-9Representationofhowthestrutsbondedtotheprimaryandsecondaryatatilt.Theviewofthesecondaryisasifonewaslookingupatthesecondaryfromtheprimary.Theviewoftheprimaryisasifonewaslookingatitfromthesecondary.Theshadedareasarewherethehydroxidebondsarethicker. ........... 179 14


........................................ 180 7-11Diagramofhowthesisterblockswereusedtobondthestrutandtile. ..... 182 7-12Representationofthesystemusedtocoolthetelescope(left)andapictureofthenishedsystem(right).NotshownontheleftgurearethetwolayersofaluminizedPET.ThealuminizedPETshieldsarepartiallyshowninthegureontherightalongwiththeSiCprimaryafterthestructurecollapsed. ...... 184 7-13Representationofhowthein-bandandabsolutelengthnoiseforthetelescopewillbefound. ..................................... 186 7-14RepresentationofhowtheMichelsoninterferometerwassetuptomeasurethedistancechangebetweentheprimaryandsecondary.. ............... 188 7-15PhotodiodeoutputoftheMichelsoninterferometerasthetelescopewascooleddown(left)andadetailedviewofthesameoutputoverashorterperiodoftime(right). ......................................... 189 7-16DetailedviewoftheoscillationsintheMichelsoninterferometerdata. ...... 190 7-17CTEasafunctionoftemperatureforCoorsTekUltraSiCSiC.DataobtainedfromDavidBathatCoorTek. ............................ 191 8-1RepresentationofthethreeLISAspacecraft,theopticalbenches,andtheback-linkopticalber. ...................................... 193 8-2DiagramofarotatingloopinterferometertoillustratetheSagnaceect.Dopplershiftedfrequenciesareshownwithrespecttotheinertialspace. ......... 194 8-3BoththePANDAandbow-tiepolarization-maintainingberstylesareshownalongwiththecoreandstresselements. ...................... 196 8-4Sagnacinterferometerusedtomeasuretheaccumulateddierentialphasenoiseinanopticalber.HWP1-3arehalfwaveplates,PBS1-3arepolarizingbeamsplitters,QWPisaquarterwaveplate,andthebalancedphotodetectorisrepresentedbyBPD. ........................................ 198 8-5Noisespectrumofthedetectorcomponentsoftheinterferometer.ThiswasfoundbyplacingamirrorinfrontofPBS1andaligningthebackreectedlight. .... 206 8-6Plotofthetimeseriesbeforeandaftermodulatingthefrequencyofthelaser.Itisclearthatmodulatingthelaserfrequencyproducessignicantlylessnoise. 208 15


..................................... 209 8-8Timeseriesofthebalancedphotodetectoratthesametimethepowersofthecounterpropagatingbeamsaremeasured. ...................... 209 8-9Noisespectrumoftheopticalberwithcounter-propagatingbeams. ....... 210 9-1ComparisonofthenoisespectrumsoftheZerodurcavitieswithbothopticallycontactedandhydroxidebondedmirrors,SiC,CFRP,andSuperInvaralongwiththeLISAtelescoperequirement. ........................ 214 A-1Noisespectrumoftheoutsidetanktemperature. .................. 219 A-2Pictureofthebuiltthermalshields(left)andamodelrepresentationofthem(right).Notethemodelisnotdrawntoanyscale. ................. 221 A-3Lumpedthermalmodelofthethermalshields. ................... 221 A-4TransferfunctionofthePETthermalshields. ................... 223 B-1TypicalsurfaceproleofaSiCsurfacebeforecleaningwithceriumoxide. .... 226 B-2TypicalsurfaceproleofaSiCsurfaceaftercleaningwithceriumoxide. .... 227 16








Space-basedinterferometricmissionssuchasSPIRIT,SPECS,TPF,andLISAwilltakemeasurementsoftheuniversewithunprecedentedresults.Todothis,thesemissionswillrequireultra-stablematerialsandbondingtechniquestobeusedforcriticalopticalcomponentssuchasopticalbenchesorsupportstructures.Asanexample,thetelescopesupportstructurefortheLISAmissionmustbemadeofamaterialthatisstabletobetterthan1pm/p TheauthorispartofagroupattheUniversityofFloridathatisdevelopingamaterialsresearchfacilitycapableoftestingthestabilityandstrengthofmaterialsandbondingtechniquesforuseinspace-basedinterferometricmissions.Thisworkdescribesthetechniquesusedtotestthestabilityofmaterialsatthefemtometerlevel.ResultsarepresentedforSiC,Zerodur,acarbonberreinforcedpolymer,SuperInvar,andk-180piezoelectricmaterial.Inaddition,severalbondingtechniquesaredescribedalongwithpotentialusesforspacemissions.Specicattentionispaidtoapromisingnewjointingmethodknownashydroxidebonding.Thebondingmethod,bondingmechanism,andshearteststrengthsarepresentedforseveralmaterialcombinations.Duringthiswork,itwasfoundthathydroxidebondingcouldbeusedtobondSiCtoSiCwithsignicant 20


Finally,investigationsweredonetodeterminethedierentialphasenoiseincounter-propagatingbeamsthroughanopticalber.ResultsusingapolarizingSagnacinterferometershowedadierentialphasenoiseof5105cycles/p 21


Inanattempttolearnmoreaboutwhywearehere,wherewecamefrom,andwhatthefutureholdsforhumankind,countlesstimeandresourceshavebeenspentonprojectsdesignedtoanswertheseandotherquestions.Ground-baseddarkmatterandgravitationalwavedetectorssearchforsignalscomingfromouruniversetolearnhowithasevolvedovertime.Gianttelescopesmaptheskyforplanetsthatmaybesimilartoourown.Althoughground-baseddetectorscanprovideawiderangeofinformationaboutboththeEarthandtheuniversewelivein,noisesourcessuchasseismicactivityoratmosphericturbulencemakeitdicult,ornearimpossible,tomakecertainmeasurementsfromtheground.Forthisreason,space-basedmissionshavebecomeincreasinglypopularinordertomakemeasurementstheEarth-bounddetectorscannot.Oftenspace-baseddetectorsworkcomplimentarytotheirground-basedcounterpartstoprovidenewandexcitinginformation.Whilemovingdetectorsintolow-Earthorbitsandbeyondcanovercomemanychallengesfacedontheground,otherdicultiessuchasdrag-freeorbitsortheinabilitytoservicecomponentsprovidenewchallengestoovercome. Asspace-basedmissionsarebecomingevermorecomplex,thesensitivityofnecessarymeasurementshasincreasedaswell.Thisrequirestheuseofultra-stablematerials,oscillators,andlasersthatpushtheirachievablelimits.Severalmaterialssuchassiliconcarbide(SiC)andSuperInvarhavebeenusedforcriticalopticalcomponentsonseveralspace-basedmissions,buttheirdimensionalstabilityhasnotbeenstudiedtodetermineiftheyaresuitableforthenextgenerationofdetectors.AlthoughthisworkprimarilyfocusesondeterminingthestabilityofmaterialsandbondingtechniquesfortheiruseontheLaserInterferometerSpaceAntenna(LISA),itcanbeusedasareferencewhendesigningothercriticalopticalcomponentssuchasopticalbenchesortelescopesupportstructuresforfuturespace-basedmissions 22


1 ].CertaingeophysicalprocessesgenerategravityanomaliesoverthesurfaceoftheEarth.MeasuringtheseanomaliesprovidesabetterunderstandingofthestructureoftheEarth.Short-termmassuctuationssuchasavariationinthewatercontentoftheEarth'scrust,globalsealevelchanges,orpolaricesheetbalancecanbemeasuredwithGRACEtodeterminetheirimpactontheglobalclimate.DatacollectedfromGRACEhasprovidedthescienticcommunitywithinformationthathasledtoasubstantialimprovementinourunderstandingofhowtheEarthchangesovertime[ 2 ]. GRACEwaslaunchedMarch17,2002andconsistsoftwoidenticalsatellitesinanearcircularorbit500kmabovetheEarth'ssurface(seeFigure 1-1 ).Thedistancebetweenthesatellitesis220kmalong-trackandlinkedbyaK-bandmicrowaverangingsystem.Inaddition,eachsatellitecarriesGlobalPositioningSystem(GPS)receivers,attitudesensors,andhighprecisionaccelerometers.Thesatelliteorbitsdecayat30m/daysuchthatthesatellitesdonothaveaxedrepeatpatternwhichreducesitsmeasurementsensitivity.Thesatellitesare3-axisstabilizedsuchthattheK-bandantennasarepointedateachother[ 1 ]. Inordertoproducethenecessaryprecision,severalpost-processingtechniquesareused.Toreducetheeectscausedbytheionosphere,thedual-frequencyone-wayK-band 23


RepresentationoftheGRACEsatellitesinorbitabovetheEarth.ImagecourtesyofNASA. Figure1-2. RepresentationofhowtheGRACEspacecraftmoveastheypassoveramassanomaly.In(a),spacecraft1feelsanaccelerationduetotheanomaly.Afterspacecraft1passestheanomaly,itfeelsanaccelerationintheoppositedirectionofspacecraft2asshownin(b).Afterspacecraft2passestheanomaly,itfeelsanaccelerationintheoppositedirectionasdepictedin(c). phasemeasurementstransmittedandreceivedbyeachsatellitearecombinedduringgroundprocessingwhichremovesmanyoftheeectscausedbyoscillatorinstability. Datagatheredfromtheprecisionaccelerometersisusedtoremovenon-gravitationalforcesonthesatellites.TheGPSreceiversallowprecisiontime-taggingofwhenthedatawastakenandisusedtodeterminetheinter-satellitedistance.Precisionestimateoftheinertialorientationisprovidedbytheattitudesensors.ThecombinationofthesemeasurementsalongwithancillarydataisusedtoproducearepresentativegravityeldoftheEarth. 24


3 ]. 4 ].Thesemeasurementsincludethetimevaryingoceanbottompressure,monitoringgroundwaterandpolaricecaps,andmeasuringthelithospherethickness.Tomakethesemeasurements,theproposedfollow-onmissionwouldconsistoftwocoplanarspacecraftincircular,polar,orbitsatanaltitudeof250-600km.Thespacecraftwouldbesmallsatellitesapproximately150-200kgandwouldrequireabout100Woftotalpowerandwouldhaveabaselinelifetimeof5years,withagoalof10years.Thedistancebetweenthespacecraftwouldbelooselymaintainedto50-100kmandmeasuredusinglaserinterferometry.MicroNewtonFEEP-typethrusterswillbeusedtoprovideadrag-freecontrolofthespacecraftinadditiontogroundtrackrepeatingwithinafewkilometers[ 4 ]. FortheGRACEfollow-onmissiontowork,thelaserinterferometrymustbecapableofprecisionphaseextractionwithlasershavingafrequencystabilityofbetterthan1partin1016rmsover1000seconds.Inaddition,thespacecraftwillneedaninternalsensoror\accelerometer"capableofmeasuringnon-gravitationalaccelerationsonthespacecrafttobetterthan1partin1013m/s2rmsover1000seconds.Itisimportantthattheinertialsensorbefullyintegratedwiththelaserinterferometerinordertoreduceerrorsrelatingtheinterferometerreferencepointtotheinertialreferencepoint.Whilethereareseveraldesignsthatcouldbeimplemented,amethodusingfreely-fallingproofmassesoneachspacecraftalongwithheterodyneinterferometrycontainingpolarizingopticsasa 25


4 ].Firstperformancelaboratorymeasurementsofaspacecraft-to-spacecraftinterferometrydesignusingametalbreadboardhaveshownasensitivityof2.5nm/p 4 ].Signicantimprovementsinthestabilityresultscouldbeachievedusinganopticalbenchmadeofaglass-ceramicsuchasZerodurandusingadvancedbondingtechniques,suchashydroxidebonding,fortheopticalcomponents. 1.4.1Introduction 5 ].Recently,theoutbursttimingoftwoblackholesintheOJ287systemhasprovidedmoreindirectevidencethatgravitationalwaves(GW)exist[ 6 ],buthaveyettobedirectlydetecteddespiteeortsfromground-basedgravitationalwavedetectorssuchasLIGOandVIRGO.Althoughground-basedGWdetectorshaveyettondasignal,theevidenceforthemiscompellingenoughtoprepareforspace-baseddetectorsthatwillsearchforcomplimentaryGWsources[ 7 ],[ 8 ].OnesuchmissionisknownastheLaserInterferometerSpaceAntenna(LISA).LISAisajointNASA/ESAmissionthatconsistsof3spacecraftthatwillformatrianglewhosearmsareseparatedby5109m.LISAwilltrailtheEarthby20,andisinclinedtotheeclipticplaneby60.ThisorientationwillallowLISAtoprovideangularresolutionofsourcestoaboutonedegree,withsub-degreeresolutionformanystrongsourceswhenaveragedovertimeperiodsof6to8months.LISAwilldetectgravitationalradiationfromsuper-massiveblackhole(SMBH)mergers,galacticbinaries,andextrememass-ratioinspirals(EMRI's)inthe30Hzto1Hzfrequencybandwithastrainsensitivityof1021/p 9 ].SinceitisthoughtthattheGWsourcesdetectablebyLISAplayanimportantroleindetermininghowthe 26


RepresentationoftheLISAorbit.ImagecourtesyofNASA. earlyuniverseformed,LISAwilleectivelyprovideuswithaglimpseintothepast,andsubsequentlystudythestructureandevolutionoftheuniverse. 10 ].Foranyappreciablesignalthathashopeofbeingdetected,thesourcemustbeextremelymassiveandcompactenoughtomoveatrelativisticspeeds.Althoughthisdoessomewhatlimitthesourcesavailablefordetection,therearestillplentyofsourcesthatwillprovideaplethoraofinformationabouttheuniversewelivein. 27


10 ].Forexample,whentheuniversewasroughly1TeV,theelectroweakinteractionseparatedintoelectromagneticandweakinteractions.Sincethisseparationwasnotuniform,someregionstransitionedbeforeothers,producingabackground[ 11 ],[ 12 ].IthasalsobeensuggestedthroughsuperstringtheorythatanetworkofcosmicstringsproducedintheearlyuniversethathaveexpandedtocosmicsizecouldproducebackgroundsdetectablebyLISA[ 13 ].AlthoughthereisanextremelysmallchanceofLISAndingGWbackgroundsignals,theinsightprovidedintooutuniversewillbetremendousandworthlookingfor. 10 ].Mostofthesearewhitedwarfbinarieswhichwillslowlyspiraltowardseachotherastheyradiategravitationalwaves.TheseobjectsareextremelyimportanttotheLISAmissionbecausetheyareguaranteedsources.TherearesomanyofthesesourcesthatcannotbedistinguishedfromoneanotherthatthereisanexpectedconfusedbackgroundthatwillbedetectedintheLISAband.Theamplitudeofthenoisewillbeofsignicantimportanceindeterminingthedistributionofstarswithinourgalaxy[ 14 ]. 28


15 ]. 10 ].Themostimportantchirpingsourceswillconsistofsystemswherebothmassesareblackholes.Blackholebinarieswithmassesrangingfrom104to107solarmassesandmassratiosrangingfrom1/20toaround1willgenerategravitationalwavesdirectlyinthemostsensitivespotoftheLISAband.ThesignalswillsweepacrosstheLISAbandintimesrangingfromseveralmonthstoseveralyears.Findingthesesignalsearlywillbeinvaluable.Withenoughwarning,telescopescanbepositionedtolookatthatpatchoftheskywhereagravitationalwaveisthoughttohavecomefrom.Inthismannerbothanelectromagneticandgravitationalspectrumcanbefoundandprovideamorecompleteunderstandingtotheconstituentsandinteractionsofblackholes.Evenwithpessimisticformationrates,atleastafewsupermassiveblackholecoalescenceshouldbeobservedoverLISA'slifetime,whilemoreoptimisticviewsputthisnumberinthehundreds[ 16 ]. 17 ].Themannerinwhichthesmallerbodyspiralsintothelargerbodywilltellagreatdealofinformationaboutthebinary'sspacetimeandessentiallymapoutthegravitationalpotentialofthelargerbody. 29


18 ]. Thedistancechangesbetweentheproofmasseswillbemeasuredusinglaserinterferometry.Lightatawavelengthof1064nmwillbesentfromonespacecrafttoadistantspacecraftwhereitwillbeinterferedwithasmallamountoflightfromthelocalspacecraft.Theprecisionofthismeasurementreliesonmeasuringthephaseoftheradiofrequency(RF)beatnoteasrecordedbyadigitalphasemeterwithmicrocycleaccuracy.AsagravitationalwavepassesthroughtheLISAconstellation,thearmlengthswillbemodulated.TheresultingchangesinthebeatnotewillberecordedandsentbacktoEarthwhereitwillbepost-processed.Theabilitytoreachtherequiredstrainsensitivitywilldependonthearmlength,theaccuracytowhichtherelativedisplacementbetweenthetwoproofmassescanbemeasured,andtheabilitytowhichtheproofmassescanbeshieldedfromdisturbances. Takingthesefactorsintoaccount,theLISAspacecraftdesignhasbeendrivenbytwosystemsknownastheDisturbanceReductionSystem(DRS)andtheInterferometryMeasurementSystem(IMS).ThepurposeoftheDRSistokeeptheproofmassasundisturbedaspossible,whilethepurposeoftheIMSistomeasurethedistancesbetweentheproofmassesusingopticalinterferometry.Bothsystemsworktogetherinamannertoreachtherequiredsensitivity. 30


18 ].TheheartoftheDRSistheGravitationalReferenceSensor(GRS)whichhousestheproofmass.Thespacecraftacttoshieldtheproofmassesfromdisturbanceforcessuchassolarwindsorcosmicdust.Ifleftunattended,theseforceswouldcausethespacecrafttoorientitselfinawaythatwouldmakemeasurementsimpossible.Toavoidthis,theGRS\senses"wheretheproofmassisrelativetothespacecraft.MicroNewtonthrustersprovidetheactuationneededinordertopositionandorientatethespacecraftinordertoprovidedrag-freeoperationaswellasprovidethenecessarypointingoftheoutgoinglaserbeamstowardsthedistantspacecraft. ThelowfrequencyendoftheLISAbandwillbelimitedbytheaccelerationnoiseoftheproofmassesandspacecraftwhicharedirectlydependantontheperformanceoftheDRS.ThenoiseallocationfortheDRSgivenasanamplitudespectraldensityforasingleproofmassisgivenby[ 18 ] s2p f:(1{1) ToconvertEq. 1{1 intoadisplacementnoise,itisintegratedtwice.TakingintoaccountthattherearefourproofmassesineachMichelson-typeinterferometerandthatallofthenoiseisuncorrelatedwilladdafactoroftwotothetotalnoise.Thisresultsin[ 18 ] (2f)2s f:(1{2) Thiserrorbudgetismadeupofsubsystems,eachwiththeirownnoisesub-allocation.Suchnoisesourcesincludeelectrostatics,Brownianmotionoftheresidualgasaroundtheproofmass,andthermaleectsamongothers.Adetailedaccelerationnoisebudgetusingallknowndisturbanceeectsisgiveninreference[ 19 ]. 31


20 ].Muchofthisworktypicallyinvolvessuspensionoftheproofmassfromatorsionpendulumtomodelandmeasureexpectednoisesources.Inaddition,theperformanceoftheGRSwillbedirectlytestedwhenitislaunchedon-boardtheLISAPathndermission[ 21 ]. 22 ].Inordertodeterminethissensitivity,knowledgeofthesystemresponsetoagravitationalwavemustbeknown.Thisresponsecanbecalculatedusingthesumsanddierencesofthephasesmeasuredineachofthesixlinksastheyarecombinedinpost-processingwiththeappropriatedelays[ 18 ].Fromthesinglelinkstrainsensitivity,theIMSmeasurementperformancecanbederivedandisgivenby[ 18 ] f4:(1{3) Withinthistotalerrorbudgetaresubsystemswiththeirownnoisesub-allocation.Forinstance,thetelescopepathlengthstabilityhasanallocationof1pm/p 18 ].AllIMSsub-allocationshavethesamefrequencydependanceasinEq. 1{3 .Manyoftheimposedrequirementsarecurrentlypushingthelimitsofwhatispossibleanditisimperativethatallsubsystemsandtheircomponentsbetestedinordertodeterminetheirstabilityandlimitations. 32


Thelight-weightbaseplateismadeofZerodurandprovidesthedimensionalstabilityofallotherpathlengths.HydroxidebondingisusedtomountthefusedsilicaopticalcomponentsonthepolishedZerodurbaseplate[ 23 ].ThebenchwillbeisostaticallymountedtoalargeinterfaceringthattheGRSisalsomountedto. Figure1-4. Simpliedviewoftheassembly(a)andtheopticalbench(b). Lightfromthelasersystemisdeliveredtotheopticalbenchthroughtheuseofaberopticcable.Themajorityofthislightisthensenttothetelescope,withtheremainderofthelightusedaslocalmeasurements.Theseparationoftheincomingandoutgoingbeamsfromadjacentspacecraftisdonethroughtheuseofapolarizingbeamsplitter.Inordertomakethenecessarymeasurementbeatnotes,lightfromoneopticalbenchneedstobesenttotheother,andviceversa.Todothis,anopticalberwillbeusedthatcontainsbothbeamsincounter-propagatingdirections.Thetransmittedpowerenteringthetelescope 33


23 ]. 23 ].Thetelescopedesignalsoincludesare-focusingmechanismtocorrectforthedierencesinalignmentongroundandinightduetotheoutgassingoftheCFRPspacer. Anotherpossiblecongurationwouldbetouseanon-axisdesignmadeofSiC.ThesecondarywouldbeconnectedtotheprimarythroughtheuseofSiCstruts.Althoughthestrutswillproduceadditionalstraylightintheon-axisdesign,theeectsneedtobestudiedmorebeforeadecisiononthebaselinedesigncanbemadesincetheprimarysourceofstraylightwillbeduetoback-reectionsfromthesecondary.FurtherdiscussionofthissubsystemispresentedinChapter 7 24 ].Todothis,twospacecraftwillbeplacedinorbitaroundtheSuninadditiontousinganinterferometerontheInternationalSpaceStation(ISS).LightfrombothspacecraftaresenttotheotherspacecraftandtotheISS.Aphasemeasurementoftheinterferedlightwilldeterminetheinternalanglesofthespacecraft-spacecraft-ISStriangle. TheanglemeasuredbythespacecraftcanthenbecomparedtotheanglecomputedusingEuclideangeometry.Adeviationinthisanglewillprovideawealthofinformation 34


CongurationoftheLATORorbits,spacecraft,andinterferometerontheISS.ImagecourtesyofNASA. aboutthesecondpost-Newtonianfeaturesofgravity,themassparameteroftheSun(thespatialmetric's\gammamass"),aswellasmeasuringthesolargravitationaleldtounprecedentedlevels[ 25 ]. 24 ],[ 25 ]. 26 ],andorbitwitha3:2resonancewiththeEarthasdepictedinFigure 1-5 .The3:2resonanceoccurswhentheEarthdoesthreerevolutionsaroundtheSun,whilethespacecraftdoesexactlytwo.Theexactperiodoftheorbitwillvaryslightlyfroma3:2orbitbylessthanonepercentdependingonthetimeoflaunch.Theadvantageofchoosingthisorbitarethatitimposesnorestrictiononthetimeoflaunch,itwillonlyrequiresmallpropulsionmaneuvers,itprovidesaveryslowchangeintheSun-Earth-Probe 35


Thesecondtelescopeemitsthedirectingbeacontotheotherspacecraft.Apairofpiezocontrolledmirrorsplacedonthefocalplanedirectsthelightintoandoutofthetelescope.Thecollimatedlightwithapowerof10Wisinjectedintothetelescopefocalplaneandproperlydirectedbythepiezomirrors[ 25 ]. Thethirdtelescopeisusedforthelaserlighttrackinginterferometryandwillbeusedtomeasuretherelativelengthchangesofthespacecrafttoanaccuracyoflessthan10pmusingheterodyneinterferometry[ 24 ].TwopiezoelectricX-Y-Zstageswillbeusedtomovethetipsoftwosingle-modeberswhilemaintainingfocusandbeampositiononthebersandotheroptics.DitheringatafewHertzwillprovidethemeanstomakethenecessaryalignmentofthebersandsubsequenttrackingofthespacecraftcompletelyautomatic. Bothspacecraftareidenticalandcontaina1Wcavityber-ampliedlaserwhosewavelengthis1064nmanddriftsat2MHzperhour.MostofthepowerispointedtotheISSthrougha20cmdiametertelescopewhosephaseistrackedbyaninterferometer.Theremainingpowerisdirectedtotheotherspacecraftthroughtheuseofanothertelescope.Inadditiontothetwotransmittingtelescopes,eachspacecraftisequippedwithtworeceivingtelescopes.ThetelescopethatfacestheISShasalaserlinelterandknife-edgecoronographtosuppressthelightfromtheSunto1partin104ofthelightlevelofthelightreceivedfromtheISS[ 25 ].Thereceivingtelescopethatpointstowardsthespacecraftdoesnotrequirealterandcoronograph. 36


1-6 and 1-7 .InFigure 1-6 ,twosidereostatsarepointedatatargetandtwoducials(representedbythecornercubes)determinetheendpointsoftheinterferometerbaseline. Figure1-6. Heterodyneinterferometryononespacecraftwithphase-lockedlocaloscillator. Stablelocaloscillatorsareinterferedwiththelightfromthetwoarms,andthephasedierenceisrecorded.Ifthelocaloscillatorswerephaselocked,thentheanglesofthe 37


Fiberlinkedheterodyneinterferometryandbermetrologysystem(left)andberlinkedheterodyneinterferometryontwospacecraft(right). targetwithrespecttothebaselinenormalisgivenby[ 25 ] whereisthewavelengthofthelaser,nisanunknownintegerarisingfromthefringeambiguity,andbisthebaselinelength[ 24 ].Unfortunately,isextremelydiculttophaselockthetwolocaloscillatorsofsuchalongbaselinewiththeprecisionneeded,soanothermethodwasproposedthatusesasinglemodeber.Thesignalfromthelocaloscillatorcanbetransmittedtobothsidereostatsthroughasinglemodeber.Inthisconguration,additionalmetrologyisneededinordertomonitorthechangesinthepathlengthoftheberasthelocaloscillatorpropagatesthroughitasshownintheleftsideofFigure 1-7 .Themetrologysystemmeasuresthedistancefromonebeamsplittertotheotherandtheresultingangleisgivenby[ 25 ] wherem1isthephasevariationsintroducedbytheber.Whenanadditionalarmisaddedtothiscongurationthephasevariationsduetotheberarecommontoboth 38


25 ] Thisisonlyageneraloverviewofhowtheanglewouldbemeasuredbetweenthetwospacecraft,andinpracticetheinterferometerwillhaveopticalpathsthataredierentlengthsandwillneedtobemeasuredwithadditionalmetrologysystems. Inordertomaketheanglemeasurements,alongbaselinewillbeintroducedontheISS.OpticalpackageswillbeintegratedontheISStotheS6andP6trusssegmentswhicharelocatedatoppositeendsoftheISS.Bothpackageswillbeseparatedby100mandwillhaveaclearlineofsightpathbetweenthetwotransceiversduringtheobservationalperiodsinadditiontohavingaclearlineofsighttobothspacecraft.LocationontheISSisidealsincealimitedamountofSuntrackingthatisneededtopointthetelescopesontheISStothespacecraftcanbedoneusingthe-gimbalsontheISS[ 25 ]. TwometrologysystemswillbeusedontheISStomaketheangularmeasurements.Therstsystemmeasurestheopticalpathdierencebetweenthetwolasersignalswhiletheothersystemmeasuresthechangesinopticalpaththroughtheber.Michelsoninterferometerswillbeusedinordertomakethenecessarymetrologymeasurements.Bothspacecraftsignalswillbemeasuredsimultaneouslyusinganelectro-opticmodulatorandmodulatingatadierentfrequencyforeachbeam.Themetrologysignalsoftheberwillbeusedinthepostprocessingtodeterminetheangularmeasurements.ThemetrologysignalssenttothespacecraftfromtheISSwillbestabilizedtobetterthanonepartin1010usingatemperaturecontrolledFabry-PerotetalonthatislockedusingaPound-Dreverscheme[ 25 ]. ThecurrentdesignforLATORwillprovideawealthofknowledgeaboutthelimitsofgeneralrelativity.Movingfroma=10cm(thecurrentstandard)toa=1mwavelengthwillgainafactorof1010reductioninthesolarplasmaopticalpathuctuations[ 25 ].Takingadvantageofcurrentdrag-freetechnologyforuseonthe 39


25 ].ThetechnologyneededfortheLATORmissionhasalreadybeendemonstratedandprovidesasolidconceptualfoundation.TheLATORmissionisuniqueinconceptanddesign,andwillleadtorobustadvancementsintheunderstandingoffundamentalphysics. 40


Satellitesforspace-basedmissionsaremadeupofmultiplesubsystemsthatallworktogethertomakecomplexmeasurements.Oftentherearestrictrequirementsonsubsystems,suchastheopticalbenchortelescopemirrors,thatmustbemetinorderforthedesignsensitivitytobeachieved.Thechoiceofmaterialsthenbecomesanimportantconsideration.Insomecases,amaterialwithanextremelylowCTEmustbechoseninordertoreducethermaleects.AlthoughtherearematerialsavailablewithCTE'sontheorderof108/K,theymayexhibitothermechanicalorthermalcharacteristicswhichareunwanted,suchasalowthermalconductivity,ferromagneticbehavior,arebrittle,orhavealowYoung'smodulus.Tofullyunderstandamaterialanditspossibleapplications,thethermalandmechanicalproperties,alongwiththeproductionprocessshouldbeknown.Forthisreason,anintroductiontoaselect,fewmaterialsthatareeithercommonlyused,orhavegreatpotentialforspace-basedinterferometricmissionsarepresented.Theproductionprocess,mechanicalandthermalproperties,andcurrentorpotentialapplicationsarediscussed. 2-1 (itshouldbenotedthatthesevaluesareforroomtemperatureandaretypicalforabroadrangeofZerodurclassesandarenotabsolute)[ 27 ]. 41


SelectedmaterialpropertiesandtheirvaluesforZerodur PropertyValue Density2.53g/cm3ModulusofElasticity13106psiThermalConductivity1.43W/mKSpecicHeat0.80cal/gCIndexofRefraction1.54 2-1 and 2-2 .TheZeroduriscastintoanannealingovenwherethetemperatureiscooledinacontrolledmannerforseveralweeks.Theouterlayersofthematerialareremovedinordertopreventuncontrolledceramizationduringthenextsteps.Duringtheceramizationprocess,the Figure2-1. Zerodurproductionprocess. materialisheatedupagaintoachievegrowthofthecrystalphaseinacontrolledmanner.TheheattreatmentprocesshasacriticaleectontheCTEhomogeneityoftheblankandcanlastseveralmonthsdependingontheZerodurmaterialneeded.Oncethisstephasbeencompleted,theZerodurcanbeprocessedintoitsnalshape.CTEsamplesaretakendirectlyfromthenalmaterialtoensurethenalCTEvalueandhomogeneity.Dependingonthesizeandshapeoftheprocessedparts,thehomogeneityoftheZerodurCTEcanbekepttolessthan0.03106/Kper18tonsofmaterial[ 28 ]. 42


Zerodurannealingprocess. 28 ].TheCTEoftheZerodurisderivedusingthelengthchangemeasurementsfromtheuseofadilatometerasthematerialisheateduporcooleddown.DependingontheZerodurclassspecied,thevariationinthemeanCTEcanbeaslittleas0.02106K1fortemperaturesrangingfrom0-50C.Incertainfabricationprocesses,thisvariationcanbeevensmaller[ 28 ].Theexcellent Table2-2. ZerodurclassvalueandcorrespondingCTE. ZerodurexpansionclassCTE 000.02106K1100.05106K1200.10106K1 28 ]. 43


28 ]. 28 ].Theslopeofthedecreaseinlengthdependsonthethermaltreatmentandtheageafterthethermaltreatment.AslowercoolingtreatmentoftheZerodurwilldecreasetheslope,makingtheZerodurappearolder.Anythermaltreatmentwillstarttheagingprocessanew. 29 ]. 44


30 ].Invaralloysarecommonlyusedinclocks,scienticinstruments,bimetalstripsinthermostats,andmanyotherapplicationswhereconstantmonitoringisnecessary.TherelativeeaseinfabricatingandmachininghavealsomadeInvaralloyspopular.WhilethereareseveralInvaralloysavailablethatonlyvaryslightlyinelementalcomposition,onlythethreemostcommonwillbediscussed.Theproductionprocess,chemicalcompositions,andapplicationsofeachalloyarepresented. Althoughtheheattreatmentprocesswillvaryslightlydependingonmanufacturer,threeannealingmethodsarepresented.Eachalloywillhaveitsownspecicheattreatment,butingeneral,theprocessissimilar. Method1: 1. Heatpartsto830Candholdatthattemperatureforone-halfhourperinchofthickness. 2. Furnacecoolataratenottoexceed100Cperhourto315C.Noadditionalmachiningshouldbedoneontheseparts. 3. Letpartsaircool. Method2: 1. Roughmachineparts. 2. Heatpartto830Candholdatthistemperatureforonehalfhourperinchofthickness. 45


Furnacecoolataratenottoexceed100Cperhourto315C.Aircoolingbelow315Cisacceptable. 4. Heatpartsat315Cfor1hourfollowedbyaircooling. 5. Heatpartsat96Cfor48hoursfollowedbyaircooling. 6. Finishmachining. Method3: 1. Roughmachineparts. 2. Heatpartsto830Candholdatthistemperatureforonehalfhourperinchofthickness,thenwaterquench. 3. Semi-nishmachiningparts. 4. Heatpartsfor1hourat315C,followedbyaircooling. 5. Heatpartsat96Cfor48hoursfollowedbyaircooling. 6. Finishmachiningparts. ThevariationinheattreatmentsandelementalpurityandconcentrationfrommanufacturertomanufacturercanproducealloysthatarenotasreliablymadeasotherlowexpansionmaterialssuchasZerodur.Thiscanproduceavariationinthermalproperties.AlthoughCTEvaluescanvary,typicallyInvar36hasaCTEofapproximately1.6106/K[ 31 ]. AspecicInvaralloywasdevelopedbyNASA/JPLfortheCassinispacecraftcamera.DubbedHighPurityInvar36(HPInvar36),ithasamuchimprovedtimevaryingCTEanddimensionalstability[ 31 ].Verypureelementswereweighed,mixed,andpressedintoamoldandsinteredinacontrolledatmosphere.AnalysisshowedthisalloytohaveaCTEanddimensionalstabilitysimilartothatofInvar32-5[ 31 ].Althoughthisspecicalloyisnotreadilyavailable,itatteststotheextenttowhichtheCTEofInvar36canbepushed.Reference[ 31 ]provideanindepthdiscussionoftheheattreatmentandstabilityofthishighpurityInvar.Itshouldalsobenotedthatinreference[ 31 ],theparticularheat 46


AllInvaralloysshouldbehandledcarefully.Droppingthemmaydisturbthegrainstructureandcauseittobecomemagnetizedtoasmallextent.TheadditionofimpuritiessuchassulfurandchromiummakeInvar36easiertomachine.Thisallowsforcomplexstructurestobemachined,whilestillkeepingitslowCTE. 2-3 .TheprimaryuseofInvar42isforglass-to-metalsealsinelectronictubes,automotiveandindustriallamps,transformerandcapacitorbushings,andotherceramic-to-metalapplications.TheseapplicationsareduetothesimilarexpansionpropertiesofInvar42andotherglasses,particularly1075glass.TheheattreatmentforInvar42issimilarthatofInvar36. Forthelowestthermalexpansionandoptimumstability,theheattreatmentrecommendedistoinitiallyheatthematerialsto840Cfor1hourandthenwaterquench.Thisisfollowedbyastressrelievingoperationat315Cfor1hour,aircoolandageat96Cfor24hours,thenaircool.Duetothehighoxidationrateofthisalloyabove538C,itisrecommendedthepartsbeheattreatedinaprotectiveenvironmentsuchasavacuumorinertgas. AlthoughSuperInvar'sinitialtemporalstabilityismuchbetterthanInvar36,ithasbeenshownthatitdegradesovertimetoratescomparabletothatofInvar36.This 47


32 ]. DuetotherarityofliteratureonthechemicalcompositionaswellasthemechanicalandthermalcharacteristicsoftheInvarfamily,thefollowingtablesareacompiledlistofselectedpropertiesforInvar36,42,and32-5. Table2-3. CompositionalvaluesofInvar36,42,and32-5. Element364232-5 Carbon0.15.05max.05Nickel36nominal4131.75Phosphorus.006{{IronbalancebalancebalanceSilicon0.400.200.09Manganese0.600.400.39Sulfur0.004{0.01Chromium0.25{0.03Cobalt0.50{5.36Aluminum{{0.07Copper{{0.08 Table2-4. Comparisonofselectedpropertiesfor3Invaralloys Property(atroomtemperature)Invar36Invar42Invar32-5 Density(kg/m3)805581108150ThermalConductivity(W/mK)10.510.7{CTE(106/K){148144TensileStrength(MPa){517144SpecicHeat(J/kgK)515500{ 33 ].In1890SiCwasaccidentallysynthesizedbyEdwardG.Achesonwhilerunninganexperimenttosynthesizediamonds.SiCoccursnaturallyinmeteorites,althoughitisfoundrarelyandoftenonlyinsmallamounts.AchesonwasabletoproduceSiCbypassinganelectriccurrentthroughamixtureofsandandpetroleumcokeatveryhightemperaturesinanelectricfurnace.Althoughthisprocesshasbeenmodiedsinceits 48


33 ]. AlthoughSiCwasdiscovered2centuriesago,notuntilrecentlyhasitbecomeapopularmaterialtoreplacetheuseofmetalsandalloysforstructuralpurposes,especiallyinextremeenvironments. 33 ]. 49


33 ].Thischunkisthenbroken,sorted,crushed,milled,andclassiedintodierentsizesandcommercialgradesofSiCpowder.Furtherprocessingofthepowdercanbedoneinordertoproducepowderswithcrystallinesizesontheorderoftensofnanometers.ThecrystallinesizeplaysanimportantroleinthedensicationprocessofSiC.TheSiCpartsmadefromthesepowderswillhavevaryingproperties(specicallyporosity)dependingonthecrystalsizeused. Figure2-3. ProcessusedtoproducesinteredSiC. 50


33 ]. TheadditionofsinteringaidscanproduceimpuritieswithinthenalSiCproduct,butmuchcareistakentoremovealmostallimpurities.High-temperaturechlorinationcanbeusedtoremoveasignicantamountofimpurities,butmaycausedepletionofsiliconintheSiC.Eachmanufacturerhastheirowntechniquesusedtokeepthelevelofimpuritiestoaminimum,andinsomecases,noteventraceamountscanbedetected. 33 ]. Theprocessbeginsbyusingagraphitethathasalreadybeenprocessedandpuried.Thisgraphiteisthenmachinedintothedesiredshape,knownasthegreenpart.TheconversionofthegraphiteintoSiCtakesplacewhentheyareexposedtosilicon-carryingspecies,typicallySiOgas,athightemperatures.Althoughagasistypicallyused,usingliquidsiliconwillalsoproducesimilarresults.Thechemicalformulaforthisstepisshownbelow.SiO+2C,SiC+CO InorderforthegreenparttoconverttoSiCproperly,itisessentialforthegraphitetohaveareasonableporosityfortheSiOtoinltratethegraphite.Thisreactionisrate-limitedbythepore-diusionresistanceandkeepingalargeSiOconcentration 51


34 ].ThereactionrateattheSiC/Cinterfaceiscontrolledbysurfacekineticswhicharespontaneousathightemperatures.ThismeanstheconversionratefromgraphitetoSiCislimitedbytheinwarddiusionofSiOandoutwarddiusionoftheCOgasthroughtheSiCshell.Forthisreason,fabricatedpartscanonlybelessthanacertainthicknessifacomplete Figure2-4. Processusedtoproducereaction-bondedSiC. infusionistotakeplace.Usingthisinltrationprocess,theresultantparttypicallyshrinksbylessthan1%ofitsoriginalshape. 33 ].TypicallytheCVDprocessproducesaSiCwithalowerpercentageofporosity. 52


NV;(2{1) whereMisthegramformulaweightofthematerial(M=40.09715forSiC),NisAvogadrosconstant(6.02213671023/mole),andVthevolumeoftheunitcell. Thetheoreticaldensityisthedensityofamaterialasiftherewerenomicrostructuralporosity.Alsoknownastheapparentdensity,itisgivenby Vs;(2{2) wheremisthemassofthematerial,andVsisthevolumeoccupiedbythesolids.Dependingonthepolytype,thetheoreticaldensityliesbetween3.166to3.249g/cm3.ThesemeasurementshavebeenveriedusingX-rayspectroscopyandX-raydiraction[ 33 ]. Thebulkdensityisthemeasuredmassinthetotalbulkvolumeofthematerial.Thebulkdensityiscalculatedby Vs+Vp;(2{3) whereVpistheporevolume. Oftenreportedwiththedensity,porosityisanotherimportantphysicalpropertythatindicatestheamountoffreespacethatisnotoccupiedbysolidmaterial.Theporosityisdetrimentaltothestrengthofthematerial.Thestrengthisinverselyexponentiallyproportionaltothetotalporosity.Openporosityreducestheoxidationresistanceandcanpresentoutgassingproblemsunderhighvacuumconditions.ForRBSiCtheopenporositycanrangefrom0.50%to20%,dependingonthegradeofSiCandprocess 53


34 ].TypicalporositiesforsinteredSiCarelessthan1%,withmosthavingclosetozeroporosity[ 35 ][ 36 ][ 37 ]. Table2-5. ComparisonofdensityandporositiesforselectedSiCmaterials. MaterialApparentDensityBulkDensityTotalPorosityOpenPorosity SuperSiC3.13g/cm32.53g/cm320%19%SuperSiC-SiC3.07g/cm33.05g/cm34%0.5%HexoloySA{3.10g/cm3<2%{CoorsTekSiC{3.10g/cm3{{ 38 ].TheexuralstrengthismostcommonlytestedusingtheASTMC-1161standardwitheithera3-or4-pointtestbeingemployed.Theequationsforthe3-and4-pointbendingtestaregivenby bd2(2{5) 54


Diagramofhow3-and4-pointloadingtestsaredone. wherelisthelengthofthespecimen,Fisthetotalforceapplied,bisthespecimenwidth,disthespecimenthickness,andaisthedistancebetweensupportingbeads.Ifamaterialisductile,likemostmetalsandalloys,thenthematerialbendsbeforefailure.Ifamaterialisbrittle,likemanyceramics,thentherewillonlybeaslightbendingofthematerialbeforefailure. Denedasthemaximumtensilestresssustainedbyamaterialwhenapullingforceisappliedalongthespecimen,thetensilestrengthdeterminesamaterial'scapabilityofloadbearinginstructuralapplications.Thetensilestrengthiscalculatedbydividingthemaximumloadneededtorupturethematerialdividedbythecross-sectionalarea.Whendeterminingthetensilestrengthofmanyceramics,thesamplepreparationiscritical.Micro-cracksinoronthematerialwillpropagaterapidlyasaloadisapplied.Forthisreason,astatisticallysignicantnumberoftestsneedtobeperformed.ThemostcommontestmethodusedtodeterminethetensilestrengthistheASTMC-1273standardmethodastheelasticmoduluscanbeobtainedaswell. Table2-6. ComparisonofexuralandtensilestrengthforselectedSiCmaterials. MaterialFlexuralStrength(MPa)TensileStrength(MPa) SuperSiC147129SuperSiC-SiC269138HexoloySA410210CoorsTekSiC410307 55


Formanyceramics,theheatcapacityisastrongfunctionoftemperature.Theheatcapacityofamaterialcanbemeasuredthroughtheuseofthecalorimetrytechnique,orforhigh-puritymaterialsitcanbetakenfrompublisheddata[ 39 ]. Table2-7. ComparisonofthermalconductivityforselectedSiCmaterials. MaterialThermalConductivity(atRTinW/mK) SuperSiC151SuperSiC-SiC218HexoloySA110CoorsTekSiC100 56


40 ].Thematrixtransferstheappliedloadtothehigh-strengthbersinadditiontoprovidingcrackandenvironmentalresistancedamagetothestructure. Thecompositepartismadefromaberarchitecturewhichisbasedonaspecic\ply."Theorientation,thickness,andbertypeoftheplywilldeterminethepropertiesofthecomposite.Thecompositepropertiesvaryfromplytoply,butitisassumedthateachspecicplyishomogeneous[ 40 ]. Compositesaretypicallymadeupofseveralplieswithmutuallyperpendicularplanesofsymmetrytocreatealaminatethatissymmetricalandbalanced.Forstrongerandlighterparts,ahighbervolumeshouldbeused.Thisisidealforpartssuchasaircraftwings,butthepriceisusuallymoreexpensivesincetheberscostmorethantheresin.Aheavier,morechemicallyresistantpart,suchaspipingatanoilrenery,canbemadefromplieswithalowerber-to-resinratio. 40 ].PitchbersaremadefrompetroleumorcoaltarpitchesandtypicallyhaveextremelyhighstinessandlowtonegativeCTE.Duetotheseproperties,pitchbers 57


Manufacturersproducecarbonbersintowsconsistingofthousandsofcontinuous,untwistedlamentsthataremicronsthick.Thelamentcountisdesignatedbyanumberfollowedbytheletter\K,"whichdesignatesamultiplicationofathousand.Forexample,a10Ktowindicatesabundlecontaining10,000continuouslaments.Towsizesrangingfrom1Kto12Karetypicallyusedforaerospaceapplications.Forapplicationintheindustrial,construction,andautomotiveindustry,lamentcountsof48Kto320Karetypicallyusedduetothelowerprocessingcost,andfasterprocessingtimes.Theproductionprocessforhigherlamentcountstendtoresultinalowertensilestrength. WhenexposedtometalsPANandpitchbersexperiencegalvaniccorrosion.Thiscanbeovercomebyusingabarriermaterialorveilplymadeofberglassorepoxyduringthelaminatelayupinordertopreventdirectcontact. InadditiontoPANandpitchbers,otherhigh-performancebersincludearamid(commonlyreferredtobyitstrademarknameKevlar)andhigh-moduluspolyethylene.Aramidbersarecomposedofaromaticpolyamideandprovideexceptionalimpactresistanceandtensilestrength,makingthemidealforapplicationssuchasballisticprotection,solidrocketmotors,andhelicopterblades[ 40 ].Polyethylenebersarelightweight,haveexcellentchemicalresistance,extremelyimpactresistant,andhavealowdielectricconstant.Manyapplicationsforthesebersarefoundasthesameofthearamidbers.Quartz,ceramic,basalt,boron,andhybridbersareallotheroptions,butnotcommonlyused. 58


40 ].Undercertaincircumstances,pressurescanbeappliedtotheparttoexpediteorimprovetheprocess. Unsaturatedpolyesterresinsarewidelyusedincommercial,mass-producedapplicationssuchasbathware,boats,orcarpartsduetotheireaseofhandling,relativelylowcost,andgoodbalanceofmechanical,thermal,electrical,andchemicalproperties.Forhigh-performanceapplications,polyesterresinsaretypicallynotusedduetotheirhighshrinkrates.Instead,thermosetssuchasepoxyresins,phenolics,cyanateesters,andpolyimidesareused. Forhighstrength,durability,andchemicallyresistantcomposites,epoxyresinsaretypicallyused.Epoxiescomeinmanyformsrangingfromliquid,tosolid,tosemisolidandaretypicallycuredbyreactingwithaminesoranhydrides.Tougheningagentssuchasthermoplasticsorrubbercompoundscanbeaddedtocounteractbrittleness.Thiscanbeusefulinaerospaceapplicationsthatuseamine-curedepoxiesthatrequiretheuseofanautoclaveathightemperaturesthatcancausebrittleness. Lowcost,ame-resistant,low-smokeproductssuchasaircraftinteriorpanelsorrocketnozzleapplicationsaretypicallymadeusingphenolicresinsduetotheirexcellentheatabsorbingandlowcharyieldqualities.Althoughphenolicresinshaveexcellentthermalqualities,theirmechanicalstrengthstendtobelowerthanwhenusingotherhigh-performanceresins. Polyimidesareamoreexoticformofresinandhavefoundtheirwayintohigh-temperatureproductssuchasjetenginenacellecomponents.Theseresinsoerlowelectricalconductanceandhighchemicalresistance.Insomecases,partsmadeusingpolyimides 59


40 ]. FabricatorsofCFRPpartsbuytheberformandresintobestsuittheneededapplication.DuetothehighvolumeofCFRPpartsnowmade,speciallyformulatedresinmatrixesarereinforcedwithcarbonbersandpartiallycuredtoproducea\prepreg."Theseprepregsarethenfrozenanddefrostedwhentheyneedtobeused.Thespecicprepregischosenbasedontheapplicationneededandcanrangefromgolfclubstoairplaneandjetengineparts. 60


41 ].Theyfoundthatifamechanicalstrainwasappliedtocertaincrystals,theybecameelectricallypolarized.Thispolarizationwasproportionaltotheappliedstrain.TheCuriesalsodiscoveredthatwhenanelectriceldwasappliedtothesematerials,theywoulddeform.Thisisknownastheinversepiezoelectriceect.Thiseectisshowntooccurinmanynaturally-occurringcrystalsincludingquartz,tourmaline,andsodiumpotassiumtartrate.Thesematerialshavebeenusedformanyyearsaselectromechanicaltransducers.Ifacrystalistoexhibitthepiezoelectriceect,itsstructureshouldhavenocenterofsymmetry.Applyingatensileorcompressivestresstosuchacrystalwillchangetheseparationbetweenthepositiveandnegativechargesitesineachelementarycell,causinganetpolarizationatthecrystalsurface.Thiseectislinearinalargeregimeandisdirection-dependantsuchthatcompressiveandtensilestresseswillgeneratevoltagesofoppositepolarity[ 41 ].Thepiezoelectriceectalsooccursinareciprocalnature.Hence,ifavoltageisappliedtothecrystal,thenamechanicaldeformationwilloccur.Applyingavoltageofoppositepolaritywillcausethecrystaltodeformintheoppositeway. 41 ].AboveaspecictemperatureknownastheCuriepoint,thesecrystallitesexhibitsimplecubicsymmetrysuchthattherearenodipolespresentinthematerial.BelowtheCuriepoint,thecrystallinestructuresformatetragonalsymmetrysuchthatanelectricdipoleisbuiltintothestructure(Figure 2-6 ). Thedipolesarenotrandomlyorientatedthroughoutthematerial,butratherformregionsoflocalalignmentknownasWeissdomains.Withinthedomainallofthedipolesarealigned,givingthedomainanetdipolemomentandpolarization,butnooverallpolarizationisexhibitedduetotherandomorientationoftheWeissdomains[ 41 ].Theceramicmaybemadepiezoelectricinanygivendirectionbyuseofapolingtreatment.In 61


a)CubiclatticeabovetheCuriepointb)TetragonallatticebelowtheCuriepointafterthepolingprocesshasbeenapplied. thisprocessastrongelectriceldisusedatatemperaturejustbelowtheCuriepointtoorientatetheWeissdomainsinthedirectionoftheelectriceld(Figure 2-7 ). Figure2-7. ElectricdipolemomentsfortheWeissdomains(a)beforepolarization(b)duringpolarizationand(c)afterpolarization. Whentheelectriceldisremoved,thedipolesremainlockedintotheirneworientation,producingaremanentpolarizationandpermanentdeformation.Thepolingtreatmentisusuallythenalstepincreatingthepiezoelectricmaterial. Thepolingprocesswillproduceapermanentpolarizationwithinthematerial,butthesepropertiescanbepartiallyoreventotallylostifthematerialissubjectedtocertainmechanical,electrical,orthermalconditions[ 41 ].Iftoolargeastaticelectriceldortoolargeamechanicalstressisapplied,orifthematerialisheatedtoatemperaturenearits 62


ThebasicbehaviorofaPZTcylinderisshowninFigure 2-8 .IfatensileorcompressiveloadisappliedtothePZT,theresultingchangeindipolemomentcreatesavoltagetoappearbetweentheelectrodes.Similarly,ifavoltageisappliedtotheelectrodes,thenthePZTmaterialwillstretchorcompressdependingonthepolarityofthevoltage. Figure2-8. Thepiezoelectriceectofacylinderfor(a)noexternalaction(b)anappliedcompressiveforce(c)anappliedtensileforce(d)anappliedvoltage(e)anappliedvoltageofoppositepolarity.Forclarityonly1dipoleisshownin(a). ThephysicalconstantsofPZTmaterialsaregiven2subscriptsduetotheiranisotropicnature.ThedirectionofpositivepolarizationischosentocoincidewiththeZ-axisofarectangularcoordinatesystem[ 42 ].TheaxesX,Y,andZaredesignatedbythenumbers1,2,and3,respectively,whiletheshearabouttheseaxesarerepresentedby4,5,and6,respectively(Figure 2-9 ). Forclarity,3exampleswillbegiven. 1. d33istheinducedpolarizationperunitappliedstressindirection3.Alternativelyitistheinducedstrainperunitelectriceldindirection3. 63


Axesdesignationanddirectionsofdeformation.Directions4,5,and6aretheshearaboutaxes1,2,and3,respectively. 2. d31istheinducedpolarizationindirection3perunitstressappliedindirection1.Alternativelyitisthemechanicalstraininducedinthematerialindirection1perunitelectriceldindirection3. 3. g15istheinducedelectriceldindirection1perunitshearstressappliedaboutaxisdirection2.Alternativelyitistheshearstraininducedinthematerialaboutaxis2perunitelectricdisplacementappliedindirection1. Fromthesenumberstheamountoftranslationandshearcanbedeterminedforacertainappliedvoltage.Afewofthesevaluesaregivenintable 2-8 forthek-180piezoelectricmaterial[ 43 ]. Table2-8. Selectedpropertiesandtheirvaluesforthek-180piezoelectricmaterial. PZTPropertyValue LongitudinalCouplingCoecient(k33)0.68PiezoelectricStrainConstant(d33)165pm/VPiezoelectricStrainConstant(d15)350pm/VPiezoelectricVoltageconstant(g33)43mVm/NCurieTemperature350CDensity7.7103kg/m3CoecientofThermalExpansion3.6106/C 64


Whendesigningcriticalopticalcomponentssuchasopticalbenchesortelescopesupportstructuresforspace-basedmissions,therearetwoimportantfactorsthatmustbetakenintoconsideration;thematerialthecomponentismadeof,andthebondingtechniqueinwhichthepiecesareputtogether.Asthecomplexityandsensitivityofspace-basedmissionsincreases,thejointingtechniqueusedneedstobestronginordertosurvivelaunchconditionsaswellasstableinordertonotspoilthemeasurementsensitivity. Whilethepreviouschapterdealtwithavarietyofmaterialsthataresuitableforseveraldierentstructures,thenexttwochapterswillfocusonseveralbondingtechniquesthatcanorhavebeenusedtoconstructopticalcomponents. Opticalcontactinghasfounditswayintomanyareassuchassiliconwaferbondingintheelectronicseld,bondingmirrorstoopticalbenchesforuseinspace-basedmissions,andbondingopticalcomponentsforuseinlasers.Althoughopticalcontactinghasbeenaroundforseveraldecadesandhasgainedpopularityinvariousindustries,literatureonthesubjectisscarceandagreementbetweentheoryandexperimentvary[ 44 ].ItisassumedtheprimarybondingmechanismisduetothevanderWaalsinteractionatthesurface.TheforceperunitareaderivedfromthevanderWaalsattractionisgivenby[ 44 ] 65


44 ]. Thereareseveralopticalbondingmethodscommerciallyavailable,butthedetailsoftheprocessareoftendiculttoascertainduetotheirproprietarynature,althoughmanyofthemrequiremoderateheatingofthesubstrate.Forthisreason,threeopticalcontactingmethodsweredevelopedforuseatroomtemperatureandhavebeenusedonBK7,SiC,andSuperInvar. Thismethodisbestusedonultra-polishedsurfaces.Ifthesurfacesareatandsmoothenough(typically/10orbetter),anopticalcontactwillformoncethe2surfacescomeincontactandcanformappreciablestrengthwithlittleornopressureneedingtobeappliedtothematerials. Thismethodworksbestwhenalignmentofthetwomaterialsisneeded.Themethanolworksasalubricantthatallowsthematerialstobemoved,butstillhassomeinitialstrength.Asthemethanolevaporates,thebondstrengthens. 66


Thismethodworksbestforsurfacesthatarenotpolishedaswell,orwhenthesurfacesaredirtierthannormal.Inmanycasesiftheprevioustwomethodsdonotwork,thismethodoftenwillsucceed,ifonlytoformapartialbond. ThesemethodshavebeensuccessfullyusedonBK7-BK7,SiC-BK7,andSuperInvar-BK7bondswithbreakingstrengthstypicallybetween0.1-1.0MPa.OtheropticalcontactingmethodsexistwhichhavereportedbreakingstrengthsrangingfromseveraltotensofmegaPascals[ 45 ].Opticalcontactingwasprimarilyusedinthisworktoattachopticalmirrorstosubstratetoformanopticalcavitywhosedimensionalstabilitywasthentested. Thebasicconceptbehindanodicbondingistousetheelectrostaticattractionbetweentwomaterialswhenavoltageisappliedacrossthebond.Sinceglasseshaveanextremelyhighelectricalresistivityatroomtemperature,thepartsmustbeheattoafewhundreddegreesbeforeasignicantdropinelectricalresistivitycanbeseen.Althoughthe 67


46 ].Thebondoriginatesatthecontactpointsbetweenthesurfacesandspreadsoutfromthere.Forthisreason,atterandsmoothersurfaceswillproduceabondthatwillformfasterandwithgreaterstrength. Figure3-1. Representationofhowanodicbondingworks. Whenoneplacesawellpolishedsurfaceontopofanother,thesurfacesusuallyonlycontactinafewspots.Thiscaneasilybeseenwhentwopiecesofwellpolishedglassareplacedontopofoneanotherandinterferencefringescanbeseenradiatingoutwardfromoneortwopoints.Withanodicbonding,thesesurfacesarebroughttogetherbytheuseofanelectrostaticforce.Thevoltageappliedacrosstheglasssetsgeneratesanelectriceldwithinthegapthat,inturn,createsanelectrostaticforce(seeFigure 3-1 )thatpullsthesurfacestogether.Ifglassbehavedasaperfectinsulator,thenthevoltageneededwouldbeafunctionoftheglassthickness.Inaddition,bondingwouldoccurirrespectiveofthevoltagepolarityused.Withmostglasses,thisisnotobserved.Glasseswiththicknessesrangingfrom0.05to25.4mmshownovoltagedependenceandonlybondifthecorrectpolarityisused[ 46 ].Thisisbecauseatthetemperaturesusedforanodicbondingmost 68


Whendealingwithanodicbonding,therearesixprocessparametersthatmustbetakenintoaccount;temperature,voltage,timeofappliedvoltageandcoolingrate,surfacenish,CTE,andatmosphere[ 46 ]. Mostworkdonewithanodicbondingoccursinthe300to600Crange,althoughthiscanvarydependingonthetemperaturedependenceoftheelectricalresistivityoftheglass.Zerodur,forinstance,hasaresistivityofabout1Mcmat300C[ 46 ]. Typicalworkingvoltagesareintherangeof200to2000V.Forglasseswithahigherresistivity,morevoltageshouldbeusedtocreatealargersurfacechargeatthebondinginterface.Theupperlimitonthevoltageissettoinhibitsparkingwhichmaydamagetheparts. Thetimerequiredinorderforthebondingtooccuristypically1-5minutes(afterthepartshavebeenheatedandtheappliedvoltageisturnedon).SlowcoolingofthepartsmaybeneededfordierencesinCTE.Theslowcoolingwillworkonlyifthetemperatureusedishighenoughthattheglassbeginstosoften.Inthiscase,astheglasscools,itdistortsitselfsuchthatitabsorbssomeofthestressesinthebond[ 47 ]. Forastrongbondtoform,itisnecessaryforthesurfacestobeasatandsmoothaspossible.Thisallowsfewerandsmallergapsbetweenthesurfaces.Typicallyasurfacenishofbetterthan50nmrmsisused. 69


Anodicbondingtypicallyoccursinroomair,althoughotherinertgassescouldbeusedaswell.Formetalsthatoxiderapidly,itmaybeusefulforbondingtotakeplaceinaninertgasastheformationofanoxidemaycausethesurfaceproletochange. UsinganaluminumcoatedpieceofULE,vanElpetal[ 46 ]wereabletoanodicallybondZerodurwithatensilestrengthof9MPa.Thepurposewastodevelopacomplicatedelectrostaticclampfortheuseinnextgenerationlithographymachines. Theinvestigationsintoanodicbondingforthisworkweredonetodeterminepotentialbondingmethodsinordertobondlowexpansionmetals,suchasSuperInvar,tootherglassesinordertomakeanopticalcavitytotestthedimensionalstabilityofthemetal.Usinganodicbondinginsteadofapasteorfritbondingwouldtestonlythedimensionalstabilityofthemetal,andnotthatofthemetal/bondsystem.SeveralinvestigationsintobondingSuperInvartoBK7glassweredone,butwithextremelypoorresults.ThematerialswerechosenbecauseoftheirsimilarCTE'sandthatBK7hasasimilarchemicalcompositiontothatofZerodurandshouldallowforasimilarcurrenttobepassedthroughit.TheSuperInvarpiecewas50mmx50mmx6.3mmandhadamirrornishonitwithaglobalatnessof/4.AvarietyofBK7pieceswereusedranginginthicknessfrom6mmto1mmandindiameterfrom25.4mmto6mm.Thepieceswereheatedinanovento300Candallowedtosoakforatleastanhour.Avoltagerangingfrom200to1000Vwasappliedandleftfortimesbetween5minutesandseveralhours.Afterthevoltagewasapplied,theovenwasturnedoandallowedtocooltoroomtemperature. 70


TheonlybondthatwasabletoformwithsignicantstrengthwasbetweenaBK7diskapproximately6mmindiameterand1mmthickandapieceofaluminumfoil.Thepieceswereheatedto300Candavoltageof1000Vwasappliedfor1hour.Aweightwasappliedonthealuminumfoilinordertokeepthesurfacesingoodcontactwitheachother.Thepieceswerecooledtoroomtemperaturebyturningtheovenoandallowingthetemperaturestoequilibrate.Althoughabonddidformandwasdiculttopullapartusingaforceinasheardirection,itcouldbepeeledoeasily. ThelackofresultsismostlikelyduetothedierenceinCTEbetweentheSuperInvarandBK7.AlthoughthereisasignicantdierenceintheCTEofaluminumandBK7,abondwasmostlikelyabletoformbecausethealuminumfoilisthinandabletostretchandcontractinordertohandlethestressesinducedfromcoolingthepieces.Becauseofthis,anodicbondingwasabandonedasawaytobondalowexpansionmetaltomirrorsinordertomakeanopticalcavity. Brazingjoinstwomaterialsthroughtheuseofallermaterial.Itisperformedatatemperaturehighenoughtomeltthellermaterial,butlowenoughastonotmeltthematerialsbeingjoined.Becauseofthis,brazingisidealforjointingdissimilarmetals.Ferrousandnon-ferrousmetalscaneasilybejoinedbybrazing,aswellasmetalsofdierentmeltingpoints.Thejointsarestronganddependantonthestrengthoftheller 71


48 ].Fillermaterialsarechosensuchthatthetensilestrengthofthejointissimilartothatofthematerialsbeingbrazed. Brazingcreatesametallurgicalbondbetweenthesurfacesofthetwomaterialsbeingjoinedandthellermaterial.Heatisappliedbroadlytothejointingmaterialstoraiseittoatemperaturegreaterthanthemeltingpointoftheller.Thelleristhenplacedincontactwiththesurfacesandisdrawncompletelythroughthejointviacapillaryaction.Incontrast,weldingusestemperatureshigherthanthemeltingpointofthetwomaterialsandfusesthemtogetherwithallermaterial.Thus,weldingtemperaturesstartatthemeltingpointsofthematerials.Thisintenseheatcancausewarpingordistortionsaroundthejoint,orstressesaroundtheweldarea[ 48 ].Often,theextremetemperaturesusedforweldingdonotmakeitsuitableforjoiningthinpiecestogether,whereasbrazingcanbedoneeasily. Agoodtofthematerialswithproperclearanceisessentialforbrazingduetothecapillaryactionneededtopullthellermaterialthroughthejoint.Typicalclearancesare25to125m.Theclearancewillalsohaveasignicantimpactonthestrengthofthejoint[ 48 ].Iftheclearanceistoosmall,thentherewillnotbeenoughforcetopulltheliquiedllerthrough.Iftheclearanceistoolarge,thenthejointstrengthmaybereducedtothatoftheller.Highlypolishedsurfacescanalsoimpedethellerow.Atypical\millnish"willprovideasurfaceroughnessallowingforthecreatingofcapillary\paths"forthellertoowthrough.Inaddition,whenbrazingmaterialswithdieringCTEs,theclearancemustbeadjustedfortheelevatedtemperaturesaccordingly.Forexample,considerbrazingabrassbushingintoasteelsleeve(asshowninFigure 3-2 ).BrasshasahigherCTEthansteelandifitistobetheinnerpartoftheassembly,alargerclearancethanifbothwerethesamematerialsshouldbeused.Conversely,ifthebrassnowbecomestheoutermemberandthesteeltheinner,thenlessclearanceshouldbeused. 72


ProperspacingforbrazingmaterialsofdieringCTE. Ofsimilarimportanceistheneedtocleanthesurfacesbeforebrazing.Capillaryactionwillworkonlywhenthesurfacesareclean.Forthisreasonitisessentialthatallsurfacestobebrazedareadequatelycleaned. Whenheatingthematerials,itispossiblethatanoxidelayerwillformonthesurface.Thisisnotidealforbrazing,andmayproduceaweakerbondsincetheoxidewillpreventthellerfromadequatelybondingtothesurface.Forthisreasonuxingofthepartsshouldbedone.Fluxisachemicalcompoundthatisappliedtothejointsurfacebeforebrazingthatpreventsoxideformation.Theuxwilldissolveandabsorbanyoxidesthatwillformduringtheheatingprocess,allowingthellertofreelyow.Althoughuxingthesurfacescanbeomittedinsomecases,eventhenitcanbeadvantageoustouseasmallamountofuxasthismayimprovethewettingactionoftheller[ 48 ]. Whenbrazing,bothmaterialsshouldbeheatedasuniformlyaspossiblesotheywillreachthebrazingtemperatureatthesametime.Whenheatingmaterialsofdierentthicknessesorthermalconductivities,attentionshouldbepaidtoapplytheheatasevenlyaspossible.Forthisreason,usinganovenisoftenthebestmethodtobrazedieringmaterials.Inmostcases,observingthecolorandviscosityoftheuxmightindicatethatthepiecesarebeingheatedappropriately.Whentheassemblyhasreachedthepropertemperature,thebrazingrodisthenplacedincontactwiththejointarea.Aportionof 73


Althoughtypicallyusedformetals,brazinghasbecomeanimportantjointingmethodforSiCpartsaswell.Itisoftendicult,orevenimpossible,tobuildsomeSiCstructuresinonepiece.AnexampleofthisistheHerschelSpaceTelescopeprimarymirror.Withadiameterof3.5m,thismirrorwasfartobigtobemadeinonesinglepiece.Toconstructthismirror,severalsmallersinteredsegmentswerebrazedtogether[ 49 ].ThellermaterialwasadopedsiliconcompoundwithaCTEthatcloselymatchedtheCTEoftheSiCpieces.ThebrazingprocessusedbyBoostectocreateHerschelprimaryproducedaverythin,uniformjoint.Thetypicalbrazethicknessislessthan50m,andinmostcaseslessthan10m.ProducingsuchathinjointresultedinajointstrengthcomparabletothatoftheSiCalone. InadditiontoproducingtheHerschelprimary,BoostecalsoconstructedtheSiCtorusfortheGAIAmissionusingtheirbrazingtechnique.Thetorusconsistedof17individualsegmentsthatwerebrazedintoonecoherentstructure.Thenalstructureisaquasi-octagonalstructurethatisthreemetersindiameterandwillsupportthetwoGAIAtelescopesandthefocalplaneassembly. 74


ConceptualdesignoftheGAIASiCtorus.PicturecourtesyofESA. one-partepoxiesiseasywithsuitableknowledgeoftwo-partepoxies.Advantagesoftwo-partepoxiesarethattheytypicallyhaveahighpeelandshearstrength,outstandingchemicalresistance,goodthermalpropertieswhichmakethemidealforcryogenicuse,lowcost,andgoodadhesion[ 50 ].Inadditiontobeingusedforbonding,epoxiesareoftenusedascoatingsbecauseoftheirexcellentchemicalandenvironmentalresistance. Atwopartepoxyadhesiveisatypeofthermosettingpolymer.AsdenedbyDaly,athermosetisamaterialthathasthepropertyofundergoingachemicalreactionbytheactionofheat,catalyst,ultravioletlight,orothermeansleadingtoarelativelyinfusiblestate[ 51 ].Two-partepoxiesconsistofaresinandahardener.Alsoknownasthecuringagent,thehardenerisrequiredtoconverttheresinintoaplastic-likesolid.Oncetheresinandhardenerarethoroughlymixed,theepoxywillbegintoset.Oftenwhenmixingthecomponents,anextremeamountofheatmaybereleased.Forthisreasoncaution 75


3-1 listsafewepoxieswithselectproperties. Table3-1. Comparisonofpropertiesforselecttwo-partepoxies. ProductApplicationWorkLifetimeShearStrengthTotalMassLoss(min)(MPaat20C)(%) 100B/AGlass-to-metal510.35Mil-BondGlass-to-metal30{1A-12Glass-to-metal18017.20.9Eccobond285/11Widetemp.range24014.50.3Hysol1CVacuumsealing30{0.8Epo-Tek375Fiberoptics30015.9{ Percenttotalmasslossisameasureofthemasslossafteraspeciedperiodoftimewhenexposedtoelevatedtemperature.Forspace-basedapplications,lessthan1%isdesirable.Somevendors,suchasMasterBondInc.,havespecialepoxiesthatareNASAapprovedforlowoutgassingandpercenttotalmassloss[ 52 ]. Manyvendorsarecapableoftakingaspecicepoxyandmodifyingitslightlytoproducethedesiredmechanical,thermal,orchemicalproperties.Forexample,smallaluminumbeadscanbeaddedtolowertheCTEofanepoxy.AlthoughthismaylowertheCTEoftheepoxy,itmaycausethebondingstrengthtochange[ 50 ]. Thereareseveralwaystoadheresurfacesusingepoxy.Thestrengthofthebondwillbedeterminednotonlybytheepoxyitself,butbythequalityofthesurfacesaswell.Forhighstrength,allpaint,oil,dust,etc..shouldberemoved.Often,aspecicsurfaceroughnessisneededforfullstrength.Althoughthevendorshouldbeconsultedtoensureproperapplicationoftheepoxy,afewcleaningmethodsaresuggestedbelowforafewcommonsurfacesasadaptedfromreference[ 50 ].Steeloraluminum: 76


Wipethesurfacesfromallcontaminantsusingasolventsuchasacetoneoralcohol. 2. Abradeusingcleannegritabrasives(180gritorner). 3. Wipeagainwithsolventstoremoveanycontaminants. 4. Applyathincoating(typicallylessthan12m)ofepoxyandcureaccordingtospecicepoxy.Slightlyelevatedtemperaturestendtoproduceastrongerbond. Plastics/rubber: 1. Wipewithisopropanol.Usingacetonemaydamagetheplasticorrubber. 2. Proceedaswithsteps2-4above. Glass: 1. Cleanusingacetone,orsolventwipe.Allowtimeforsurfacestocompletelydry. 2. Applyathincoating(typicallylessthan12m)ofepoxyandcureaccordingtospecicepoxy.Slightlyelevatedtemperaturestendtoproduceastrongerbond. Tobreaktheepoxybond,mechanical,chemical,orthermalmeansmaybeemployed,althoughthismaycausedamageordestructionofthepieces.Thesemethodsincludeusingapick,screwdriver,razorblade,heatingthebond,orsoakinginstrongsolventssuchasMEK. 77


Chapter 3 describesafewselectbondingtechniquesthathavebeenusedinavarietyofapplications.Unfortunately,thesetechniquessuerfromdrawbackswhichmaynotmakethemidealforsomespace-basedinterferometricmissions.Forexample,opticalcontactingworkswelloveralargetemperaturerangewhensmallareaswithsimilarCTE'sareused.AlthoughopticalcontactingcanbedonebetweenmaterialswithdieringCTE's,changingthetemperatureatwhichthematerialswerebondedwillinducestressesthebondmaynotbeabletohandle.Brazingessentiallymakesthebondedcomponentsasone,buttheelevatedtemperaturesneededmaycausethematerialpropertiestochange,ormayeectpossiblecoatingsappliedtothematerial(aswithatelescopeforexample).Achemicalbondingtechnique,colloquiallyknownashydroxidebonding,providesstrongbonds,iscapableofusingsmallersurfaceareasthatdonotneedtomeetthehighpolishingtolerancesofopticalcontacting,andcanbeusedforprecisionalignmentneededforopticalbenches[ 53 ].Thischapterwilldetailthehydroxidebondingmechanismandpresentstrengthresultsforseveralmaterialsthathavebeenbondedusingthehydroxidebondingtechnique. 54 ].Thecharacteristicsofhydroxidebondingmakeitidealforopticalbenchesduetothehighbondstrengthwhilestillallowingforprecisionalignment.AsdescribedbyGwo,thehydroxidebondingprocessiscapableofformingbondsaslongasasilicate-likenetworkcanbecreatedbetweenthesurfaces.ThemostobviousofmaterialsforuseofhydroxidebondingisZerodur.ExtensivemeasurementshavebeenmadeinvolvingthemechanicalandthermalpropertiesofhydroxidebondedZerodur,specicallyfortheLISAmission[ 53 ].Overthepastfew 78


1. Hydrationandetching:Inthisstep,ahydroxidesolution(typicallyNaOHorKOH)isaddedtooneofthesurfaceswherethefreeOHionsacttoliberatesilicateions. 2. Polymerization:Thesilicateionsdissociateandformsiloxanechains. 3. Dehydrationandbonding:Asthewaterevaporates,thesiloxanechainsintertwineandbondthe2surfacestogether. Althoughtheoverallprocessforhydroxidebondingisthesame,itvariesslightlydependingonthematerialsandbondingsolutionused.Inthecaseofglass-glassbonds,thehydrationandetchingstepiseasilydonesincesilicamaterialseasilyliberatesilicateionsinthepresenceofOH.Formetals,therearenosilicateionsforthesiloxanechainstoattachto,sothesecondstepmustoccurinadierentmanner[ 53 ].InthecaseofSiC,whichcontainssiliconbutisextremelychemicalresistant,themannerinwhichthesiloxanechainsattachtotheSiCsurfacemustbedierentthanthatofglass-glassbonds.Inthefollowingsubsectionsamorecompleteanalysisofthehydroxidebondingmechanismwillbediscussedforthecasesofglass-glass,glass-metal,metal-metal,andSiC-SiCbonds. SiO2+OH+2H2O!Si(OH)5:(4{1) Astheetchingtakesplace,thenumberofOHionsarereducedandthepHofthebondingsolutiondecreases.OncethepHisbelow11,thesilicateionsdissociatetoform 79


Si(OH)5!Si(OH)4+OH:(4{2) TheSi(OH)4moleculesthencombinetoformsiloxanechainsandwaterasdescribedby 2Si(OH)4!(OH)3SiOSi(OH)3+H2O:(4{3) Thesiloxanechainsstarttoformasthewaterevaporatesandprovidetheoverallrigidityofthebond[ 54 ].Asthedehydrationcontinues,a3Dnetworkofsiloxanechainsformandeventuallyjointhetwosurfacestogether. 54 ]. 55 ]thathydroxidebondingcantakeplacewithlimitedresultsbygrowinganSiO2layerontheSiCsurfacebyheatingtheSiCtotemperaturesinexcessof1000C.TheoxidizedSiClayerallowsthesiloxanechainstobondthetwosurfacesinthesamemannerthatglass-glassbondingoccurs. 80


InordertodeterminethebondingmechanismbetweenSiCandSiC,thechemicalcompositionsofthesurfaceswereanalyzedbothbeforeandafterbonding.ThiswaseasilydoneastheSiC-SiCbondalmostalwaysbrokeatthebondline.TodeterminethesurfacecompositionofthesiliconcarbidesurfacesduringSiC-SiCbonding,anX-RayPhotoelectronSpectroscopy(XPS)systemwasused.TheXPSsystemisbasedonthephotoelectriceectandistheleastdestructiveofalltheelectronorionspectroscopytechniques.TheXPSisroutinelyusedinindustryandresearchwhenelementalorchemicalstateanalysisofasurfaceneedstobedone.Forasolid,XPSprobesapproximately5-500Adeep,dependingonthematerial,theenergyofthex-raysandemittedphotoelectrons,andtheanglewithrespecttothesurfaceofthemeasurement. 81


whereKEisthekineticenergyofthephotoelectron,histheenergyofthex-ray,andBEisthebindingenergyoftheparticularatomconcerned.Thus,elementidenticationcanbedonebyanalyzingthephotoelectronenergydistribution.Themeasuredspectrumwillexhibitphotoelectronsignalsattributedtocore-level(O1s,C1s,Fe2p,etc...),valencelevel(Fe3s,Fe3p,etc),andAugeremission.Augeremissionoccursfromthellingofthecore-levelvacancybyanelectroninanouterlayerstate.Asthiselectronrelaxes,energyisemitted.Iftheenergyisabsorbedbyanotherelectron,thiselectronwillbeejectedfromthematerialanddetectedonthespectrometer.TheAugerpeaksaremuchbroaderthanthephotoelectronpeaksduetothemultiplepathstheelectronmayescapefromwithintheorbital. Thetightlyboundcore-levelelectronsofanatomhaveenergiesthatarenearlyindependentofthechemicalspeciesinwhichtheatomisboundbecausetheyarenotinvolvedinthebondingprocess.Thus,acompoundwillhavecore-levelenergiesthataresimilartotheelementalenergyvalues.Forexample,innickelcarbide,theC1sBEiswithinafeweVofitsvalueforelementalcarbon.Inaddition,theNi2pBEiswithinafeweVofitsnaturalvalueforNimetal.Identifyingthecore-levelBE'sofcertainelementswillprovidesignaturesoftheelementsinthecompound.Althoughcrosssectionvaluesareneededtodeterminerelativeatomicconcentrations,whencomparingrelativeintensitiesofthesameatomiccore-leveltogetcompositionaldata,thecrosssectionvaluesdonotneedtobetakenintoaccount.Thustherelativeabundancescanbetakendirectlyfromtherelativepeakheights.Forexample,gure 4-1 showsthesampleconstituentsaswellastheirrelativeabundances.Thesenumbersarecalculatedusingthecrosssectionsoftheelementsandtheirrelativepeakintensities.Figure 4-5 ontheotherhandonlyrepresents 82


Figure4-1. SpectrumofanunbondedCoorsTekSiCsample.ThepeaksfromlefttorightaretheAugercarbonpeak,Augeroxygenpeak,oxygen1speak,carbon1speak,andthesilicon2sand2ppeaks. Thex-raysourceoftheXPSwillproducephotoelectronstothedepththatitpenetratesthesample.Photoelectronsthatareejectedfromthetoplayersofthesamplewillbedetectedwithminimalscatteringandshowupaspeaks.Theelectronsthatarereleaseddeeperwithinthesamplewillinteractwiththesampleandmultiplescatteringwilltakeplace,reducingtheirchancestomakeittothedetector.TheseelectronsshowupatlowerKEandmostlyendupinthebackgroundaftertheXPSpeaks(Figure 4-2 ). Thus,peakscomeprimarilyfromthesurfacephotoelectrons,whilethebackgroundisaresultofthescatteredelectronswithinthematerial.Theintensityoftheuxof 83


Kineticenergydistributionobtainedduetotheinelasticscatteringofelectrons.Thepeakcorrespondstothephotoelectronsemittednearthesurface,whilethesignalatlowerkineticenergiesresultsfromthescatteringoftheelectronsdeeperwithinthesamplesurface. electronsasafunctionofdepthandelectronangleemissionisgivenby sin()0(4{5) whereI0istheuxofelectronsoriginatingatdepthD,IDistheuxemergingwithoutbeingscattered,D/sinisthedistancetraveledthroughthesolidatthatangle,istheangleofelectronemission,andistheinelasticmeanfreepath.ThevalueofdependsontheKEoftheelectronandthematerialthroughwhichittravels.Italsodetermineswhatthesurfacesensitiveofthemeasurementis.Thevaluesofcanbefoundinstandardtablesfrombooks[ 56 ].Todetectthesurfacecomposition,thegrazingemissionangle,,canbelowered,whiledetectingdeeperlayersrequiresalargerangle. 84


SamplesofHexoloySAfromSaintGobain,POCOSuperSiCfromPOCOgraphite,andUltraSiCfromCoorsTekwereanalyzedintheXPS.AnintegratedbackgroundwassubtractedfromtherawdataandGauss-LaguerrepeakswerettedusingAugerScansoftware.Oncetheenergyofthepeakswerefound,theywerematchedtotheNISTonlinedatabasetodeterminethechemicalcomposition.AlthoughsomeofthepeaksdidnotmatchupexactlywiththeNISTdatabase,allpeakswerewithin0.10-0.15eVofthedatabase,withthenextsuitablecandidatebeingsignicantlyfurtheraway.Fromthisinformationabondingmechanismwasformulatedconsistingofthefollowingsteps: 1. ThehydroxideionsinthesodiumsilicatesolutionbreaktheSiObondsontheSiCsurface.TheSiOispresentonthesurfacepossiblyduetoeitherthepolishingprocessorfromoxidationduringthemanufacturingprocess. 2. ThebrokenSiObondsthenattachtotheSiO2inthesodiumsilicatesolution. 3. Asthewaterinthesolutionevaporates,theaqueousSiO2formsasolidlayerthatbindsthetwoSiCsurfacestogether. TheendresultisthattwopiecesofSiCarebondedtogetherbyalayerofSiO2[ 57 ].ThiscanbeseenfromthesurfacecompositionoftheSiCbeforeandafterbonding.Figure 4-3 showstheO1sorbitalenergiesbothbeforeandafterbonding.Beforebonding,onlytheSiOpeakispresent.Afterbonding,anSiO/SiO2peakispresent[ 57 ].ThiswouldindicatethatabondbetweentheSiOandSiO2hasoccurred.Duetothedierentnumberofdetectedcountsbeforeandafterbonding,theguresshowthenormalizeddata.Whilethiswillnotshowtherelativeabundanceofcompoundsbothbeforeandafterbonding, 85


4-4 ).Inasimilarmanner,analyzingtheC1sorbital,wendthatonlySiCandgraphitearefoundbeforebonding,butanadditionalSiO2+SiCpeakispresentafterbonding(Figure 4-5 ).Thissupportsthebondingprocesspresentedabove. Figure4-3. O1sorbitalbothbefore(bluecurve)andafter(redcurve)bonding.ThedashedanddottedlinesrepresentthepeaktvalueswhilethesolidlinesaretherawdatatakenfromtheXPS.Theunbondedpeakisat532.61eVandthebondedpeakisat532.96ev. UsingtheXPStoprobedeeperintothesamples,itwasfoundthatasthedepthincreased,theeectsoftheSiO2diminished,andthechemicalcompositionsbecameidentical.ThismechanismalsoexplainsthepoorbondingresultswhenanNaOHsolutionisused.SinceSiCisextremelyresistanttobothacidsandbasesaswellasbeinganextremelypooroxidizeratroomtemperature,thereisnointermediatematerialthatiseitherproducedorpresentduringbondingwhenanNaOHsolutionisused. 86


Si2p3orbitalbothbefore(bluecurve)andafter(redcurve)bonding.ThedashedanddottedlinesrepresentthepeaktvalueswhilethesolidlinesaretherawdatatakenfromtheXPS.Theunbondedpeaksareat102.87eVand100.60eV.Thebondedpeaksareat100.42eV,103.82eV,and102.21eV. 1. Preparationofthesurfaces:ThesurfaceswerecleanedasdescribedinAppendixB. 2. Applicationofsolution:Afterthesurfaceshavebeencleanedandsonicated,theyarethentakenoutofthesonicbathandcleanedagainwithanopticalwipeandmethanol.ThisstephelpstoremoveanyparticulatethatmaystillbeonthesurfacesaswellasremovingthesurfaceofanyDIwaterleftfromthesonicbath. 3. Jointingofthe2surfaces:Afterthesurfaceshavebeencleaned,onepieceisplacedonalevelingtableandthesolutionisthenapplied.Theotherpieceisthenplacedontop.Alevelingtableisusedtoavoidanyslippingofthebondingpiecesafterthebondingsolutionhasbeenused.Asmallweight(typicallyafewtoseveralounces)canthenbeplacedontopofthepieces,butthisisusuallyonlyneededforsurfaces 87


C1sorbitalbothbefore(bluecurve)andafter(redcurve)bonding.ThedashedanddottedlinesrepresentthepeaktvalueswhilethesolidlinesaretherawdatatakenfromtheXPS.Theunbondedpeaksareat282.79eV,and285.12eV.Thebondedpeaksareat285.48eV,283.22eV,and282.11eV. thatarenotveryat.Ifaweightisnotideal,thenpressurecanbeadministeredbypressingdownonthetoppiece.Thepiecesarethennotdisturbedforatleast18hours. Whileallpreparationandbondingwasdoneinacleanroom,anycleanarea,suchasaventedhood,willsuce.Acleanroomreducesthechanceofparticulatescontaminatingthesurfacewhichmaylowerthebondingstrength. 4-6 .Theleverarmisapproximately39cmlongandhasa5:1torqueadvantage. 88


Thewirewasthenwrappedaroundthebondedpieceandadownwardforcewasappliedontheotheruntilthetestpiecebroke.Theappliedforcewasmeasuredbyreadingthemaximumforceindicatorfromthespringscale.Itshouldbenotedthatsimilartoexperimentsinreference[ 53 ],byplacingthewireafewmillimetersfromthebond,aslighttorqueorpeelingeectwascreatedonthebondingareaperpendiculartothedirectionoftheshearstress.Afterextensivetestingusingthisapparatus,promisingresultswereobtained.However,thevariationofthemeasuredbondstrengthwashigherthanexpected,soamoreprecisemethodofmeasuringbreakingstrengthswasused.Thismethodusesamodiedcompressionsheartest(MCST)asdescribedinreference[ 58 ](Figure 4-6 ).Alinearactuatorwithan8mm/stravelspeedwasusedtoapplytheforcewhileaLabVIEWprogramsampledtheampliedoutputofanMLPseriesloadcellfromTransducerTechniquesat1kHz.Samplingatthisrateshowedaclearlydenedpeakandproducedanerrorintheappliedforcemeasurementsoflessthan1.5lbs.Ineveryinstance,therangeofthemeasurementswasgreaterthantheerrorproducedfromthemeasuringdevice.TocomparetheleverarmapparatusandMCST,eightBK7-BK7bondsweremadeusing0.50Lofsolutionandthenhalfwerebrokenineachdevice.Foreachapparatus,therangeofbreakingstrengthsweresimilar,buttheaverageusingtheMCSTwashigher.Thisismostlikelydueaslightcuttingintheglassfromthewireusedintheleverarmapparatusastheforceisbeingapplied,whichcausestheglasstobreakatlowerstrengths.InourresultsboththedataobtainedusingtheleverarmapparatusandMCSTareused. 89


Leverarmapparatususedtoinitiallytestthehydroxidebondedshearstrengths(left)andMCSTapparatususedforlatertests(right). mass10mg,andsubjectto35gaccelerationsduringlaunchconditionsresultsinarequirementthatthebondsbeabletowithstandontheorderoftensofkiloPascals.ResultsfromtheUniversityofGlasgowhaveshownhydroxidebondstobesignicantlystrongerthanwhatisrequired[ 53 ],althoughmoreresearchisneededtobetterunderstandtheprocessanditslimitations.InthissectionIelaborateonexistinghydroxidebondingresultstogainamorecompleteunderstandingofthisprocess. WhileZerodurorULEareusuallyusedforopticalcomponentsinspace-basedinterferometricmissions,BK7glasshasasimilarmolecularstructureandcompositiontothesematerialsandbondsinthesamemannerwithsimilarbreakingstrengths[ 45 ].Consequently,BK7providesuswithacost-eectivealternativetoZerodurorULEto 90


Thedehydrationofthewatermoleculesinthehydroxidesolutionplaysanimportantroleinthehydroxidebondingprocess.Asthewaterevaporatesormovesintothebulkofthematerial,thesiloxanechainsforma3Dnetworkandjointhetwosurfaces.Ifittakesappreciableamountoftimeforthebondstoreachtheirmaximumstrength,thenthebondsmustbelefttocureforthistimebeforetheirbreakingstrengthshouldbemeasured.Inaddition,ifthedehydrationprocessoccursoveranextendedperiodoftime,thenthismaycausethebondedpiecestoshiftslightlyandmaycauseproblemswhenusedforprecisionalignmentforusessuchasopticalbenches.Todeterminetheeectsofdehydrationtimeontheshearstrength,multiplesampleswerepreparedusing0.40L/cm2ofavolumetric1:4sodiumsilicatetodeionizedwatersolutionbetweentwoBK7piecesusingtheunpolishedside.Eachsamplehad1.27cm2ofbondingareaandstayedinthecleanroomapproximately18hours.Theshearstrengthofonebondedpiecewasthentestedusingtheleverarmapparatus.Theotherpieceswerekeptinanovenat50Cuntilbrokeninthesamemanner.Twobondedpieceswerealsoleftinthecleanroomfor11daystodetermineiftheelevatedtemperaturewithintheovenhadanyeectonthebondingstrength(seeFigure 4-7 ).Thetemperaturetheovenwassetatwasbasedonpreviouslybondedsamplesthathadbeenheatedtovarioustemperaturestodeterminewhattemperaturerangethehydroxidebondscouldsurvive[ 59 ].Mosteitherfailedorweakenedsignicantlyabove100C.Usinganoventemperatureof50Cwillexpeditethedehydrationprocess,butshouldn'tcausethebondtoweakenorfail. Itappearsthedehydrationtimehasnoappreciableeectonthebreakingstrengthwhenusingasodiumsilicatesolutionandthebondreachessignicantstrengthafterapproximately18hours[ 59 ].Afterbreakingthepiecesthatwereleftinthecleanroom 91


Figure4-7. BreakingstrengthsofBK7-BK7bondsafterbeingkeptinanovenat330K.Thezeroonthex-axisrepresentsthetimeafterthesampleswereinthecleanroomforapproximately18hours.Alldatapointsweretakenusingtheleverarmapparatus. Itwassuggestedinreference[ 53 ]thatthewaterinthehydroxidesolutionmaybeweakeningtheglassmaterial,causingthefracturetooccurintheglassandnotalongthebondinterface.Throughextensivetestingofmultiplesamples,itisbelievedthatthiseectismostlikelycausedbythesmallsurfaceareaofthewirethatmakescontactwiththeglass,whichinturn,createsamuchhigherlocalizedpressureatthewire-glassinterface,andnotoverthebondarea.Thisresultsinthewirecuttingorscoringtheglass, 92


Totestthis,multiplesamplesthathadbeenbrokenattheglass-glassinterfaceafterhavingtheirshearstrengthtestedhadtheirbulkshearstrengthtestedusingtheMCST.ThevaluesofthesedatapointswerecomparedtothebulkshearstrengthsofglasspiecesbrokenusingtheMCSTthathadnotbeenbonded.Bothbondedandunbondedsampleshadsimilarshearstrengths.Fromthesimilarbreakingstrengthvalues,itisassumedthatthewaterinthehydroxidesolutiondoesnotweakentheglassinanappreciablemanner,ifatall[ 59 ]. TheprimaryadvantagetheMCSThasovertheleverarmapparatusisthattheMCSTappliestheforceevenlyovertheglassrod,whichreduceslocalizedforcesandscratching.ComparingthebreakingstrengthsofthebondedBK7-BK7samplesusingtheMCSTandtheleverarm,itwasfoundthattheMCSTbreakingstrengthswerehigherthanthoseusingtheleverarm.Inaddition,itwasalsofoundthatallpiecestestedusingtheMCSTbrokeatthebond,anddidnotbreaktheglass.Thissupportstheassumptionthatthereasontheglasspieceswerebreakingattheglass-wireinterfacewhenusingtheleverarmdeviceisbecauseofacuttingeectproducedwhenusingthewire,andnotduetoaweakeningoftheglassfromthewaterusedinthebondingsolution. Itwasshowninreference[ 53 ]thatthemolarityofthehydroxidesolutionwasvaried,amaximumbreakingstrengthwasfound.Thisvaluewasdierenteachalkalimetalused.Anotherparameterthatcouldeectthebondingstrengthistheamountofsolutionusedpersquarecentimeter.Usingavolumetric1:4sodiumsilicatetodeionizedwatersolutionandvaryingtheamountsofsolutionusedfrom0.20Lto1.25Lwhilekeepingthebondareaconstantandusingthepolishedsurfacesoftheglassrods,multiplesamplesweremadeandkeptinacleanroomatleast18hoursbeforebeingtested.Thedataconsistedofthreebondedpiecesusing0.20Lofsolution,sevenpiecesusing0.50L,fourpiecesusing0.75L,vepiecesusing1.00L,andtwopiecesusing1.25L. 93


59 ].Inthiscase,itwasasingledatapointthatwaswellabovetheaverage.Fromthedataitwasfoundthattherewasnosignicantcorrelationbetweentheshearstrengthofthebondsandtheamountofsolutionused.Itshouldbenotedthatastheamountofsolutionusedwasincreased,moresolutionwouldspilloutofthesides.Unfortunatelytherewasnowayofdetermininghowmuchuidwasspillingout.Theaverageofthe21sampleswas3.2MPawitharangeof1.7to7.6MPa.Comparingtheaveragebreakingstrengthofthehydroxidebondstothoseobtainedinreference[ 45 ]byopticallycontactingsimilarmaterials,theaveragebreakingstrengthofthehydroxidebondsisapproximately2.5timesgreaterusingthehydroxidebondingtechniquewhencomparedtoopticalcontacting.Inaddition,thegreatestbreakingstrengthfoundinreference[ 45 ]wascomparabletothelowestbreakingstrengthwhenhydroxidebondingisused. 60 ][ 61 ].TodeterminetheshearstrengthforSiC-BK7hydroxidebonds,avolumetric1:4dilutionofsodiumsilicatetoDIwaterwasused.Theamountofsolutionwasvariedwhilekeepingthebondingareaconstantandusingthepolishedsideoftheglassrods.TheSiCtilesaremadeofHexoloySA(CTEof4106/C)are50.8mmoneachsideand6.4mmthickwitha/8globalatnessand60-40scratch-digonbothlargesides.TheSiCtileswerecleanedusingceriumoxideandsodiumbicarbonateasdescribedinAppendixB.Althoughthesurfaceswerecleanedusingceriumoxide,similarresultsshouldbeobtainedusingonlyasonicbathtocleantheSiC.Thedataconsistedofthreeglassrodsbondedusing0.12,0.35,0.60,0.75,1.00,and1.20Lofsolution,verodsat0.25L,andfourrodsat0.50Lofsolution.Thetypicaldeviationfromtheaverageforeachamountofsolutionusedwasbelow50%.As 94


62 ].ForacomparisonbetweenthebreakingstrengthofSiC-BK7hydroxidebondsandSiC-BK7opticallycontactedbonds,thebreakingstrengthofoneopticallycontactedSiC-BK7bondwasbrokenandfoundtohaveabreakingstrengthofalittleunder1MPa.MoredataisneedtodrawaconclusiononthecomparisonbetweenopticallycontactedandhydroxidebondsbetweenSiCandBK7.ItshouldbenotedthatopticallycontactingSiCtoBK7isdiculttodowithmuchsuccess. Figure4-8. Averagesofthedataobtainedusingtheleverarmandmodiedcompressionsheartest(MCST)devicesforbotha1:4and1:6dilutionfactor. 95


4-8 ).Twoglassrodseachbondedusing0.12,0.40,0.60,and1.30Lofsolution,sixrodsusing0.25L,andthreerodsat0.50,0.75,and1.00L.Thevariabilityinthemeasurementsstayedconsistentaswithwhatwasobtainedaboveandnosignicantcorrelationbetweentheamountofsolutionusedandtheshearstrengthcouldbeobserved[ 59 ].Theaverageofthe23datapointsis3.4MPa,witharangeof1.0to6.3MPa. 53 ].Totestthis,threeglassrodswerebondedusingtheroughsideoftheglasswith0.25Lof1:4sodiumsilicatesolution,tworodsat0.40,0.75,and1.00L,andsixrodswerebondedusing0.50Lofsolution.Thevariationwasagaincomparabletopreviousmeasurementswithanoverallmeanof3.0MPaandarangeof1.1to7.3MPa.ThisiscomparabletowhatwasachievedwhenusingthepolishedsideoftheBK7[ 59 ].Thefactthatagroundnishwillsucetobondthesematerialsandstillproduceasubstantialbondhasasignicantcost-reducingeect.ItisoftenexpensivetopolishglassorSiCatandsmoothenoughtoobtainastrongopticalcontact,especiallyasthesurfaceareaincreases.Usinghydroxidebondingallowsthefacesofthepiecestobebondedtohavesignicantlylessdemandingpolishingtolerances,which,inturn,reducesthecostofthematerials. 96


63 ][ 64 ][ 65 ],ortheglassandmetalhavecloseCTE's.ThismakesitvirtuallyimpossibletobondbulkmaterialswithdieringCTE'sinanon-destructivemanner.Usingthehydroxidebondingtechniqueallowsmetalsrangingfromaluminumtocopper[ 66 ][ 67 ]toSuperInvartobebondedtoglasswhileretainingtheirthermalandmechanicalcharacteristics.TotesttheshearstrengthoftheSuperInvarBK7hydroxidebond,twobondsusing0.20,0.40,0.60,0.80,and1.00Lof1:4sodiumsilicatesolutionweremadeusingthepolishedsideoftheglassrods.TheSuperInvartiles(CTEof0.7106/C)are38.1mmoneachsideand6.4mmthickandpolishedtoa/4globalatnessand60-40scratch-digonbothlargesides.TheSuperInvartileswerecleanedusingceriumoxideasdescribedinAppendixB.AllbondsweretestedusingtheMCSTandnosignicantcorrelationbetweentheamountofsolutionusedandtheshearstrengthcouldbeobtained.Theaverageofthemeasurementsis5.8MPawitharangeof2.3to8.1MPa. Inadditiontothesetest,anecdotalexperimentsweredonetodeterminewhatthermalstressestheSuperInvar-BK7bondscouldendure.Inoneinstance,aBK7glasswindow25mmindiameterand6mmthickwasbondedtoaSuperInvartileandthenheatedto100Cat0.70C/min,leftfor1hour,andthenquenchedinicewater.Althoughtheglasswindowbroke,piecesoftheglasswerestillbondedtotheSuperInvar.Inanothercase,a12mmdiameterplano-convexlenswasbondedtoanotherSuperInvartileandheatedinthesamemanner.Afterbeingquenchedinicewater,novisibledegradationofeitherthebondorthematerialscouldbedetected.Thisremarkablepropertyneedstobestudiedingreaterdetailsinceitmaybepossiblethesiloxanechainsinthehydroxidebondmaybereducingthethermallyinducedstressesofthematerials. 97


68 ].TheSiCwasthenallowedtoreturntoroomtemperatureandthebondstrengthwastestedandfoundtobe6MPa. Anothersectionofthesametile10mmx10mmwascutandtheatfaceswerebondedusing1.0Lof1:4sodiumsilicatesolution.Thebondedpiecewasthenlefttocurefor10daysandwascutwithadiamondtilesawrotatingat3600rpm.Thecutwasmadeapproximately3mmfromtheoutsideandtookroughly45minutestoperform.NodegradationofthebondcouldbeseenandtheresultingcutSiCcouldnotbepriedapartusinghandstrength.Althoughthebondretainedsignicantstrengthaftercutting,itmaybepossiblethatprolongedexposuretowaterorothercuttinguidsmaycausethebondtofail.Thebondedpieceswerethendippedinliquidnitrogenforseveralsecondsandthentakenoutforafewseconds.Thiswasdonemultipletimes.Thebondcouldnotbebrokenusinghandstrengthafterthisprocess. 98


55 ]thatthesurfacesneedapeak-to-valleyatnessof60nm,itwasfoundthatitispossibletobondwithsurfacesasroughas24movera2mmlength.Anunpolisheddisc3mmthickand50.8mmindiameterofSuperSiCwasobtainedfromPOCOGraphite.Thesurfaceroughnesswasfoundtobeaslargeas24mover2mm(gure 4-9 ).A10mmx10mmsectionwascutfromthispieceandbondedusing3.0Lofsolution.Theresultingshearstrengthwasfoundtobe2.45MPa[ 68 ]. ThesmallerSiCpieceswerebondedtothelargerSiCtilesusingdieringamountsofsodiumsilicatesolutionandtheshearstrengthsweremeasured.Twopieceswerebondedusing1.00Lofsolutionandhadbreakingstrengthsof7.12MPaand4.00MPa.Twopieceswerebondedusing0.50Lofsolutionandhadbreakingstrengthsof5.78MPaand Figure4-9. SurfaceprolesofareasonthePOCOSuperSiCusingaTencorsurfaceproler.ImagecourtesyofPetarArsenovicatGoddardSpaceFlightCenter. 99

PAGE 100

SurfaceproleofoneofthelargeCoorsTekSiCtileusingaTencorsurfaceproler.ImagecourtesyofPetarArsenovicatGoddardSpaceFlightCenter. Figure4-11. SurfaceproleofoneofthesmallSiCpiecesusingaTencorsurfaceproler.Thelargespikearound750mismostlikelyaspeckofdirt.ImagecourtesyofPetarArsenovicatGoddardSpaceFlightCenter. 3.25MPa.Onepiecewasbondedusing0.25Lofsolutionandhadabreakingstrengthof2.22MPa[ 68 ]. Itwaspreviouslyshownthatbyusingthistestingmethodoftheshearstrengths,thetypicalvarianceinbreakingstrengthswasapproximately35%.Thevariationinour 100

PAGE 101

AnotherpieceofSiCwasbondedusing1.50Lofsolution.Thebondedpieceswerecuredovernightandthencooledto77Kina2-stepprocess.Intherststep,theSiCwasplacedonanaluminumblockthatsatinasmallreservoir.Thereservoirwaslledwithliquidnitrogenwhichslowlycooledthealuminumblock.Asthetemperatureofthealuminumblockwaslowered,thetemperatureofthebondedSiCpiecefellaswell.Oncetheboilingrateoftheliquidnitrogenappearedtobeconstant,thebondedSiCpiecewastakenothealuminumblockanddippedafewtimesintotheliquidnitrogenforafewseconds.Thebondshowednonoticeablesignsofdegradation.AlthoughtheSiCpiecewasnottestedforitsshearstrengthafteritwarmedbacktoroomtemperature,itcouldstillnotbebrokenusinghandstrength,andisassumedtohavesimilarbreakingstrengthtothepiecestestedabove. Inadditiontocoolingthebondedpieces,thesamplethatwascooledwasalsoheatedtodeterminetheeectsofheatingonthebond.Thesamplewasheatedinanovenfromroomtemperatureto80Cat5C/minandallowedtosoakat80Cfor1hour.Afterthiswasdone,thebondcouldnotbebrokenusinghandstrength.Thebondwasthenheatedto100Cat5C/minandallowedtosoakat100Cfor30minutes.Thebondcouldnotbebrokenusinghandstrength.Thesamplewasthenfurtherheatedto130Cat5C/minandallowedtosoakfor30minutes.Thebondwasthenbrokenusingasignicantamount 101

PAGE 102


PAGE 103

Continuous-wave(cw)lasershaveplayedamajorroleinadvancingprecisionmeasurementsinmanyelds.Frequently,thelinewidthofsuchlasersisnotadequateformanyapplicationswithoutactivestabilizationofthelaserfrequency.Forexample,someatomicclocksrequireextremelystablelasersourcestoprobesub-Hertzlinewidthsavailableinlasercooledsamples[ 69 ].Interferometricgravitationalwavedetectors,suchasLIGOandLISA,requirelasersystemswithextremelylowfrequencyandamplitudenoise.Throughtheuseofanopticalcavityandfeedbackcontrolsystem,thestabilityofthecavitycanbetransferredtothelaserwhichcanthenbemeasuredagainstanotherstablelasersource.InthischapteraversatilefeedbackcontrolknownasPound-Drever-Hall(PDH)frequencystabilizationwillbediscussedforitsapplicationtomeasurethestabilityofselectmaterialsonthefemtometerlevel.Abriefdiscussionofopticalresonatorsispresentedaswellashowtheopticalcavitiesweremadeinordertotestthedimensionalstabilityofseveralmaterials.Lastly,therelativedimensionalstabilityofZerodur,SuperInvar,SiC,CFRP,andatunablePZT-actuatedcavityarepresented. 5-1 ,thelasereldsofanopticalcavityaredescribedby 103

PAGE 104

Figure5-1. Electriceldsinsideandoutsideofanopticalcavity. reectedtransferfunctions,Tt(f)andTr(f),canbefoundbysolvingEq. 5{1 intermsofEt/EinandEr/Ein,respectively.Doingthisresultsin Foralosslesscavity(r21+t21=1)thetransferfunctionofthereectedeldsimpliesto Eq. 5{3 willpresentresonantbehaviorwhenTr(f)isaminimumandoccurswhen 2(2f)L c=2n;nN:(5{4) Thisequationcanalsobewrittenas Thefrequencydierencebetweenneighboringresonantmodesisconstantandisknownasthefreespectralrange(FSR).Itisgivenby 2Lcn 104

PAGE 105

Plotofthereectedtransferfunctionforalosslesscavity30cmlongwithbothmirrorshavingreectivitiesofr=0.90.TheFSRis500MHz. InadditiontotheFSR,anotherimportantparameterofopticalcavitiesisthenesse,F.ThenesseisdenedastheratiooftheFSRtothelinewidth.Thiscanalsoberepresentedintermsofthereectivitiesortransmissivitiesby linewidth=p t1+t2:(5{7) Themirrorsusedfortheopticalcavitiesweremadefromfusedsilicaandwerequotedbythevendor(ATFilms)ashavingtransmissivitiesnear360ppm. 70 ].Acavitywithsphericalmirrorsisanexampleofaperiodicfocusingthatcanresultineitherstableorunstablebehavior.Usingwhatareknownasthegparameters,thestabilityof 105

PAGE 106

Theresonatorgparameters. theresonatorcanbedetermined.Eachmirrorisgivenag-valuedenedby Ri;(5{8) whereListhelengthofthecavity,andRiistheradiusofcurvatureoftheithmirror(i=1isthefrontmirrorandi=2isthebackmirror).Foracavitytobeastableresonator,thecondition 0g1g21(5{9) mustbemet.Thewaistspotsize,!0,oftheresonatorcanthenbewrittenintermsofthegparameters,namely s (g1+g22g1g2)2:(5{10) Thespotsizesattheendsoftheresonatoraregivenby r r 106

PAGE 107

5{11 reducesto r s Knowingonlythelengthofthecavityandtheradiusofcurvatureofthetwomirrorswilldeterminetheresonantmodeofthecavity.Fromthisdatathelaserbeamcanbemode-matchedusinglensestotthemodeofthecavity. 5-4 as Figure5-4. Geometryforanalyzingthemisalignmentofacavity. aguide,whereListhelengthofthecavity,1and2arethesmallangularrotationsofthetwoendmirrors,x1andx2arethesmallsidewaystranslationsofthenewopticalaxisatthepointwhereitinterceptstheendmirrors,andC1andC2arethecentersof 107

PAGE 108

5-4 andsomegeometry[ 71 ],itcanbefoundthatthemisalignmentsaregivenby x1=g2 1g1g2L2(5{13) x2=1 1g1g2L1+g1 Thefrontfacecanbeassumedtobethereference,so1canbeassumedaszero.Thisholdsaslongasthecenterholeislargeenoughforthebeamtoeasilypassthroughandtheperpendicularityofthefrontfacetothespacerisreasonable.Inthiscase,Eq. 5{14 reducesto x2=g1 Sinceallthefrontmirrorsareatforthecavitiesproducedinthiswork,g1=1,andEq. 5{15 thenreducestox2=L 72 ].Thethirdcavitywithhydroxidebondedmirrorswasconstructedbytheauthor.Table 5-1 speciesthelengthoftheZerodurspacer,theradiusofcurvatureforthemirrorsused,theFSR,linewidth,andnesseofall3cavities.Thebiggestdierence 108

PAGE 109

ParametersofthethreeZerodurcavitiesconstructedbyeitheropticallycontactingorhydroxidebondingthemirrors. CavityLR1R2FSRlinewidthFinesse Z1260mm10.5m576MHz50kHz11,500Z2225mm12.0m647MHz140kHz4,600Z3210mm11.0m715MHz60kHz11,700 betweenthethreecavitiesisthatZ2hasamuchhigherlinewidththantheothertwocavities.Thisismostlikelyfromthemirrorsthatwereusedtomakethecavities.BothZ1andZ3weremadeusingmirrorsthathadbeenmadeatWavePrecisionandthencoatedatATFilms.ThemirrorsonZ2wereoldermirrorsfromadierentvendorthathadpossiblecontaminationsofthecoatings.ThespecicsoftheconstructionofZ1andZ2canbefoundinreference[ 72 ]. ToconstructZ3,theZerodurspacerwasrstcleanedasdescribedinAppendix B .Itwasallowedtodryovernightinordertomakesurethesurfaceswerecompletelydry.Thecurvedmirrorwasthencleanedinthesamemanner.Afteritwascleaned,itwasallowedtodryforapproximatelyveminutes.Duringthistime,thesurfaceoftheZerodurspacerwascleanedusingacleanroomwipethathadbeenwettedwithmethanol.Bothsurfacestobebondedweretheninspectedforanycontaminatesontheirsurfaces.TheZerodurspacerwasthenplacedsuchthatthesurfacetobebondedwasfacingstraightup.1.90Lofhydroxidesolutionthathadbeendilutedbyafactorof1:128KOH:H2OwasthenpippettedontheZerodursurfaceandthemirrorwasthenplacedontop.Themirrorwasthenalignedbyeyetobeintheapproximatecenter.Thebondwaslefttocurefortwodays.Tobondtheatmirror,itwascleanedinthesamemannerasthecurvedmirror.TheZerodurspacerwasippedaroundandthesurfacetobebondedwascleanedwithacleanroomwipewettedwithmethanol.Thesamesolutionandamountwasusedtobondtheatmirrorasthecurvedmirror.Asmallamountofthehydroxidesolutionleakedintothecenterbutdidnotspreadenoughsuchthatitwouldbeinthepathofthebeamasitpassedthroughthecavity. 109

PAGE 110

TomaketheSuperInvarcavity,asimilarmethodwasusedwhenconstructingtheZerodurcavity.TheSuperInvarspacerwascleanedasdescribedinAppendix B .Itwasthenallowedtodryovernight.ThecurvedmirrorwascleanedinthesamemannerastheZerodurmirrorswere.1.70Lofasodiumsilicatesolutionthathadbeendilutedbyafactorof1:4volumetricallywithDIwaterwasused.Thebondcuredfortwodays.Theatmirrorwasthencleanedinthesamemannerandbondedusingthesameamountofsolutionusingthesolutionfromthepreviousbonding. ThecavityparametersfortheSuperInvarcavityaregivenintable 5-2 .Although Table5-2. ParametersoftheSuperInvarcavitywithhydroxidebondedmirrors. LR1R2FSRlinewidthFinesse 178mm11.0m843MHz167kHz5030 themeasurementforthelinewidthoftheSuperInvarcavitywasdoneinair,measurementsofthelinewidthforZ3weredonebothinairandinvacuumandwerewithin10%ofeachother.SimilarvaluesshouldholdfortheSuperInvarmeasurements. 110

PAGE 111

5-5 .Itconsistsofthreestrutsthatrepresentthestrutsthatwouldholdthesecondaryinplace.TheoverallstructureoftheSiCspaceristhatofatube250mminlengthand28.6mmindiameterwitha12.7mmholethroughthecenter.Thestrutsare6.35mmthickandspaced120apart.TheendswerepolishedatSurfaceFinishesCo., Figure5-5. CompletedSiCcavitymadeofHexoloySA. Incwithasurfaceatnessof/8,mirrornish,andendparallelismtobetterthan0.05mm.DuetotheextremehardnessofSiC,polishingtothesespecicationscanbedicult.Inparticular,theatnesscanbehardtoachieve.Althoughthepolishingwasstillkeptwithinthestatedspecications,asmalldipscouldbeseenonthepolishedendsandmadeopticalcontactingdicult(Figure 5-6 ).TheSiCspacerandmirrorswerecleanedinthesamemannerastheprevioustwocavities.Inadditiontobeingcleanedusingasonicbath,theSiCspacerwascleanedthoroughlyusingcleanroomwipeswettedwithmethanolbecauseoftheextremelydirtynatureofthestrutsduetothefabricationprocess.Afterthiswasdone,thespacerwassonicatedagain.Thepolishedfacesofthespacerweresignicantlycleaneranddidnotneedtheextracleaningusingthemethanol.Bothmirrorswereopticallycontactedusingacombinationofmethods2and3aspresentedinChapter 3.1 .Asmallamountofmethanolwasputonthesurfaceandthemirrorwasthenplacedontop.Themirrorwasthenrotatedinacircularfashionuntilthemirrorstuck.The 111

PAGE 112

WYKOimagesofthepolishedendsoftheSiCspacer.ImagesproducedbySurfaceFinishesCo.,Inc. mirrorwasthenpushedandrotateduntilitwasclosetocenteredonthespacer.Thecavitywasleftfortwodaystoallowallofthemethanoltoevaporate.Thelinewidthwasthenmeasuredinairandfoundtobe97kHz. Table5-3. ParametersoftheSiCcavitywithopticallycontactedmirrors. LR1R2FSRlinewidthFinesse 250mm11.0m600MHz97kHz6200 73 ].Thespacerconsistedofathinoutershellthatwasusedtoprovideadditionalsupporttoaninnertubewhosethicknessislargerthantheoutershell.Strutsconnectedtheoutershelltotheinnertube(seeFigure 5-8 ).Theoutershellhasadiameterof200mmandisafewmillimetersthick.Theinnertubehasanouterdiameterof45mmandaninnerdiameterof30mm.Thelengthofboththetubeandshellare230mm.Thestrutsconnectingtheshellandtubeareafewmillimetersthickandspanthelengthofthespacer.TheprocessowusedtocreatetheCFRPcavityisshowninFigure 5-7 112

PAGE 113

FabricationowchartoftheCFRPspacerattheUniversityofBirmingham. InadditiontoprovidingtheCFRPspacer,theUniversityofBirminghamgroupalsosentZerodurtubesthathadbeenpolishedononeface.OneatmirrorandonemirrorwithROCof1mwereopticallycontactedtotheZerodurtubes.TheseZerodur/mirrorplugswerethenepoxiedintothecenteroftheCFRPtube.Todothis,analignmentjigwasmadesuchthattheplugscouldbeslidinandoutofthetubewhilestillkeepingthealignment.Alaserwasthenalignedthroughthecavitysuchthatthecavityhadan80%visibility.Theplugsweretakenoutandepoxiedusingeccobond285epoxythatwasmixedwithcatalyst9andasmallamountofacetonetomakethemixturelessviscous.Eccobond285wasrecommendedbycolleaguesattheUniversityofBirminghamsinceitistheepoxyusedtocreatetheCFRPstructure.Theamountsofepoxy,catalyst9,andacetoneusedwere100g,3.5g,and2.0g,respectively.Thecontentsweremixedthoroughlyusingaglassrodinaglassbeakerforapproximately5minutes.Afterabout15minutesthepastebecomeslessviscous.After30minutestheepoxyhadaconsistencythatcouldbespreadontotheplugsandtheninserted. TheCFRPspacerhadbeenleftinatmosphereapproximately4monthswhileworkingonthealignmentjigandthealignmentprocess.Duringthistime,itisspeculatedthattheCFRPspacershouldhaveabsorbedasmallamountofwatervaporfromthesurroundings.Forthisreason,theCFRPspacerwasbakedinanovenat100Cfor7daystogetridofanyabsorbedwater.Afterthiswasdone,theplugswereepoxiedintotheCFRPspacer. 113

PAGE 114

CFRPspacer,Zerodur/mirrorplugs,andthealignmentjig(left).FinishedCFRPcavityafterepoxyingtheplugs(right). Thecurvedmirrorwasrstepoxiedintoplaceusingthealignmentjigandallowedtocurefor2days.Theatmirrorwasthenepoxiedandthelaserwasalignedsuchthatthecavityhadavisibilitycloseto80%.Thecavitywasthenleftfor1weekandwhenthevisibilitywascheckedagain,itwascloseto40%.Thelaserwasthenrealignedandavisibilityof80%wasachievedagain.Thechangeinvisibilityismostlikelyfromtheepoxysettlingandcausingthemirrorstomoveslightly. TheCFRPcavitywasthenleftinatmosphereanother3months.Itcouldnotbemovedintovacuumduetoanotherongoingexperiment.TheFinesseandlinewidthofthecavitywasmeasuredtobe2900and224kHz,respectively,inair.AftertheCFRPwasmovedintohighvacuumandallowedtostaythereforapproximately5months,theFinesseandlinewidthweremeasuredtobe4060and160kHz,respectively.Theerrorinbothmeasurementswas10%.ThecauseofthechangeinlinewidthandFinesseisunknown. Table5-4. ParametersfortheCFRPcavity. LR1R2FSRlinewidthFinesse 230mm11.0m650MHz225kHz2900 114

PAGE 115

Tousearm-locking,itwillbecomenecessaryforthepre-stabilizedlasertobetunedtofollowthechangingLISAarms.Thisrequiresacontinuoustuningofthepre-stabilizedlasersover20MHz.Thereareseveralmethodsavailableforthisincludingsidebandlockingandphaselockedlasers[ 74 ].Anotherpossiblemethodistouseatunablecavity.Bychangingthelengthofacavity,theresonantfrequencyofthecavitywillchangeaswell.ThiscanbedonebybondingaPZTactuatortoastablespacermaterialandapplyingavoltagetothePZT.Tomakethiscavity,aPZTringmadeofk-180materialfromPiezoTechnologieswasusedastheactuator.ThespecicPZTmaterialwaschosenduetoitslowCTEandhighd33coecient.ThePZTringhasa41mmO.D,12.7mmI.D,is3.2mmthick,andthewidefacesarecoveredwithasilvercoating.ThecapacitanceofthePZTis1.3nFandwasfoundusingamultimeter.ThespacermaterialwaschosenasZerodurforitshighstabilityandhasa25.4mmO.D,6.4mmI.D,andisapproximately51mmlong.TheZerodurspacerwaspolishedbyMindrumPrecisiontoa/10globalsurfaceatness,40/20S-D,and12.7mparallelism.ThesilvercoatedfacesallowhydroxidebondingtotakeplacebetweenboththeZerodurspacerandthemirror(seeFigure 5-9 ). Toconstructthecavity,theZerodurspacerwascleanedasdescribedinAppendix B .Itthendriedovernight.ThePZTwascleanedinthesamemannerastheZerodur.AcleanroomwipewettedwithmethanolwasusedtowipethesurfaceoftheZerodurto 115

PAGE 116

RepresentationofthePZT-actuatedcavity.Notethelengthsarenottoscale. bebonded.1.75Lofasodiumsilicatesolution(SigmaAldrich)thathadbeendilutedvolumetricallywithdeionizedwaterbyaratioof1:4wasappliedtotheZerodurtubeandthePZTmaterialwasplacedontopandallowedtocureovernight.Thenextdayaatmirrorwascleanedinthesamemannerandbondedusing1.75Lofthesamesodiumsilicatesolution.Thebondcuredfor2days.AmirrorwithaROCof1.0mwasthencleanedandopticallycontactedtotheZerodurspacerusing0.75Lofmethanol.KaptoncoatedwiresweresolderedtothePZTusingasilversolder. TheFinesseandlinewidthweremeasuredinairandfoundtobe6900and400kHz,respectively. Table5-5. ParametersforthePZTcavity. LR1R2FSRlinewidthFinesse 54mm11.0m2780MHz400kHz6900 116

PAGE 117

75 ].Asimilartechniqueusedextensivelyinatomicphysicsisknownasfrequency-modulation(FM)spectroscopyandisdiscussedinfurtherdetailinChapter 6 .Thefoundationsofbotharesimilarandbyunderstandingone,theotherisofteneasytounderstand. Figure5-10. DiagramofthePDHlockingtechnique. 5-10 .ThelaserbeamispassedthroughaFaradayisolatortoreduceback-reectionsfromcouplingintothelaser.Thebeamthenpassesthroughanelectro-opticmodulator(EOM)whichmodulatesthecarrierbyfrequencytoproducesidebandsonbothsidesofthecarrier.Thebeamthenpassesthroughapolarizingbeamsplitterand/4waveplate.Thesecomponentsactasanopticaldiodeforthereectedlightfromthecavity.Ifthesidebandsofthebeamarewelloutsidethecavityresonance,thentheywillbereectedfromthecavityinputmirrorwithnosignicantphaseshift.Whenthelasercarriermatchesthecavityresonance,thereectedlightfromthecavityatthecarrierfrequencywillbein 117

PAGE 118

Whenthelaserisslightlyoresonance,thereectedlightcarrierwillexperienceaphaseshift.Thesignofthephaseshiftisdeterminedfromthedeviationoresonance.Forexample,ifthecarrierfrequencyishigherthanresonanceandproducesapositivephaseshift,thenafrequencylowerthantheresonancewillproduceanegativephaseshift.Inthiscase,theheterodynebeatswillnolongercancelsincetheyarenot180outofphase.Figure 5-10 showshowtheamplitudeofthebeatisextractedfromthephotodiodesignal.ThephotodiodesignalismixedwiththelocaloscillatorthatdrivestheEOMtoproduceanerrorsignal.Thephaseshifter(typicallydieringlengthsofcoaxialcable)isusedtoensurethemixerinputshavethecorrectphase.Themixedsignalcanthenbefedintocontrolelectronicsthatproducesignalsthatarefedbacktothelaser. FromEq. 5{16 itbecomesapparentthattheEOMaddstwosidebandsinfrequencyspace,eachonelocated/2HzawayfromthecarrierfrequencyasshowninFigure 5-10 .The 118

PAGE 119

AfterthelaserhaspassedthroughtheEOM,itpassesthroughthepolarizingbeamsplitterand/4waveplate.Thebeamisthenincidentonthecavityandeitherreectedortransmittedthroughthecavity.Thereectedbeampassesthroughthe/4waveplateandbeamsplitteragainandisreectedontothephotodiode.Sincethereectedbeamisusedforactuatingthelaser,thereectedeldwillbeconsidered,andthetransmittedsignalwillbeignored.UsingthetransferfunctioninreectionasdenedbyEq. 5{3 ,where!=2f,thereectedelectriceldis Sincepowerismeasuredbyaphotodiodeandnottheelectriceld,Pref=jErefj2isthesignalofinterest.Aftersomealgebra,thiscanbeexpressedas Eq. 5{18 representsthreewavesofdierentfrequenciesbeingmixedtogether.The2termsrepresenttheinterferenceproducedfrommixingthetwosidebandswitheachother.TheremayalsobesomecontributionfromhigherordertermsthatwereneglectedwhenEq. 5{16 wasexpandedintermsofBesselfunctions.Thesemaymakeasignicantcontributiontothe2terms,butarenotconsideredhere[ 75 ].WhenusingaphotodiodetomeasureEq. 5{18 ,alltermswillshowup.Nearresonance,thersttwotermsinEq. 119

PAGE 120

arenegligiblesincemostofthepowerinthecarrierwillbetransmittedthroughthecavity,andthepowerinthesidebandsistypicallysmall.Thisleavesonlythesineandcosineterms. MixingthephotodiodesignalwiththedrivingsignalfromtheEOMisolatesthesineterm.Thus,theerrorsignalbecomes (1r1r2)2!L c: (1r2)2!L c=8p T!L c: T=c=2L Fromthis,itiseasytoseethattheerrorsignalislineararoundresonance.Figure 5-11 showstwoerrorsignalsforthesamecavitylengthandEOMmodulationfrequencies, 120

PAGE 121

Twoerrorsignalsproducedfromcavitiesofdierentnesse.Thegreencurveisforacavityoflength200mmandnesse500.Theredcurveisforacavityofthesamelengthbutnesseof5000.BothhaveEOMmodulationfrequenciesof50MHz. butdierentFinesse.Threeerrorsignalscanbeseen.Thecentererrorsignalisproducedfromthecarrierinterferingwiththesidebandswhilethesignalsoneithersidecorrespondstotheerrorsignalsproducedfromthesidebands.Figure 5-12 showsacloserviewofthecarrierbeatnotefromFigure 5-11 .ThenarrowercurvecorrespondstothecavitywithahigherFinesseasisexpectedfromEq. 5{19 121

PAGE 122

ErrorsignalsforonlythecarrierfrequencyfromFigure 5-11 .Thegreencurveisforacavitywithalowernesse. 5-13 ).Tank1usesvecylindricallayersofgold-coatedstainlesssteelshellswhichareseparatedbyMacorspacerstoprovidepassivethermalisolationfortwocavitiesthatareplacedinthemiddle.Tank2usesvelayersofaluminizedPET(basicallyathickMylar)separatedbyMacorspacerstoprovidepassivethermalisolationtoonecavitythatisplacedinthecenterofthethermalshields(seeAppendix A foracompleteanalysisofthealuminizedPETthermalshields).Inthismanner,thedimensionalstabilitycanbetestedundercommonanduncommonthermalconditions. Laser1islockedtoCavity1andLaser2islockedtoCavity2usingthePDHstabilizationmethod.Pick-osfromeachlaserarethensuperimposedoneachothertoproduceabeatnote.Thebeatnoteissentintoaphotodiodewherethefrequencyisreadby 122

PAGE 123

Electro-opticalsetupusedtotestthedimensionalstabilityofselectmaterials. afrequencycounter.Usingtherelationship =l l Laser3islockedtoCavity3inTank2usingthePDHmethod.Apick-ofromthislaserissuperimposedonthepick-ofromLaser2todeterminetherelativestabilitybetweenCavity2andCavity3. 123

PAGE 124

5-13 .Thebeatnotebetweenthelaserswascontinuouslymonitoredfor28daysusingafrequencycountersampledevery20seconds.TheresultantbeatnoteisshowninFigure 5-14 .Inadditiontothismeasurement, Figure5-14. Timeseriesofthebeatnotetakenover28days. frequencymeasurementsweretakenusingaphase-metersamplingat98kHzand15Hz.Todeterminethefrequencystability,thefrequencytimeseriesisbrokenintosmallerpieces.Thesepieceshavealinearregressionttothedatasetsthatissubtracted.Afouriertransformisdoneontheresidualsandtheresultantlinearspectraldensity(LSD)infrequency-spaceisgiven.Foradiscussionontheprocessthatwasused,seeAppendix C .TheresultantLSDforthefrequencycounterandphasemetermeasurementsisshown 124

PAGE 125

5-15 .Althoughmeasurementslessthan10Hz/p 72 ],typicalresultsarebetween20-100Hz/p Figure5-15. NoisespectrumoftheZerodurcavitieswithopticallycontactedmirrors.Measurementsweretakenusingthefrequencycounter,phasemetersamplingat15Hzand98kHz. LISAmission,theprestabilizationstepcouldbedonebylockingtheNd:YAGlaserstocavitiesmadeofZerodur.Therequiredstabilityforthisstepis[ 76 ] TherelativestabilityoftheZerodurcavitiesisshowninFigure 5-16 alongwiththeLISApre-stabilizationrequirement.TheresultsinFigure 5-16 weretakenfromadatarunthatlastedfor7daysandwassampledat1Hz.FromFigure 5-16 ,itcanbeseenthattheZerodurcavitieseasilymeettheLISApre-stabilizationrequirementintheentirefrequencyband. 125

PAGE 126

NoisespectrumoftheZerodurcavitieswithopticallycontactedmirrorsandtheLISApre-stabilizationrequirement.Theresultsarefromatimeseriestakenoversevendaysandsampledat1Hz. ThereareseveralnoisesourceswithintheLISAbandthatmaybeaectingtheachievedsensitivity.Amongthemarethetemperatureuctuationsofthecavityduetochangesinthelabtemperature,theRFAMduetotheEOM,andtheheatingofthemirrorsduetothelaserpowerstoredwithinthecavity.Noisefromthefeedbackcontrollersisalsoapotentialsource,althoughmeasurementsofthesenoisesourcesaretypicallyanorderofmagnitudeormorebelowthemeasurednoise.Anadditionalnoisesourceduetoaccelerationfromtheopticaltableisalsosignicant,althoughitisfoundoutsidetheLISAbandandistypicallyaveragedoutbythefrequencycounterwhensampledatalowenoughfrequency.Thepeakbetweenapproximately8-60HzshowninFigure 5-15 ismostlikelyduetothisaccelerationnoise[ 77 ]. Modelshaveshownthatfrequencynoisebelow1mHztendtobedominatedbytemperaturechangesinthelabpropagatingthroughthethermalshieldsintothecavitywhichcauseanexpansionandcontractionofthecavity.Thevelayersofgold-coatedstainlesssteelthermalshieldsprovideanexcellentlowpasslterfortemperature 126

PAGE 127

5-17 .ThetemperaturewasmonitoredusinganOM-PLTTtemperatureloggerfromOmegaEngineering,Inc.andprovidesatemperatureresolutionof0.10Fusingathermocoupleprobe.Fromthegure,itisclearthatasthetemperatureofthetankincreased,anotabledecreaseinthebeatnotecouldbeobserved.The3.5MHzbeatnotechangeoverthe28daysisalmostthelargestchangeinbeatnotefrequencyobserved.ThebeatnotebetweenthetwoZerodurcavitieshaschangedbylessthan5MHzover3years.Thisismostlikelyduetoseasonaltemperaturechangesinthelabandelectronicoseteectsinthefeedbackcontrols. Thelargetemperatureswingoverarelativelyshorttime(typicallythesetemperaturechangesoccuroversixmonths,notone)meansthetemperatureofthelabwasnotinasteadystate.AlthoughitisobviousfromFigure 5-17 thereisatemperaturedependanceonthebeatnote,byhighpasslteringthedataamoresteadystatetransientcanbeseen.Boththebeatnoteandtemperaturedatawerehighpasslteredusinga2-poleButterworthlterwhose3dBfrequencywassetat5106Hz.Theresultinghighpassltered(HPF)dataisshowninFigure 5-17 .Again,itisclearthereisadenitivecorrelationbetweenthechangingtemperatureofthetankandthechangeinbeatnotefrequency.Althoughthecavitiesareinthesamethermalenvironment,thisshowsthereissomenon-commontemperatureeectsduetothechangingtemperatureinthelab.Thismaybeduetothedierentlengthsandorientationofthecavities. 127

PAGE 128

PlotsoftheZerodurbeatnoteandtanktemperature(left)andplotsofthebeatnoteandtemperaturedataafterbeinghighpassltered(right). missionskeeppushingmeasurementlimits.ThefrequencystabilityofthehydroxidebondingtechniquewastestedbyhydroxidebondingmirrorstoaZerodurspacerasdescribedinSect. .ThiscavityreplacedZ2.Theresultantnoisespectrumisshowningure 5-18 [ 78 ]. Figure5-18. ComparisonofthenoisespectrumsoftheZerodurcavitieswithopticallycontactedandhydroxidebondedmirrors. 128

PAGE 129

5-18 comparesthenoisespectrumfromthecavitieswithopticallycontactedmirrorsandthatwhenoneofthecavitieshashydroxidebondedmirrors.Aboveapproximately3mHz,thenoisespectrumisthesame.Belowthisfrequencythenoisespectrumisslightlyhigher.Thismaybeduetodierenttemperatureuctuationswithinthelabwhenthemeasurementsweretaken,oritcouldbenoiseattributedtothehydroxidebonds.Unfortunately,thesenoisesourceswerenotstudiedatthetimethemeasurementshadbeentaken.Despitethis,itappearsthatthehydroxidebondingprocessaddslittleornoadditionalnoiseand,moreimportantly,thenoisespectrumisstillbelowtheLISApre-stabilizationrequirement. 62 ].Inaddition,usingthehydroxidebondingtechniqueonSiCopenupnewpossibilitiesforitsuserangingfromtelescopesupportstructurestoopticalbenches.TotestthestabilityofHexoloySA,acavitywasconstructedasdescribedinSect. andlockedtoLaser1inTank1.AtimeseriesofthebeatnotebetweenLaser1andLaser2takenoveralongweekendisshowninFigure 5-19 (left).Thetemperaturestabilityinthelabistypicallymuchbetteroverweekends,sothereisalackofanoscillatorynatureinthebeatnoteasiscommonlyseen.Forlongertimeperiods,orwhenthebeatnoteistakenduringtheweek,thebeatnotewouldoscillatewithanamplitudebetween80and120MHzoverseveraldaysasshowninFigure 5-19 (right). TheresultantnoisespectrumforthebeatnoteshownontheleftsideofFigure 5-19 isshowninFigure 5-20 andmeetsboththeLISAtelescopeandpre-stabilizationrequirements.Thespikeinnoisearound0.20Hzismostlikelyfromhavingtoomuchgaininthetemperaturefeedbackcontrollertothelaserandisnotpresentinothermeasurements.Thenoiseabove0.01HzissimilartothatoftheZerodurcavities[ 79 ]. 129

PAGE 130

PlotsoftheSiCbeatnotetakenoveralongweekend(left)andplotsofthebeatnotetakenover11days(right). Thissuggestsalimitationofthemeasurementsystemisbeingreached.AlthoughthemirrorswereopticallycontactedtotheSiCspacer,itisassumedthathydroxidebondingthemirrorsshouldnotproduceanyadditionalnoisethatwouldbedetectable. Figure5-20. NoisespectrumoftheHexoloySASiCcavitywithopticallycontactedmirrorsalongwiththeLISApre-stabilizationandtelescoperequirements. 130

PAGE 131

5-21 .At1mHz,thetemperaturestabilityis3K/p Figure5-21. InferredtemperaturestabilityatthecenterofTank1. ThesurprisingresultisthattheSiCcavityperformedsowellathighfrequenciesandhadnoiselevelsthesameastheZerodurcavities.Temperaturestabilizationofthecavitycouldbringdownthelowfrequencynoise,whileusinglessinputpowerwouldlowerthenoiseinthe103to102frequencyband(aswillbediscussedintheCFRPcavitysection).Thus,itwouldappearthatSiCwouldbeaviablecandidatefortheLISAtelescopesupportstructure. 131

PAGE 132

65 ].Itshighstrength,lowCTE,andeaseofmachiningmakeitidealforsupportstructuressuchasthenarrowanglecameraontheCassinispacecraft.AlthoughthedimensionalstabilityofSuperInvarhasbeenmeasured,ithasnotbeenmeasuredatthefm/p TotestthedimensionalstabilityofSuperInvar,acavitywasconstructedasdescribedinSect. andplacedinTank2.Laser3waslockedtothecavityandthebeatnotebetweenLaser2andLaser3wasmonitored.A28daymeasurementwasmadeatthesametimetheZerodurmeasurementinSect. wasmade.ThetemperatureofthetankwasmeasuredatthesametimeusinganotherOM-PLTTtemperaturelogger.TheresultantbeatnoteandcorrespondingtanktemperaturesareshowninFigure 5-22 (left).ThecorrespondingdimensionalstabilityisshowninFigure 5-23 .NotonlydoestheSuperInvarmeettheLISAtelescoperequirements,butitmeetsthepre-stabilizationrequirementaswell,andissimilartowhatwasfoundwiththeSiCcavity.Above10mHz,thenoiselevelisalmostidenticaltothatobtainedfromtheZerodurcavities.Thiswouldalsoimplythatanexperimentalnoiseoorforoursystemisbeingmet. TheresultingnoisespectrumathighfrequenciesoftheZerodur-SuperInvarmeasurementagainshowsthatthereisnodetectableadditionofnoisefromusingthehydroxidebondingtechnique.Twoseparatesolutionswhichproducedtwodierentbondthicknessesbothshowednoadditionalnoise. FromFigure 5-22 ,itisapparentthatasthetemperatureofthetankchanges,sodoesthebeatnotefrequency.Tofurthertestthis,boththebeatnoteandtemperaturedatasetswerehighpasslteredasdescribedinSect. tolookatamoresteadystateofthedata.TheresultsareshowninFigure 5-22 (right).Itisevidentthebeatnoteisinuenced 132

PAGE 133

PlotsoftheSuperInvarbeatnoteandtanktemperature(left)andplotsofthebeatnoteandtemperaturedataafterbeinghighpassltered(right). bythetanktemperatureandthebeatnotechangelagsthetanktemperaturechangebyapproximatelyoneday. Figure5-23. NoisespectrumoftheSuperInvarcavitywithhydroxidebondedmirrorsalongwiththeLISAtelescopeandpre-stabilizationrequirement. KnowingthetemperaturestabilityinsideTank2willbecriticalinordertodeterminetemperatureeectsonthematerial.Thegold-coatedstainlesssteelthermalshieldsin 133

PAGE 134

A forafulldescription).Ifitisassumedthatthelimitingnoisesourceatlowfrequenciesisfromtemperatureuctuationsinthelabthatcoupleintolengthchangesofthecavity,thenanupperlimitonthetemperaturestabilityofTank2canbefound.Althoughthisassumptionwillbevalidatlowerfrequencies,athigherfrequenciesonlyanupperboundcanbedeterminedsinceitisassumedthetemperatureeectsaresignicantlylowerthanothernoisesourcessuchaselectronicoraccelerationnoise.TheestimatedtemperaturestabilityoftheinsideofTank2isshowninFigure 5-24 Figure5-24. InferredtemperaturestabilityinsideTank2. Itshouldbenotedthattomakethiscalculation,theCTEoftheSuperInvarcavityneedstobeknown.UnfortunatelytheCTEoftheSuperInvarisnotknownsinceitwasboughtbyapreviousgroupmemberandrecordsofwhereitwaspurchasedcannotbefound.Aftersearchingseveralsources,arangeofCTEvalueswasfound.Althoughthe 134

PAGE 135

AnestimationoftheCTEoftheSuperInvarcavitycanbedoneusingtheoutsidetanktemperatureandthebeatnotetimeseries.ToestimatetheCTEoftheSuperInvarcavity,measurementsofboththetanktemperatureandbeatnoteweresimultaneouslytakenforspansofseveraldays.Sincethethermalshieldswillhavelittleattenuationofthetemperaturechangesinthetankatlowfrequencies,ifthetanktemperatureisconstantlychanginginthesamedirectionoveralongperiodoftime,thenthebeatnotewillbecontinuouslychanginginasimilarmanner.Bycuttingotherstandlast1-2daysofdata,andlookingatthechangeinbeatnotecomparedtothechangeintanktemperature,theCTEoftheSuperInvarcanbefound.Oncethechangeinbeatnotefrequencyperdegreeisfound,theCTEcanbefoundvia FSR1 2L;(5{21) whereisthechangeinbeatnotefrequency,Tisthecorrespondingchangeintanktemperature,isthewavelengthofthelaser,FSRisthefreespectralrangeofthecavity,andListhelengthofthecavity.Doingthis,theCTEoftheSuperInvarisfoundtobeapproximately0.2-0.3ppm,whichiswithintheCTEvaluespreviouslystated.Thisis,infact,alowerboundontheCTE,sinceitisexpectedthatthereissomeattenuationatfrequenciesof105HzasshowninFigure A-4 andwouldproducealargerCTEvalue.Forthisreason,aCTEvalueof0.7106/KisanappropriateapproximationtotheCTEoftheSuperInvarcavity. FortheinferredtemperaturestabilityinsideTank2,aCTEof0.7106/Kwasassumed.ComparingthethermalstabilityofTank1andTank2,Tank2appearstohaveapproximatelyanorderofmagnitudeworsetemperaturenoiseatlowfrequencies.AlthoughthetemperaturestabilityofTank2isworsethanthatofTank1,thedierence 135

PAGE 136

80 ].ForLISA,aPZTcavitycouldbeimplementedforuseinarm-locking.Toutilizearm-locking,itwillbecomenecessaryforthepre-stabilizedlasertobetunedtofollowthechangingdistancesoftheLISAarms.Thisrequiresacontinuoustuningofthepre-stabilizedlasersover20MHz.DuetothelengthstabilitynecessaryforLISA,thePZTcavitymustbeextremelystableaswell. TotestthestabilityofthePZT-actuatedcavity,itwasplacedinTank2andLaser3waslockedtoit.ThestabilityofthetwoZerodurcavitieswithopticallycontactedmirrorswereputinTank1andtherelativestabilitybetweenthetwoweremeasuredwhilethePZTcavitymeasurementsweretaken.ThestabilityofthetwoZerodurcavitieswerefoundtobesimilartowhatwasmeasuredaboveandtypicallyshowedanoisespectrumof20to50Hz/p TodeterminethetuningcoecientofthePZT,thePZTcavitywaslockedandan8Vpeak-to-peaktrianglewaveat0.10HzwasappliedtothePZT.Themeasuredtuningcoecientwasfoundtobe1.5MHz/V[ 81 ].Thecurrentestimatesforthenecessarytuningfrequencyforarm-lockingis20MHzandisbasedonthevariationsintheLISAarms.Duetothedriftsofthecavityitself,whichareontheorderofHz/s,a 136

PAGE 137

CommerciallyavailableampliersrarelyspecifytheirperformanceintheLISAband.Forthisreason,alowcost(lessthan$100),lownoisealternativetoacommerciallyproducedamplierwasbuilt.ThedesignshowninFigure 5-25 isalmostidenticaltothedesigninreference[ 82 ],exceptfora60mHzlow-passlterthatisplacedontheoutputoftheamplier,andthe1000Vsourceisprovidedbyacommercialhigh-voltageDCsource.Theinputvoltageisprovidedbya12Vvoltagereferenceandvoltagedividerinwhichoneoftheresistorsisapotentiometer.Theinputvoltagecanbetunedthroughthepotentiometer. Figure5-25. VoltageamplierusedtoprovidetheappliedvoltagetothePZTonthePZTcavity. Whentheamplierneedstochargethepiezo,thechargerateisxedatthevalueofthecurrentsource,whichisdeterminedbyQ2andR2.SincetheQ2-R2valuealsodeterminesthequiescentcurrentoftheamplier,thequiescentcurrentmustbelow 137

PAGE 138

Whentheamplierneedstodischargethepiezo,theoutputoftheopampgoesupandQ1actsasasinkwhosecurrentislimitedbythesetvalueofR1.ThediodeprovidesapathforthedischargecurrentthroughQ1. Thegainissetat100throughtheR3andR4resistors.BothR3andC1mustsupporta1kVdropandareratedsuchthattheoperatingvoltageandwattagecansupportthis.R5isusedtoisolatethecapacitiveloadinadditiontoimprovingthestability.ResistorsR6andR7actasavoltagedividertomonitortheoutputvoltage. ThebeatnotebetweenLaser2andLaser3wasmeasuredusingaphasemetersamplingat98kHzand15Hz,aswellasusingafrequencycounterwhennovoltagewasappliedtothePZT.ThenoisespectrumisshowninFigure 5-26 .Thetwopeaksinthe10Hzregimearemostlikelyduetotheseismicnoiseandthedierentdampingmechanismsofthetwovacuumtanks.Thepeakaround20kHzismostlikelyfromtheresonantfrequencyofthePZT[ 83 ].Thebeatnotefrequencywasalsomeasuredusingafrequencycounterforseveraldays.Theresultantnoisespectrumsfor0Vand200VappliedtothePZTareshowninFigure 5-27 .Atypicalnoisespectrumforappliedvoltagesupto200VshowsthatthePZTcavitymeetsorexceedstheLISApre-stabilizationrequirementinallbutasmallfrequencyband.Forappliedvoltagesrangingfrom0-100V,thenoisespectrumsshowedsimilarlevels.Forappliedvoltagesbetweenapproximately100-200V,thenoiseincreasedslightly.Above200V,thenoiseincreasedastheappliedvoltageincreased.Thevisibilityofthecavityandpeak-to-troughvoltageoftheerrorsignalwerefoundtobesimilarupto300V.Adecreaseoflessthan10%wasfoundforanappliedvoltageof450V.ThismightbeduetotheshearingeectsofthePZT. AtypicalbeatnotebetweenthelaserlockedtotheZerodurcavityandthelaserlockedtothePZTcavitywilldriftat50-350Hz/sandisassumedtobefromthechanging 138

PAGE 139

NoisespectrumofthePZTcavitywithnoappliedvoltage.Measurementsweretakenusingthefrequencycounter,phasemetersamplingat15Hzand98kHz.Thespikearound20kHzismostlikelytheresonanceofthePZTring. temperaturewithinthelabthatcausethecavitytoexpandorcontract[ 81 ].Whenavoltagewasapplied,changed,orremovedfromthePZT,atemporalchangeinthebeatnotedriftcouldbeobserved.Theamountoftimeforthebeatnotetoreturntonormalincreasedasthevoltageincreased.Forexample,when10VwasappliedtothePZT,theinitialdriftofthebeatnoteincreasedtoapproximately2000Hz/s.Afterabout30minutes,thedriftinthebeatnotereturnedto100-400Hz/s.When200Vwasapplied,theinitialdriftincreasedtoapproximately18000Hz/s,andthetimeforthebeatnotetoreturntonormalincreasedtoseveralhours(Figure 5-28 ).Monitoringthevoltageoutputofthevoltageampliershowedthechangeinbeatnotefrequencycouldnotbeattributedtoanydriftinthevoltageamplier.TheslopeofthebeatnotedecaysexponentiallywhenavoltagewaschangedonthePZT,andisbelievedtobeattributedtotherelaxationphenomenonofthechargeddomainsslowlyreturningtotheirnalstateinthePZTmaterial[ 84 ].Thiscouldbeseensincethebeatnotewouldchangeinonedirectionasthe 139

PAGE 140

NoisespectrumofthePZTcavitywith0V,and200VappliedtothePZTalongwiththeLISApre-stabilizationrequirement. voltagewasapplied,andthenchangeintheotherdirectionafterapplyingthevoltage.Thisimpliesthattherelaxationprocessoccursintheoppositedirectionofwhenavoltageisapplied. Arm-lockingusesthestabilityoftheLISAarmstofurthersuppressthefrequencynoiseofapre-stabilizedlaser.Anerrorsignalcanbegeneratedfromapre-stabilizedlaseranditstranspondersignalfromthefarspacecraft.Thiserrorsignalisthenfedintoanarm-lockingcontroller(ALC)whichservesasatuningmechanismforthefrequencyofthelaser[ 74 ]. IthasalreadybeenexperimentallyveriedthatbyusinganALCthefrequencynoiseofapre-stabilizedlasercanbefurthersuppressedandthattheexperimentalresultsareincloseagreementtowhatistheoreticallyexpected[ 74 ]. Totestthis,thebeatsignalbetweentheLaser2andLaser3wasmeasuredwithaphasemeter.Thesignalwasdelayedby0.98sandDopplershiftedusingadelayunit.ThedelayedandDopplershiftedsignalisthenmixedwiththepromptsignaltoformthearm 140

PAGE 141

Theslopeofthebeatnoteovertimewith200VappliedtothePZT.Thetimeserieswastakenapproximately10minutesafterthevoltagewasappliedtothePZT. lockingerrorsignalupconvertedbytheDopplershiftsimilartotheupconvertedsignalinLISA.AsecondphasemeterisusedtodemodulatetheDoppler-shiftedsignalwiththeDopplerfrequency.Anydierentiallaserfrequencyuctuationswillchangethephasemeteroutput.Thissignalisappropriatelyltered,amplied,andappliedtothePZTtochangethelengthofthePZTcavityandstabilizethebeatfrequencyofthelaserstothedelayline.NotethatLaser2isonlyprovidingareferencetobeabletoelectronicallydelaythephasenoiseofthemainlaser. ThefrequencynoiseofthebeatsignalisshowninFigure 5-31 .Thefrequencynoisebetween104and102Hzis4to6ordersbelowtherequirementsandthenrollsupwithf2upto0.5Hz.Thepeaksatmultiplesoftheinversedelaytimearecausedbythefactthatthearmlockingsensorisinsensitiveatthesefrequencies[ 81 ].Usingthismethod,thecavitycanbestabilizedfromnoiselevelsthatarejustattheLISApre-stabilizationrequirement,tolevelsthatareseveralordersofmagnitudebelowtherequirement.Using 141

PAGE 142

ThemaindierenceinthissetupwhencomparedtowhatwillbeexperiencedonLISA,isthatthereisa0.98sdelayinsteadofthe32sdelaythatthelightwillexperiencebetweentheLISAspacecraft.However,LISAwillusealinearcombinationofthesignalsfrombotharms.Thislinearcombinationisinsensitivetolaserfrequencynoiseatmultiplesoftheinverseofthelighttraveltimedierence.Thesenullsintheresponsefunctionareatorabove1Hzformostofthetimeofthemission,whichisequaltotheresponsewehaveinoursystem. Figure5-29. Arm-lockingsetupusedtoarm-lockthePZTcavitytoaZerodurcavity.Thefeedbackcontrolsusedtoinitiallylockthelaserstothecavitiesarenotshown.PD:photodiode,PM:phasemeter,FM:frequencymeter AlthoughthenoisespectrumofthePZTcavityalmostmeetstheLISArequirementsinmostofLISAband,suddenjumpsinthetimeseriesofthebeatnotedooccur.Dependingonthesizeandfrequencyofthesedataglitches,theyareoftennotseeninthenoisespectrumduetotheaveragingofthetimeseriesbeforetheFFTisperformed.Theseglitchesmaycauseproblemsinkeepingthelaserproperlylockedifarm-lockingistobeused.AsuddenjumpinfrequencymayhavealargeenoughslopesuchthattheALCcontrollerisnotabletotrackthebeatnote,subsequentlycausingthelasertoloselock. 142

PAGE 143

Modeledopen-looptransferfunctionofthearm-lockingcontrollerusedtosuppressthelasernoise. Figure5-31. NoisespectrumofthePZTcavitywithoutanappliedvoltage,thenoisespectrumofthePZTcavityusingarm-locking,andtheLISApre-stabilizationrequirement. 143

PAGE 144

Thereare2distincttypesofdataglitchesthatoccurwithinthedata.Therstisthestep-glitch(Figure 5-32 ,top)andischaracterizedbyasuddenjumpinfrequencythatdoesnotreturntothenormalbeatnotefrequency.Thesecondtypeofglitchisthespike-glitch(Figure 5-32 ,bottom)andischaracterizedbyajumpinfrequencythatreturnstothenormalbeatnotefrequency. Figure5-32. Exampleofastepglitch(top)andspikeglitch(bottom). Threedierentcausesforthestep-glitcheswereinvestigated.Suddenchangesinthealignmentofthelaserbeamcausedbyhumanactivityinthelabisalwaysanissuebutcanusuallybeaccountedfor.Electronicdisturbancessuchassuddenchangesinanosetvoltagewithinthelaserfeedbacksystemproducesstep-glitches.Inaddition,internalrelaxationprocesseswithintheZerodurorPZTmaterialcannotberuledoutasacauseforstep-glitches.However,oncethefeedbacksystemwasoptimized,stepglitchesoccurred 144

PAGE 145

81 ]. Theoverwhelmingmajorityoftheglitchesthatoccurarespike-glitches.Thedetectionrateofspike-glitchescandependonthesamplingrate.Onlyspike-glitcheswheretheaveragefrequencyvariationduringasampletimeislargecomparedtothenormalrmsnoisecanbedetected.Consequently,shorterspikesrequirealargeramplitudetobedetectable.However,whenweincreasedthesamplingratefrom1Hzto8Hz,theratestayedessentiallythesame.Incontrasttothis,whenthesamplingratewasreducedto0.1Hz,mostofthespike-glitchesdisappearintothermsnoise.Theonlyidentiablesourceforthesespike-glitcheswasthePZTvoltage.At50V,thespike-glitchratewasaboutonceevery3-4hours.Atypicalamplitudeof1000Hzata1Hzsamplingratewastypicallymeasured,whiletheamplituderarelyexceeded4000Hzwhensampledat8Hz.Thescalingindicatesthattheamplitudeisprobablynotlowpasslteredbythesamplingrate. Althoughspike-glitcheswereshowntoincreaseastheappliedvoltageincreased,theywerenotabletobecompletelysuppressedwithnovoltageappliedtothePZT.Boththebeatnoteandthevoltageappliedtothefastactuatorofthelaserweresimultaneouslymeasuredandshowedthatspikesinthebeatnoteoftenshowedaspikeinthelaseractuatorvoltageaswell.Thesespike-glitchesaremostlikelycausedbyelectronicnoiseinthelaserfeedbackcontrollersand/orhumanactivityinthelab[ 81 ]. TotestthemechanicalfeasibilityofusingtheconstructedPZTcavityontheLISAmission,a\shaketest"wasdoneatGSFC.TomeasurethevibrationalsignatureofthePZTcavityanddetermineifthecavitywouldsurviveexpectedlaunchconditions,twoaccelerometerswereused.OnesensorwascementedtothePZTtosensemotionalongthecavityaxisonly.Theothersensordetectsmotioninallthreeaxesandwascementedtothetopofthemountingclampthatholdsthecavity.Tworoundsoftestingweredoneforatotalofsixtests.OneroundoftestingconsistedofvibratingthePZTcavitytransverse 145

PAGE 146

23 ].Totheauthorsknowledge,thereisnoinformationonthestabilityofCFRPmaterials,althoughitisclaimedbymanufacturersofspace-qualiedCFRPthattheoutgassingiswellunderstoodandcanbeaccountedforthroughtheuseofopticstoadjustthebeamasthestrutoutgasses.TotestboththedimensionalstabilityandoutgassingpropertiesofCFRP,theCFRPcavitydescribedabovewasplacedinTank2andLaser3waslockedtothecavity.AtimeseriesbetweenLaser2andLaser3takenover4daysisshowninFigure 5-33 [ 68 ].Typicalbeatnoteuctuationsrangefrom250-600MHz,buthavebeenmeasuredtobeover1GHz.Forthisreasonitisdiculttogetdatarunsthatlastseveralweeksbecausethebeatnotewilldriftoutofrangeofeitherthephotodiodeorthefrequencycounter. TypicaldriftsinthebeatnotefrequencyareafewkiloHertzpersecond.WhentheCFRPcavitywasinitiallyplacedinthevacuumtank,thetankwaspumpeddownusingaroughingpumpforapproximatelyoneday.Theturbopumpwasthenturnedonandleftonforseveralweeksbeforetheionpumpswereturnedon.TherstmeasurementsoftheCFRPbeatnoteweretakenapproximatelyonedayaftertheturbopumpwereturnedon 146

PAGE 147

BeatnoteoftheCFRPcavityinTank2takenover4days. andshowedadriftinthebeatnoteof12kHz/s.Thenextday,thedriftwasobservedtobe6kHz/s,andthefollowingdayitwasobservedtobe3kHz/s[ 68 ].Itthencontinuedtodriftatthisrateforseveraldays.ItisassumedthisdriftismostlikelyduetotheoutgassingoftheCFRPcavity. AlthoughtheCTEoftheCFRPcavityisnotknown,itisassumedtobelargerthanthatoftheSuperInvarcavitysincethebeatnoteoscillationsaresignicantlyhigherthanthoseoftheSuperInvarcavity.ComparisonofthetanktemperatureuctuationswhentheSuperInvarcavitywasmeasuredandwhentheCFRPcavitywasmeasuredshowtobewithinafactoroftwo.ThiscannotexplaintheorderofmagnitudelargerfrequencyuctuationoftheCFRPbeatnoteanditisassumedthelargeswingisduetoalargerCTEthanthatoftheSuperInvarcavity.Basedonthis,itisexpectedthattheCTEoftheCFRPshouldbeaboutanorderofmagnitudelargerthanthatoftheSuperInvarcavity. 147

PAGE 148

5{21 theresultantCTEwasfoundtobe1-3ppm.Asexpected,thisCTEisaboutanorderofmagnitudehigherthanthatoftheSuperInvar. Figure5-34. StabilityoftheCFRPcavityalongwiththeLISAtelescopeandpre-stabilizationrequirements. AlthoughtheCFRPcavitymeetstheLISAtelescoperequirementathigherfrequencies,thecausesofthehighnoiseatlowerfrequenciesisofinterestinordertodeterminewhatprocessesareresponsible.ItisknownfromthepreviouscavitiesthatthelowfrequencynoiseisduetothehighCTEoftheCFRPmaterial.Thefrequencythesetemperatureeectstypicallyoccuratarebelow1103Hz.Theadditionalnoisebetween103and102Hzwasofinterestanditwasspeculatedthatthemirrorcoatingscouldbethecause. Thehighreectivitiesofthecavitymirrorswillcausethelighttobouncebackandforthwithinthecavity,causingapowerbuild-up.Thiswillcausealocalheatingofthemirrorsandmirrorcoatingsthatwilltranslateintoalengthexpansion.Theeectoftheincidentpoweronthenoisespectrumcanbetestedbychangingtheinputpowerofthe 148

PAGE 149

5-35 .Itisclearthatbydecreasingtheincidentpower,thenoisespectrumdecreasesinthe104to102frequencyrange.Withalowenoughincidentpower,theCFRPcavitynoisespectrumwillmeettheLISAtelescoperequirement. Figure5-35. DimensionalstabilityoftheCFRPcavityfordieringincidentintensitiesalongwiththeLISAtelescoperequirement. Fromthenoisespectrumalone,itwouldappearthatCFRPwouldbeaviableoptionforuseontheLISAmission.Unfortunately,theoutgassingpropertiesandabsolutelengthstabilitystillneedtobestudiedinordertodetermineiftheymeettheLISArequirements. 149

PAGE 150


PAGE 151

Insomecases,itwillbenecessarytoknowtheabsolutestabilityofamaterialinadditiontoitsrelativestability.AnexampleofthisisthetelescopesupportstructurefortheLISAmission.Thematerialchosenmusthavearelativestabilityofbetterthan1pm/p 6-1 showsalaserbeamtransmittedthroughbothagascellcontaininganabsorbingmolecularvaporandalasertransmittedthroughaFabry-Perotcavity.Thetransmittedintensityfortheopticalcavitywillpeakaroundresonance,whilethetransmittedlightforthegascellwilldiparoundthemolecule'stransition.Derivinganerrorsignaltolockalasertotheresonantfrequencyofanopticalcavityhasalreadybeenpresentedandissimilartousingamoleculartransition[ 85 ].Inthecaseofamoleculartransition,thelinewidthofthetransitionissignicantlylargerthanthatofahighFinesseopticalcavity. 151

PAGE 152

85 ].Anotherwayofstatingthisisthatthefrequencymodulationonthelaserconvertstoanamplitudemodulationbytheabsorption. Figure6-1. Intensityasafunctionoffrequencyforbothanopticalcavityandgascell. Nearthefrequencyofmaximumabsorption,theconversionfromfrequencymodulationtoamplitudemodulationwillbesmall,andgoestozeroatthetransitionline'scenter.Afteritpassesthroughtheline'scenter,theamplitudeconversionincreases,butthephaserelationshiphasnowreversed.Ifthelaser'sfrequencyiswelloutsidetheabsorptionline,theamplitudemodulationgoestozeroduetothelackofabsorption[ 85 ].Thephenomenonisknowasfrequencymodulationspectroscopyandisextremelyusefulinlockingalasertothefrequencyofanabsorptionlineofagaseoussubstance[ 85 ]. Whenthebeampassesthroughthegascell,thelaserfrequencydoesnothavetobeinexactresonancewiththetransitiontoexperienceanabsorptionoflight.ThisisduetotheDopplereect.TheDopplerbroadeningofthetransitionlineoccursbecausetheatomsormoleculesareinmotion[ 85 ].Ifanatomormoleculeismovingsuchthatithasavelocity 152

PAGE 153

85 ] whereistheresonantfrequencyofthemovingmolecule,0istheresonantfrequencyofthemoleculewithnovelocity,cisthespeedoflight,andvjjistheparallelcomponentofthemovingparticlewithrespecttothelaserbeam.ItisclearfromEq. 6{1 thatasthegasisheatedup,theDopplerfrequencywillbelarger,resultinginabroadertransitionline[ 85 ]. Let'snowconsideraweakbeampassingthroughagascellthatisincidentonaphotodiodeandastrongbeampropagatingintheoppositedirectionastheweakbeamsuchthatthetwobeamsnearlyoverlap.Itwillbeshownthatthisresultsinsignicantlynarrowingthewidthsoftheabsorptionlines.Astheweakbeampassesthroughthegascell,itwillexperienceaDopplerbroadeningofthetransitionline.ThestrongbeamwillexperienceaDopplershift,althoughintheoppositedirection,andsobothbeamsinteractwithacompletelydierentgroupofatomsaslongasthelaserfrequencyisfarenoughoresonance.Ifthelaserfrequencyisnear0,thenbothbeamsarenowcompetingforthesameatoms.Thestrongbeamwillquicklydepletethenumberofatomswithvjj=0,leavingveryfewatomstointeractwiththeweakbeam[ 85 ].Thiscausestheweakbeamtopassthroughthecellwithlittlelossinintensity,resultinginaspikeinintensityatresonanceonthephotodetector.ThiseectcanbeseeninFigure 6-6 andcanbeusedtogenerateanerrorsignalthatthelasercanlockto.Inthismanner,anabsolutefrequencyreferencecanbeestablished. 86 ][ 87 ].Oftheseabsorbers,onlymolecularcesium(133Cs2)canbeinvestigatedbystandardsub-Dopplertechniques.Theotherabsorbershavelowcross-sectionsandrequire 153

PAGE 154

86 ].Manydicultiesassociatedwithlaserlockingtomolecularcesiumcanbecircumventedbyusingsecond-harmonicgenerationandlockingtosub-Dopplerlinesofmoleculariodine(122I2)asdiscussedinChapter 7 Themajorityofthedevelopmentoffrequencystandardsat1064nmhasbeendoneusingCs2[ 88 ].Cesiumspectroscopyinthe0.90to1.14mregimeoccurintheA1+X1+gbandandhavebeenreportedbyOrlovandUstyugov[ 86 ].IthasalsobeenshownbyMaketal.thatlaserfrequencystabilizationispossibleusingsub-Dopplerlinesofacesiumcellwhichwasheatedto220Candachievedastabilityof61011atameasurementtimeofonesecond[ 89 ].Investigationsintothesub-Dopplerspectroscopyofmolecularcesiumweredoneinadditiontousingtwodierentlaserlockingmethodsknownasmodulationtransferspectroscopyandfrequencymodulationspectroscopytoprovideafrequencyreferenceforuseindeterminingtheabsolutestabilityofamaterial. 6-2 .ThelaserpassesthroughaFaradayisolator,half-waveplate,andpolarizingbeamsplitter.Byrotatingthehalf-waveplate,thepowerinthepumpandprobebeamscanbeeasilychanged.Theprobebeamthenpassesthroughtheovenandcesiumcell.AdetaileddrawingoftheovenandcesiumcellareshowninFigure 6-3 Thecesiumcellis150mmlong,19mmindiameter,andhasBrewsterwindowsateachendtopreventback-reectionsfrominterferingwiththeincidentbeam.Thecesiumcellisheldinplacebytwocircularcoppersupportsateachendofthecell.Thesesupportsarethenheldinplacebystainlesssteelrods.Thesupportstructureisabletoslideinandoutofathincoppertube200mmlongand50mmindiameter.Thecoppertubeiswrappedinaheatingmeshwhosepoweriscontrolledbyatemperaturecontroller(model 154

PAGE 155

Experimentalsetupusedformodulationtransferspectroscopy. BriskONEfromBHThermalCorporation).AK-typethermocoupleisattachedtotheendoftheglasstubeusinghightemperatureadhesivetapeandisfedintothecontrollertoregulatethetemperature.Theheatingmeshiswrappedwithahightemperatureinsulator,andthisisthensurroundedbyadditionalinsulation.Thetubeandinsulationareenclosedinahightemperatureglasswithahightransmissionat1064nm.Thetemperaturestabilityofthecesiumcellisbetterthan0.10Cover10sasmeasuredbythethermocoupleandtemperaturecontroller. Aftertheprobebeampassesthroughthecesiumcellitpassesthrougha50/50powerbeamsplitterandisincidentonahigh-bandwidthphotodetector.Simplyusingtheprobebeaminthissetupwillallowspectroscopyofmolecularcesiumtobedone,althoughresultswithalowsignal-to-noiseratioweretypicallyachieved.Thisisduetothefactthatcesiumisapoorabsorberat1064nm,sotemperaturesinexcessof220Careneededinordertoseesignicantabsorptionwhichalsocausesadditionalnoise.Typicalcesiumlinesat240CareshowninFigures 6-4 and 6-5 .Todeterminethewavenumberofthedips,apick-ofromthelaserwassentintoawavemeterandthelocationofthedipswerematchedtolinenumbersinreferences[ 90 ]and[ 89 ]. 155

PAGE 156

Diagramoftheovenusedtoheatthecesiumcell. Thepumpbeamwasphasemodulatedat12MHzusinganEOMandthenshiftedinfrequencyby20MHzusinganacousto-opticmodulator(AOM)topreventinterferencebetweentheprobeandscatteredpumpbeams.Thepumpbeamhadadiameterof2.0mmwithapowerof5mWandtheprobebeamadiameterof1.3mmandpowerof1mWastheypassedthroughthecesiumcell.Afour-wavemixinginteractionofthepumpandoneofitssidesidebandsandtheprobebeaminducetwoneweldsinthedirectionoftheprobebeamatfrequenciesof!,where!istheprobecarrierfrequencyandisthemodulationfrequencyoftheEOM[ 90 ].Inthismanner,modulationtransferspectroscopyhasnoDoppler-inducedbaselineosets,sincethefour-wavemixingisnotinuencedbytheDopplerbackground[ 90 ].Forthisreasonadditionalmodulation,suchasusingachopper,isnotneededtoremovetheDopplerbackground. Asub-DopplererrorsignalwasobtainedinasimilarmannerasthatofthePDHsignaldescribedinsection 5.2 bymixingthephotodetectorsignalwiththeEOMmodulationfrequency.Duetothelowsignaltonoiseratioproducedfromthelackofhighabsorbingcesiumlines,asignicantamountofsignalconditioninghadtobedone 156

PAGE 157

Intensityproleasthelaserfrequencyisscannedovermultipleabsorptionlines. toproduceasuitableerrorsignal.TheACsignalfromthephotodiodewasampliedbyapproximatelytwoordersofmagnitudebeforeitwasmixedwiththefunctiongeneratorusedtopowertheEOM.Afterthephotodiodeandoscillatorsignalweremixed,theerrorsignalwaslteredusinga30Hzlowpasslterandampliedagain,typicallybyafactoroftentoftydependingonthecesiumlineused. Molecularcesiumisapoorabsorberat1064nmatroomtemperatureandrequiressignicantheatingbeforenoticeableabsorptionoccurs.Todeterminethetemperaturedependenceontheabsorptionpercentage,thecesiumcellwasheatedto200C,220C,and250Cwhilescanningoverthesameline.TheresultsareshowninFigure 6-6 andclearlyshowthatasthetemperatureincreases,theabsorptionpercentageincreasesaswell.Inadditiontodeterminingtheabsorptionpercentage,thelinewidthsofthetransitionscanbedeterminedaswell.ItwasfoundthatmosttransitionshadFWHMlinewidthsof300to500MHz. Fromgure 6-6 itwouldseemthattoobtainthebesterrorsignal,temperaturesinexcessof250Cshouldbeusedtoincreasetheabsorbtion.Unfortunatelyatthese 157

PAGE 158

Detailedintensityproleofthreeabsorptionlines. Figure6-6. Temperaturedependanceonabsorptionofmolecularcesium. temperatures,thecesiumreactsviolentlywiththeglasscellandwilloftendegradethecelltoapointwheremicro-cracksareinducedinthecellandairentersandreactswiththemolecularcesiumandbecomesuseless[ 90 ].Evenifthecesiumcellswerekeptattemperatureslessthan240C,theywouldoftenhavelifetimeslessthaneightmonths. 158

PAGE 159

Usingthemodulationtransferspectroscopytechniqueandscanningoverawiderangeoffrequencieswillproduceasetoferrorsignals.Inadditiontoprovidingasetofsuitablelockingpoints,theerrorsignalsalsoprovideinformationaboutthespectroscopyofcesium.Alargerpeak-to-peakvoltageimpliesthelineisabetterabsorber.Usingtheerrorsignalsalsomakesiteasiertoseewherethecesiumlinesare.ThisiseasilyseenbycomparingFigures 6-4 and 6-7 .ThedipsinFigure 6-4 areoftenhardtospotwithouthavingareferencetomatchthemupto.Alternatively,thelinesareeasily Figure6-7. Errorsignalsproducedforarangeofcesiumlines. determinedfromFigure 6-7 .Inthismannerthecesiumspectracanbeobtainedwith 159

PAGE 160

90 ]thatbetterresultswereobtainedusingFMspectroscopy.Forthisreason,bothmodulationtransferandFMspectroscopymethodswerestudiedforlaserlocking. TheexperimentalsetupusedforFMspectroscopyisshowninFigure 6-8 andissimilartothatofmodulationtransferspectroscopy.InthecaseofFMspectroscopy,thepumpisphasemodulatedandachopperisusedtoextractthesignalfromtheDopplerbackground.ThemodulatedprobebeamisdetectedonthephotodiodeandismixedwiththesignalgeneraterthatisfedintotheEOM.Thissignalisthenfedintoalock-inamplierwhosereferenceisprovidedbythechoppersource.Mostoftheamplicationandlowpasslteringwasdonebythelock-inamplierandonlyasmallamountofsignalconditioningwasneededtoproduceanerrorsignalwithsimilarpeak-to-peakandnoiselevelsasthoseobtainedusingmodulationtransferspectroscopy.ArangeofchopperfrequencieswerestudiedanditwasfoundthatusingchopperfrequenciesontheorderofhundredstoafewthousandsHertzproducedbettererrorsignalsthanusingslowerfrequencies. 5-13 .Inthismanner,theabsolutestabilityofthematerialLaser2islockedto 160

PAGE 161

Experimentalsetupusedforfrequencymodulatedspectroscopy. canbedetermined.Inthiscase,thematerialchosenwasaZerodurcavitywithopticallycontactedmirrors. Althoughitwasreportedinreference[ 90 ]thatbetterresultswereachievedwhenFMspectroscopywasused,theoppositewasfoundtobetrueformysetup.Theresultsforusingmodulationtransferspectroscopy,FMspectroscopy,andafree-runninglaserareshowninFigure 6-9 .Foralmostallfrequencies,usingmodulationtransferspectroscopywasbetterthanusingafree-runninglaser,butlockingalaserusingFMspectroscopywasworsethanusingafree-runninglaser.ThecauseoftheexcessivenoiseusingFMspectroscopyisunknownandseveralnoisesourceswereinvestigatedforbothexperimentalsetups.ThemostplausiblenoisesourcesinbothsetupsareRFAMnoiseproducedfromtheEOM,noiseproducedfromtheAOM,noiseproducedfromachangingchopperfrequency,andtemperaturechangesinthecesiumcell. 161

PAGE 162

Stabilityresultsofthemodulationtransferandfrequencymodulatedspectroscopytechniquescomparedtoafree-runninglaser. ismodulatedat.Asthebeamalignmentischanged,thepeakamplitudechangesaswell.Achangeintheinputpolarizationofthebeamwillalsocausetheresidualamplitudemodulation(RAM)tochange.Inaddition,ifthealignmentandpolarizationareleftconstant,theRAMwillchangeovertimeduetothelengthchangeoftheEOMcrystalwhichiscausedbytemperatureuctuationsofthecrystal.AllofthesesourceswillcauseachangingRAMwhichisthendetectedonthephotodiode. TodetermineiftheRFAMproducedfromtheEOMwasalimitingnoisesource,theincidentbeamwasmisalignedsuchthattheRFAMdetectedonaspectrumanalyzerwastentimesmorethannormal.Thenoisespectrumsbetweenthealignedandmisalignedbeamswerethencomparedandshowednodierence. 162

PAGE 163

TotesttheeectsofachangingAOMfrequency,thelaserwaslockedandthebeatnotewasmonitoredastheAOMfrequencywaschanged.Thiseectwasnoticeablebutsmall,anditwasconcludedthatthiswasnotasignicantenoughnoisesourcetobecausingthelowfrequencydriftinthebeatnote. Totesttheeectsofchangingthechopperfrequency,thelaserwaslockedandthebeatnotewasmonitoredasthechopperfrequencywaschanged.Theeectwasnegligibleanditwasconcludedthatthiswasanextremelysmallcontributiontothenoise. 6-9 isassumedtobefromnoiseproducedfromatom-moleculecollisionsinthecesiumcell.Thisisbecausetheatomicpressureisalmostthreeordersofmagnitudehigherthanthemolecularpressureat220C[ 90 ].Thus,operatingatlowertemperatureswouldbeideal,butthisbecomesdicultsincethe 163

PAGE 164

90 ],theyarenotdiscussedhere. Althoughthehighfrequencynoiseisassumedtobefromoperatingthecesiumcellatelevatedtemperatures,itwasfoundthatthelowfrequencynoisecouldnotbeattributedtodriftsinthetemperatureofthecesiumcell.Figure7inreference[ 90 ]showshowthebeatnotefrequencychangesasthetemperatureofthecellisincreased.Similartrendswereobtainedfrommyexperimentalsetup,withachangeinfrequencyperdegreeCelsiusfoundtobe250kHz/C.Theassumedtemperaturestabilityofthecesiumcellis0.10Cwhichwasdeterminedbymonitoringthefeedbackcontrollerusedfortemperaturestabilization.Fromthesetwonumbers,itwasdeterminedthatthelow-frequencynoisecouldnotbeattributedtothedriftsinthecesiumcelltemperature.Inaddition,ifthelowfrequencynoisewasduetothechangingcelltemperature,thebeatnoteshouldshowanoscillatorynaturethatwouldarisefromthecellheatingandcooling.Thiseectwasnotseen. Afterananalysisofthepotentialnoisesources,itwaspostulatedthatthelowfrequencynoiseseeninthenoisespectrumwasduetotheexpansionandcontractionoftheZerodurcavity,andnotfromaninherentnoisesourceinthesystem.Totestthis,thecesium-Laser2(orcesium-Zerodur)beatnotewasrecordedatthesametimetheLaser1-Laser2(orZerodur-Zerodur)beatnotewasrecordedandtheresultsareshowninFigure 6-10 .ThekinkintheZerodur-Zerodurbeatnoteisduetooneofthelaserscomingunlockedduringtherun.Thekinksinthecesium-Zerodurbeatnoteisunknown,butwerecommonlypresentinmanydataruns.Itispossiblethekinkindataisfromthecontrolelectronicslockingontoadierentpointduetothehighlevelofnoiseintheerrorsignal.Despitethekinksinbothdatasets,itisevidentthatbothbeatnotesaredriftinginthesamedirection.Acomparisonofother,shorter,datarunsshowsthatthecesium-Zerodurbeatnoteconsistentlydrifted1MHzforevery10kHztheZerodur-Zerodurbeatnote 164

PAGE 165

Beatnotesofboththecesium-ZerodurandZerodur-Zerodurlockedsystems. drifted.Theexactcauseofthisisunknown,butitmaypossiblybefromtemperaturedriftsinthecavity,orlaserintensitydriftsovertime. 165

PAGE 166

InorderfortheLISAmissiontoachieveitsgoalofdetectinglowfrequencygravitationalwaves,thedistancesbetweenproofmassesonadjacentspacecraftmustbemeasuredtopicometerprecision.Todothis,threedistancesmustbemeasured;Thedistancefromproofmasstoopticalbenchononespacecraft,thedistancefromtheopticalbenchononespacecrafttotheopticalbenchontheotherspacecraft,andthedistancefromtheproofmasstotheopticalbenchontheotherspacecraft.Inordertomakethesemeasurements,thelightonadjacentspacecraftmustbesenttoeachother.Thiswillbedoneviaatwo-mirrorreectingtelescope.Thelightontheopticalbenchwillbefocusedandsenttoasecondarymirrorwhereitwillbemagniedandreectedtotheprimarymirror.Theendresultwillbeahighlycollimatedbeamthatwillbesenttotheopticalbenchontheotherspacecraft.Throughtheuseofthesametelescopemechanismmountedonthesecondspacecraft,thelightwillberefocusedandusedonthesecondopticalbench.Inthismannerthenecessarymeasurementscanbetakentodeterminethedistancebetweenproofmassesonopposingspacecraft.Far-eldwavefrontdistortionscausedbymisalignmentofthemirrors,achangeinmirrorseparation,clippingfromthesecondarysupportstructure,orerrorsinpointingthetelescopewilldecreasethesensitivityofthelengthmeasurementsneeded.Insomecases,suchasthede-focusingofthemirrorsorbeampointing,iftheeectislargeenough,thenthemeasurementsmaynotbeabletobetaken,andthelinkwillessentiallybelost.ThismakesthetelescopeanditssupportstructurecriticalopticalcomponentsfortheLISAmission.DespitetheimportanceofthetelescopetotheLISAmission,adesignforeitheranon-oro-axistelescopehadyettobedesigned,fabricated,andtested.ThroughcollaborationwithseveralgroupmembersattheGoddardSpaceFlightCenter(GSFC)andtheUniversityofFlorida,anon-axistelescopewasdesigned,fabricated,andcurrentlyisintheprocessofdeterminingitsfeasibilityfortheLISAmission. 166

PAGE 167

7-1 .The Figure7-1. Representationof(a)anon-axisCassegraintelescopeand(b)ano-axisCassegraintelescope. choiceofmaterialusedforthesecondarysupportstructurewillhavedirectinuenceontheoveralltelescopedesign.Thein-bandlengthrequirementof1pm/p 18 ].Currently,thetwomostappealingmaterialsforthetaskareSiCandCFRPduetotheirhighstrength,extremestiness,andlowCTE.Bothmaterialscanbeimplementedinbothon-ando-axisdesigns,butCFRPistypicallyfavoredmorefortheo-axisdesign,andSiCisfavoredmorefortheon-axisdesign.ThecurrentbaselinefortheLISAmissionisano-axisdesignusingCFRPforanf/1telescope[ 23 ].Althoughtheviabilityofbothon-ando-axisdesignsusingboth 167

PAGE 168

91 ].Themagnitudeofthisnoisewilldependonseveralfactorsincludingtheintensityofthedetectedlaserlight,thebeamwidthattheoutputofthetelescope,andhowwellthebeamisfocused.Thequalityofthetelescopewilldirectlyeectthesefactorsandmustbetakenintoaccount. Asanexample,takeintoaccountthedefocusingofthetelescope.Thedistancebetweentheprimaryandsecondarymirrorsisacriticalcomponentofthetelescopedesign.Ifthedistancebetweentheprimaryandsecondarychangebytoomuch,thenthef/1telescopemaydefocusthelaserbytoomuch.Thefollowingisanbriefanalysisthatdescribestherequirementontheprimary-secondarydistancechange. Thetelescopedesigncallsforaatwavefrontoftheoutgoingbeamofsize!immediatelyaftertheprimary.Ifthewavefrontisnotatandhasasagittaofsizeh,thebeamappearstohavecomefromafocuswitharadiusgivenby[ 91 ] )2:(7{1) Ananalysishasshownthatasagittaofh=/2willdecreasetheapparentbeamradiusbyp 91 ].Thef/1designofthetelescoperesultsinastronglydivergingbeam.Attheprimary,thebeamisalreadyinthefareldregimeandtheradiusofcurvatureisequaltothedistancefromtheapparentfocus,whichisbehind 168

PAGE 169

whichattensthewavefront[ 91 ]. Ifweassumethatwemovetheprimarybyz,thenthiswillcauseachangeintheradiusofcurvatureofthewavefrontbythesameamountwhichintroducesamismatchtotheprimaryandtheresultingsagittaafterreectionattheprimarybecomes[ 91 ] R:(7{3) Forthecurrentdesign,estimatesofh/20,R=0.6m,and!=17.8cmareappropriate[ 91 ].Avalueof!=17.8cmistheGaussbeamradiuswhichoptimizestheintensityatthefarspacecraftfora40cmapertureGaussianbeam.SubstitutingthesenumbersintoEq. 7{3 resultsinanallowablelengthchangeof Thismeansthatinorderforthetelescopetomeetthewavefrontdistortionrequirementof/20,thedistancebetweentheprimaryandsecondarymirrorscannotchangebymorethan1.2moverthelifetimeofthemission.Othernoisesources,suchasthetiltofthereceivedwavefront,beamdivergence,andpointingstability,willplayacriticalroleintheachievedsensitivityofLISA[ 92 ]. Takingthesepotentialnoisesourcesintoaccountandcurrentmaterialfabricationlimitations,adesignforthetelescopewaschosenasshowninFigure 7-2 .Itconsistsofa400mmdiameterprimary,a135mmdiametersecondary,andfour5mm30mmstrutsthatconnectandseparatetheprimaryandthesecondaryby600mm.SiCwaschosentomaketheentirestructureduetoitshighstiness,lowCTE,andhighthermalconductivity.Thehighstinessandthermalconductivitywillreducetemperaturegradientsinthetelescopewhichwillreducetheeectsoftiltingandpointingerrors,while 169

PAGE 170

Isometricviewofthetelescopedesign.FigurecourtesyofJosephGenerieatGoddardSpaceFlightCenter. thelowCTEwillkeepthein-bandnoisebelowtherequired1pm/p Althoughathreestrutdesignseemslikeamorelogicalchoicethanfourstrutsfromamechanicalpointofviewsinceafourstrutsystemwillbeanover-constrainedsystem, 170

PAGE 171

Finally,theholesintheprimaryandsecondaryweredesignedtoallowalargeexibilityinmeasurements.TheuseofeitheraFabry-PerotcavityorMichelsoninterferometercaneasilybeimplementedorchangeddependingonthespecicmeasurementthatneedstobemade.TheouterholesareidealforMichelsoninterferometersthatwillmeasuretheinducedtiltofthetelescopeasitcoolsdown,whilethecenterholesareidealforusingaFabry-Perotcavitytodeterminethein-bandnoise. Oncethetelescopedesignhadbeennalized,thenexttaskwastodeterminewhattemperatureswouldbeexpectedonthetelescope.Thecurrentdesignforanon-axistelescopehasthesecondaryeectivelyseeingopenspacetemperatures,whiletheprimaryexperiencestheheatgeneratedfromtheopticalbench.AniteelementanalysisoftheLISAspacecraftwasdoneattheGoddardSpaceFlightInstituteanditwasfoundthattheexpectedoperatingtemperatureofthetelescopewas-71Candthattherewasanaxialgradientofabout1.5Candnegligibleradialgradient.ThelowtemperaturegradientisduetothehighthermalconductivityoftheSiC.Thisisidealsincetheuniformtemperaturewilldecreasethetiltingeectsofthetelescope.Inordertomeet 171

PAGE 172

Payloadassemblyforoneoftheopticalbenches,on-axistelescope,andthermalshieldforone\tube"oftheLISAspacecraft.Eachspacecraftwillconsistoftwotubes.FigurecourtesyofNASA. theparallelismtolerancesneededinthedesignofthetelescopesupportstructure,ahigh-stability,strongbondingtechniqueneededtobechosenthatalsoprovidedprecisionalignment.Thehydroxidebondingtechniquewaschosenduetoitsexibilityandeaseofuse,strongbondstrength,anditabilitytosurvivethermalcyclingfromroomtemperaturetoliquidnitrogentemperature. 172

PAGE 173

ExpectedoperatingtemperaturesontheLISAtelescope.FigurecourtesyofAngeliqueDavisatGoddardSpaceFlightCenter, tolerancesrequired.Brazingandusinganepoxywereconsideredasoptionstoassemblethetelescope,butwereeventuallydiscarded.Brazingandsinteringwouldrequireelevatedtemperaturesandcouldruinanyopticalcoatingsthatwouldbeonthenaltelescopedesign,andsoitwasconsideredanunrealisticoption.Usingepoxywasdiscardedasanoptionduetotheextremedicultyingettingtherightbondthicknessinordertokeeptheparallelismrequirementbetweentheprimaryandsecondarymirrors.Itwasdecidedthathydroxidebondingwouldbeusedbecauseitallowedforprecisionbonding,hassignicantbondingstrength,andcouldsurvivethermalcyclingbeyondtheoperatingtemperaturesofthetelescope.Theuseofhydroxidebondingalsorequiredthatsignicantdimensionaltolerancesbeplacedonthetelescopepartssincelargegapscannotbelledusingasodiumsilicatesolution. 173

PAGE 174

68 ].Estimatesofanupperlimitonthegapsizehydroxidebondingcouldllwereplacedas8mandwerebasedondatafromtheUniversityofGlasgow.AnanalysisbythegroupmembersatGSFCshowedthatifthefollowingpolishingtolerancesweremet,thentherewouldbegapssmallenoughthatthehydroxidebondingprocesswouldbeabletobridge: Primarydisktolerance: 1. Flatnessonfourstrutmountpads-1m 2. Flatnessonfoursmallmirrormountpads,andcentermirrormountsurface-5m 3. Flatnessonprimarydiscbackside-0.5mm 4. Surfacenishonfourstrutmountpads,fourmirrormountpads,andcentermirrormountsurface-0.1m Secondarydisktolerances: 1. Flatnessonbottomside-2micron 2. Surfacenishonbottomside-0.1m Struttolerances: 1. Flatnessonbottomsurface-1m 2. Parallelismbetweentopsurfaceandbottomsurface-2m 3. Perpendicularitybetweenbottomsurfaceandsidesurfacesandangularity-1mm(Thismeansthetopofthestrutcouldbe1mmoutoftoleranceonallfoursidesattopofstrut.The1mmtoleranceispeak-to-valley) 4. SurfaceFinishontopandbottomends-0.1m 5. Variationinlengthbetweenallfourstruts-2m 174

PAGE 175

Theprimary,secondary,andstrutswereallfabricatedandpolishedattheCoorsTekfacilityandshippedtoGSFCwherethetoleranceswerechecked.Itwasfoundthattherequirementshadbeenmet.AnalignmentjigwasdesignedatGSFCandbuiltbyanoutsidecompany(Figure 7-5 ).Thealignmentjigheldthestrutsinplaceandallowedthemtoslideupanddowninareproduciblemanner.Inthreeofthefourcasesthealignmentjigallowedthestrutstobeliftedandthenplacedbackontheprimarywithnochangeinthesurfacecontactbetweenthestrutandprimary.Intheothercase,asmallweight Figure7-5. Diagramofhowthetelescopepartstusingthealignmentjig.Forclarity,notallofthestrutclampsareshown.CourtesyofJeLivasatGoddardSpaceFlightCenter. 175

PAGE 176

Figure7-6. Pictureofworkstationtodeterminethegapsizesbetweenthestrutsandeitherprimaryorsecondarymirrors. Toproducethisalignment,aweightofapproximately5kgneededtobeplacedontopofthesecondarytomakesurethestrutsstayedushwiththeprimaryandsecondaryasmuchaspossible.Frompreviousbondingexperience,theadditionalweightshouldnotcauseaweakeningofthebond.Infact,severalSiC-BK7bondshavebeenproducedwithsignicantstrengthbyapplyingadownwardforceontheBK7rodwhenbonding.AllSiCpieceswerecleanedthoroughlybeforebonding.ThestrutshadbeencleanedusingmethanolandthestrutsendsweresonicatedinDIwaterfor30minutesbeforebeing 176

PAGE 177

Gapsizewhennotastrutisnotproperlyaligned(left)andgapsizewhenastrutisalignedproperly(right).Insomecasesadownwardforcewasneededinorderforthegaptoclose.Gapsshownontheleftsideofthegureweretypically10to20m. cleanedwithmethanolandcleanroomwipes.Theprimaryandsecondaryweretoobigtobeplacedinthesonicator,sotheywerecleanedusingmethanolandcleanroomwipes.ThebondingwasdoneinacleanroomthathadbeenusedafewweeksbeforetoassembletheLunarReconnaissanceOrbiter(LRO). Inordertobondthetelescopepiecestogether,allfourstrutswereliftedandapproximately3.0Lofsodiumsilicatesolutionthathadbeendilutedvolumetricallybya1:4ratiowasplacedunderthestrutoneatatime.Afterthesolutionhadbeenappliedthestrutwasthenloweredandkeptinplacebythealignmentjig.Onceallfourstrutswereloweredontotheprimary,thesamesodiumsilicatesolutionwasappliedtothetopofthestrutsandthesecondarywasplacedontopalongwiththeweight.Thebondscuredovernightinthecleanroomandwereanalyzedusingthesamemethoddescribedforthealignmentofthestruts.Analysisofthebondsshowedthatallgapshadbeenlledbythesodiumsilicatesolution,andinsomecasesagapof5mwasseentobelled. ThetelescopewasthentransportedbycarfromGSFCtotheUniversityofFloridawhereitwaskeptinalow-trac,low-dirtroom.Itstayedinthisroomwhilepreparation 177

PAGE 178

Pictureofthesodiumsilicatesolutionllingingapsbetweencontactpointsofthestruts. weremadetotestthetelescope.Itwaskeptinthisroomforseveralweekswithnovisibledegradationtothestructure.Withinamonth,itwasnoticedthatoneofthestrutshadcomelooseandwasleaningagainsttheoppositestrutbutwasfullyintact.Themostplausibleexplanationwasthatithadbeenbumpedbyworkersthathadbeenintheroomalthoughnoconcreteevidencecouldbefoundtosupportthis.Therewasalsonoevidencethatwouldsuggestthebondbrokeduetoanyotherreason.Thebondswerevisuallyinspectedanditwasdecidedthatthetelescopeshouldbecooledto-70Ctoseeifthedamagedstructurewouldsurvivethethermalcycling.AMichelsoninterferometerwassetupinthecenterofthetelescopetomonitorthedistancebetweentheprimaryandsecondarytodetermineifthebondsfailedasthetelescopeiscooled.BoththevacuumchamberandMichelsoninterferometeraredescribedindetailinsection 7.4 and 7.6 ,respectively.Temperaturesensorsontheprimaryandsecondaryindicatetheywerecooledto-70Cand-60Crespectivelybeforebeingbroughtbackuptoroomtemperature.Avisualanalysisofthetelescopeshowedthatitwasstillin-tact,butcollapsedafterapplyingasmallamountofforcetothesecondaryinthedirectionperpendiculartheopticalaxis.OneofthestrutsfellontothemountingstructurefortheMichelsoninterferometerand 178

PAGE 179

Togainaninsightastowhythestructurecollapsed,thehydroxidebondswereinvestigated.Itwasdeterminedthatthealignmentjigconstrainedthestrutswellintheintheradialplane,butnotitatorsionalmotion,andbecauseofthisthetelescopeexperiencedatwistingmotionasitwasbonded(Figure 7-9 ).Thiswasveriedbyavisualinspectionofthebondsusingalow-powermicroscopecamera.Usingthecamera,it Figure7-9. Representationofhowthestrutsbondedtotheprimaryandsecondaryatatilt.Theviewofthesecondaryisasifonewaslookingupatthesecondaryfromtheprimary.Theviewoftheprimaryisasifonewaslookingatitfromthesecondary.Theshadedareasarewherethehydroxidebondsarethicker. wasclearthebondwasthickerononesideofthestrutsthanitwasontheother.Whenlookingattheprimaryandsecondary,thethickerpartofthebondwasseenonoppositesides. Afteravisualinspectionofthebondsitwasthoughtthatanotherreasonthetelescopecollapsedcouldhavebeentothebondnotfullyformingoverthebondingsurface.Totestthis,apieceofthestrutthathadbrokenwascutandanalyzedusinganXPS.Onepiecewascutwhereitlookedlikethebondingoccurred,andonepiece 179

PAGE 180

Pictureofthesecondarywherethestruthadbondedtoit(left)andapictureofpartofastrutthathadbeenbonded(right).Thebondsappeartobemuchthickerononeside,whichleadstoevidencethatthestructuretwistedduringbonding. wascutwhereitappearedlittleornobondingoccurred.Ifnobondingtookplace,thenthereshouldbelittleornoSiO2peaksinthespectrum.AnalysisoftheXPSdataprovedinconclusiveevidenceforthistheory.WhilethesamplethatappearedtobondshowedahigherrelativeabundanceofSiO2thanthatofthepiecethatappearednottobond,thepiecethatappearednottobondshowedaspectrumsimilartothatofpreviouslypiecesthathadbonded.Themostlikelytheoryforthisisthatthebondthicknessofthesamplethatappearednottobondisverythin,andsoappearstonothavebondedsincethebondingagentisglass.Inthesamplethatappearedtohavebonded,thebondismostlikelysignicantlythickerandiseasiertosee.Thus,itwouldappearthatthemajorityofthesurfacedidbond,butthestrutbondedatatiltwhichproducedathickerbondononesideofthestrutthanitdidontheother[ 68 ]. Aftertheanalysisofthebondsweredone,itwasdecidedthatthetelescopepartswouldberepolishedandrebonded.TheprimaryandsecondaryweresentbacktoberepolishedattheCoorsTekfacility.Whentheoriginaltelescopepiecesweremade,copiesweremadebyCoorsTekincasesomethinghappenstotheoriginal.Sinceoneofthestruts 180

PAGE 181

Sincenoconclusiveevidencecouldbefoundastowhyoneofthestrutscameloose,acombinationofhydroxidebondingand\sister-block"bondingwillbeused.Theideabehindsister-blockbondingisthatasmallrectangularpieceofSiC(roughly3mm4mm20mm)willbeepoxiedtobothsidesofthestrutwherethestrutmeetstheprimaryandsecondary.Usingthesister-blockbondingwillprovideextrastrengthinadditiontothehydroxidebonds.Afterextensiveresearchwasdoneintodierenttypesofepoxies,itwasdecidedthatEP21TCHT-1fromMasterBondwouldbeusedforthesister-blockbonding.Thischoicewasprimarilyduetothefastcuretimeatroomtemperature,lowoutgassing,lowCTE,highstrength,andabilitytobeusedatcryogenictemperatures.AsdiscussedinChapter 3 ,foranepoxytoformappreciablestrength,aroughsurfaceandelevatedtemperaturesareusuallyrequiredtoformthebestbond.Toassemblethetelescopeneitheroftheseconditionswouldbeideal.Thesurfacerecommendedwouldbetooroughtokeeptherequiredparallelismofthetelescope,andheatingthetelescopetoomuchmaycausethehydroxidebondstofail.Todeterminewhatbondstrengthwouldbeobtainedusingthesister-blockbonding,astrutwasbondedtoatilewiththesamepolishasthetelescopeusingthesisterblocksandepoxy(Figure 7-11 ).Thestrutusedforthistestwascutusingatilesawfromthestrutthathadbrokenafterthermallycyclingthetelescope.Theuseofthetilesawmadethestrutsuchthatitdidn'tsitushwiththetile,andasmallgapcouldbeseen. Thesisterblockswerebondedandallowedtocurefortwodaysatroomtemperature.Theshearstrengthofthesister-blockbondingwastestedusingtheMCSTandfound 181

PAGE 182

Diagramofhowthesisterblockswereusedtobondthestrutandtile. tobe8MPa.ThisishalfofwhatisquotedbyMasterBondandismostlikelyduetothesmoothersurfacenishandlackofusinganelevatedtemperaturetocuretheepoxy.AlthoughtheEP21TCHT-1epoxyiscapableofcuringatroomtemperature,itisrecommendedthatanelevatedtemperaturebeusedtoformthestrongestbond.Itismostlikelythesetwofactorsthatcontributedtothelowershearstrength.Despitethelowerthanexpectedshearstrength,thesister-blockbondingprovedtobeamethodthatproducedsignicantstrength(largerthanusinghydroxidebonding)andwouldbefeasibleforusewhenassemblingthetelescopeagain[ 68 ].Withthisresult,itwasdecidedthathydroxidebondingwillbeusedtoallowfortheprecisionalignmentneededtokeeptheprimaryandsecondarywithinparallelismtolerances,andthesister-blockbondingwillbeusedtoprovideadditionalstrengthtothetelescopestructure. 182

PAGE 183

7-12 ).Toimpedetheowofheatfromthetankbasetothetelescope,aseriesofstainlesssteeltubesandplateswereused.Theplatesare1.2mindiameter,3.1mmthickandthebottomfaceiscoveredinalayerofaluminizedPET.Thetubesare25.4mmlong,25.4mmindiameter,andhaveawallthicknessof1mm.Tentubesareplacedonthetankbasefollowedbyastainlesssteelplate.Thislayeringisthenrepeated.Fouraluminumrods38mmindiameterand1mlongarescrewedintothetopstainlesssteelplateandasupportstructurethatholdsthereservoirinplace.Thetopofthereservoirisequippedwithaangethatallowsallingtubetoberemovedsothetanklidcanbeplacedovertheentirestructure.Inthismanner,thereservoirandthermalshieldingareenclosedinvacuum,butllingthereservoircanbedoneoutsidethetank.FourMacorrods50.8mmindiameterand25.4mmlongareusedtosupportanaluminumbaseplate.Thisbaseplatehasholesdrilledandtappedina25.4mmsquaregridtoallowopticstobeplacedonit.Thealuminumbaseplateisconnectedtotheliquidnitrogenreservoirthroughtheuseoffourcopperrods25.4mmindiameterand38cmlong.Inaddition,twolayersofaluminizedPETthataredirectlyconnectedtotheliquidnitrogenreservoirareusedtoshieldthetelescopefromthermalradiation.Therstlayersurroundsthereservoir 183

PAGE 184

Representationofthesystemusedtocoolthetelescope(left)andapictureofthenishedsystem(right).NotshownontheleftgurearethetwolayersofaluminizedPET.ThealuminizedPETshieldsarepartiallyshowninthegureontherightalongwiththeSiCprimaryafterthestructurecollapsed. andsecondstainlesssteelplate.Thesecondlayersurroundsthetelescopeandaluminumbaseplate.Coolinganaluminummock-upofthetelescopeusingthiscongurationproducesatemperatureofapproximately-95Contheprimary.Toraisethetemperaturetothenecessary-70C,afeedbackcontrollerusingheatersandtemperaturesensorsareusedtoproducethedesiredtemperature.Inthismannerthetemperatureofthetelescopecanbechangedoverawiderangeoftemperatures.AlthoughthecomputersmodelsshowedthenominaloperatingtemperatureoftheLISAtelescopewas-70C,thisnumbervarieddependingonthemodelused,materialproperties,andwerebasedonestimatesofheatgeneratedfromtheopticalbench.TestingoverawidetemperaturerangewillgiveagreaterinsightintohowwellthetelescopedesignwillworkontheLISAmission.Usingthecombinationofthermalshields,temperaturesensors,andheaters,atemperaturerangeof-90to-50Ccanbeobtainedwithatemperaturestabilityofbetterthan100K/p 184

PAGE 185

5 ,bybeatingalaserlockedtoacavityonthetelescopeandalaserlockedtoaZerodurcavity,therelativein-bandnoisecanbefound(Figure 7-13 ).Byvaryingthetemperatureofthetelescope,thein-bandnoisecanbecharacterizedasafunctionoftemperature. Totesttheabsolutelengthstabilityofthetelescope,alaserlockedtoamoleculariodinetransitionwillbeused.Iodinewillbeusedinsteadofcesiumduetothedicultiesassociatedwithoperatingthecesiumcellinanoven.TheprocessissimilartothatdescribedinChapter6,althoughaniodinecellandfrequencydoublingcrystalareusedinsteadofanovenandcesiumcell.Thefrequencydoublingcrystalisneededinordertoconvertthe1064nmlightinto532nmlightsinceiodinehasstrongabsorptionlinesat532nm.ThedetailsoftheexperimentalsetupwillnotbediscussedduetothesimilarityofthecesiumlockingexperimentpresentedinChapter 6 .Bylockingalasertoamoleculartransitionofiodineandbeatingitwithalaserlockedtoacavityonthetelescope,theabsolutestabilityofthetelescopecanbemeasured.Thisworkiscurrentlyunderwaybyothermembersofthegroup 185

PAGE 186

Representationofhowthein-bandandabsolutelengthnoiseforthetelescopewillbefound. Althoughitwouldbeidealtodeterminetheabsolutestabilityofthetelescopeat-70C,itmaybemorefeasibletodeterminetheabsolutestabilityatroomtemperature.Determiningtheabsolutestabilityathighertemperaturesshouldprovideanupperlimitonthecompacticationofthematerialsincedislocationsarelesslikelytopropagatethroughamaterialatlowertemperatures,andarethoughttobetheleadingcauseoftheshrinkingofamaterial.Oncethecompacticationhasbeencharacterizedathighertemperatures,thetemperatureofthetelescopecouldthenbeloweredforseveralweekstodeterminetheeectsoflowertemperaturesonthedensicationprocess. Thetelescopewasdesignedsuchthatthetiltbetweentheprimaryandsecondarycouldbemeasuredinseveralways.OnewaywouldbetouseFabry-Perotcavitiesplacedaroundthecentralcavityusedtomeasurethein-bandnoise.Lockingalasertoeachofthecavitiesandmonitoringthebeatnotesbetweenthemwouldprovideaprecisewaytodeterminetheinducedtiltinthetelescope,butwouldrequiretheadditionoffourextralasers.AmorefeasiblewaywouldbetosetupMichelsoninterferometersaroundthe 186

PAGE 187

7-15 ).Tosetuptheinterferometer,anon-polarizingbeamsplitterwasmountedontheprimaryandamirrorwasmountedonthesecondary(seeFigure 7-14 ).Amirrorwasplacedontheprimaryafewinchesfromthebeamsplitter.Sincethedistancebetweentheprimaryandsecondaryisaboutafactoroftengreaterthanthedistancebetweenthemirrorontheprimaryandthebeamsplitter,theinterferometerwillessentiallymeasurethechangeindistancebetweentheprimaryandsecondary. Fromthedatatherearefournoticeablecharacteristicsinthedataandeachonewillbediscussed.Therstcharacteristicisthefactthatthetimebetweenpeaksslowlyincreasesovertime.Thisisexpectedandisduetothetemperatureofthetelescopereachingitsequilibriumvalue.Initiallythetelescopecoolsdownrapidly,causingthepeakstobeclosertogether.Astimeincreases,thetelescopetemperaturereachesitsequilibriumtemperatureslowly,causingthepeakstospreadoutintime.Analysisofthetelescopetemperaturedatashowsthatastheslopeofthetemperaturedecreases,thepeaksspreadoutmore.Towardstheendofthedata,thepeaksbegintogetcloseragain,whichisduetothetelescopeheatingup. 187

PAGE 188

RepresentationofhowtheMichelsoninterferometerwassetuptomeasurethedistancechangebetweentheprimaryandsecondary.. Thesecondtrendintheinterferometerdataisthatthepeak-to-peakvaluesslowlydecreaseovertime.Thisisduetothemisalignmentoftheinterferometerastherodsthatholdthealuminumbaseplateshrink.Realignmentoftheinterferometerwasdoneasthetelescopewascoolingdownandthiscanbeseenfromtheosetofthepeaksincreasinginadditiontothepeak-to-peakvaluesincreasing. Anothercharacteristicofthedataarethespikesthatcanbeseen.Thespikeswherethevoltagesdropstozerocanbeaccountedforfromoccasionallyblockingthelaserwhenrealignmentoftheinterferometerwasdone.Thesearealmostalwaysseenbeforeachangeinamplitudeofthevoltagecanbeseen.Theotherspikeswereseenwhentheliquidnitrogenreservoirwasrelled.Theexactcauseofthisisunknown,butitismostlikelyduetoasettlingorshiftingofthethermalshieldsthatslightlymisalignedtheinterferometer. ThelastcharacteristicofthedataisthehighfrequencynoiseofthedataasshowninFigure 7-16 .Thisisbelievedtobeduetoaspuriousinterferometerthatwasinduced 188

PAGE 189

PhotodiodeoutputoftheMichelsoninterferometerasthetelescopewascooleddown(left)andadetailedviewofthesameoutputoverashorterperiodoftime(right). fromusinganNDlterbeforethedetectionphotodiodetoavoidsaturation.Subsequentinterferometersonthenaltelescopedesignwillusesubstantiallylesspower,andsothisproblemshouldnotbepresent.Noiseassociatedfromthedriftinthelaserwasruledoutduetotheperiodicnatureofthesignal,andthefactthatthelaserdriftsbylessthan50MHz/dayandissignicantlylessthantheFSRofthelongarmoftheinterferometer. Bycountingthenumberoffringepeaksandusingtheequationd=n/2,wheredisthechangeindistance,nisthenumberoffringepeaks,andisthewavelengthofthelaser,thetotaldistancethetelescopeshrunkcanbefound.Fromtheinterferometerdata,itwascalculatedthatthetelescopeshrunkby60m.ThedistancethetelescopeshrunkcanalsobefoundusingthetelescopetemperatureandtheCTEasafunctionoftemperatureprovidedbyCoorsTek.Thetelescopeinitiallystartedat20Candcooleddownsuchthatthenaltemperatureofthesecondarywas-60Candtheprimarywas-70C.Takingtheaverageofthesevalues,theassumednaltemperatureofthetelescopeis-65C.TheplotoftheCTEversustemperatureisshowninFigure 7-17 andisroughlylinearbetween200and300K.FittingastraightlinebetweenthistemperaturerangeresultsinaCTEgivenbyCTEppm=0.011T-1.11,whereTisthetemperatureofthe 189

PAGE 190

DetailedviewoftheoscillationsintheMichelsoninterferometerdata. telescopeinKelvin.UsingtheaverageoftheCTEat20Cand-65C,andmultiplyingbythechangeintemperature(85C)andlengthbetweentheprimaryandsecondary(0.6m),resultsinalengthchangeof85m. AlthoughtheamountthestructureshrankusingtheMichelsoninterferometershouldbelessthanthevaluefoundusingtheCTEvaluessincetheinterferometerarmontheprimarywillshrinkasthetelescopecoolsdown,thediscrepancybetweenthetwoshouldnotbeaslargeasitis.Thecauseofthediscrepancyisnotknownandfurthertestingwillneedtobedoneonthealuminummock-upbeforebeingusedontheSiCtelescope. Currently,workisunderwaytorebuildthetelescopespacer.Thealignmentjigwasredesignedandusedtobondanaluminummock-upusingbothhydroxidebondingandsister-blockbonding.Themock-upwascooledto-80C,warmedbacktoroomtemperature,andnodegradationofthebondscouldbefound.Thus,thecombinationofusinghydroxidebondingandepoxyingthesisterblocksshouldprovidemorethanenough 190

PAGE 191

CTEasafunctionoftemperatureforCoorsTekUltraSiCSiC.DataobtainedfromDavidBathatCoorTek. strengthtothestructure.Itremainstobeseenhowtheepoxywillbehaveduringtherelativestabilitymeasurements. 191

PAGE 192

TheLISAmissionconsistsofthreespacecrafteachseparatedbyvemillionkilometers.Eachofthesespacecraftcontaintwoopticalbenchesthatcontainlaserswhichpointtotheadjacentspacecraft.Eachopticalbenchcontainsaproofmassthatthespacecraftorientatesitselfaroundtoprovideadrag-freeenvironment.Duetothegeodesicsoftheproofmasses,thelengthsbetweenthespacecraftwillchangebyasmuchas1%overthecourseofayear.Totrackthebreathingofthespacecraft,theanglebetweentheopticalbenchesonaspacecraftwillhavetochangebyasmuchas1. IftheLISAconstellationisthoughtofasaMichelsoninterferometer,thenonespacecraftactsasabeamsplitter,whiletheothertwospacecraftactlikemirrors.Becausetwoseparatelasersareusedatthebeamsplitterspacecraft,theywillneedtobeinterferedinordertomaketheappropriatephasemeasurements.Thisthenbecomesaproblemsincetheanglebetweenopticalbencheswillchangeovertime. Theproposedsolutionistouseanopticalbertotransferthebeamfromonebenchtotheother,andvice-versa.Theberwillcontainbothcounter-propagatingbeamsandwillinduceadditionalphasenoiseasittravelsthroughtheberbutcanbesubtractedoutthroughtheuseofappropriatemetrology.IfthesensitivityofLISAisnottobespoiledfrominducedbernoise,thenthedierentialphasenoiseofthecounter-propagatingbeamsmustbelessthan106cycles/p Thecurrentprojectplanistouseasingle-modepolarization-maintainingberapproximately0.5to1.0mtoaccommodatethecounter-propagatingbeams.ExperimentsatboththeJetPropulsionLaboratory(JPL)andtheAlbertEinsteinInstitute(AEI)inHannoverhavefailedtoconrmthatabi-directionalbermeetstheLISArequirementandbothappeartobelimitedbysystematicnoise[ 93 ],[ 94 ]. Inordertotestthedierentialphasenoiseinducedincounter-propagatingbeamsthroughanopticalber,aSagnacinterferometerwasconstructedandanopticalberwas 192

PAGE 193

RepresentationofthethreeLISAspacecraft,theopticalbenches,andtheback-linkopticalber. placedinitspath.TheSagnaccongurationtakesadvantageofthefactthatallnoisewillbecommonexceptthatofthenoiseinducedfromthebersincetheopticalpathsarethesame,butwithoppositedirections.ThischaptercontainsanintroductiontotheSagnaceectandinterferometer,theexperimentalsetupusedtotestthedierentialnoiseinanopticalber,andtheresultsobtainedfromthissetup. 193

PAGE 194

8-2 Figure8-2. DiagramofarotatingloopinterferometertoillustratetheSagnaceect.Dopplershiftedfrequenciesareshownwithrespecttotheinertialspace. Whentheinterferometerisnotrotating,thenthecounterpropagatingbeamswillhaveequalpathlengthssincethelighttravelsatthesamevelocityinbothdirections.ThiswillresultinbothbeamsreturningtothesamepointPiofinjectionafteratimeof=2R/c,whereRisthepathradiusandcisthespeedoflightinvacuum.Whentheinterferometerisrotatingatarateandtheobserverisstationaryintheinertialframe,thenthepointPimovesthroughanangleduringthepropagationtime.Hence,thedierenceinpropagationtimestolowestorderinR/cisgivenby[ 95 ] =(2+)R c(2)R c=4R2 Foracontinuouswaveoffrequency!,thecorrespondingphaseshiftbecomes =!=4R2! c2:(8{2) 194

PAGE 195

96 ]. TheSagnacphaseshiftmaybecalculatedusingtheDopplerfrequencyshift,butwillleadtothesameresultasinEq. 8{2 sincetheDopplershiftwillcausethewavelengthfromthewavepropagatinginonedirectiontobedierentfromthatofthewavetravelingintheotherdirection.Thus,thetwointerpretationsoftheSagnaceectareequivalent(tolowestorderinR/c). AmorerigorousapproachtotheSagnacphasedierenceistosolvetheelectrodynamicpropagationequationsintherotatingmedium[ 97 ],butwillnotbeconsideredhere.Whatcanbederivedfromtheelectrodynamicformulismisthatthephaseshiftdependsonlyonthelightfrequency!andonthedotproductoftheequivalentareavectorofthemeanopticalpathandtherotationratevector.Thisresultsinaphasedierenceof =4! c2A:(8{3) Thus,theSagnaceectisduetothedierenceinpropagationtimesofthewavesastheytravelaroundtherotatingloop.Thewavesfrequencyactsasaclocktomeasurethistimedierenceandisindependentofthevelocityofthephysicalsubstrateused.Usingahigherfrequencylightwillresultinalargerphasedierence. Tocircumventthisproblem,aberwithastrongbuilt-inbirefringencewillpreservethepolarizationstateeveniftheberisbentaslongastheinputpolarization 195

PAGE 196

8-3 .Thestressrodsdonotnecessarilyneedtobecylindrical,butthemodiedglasscompositionmusthaveadierentcoecientofthermalexpansion. Figure8-3. BoththePANDAandbow-tiepolarization-maintainingberstylesareshownalongwiththecoreandstresselements. Thetwomostcommonstylesofpolarization-maintainingbersarethePANDAandbow-tie,althoughothersareavailable.Thetwostylesarebasedonthestressrodsusedthatrunparalleltotheber'score.PANDAstressrodsarecylindrical,whilethebow-tiedesignusestrapezodialprismstressrods.Formostapplicationsotherthanthetelecommunicationsindustry,eitherstylecanbeused,althoughPANDAstylebersaretypicallyusedmoreoftensincetheuniformityoftheberismucheasiertomaintain. InthecaseofthePANDAdesign,theopticalcoreoftheberismanufacturedindependentlyofthestressrods,allowingittobemadewithhighchemicalpurity.Corecompositionsareeitheragermanium-dopedorpuresilicacore.Thestressrodsaretheninsertedintocircularholesdrilledintothepristinepreformallowingforahigherdegreeof 196

PAGE 197

Polarization-maintainingbersshouldnotbeconfusedwithsingle-polarizationbers,whichcanonlyguidelightofacertainpolarization.Polarization-maintainingbersareusedinapplicationswherethepolarizationstatecannotbeallowedtodriftasaresultoftemperaturechanges,suchasforuseinberopticgyroscopes,berinterferometers,andcertainberlasers[ 98 ].Thesebersusuallyrequireamuchgreaterdegreeofalignmentthanforastandardber,andthepolarizationdirectionisalsorequiredforoptimalperformance.Propagationlossesarealsohigherthanforstandardbers,andnotalltypesofbersareeasilyobtainedinpolarization-preservingform. 8-4 .Bothaqualitativeandquantitativedescriptionofthesetuparegiven. 197

PAGE 198

Sagnacinterferometerusedtomeasuretheaccumulateddierentialphasenoiseinanopticalber.HWP1-3arehalfwaveplates,PBS1-3arepolarizingbeamsplitters,QWPisaquarterwaveplate,andthebalancedphotodetectorisrepresentedbyBPD. 198

PAGE 199

8-4 matchupwiththenumbersbelow): 1. ThebeamfromthelaserpassesthroughalensthatproducesabeamwithaRayleighrangeof1m.Thisallowsthebeamtobewellcollimatedthroughouttheinterferometer.Thebeamthenpassesthroughapolarizingbeamsplitter(PBS)toproduceabeamwithalmostallp-polarization. 2. Thebeamthenpassesthroughanotherpolarizingbeamsplitter(PBS1)thatactsasbothanadditionalpolarizationcleanerfortheincidentbeamandthepick-otothedetectorfortheoutgoingbeamfromtheinterferometer. 3. Thebeampassesthroughahalfwaveplate(HWP1)suchthatthereisnowanequalamountofs-andp-polarizationinthebeam. 4. Thebeamisincidentonapolarizingbeamsplitter(PBS2)thatsplitsthebeamintoitspolarizationcomponents.Thetransmittedbeam(green)containsp-polarizationonly,whilethereectedbeam(red)containsprimarilys-polarization,withasmallamountofp-polarizationduetothebeamsplitters. 5. Thetransmittedbeam(green)continuestowardstheber,whilethereectedbeam(red)passesthroughahalfwaveplate(HWP2)orientatedsuchthatthepolarizationsarerotatedby90. 6. Thetransmittedbeam(green)passesthroughacollimatinglensandisincidentontheber.Thereectedbeam(red)passesthroughanotherpolarizingbeamsplitter(PBS3)toisolatethespuriouss-polarizationthatisproducedfromPBS2.Inthismannerbothbeamsincidentontheberhavethesamepolarizationastheygothroughtheber. 7. Bothbeamsemergefromtheberwith80%eciency. 8. Thetransmittedbeam(green)passesthroughPBS3cleanly,although5%ofthepowerislost.Thereectedbeam(red)continuestowardsPBS2. 9. Thetransmittedbeam(green)passesthroughHWP2andisrotatedby90,whilethereectedbeam(red)continuestowardsPBS2. 10. ThebeamsthenrecombineatPBS2.Iftherewasnodierentialphasenoiseinducedfromthesystem,thentheresultingbeamwouldemergefromtheSagnacinterferometerwithlinearpolarizedlightinequalamounts.Ifanydierentialphasenoiseispickedupfromtheinterferometer,thenthebeamswillrecombinetoproduceellipticallypolarizedlight.Theminoraxisoftheellipticallypolarizedlightservesasameasureofthedierentialphaseaccumulatedintheinterferometer. 199

PAGE 200

TheellipticallypolarizedlightpassesthroughHWP1andchangesthe\handedness"ofthebeam.Thismeansthatifthebeamwasrotatingclockwise,itwouldnowberotatingcounter-clockwise,andvice-versa. 12. OncetheellipticallypolarizedlightisincidentonPBS1themajorityofthes-polarizedlightwillbereected,whileonlyafractionofthep-polarizationwillbereected.Theresultissuchthatanellipticallypolarizedbeamisreected,butwithamuchlowerpower.IfPBS1wasidealandonlythes-polarizationwasreected,thentheoutputofthisportwouldserveasameasureoftheaccumulateddierentialphase. 13. Thequarterwaveplate(QWP)isorientatedsuchthatthebeamislinearlypolarizedaftertheellipticallypolarizedbeampassesthrough. 14. Thelinearlypolarizedlightthenpassesthroughahalfwaveplate(HWP3)withanorientationsuchthattheemergingbeamhasequalamplitudesofs-andp-polarizations. 15. Thebeamthenpassesthroughacalcitewedgewhichseparatesthepolarizationsthatarethenseparatelyincidentonabalancedphotodetector(BPD).Acalcitewedgeisusedtoseparatethepolarizationmorecleanlythanthroughtheuseofapolarizingbeamsplitter.TheoutputontheBPDwillreadzero(orpossiblyasmalloset)iftheorientationsofthewaveplatesarecorrect,andthebeamisproperlyalignedintotheber.Anyaccumulateddierentialphasefromtheinterferometerwillcausethes-polarizationatstep12tobelargerandwillpropagatethroughthewaveplatessuchthattheamountofpowerontheBPDisnotthesameandanon-zerovoltagewillbemeasured. TheadvantageofthissetupisthatboththereectedandtransmittedbeamsfromPBS2seethesameopticalcomponentsandbothexperiencetwotransmissionsandonereectionthroughapolarizingbeamsplitter.Thesenoisesourcesarecommontoeachbeamandwillcancelonthebalancedphotodetector,resultinginonlythedierentialnoiseinducedfromtheopticalber. 200

PAGE 201

8-4 ,andERandRaretheelectriceldandphaseofthebeamaftertravelingtheredpath.IfweusethefollowingdenitionsE=1 2(EG+ER)E=1 2(EGER) (8{4)=1 2=(G+R)=1 2(GR); thenourelectriceldafterrecombiningbecomes Thecommonphaseshift(ei)canbeignoredsinceitisafactorthatwillbecarriedaroundandeventuallycanceledwhenndingthephotodiodesignals.SortingEq. 8{6 intoamoremanageableformresultsin Asthecombinedelectriceldinstep10passesthroughHWP1,thehalfwaveplatewillrotatebothpolarizationsby45.Thiswilltransformtheelectriceldsinthefollowingmanner:1 whichtransformstheelectriceldatstep11into 201

PAGE 202

AligningQWPsuchthatitsfastaxisisalignedwith^panditsslowaxiswith^sistheidealcasesincethiswillonlyproduceanadditionofa90phaseshifttothes-polarizedlight.Thisresultsinanelectriceldatstep13of Intheidealcase,HWP3wouldbealignedsuchthatitrotatesbothpolarizationsby45.Iftherewerenodierentialphaseshiftsaccumulatedintheinterferometer,thentherewouldbenos-polarizedlight,andHWP3wouldrotatethep-polarizationsuchthatanequalamountoflightfellonbothphotodiodesonthebalancedphotodetector.RotatingEq. 8{11 by45resultsin RearrangingEq. 8{12 intoitss-andp-polarizationsresultsinE14=[(rEcos+Esin)+i(rEsinEcos)]^p+[(rEcosEsin)+i(rEsin+Ecos)]^s: 202

PAGE 203

TorstorderinandsecondorderinE,cos1andsin,andthepowerlevelsaregivenby andthedierenceinthephotodiodesignalsisgivenby Eq. 8{15 istheresultingbalancedphotodetectorsignalunderidealconditionsfortheinterferometer.Thep-polarizationleakagefromPBS1actsasalocaloscillatorandifnodierentialphaseisintroduced,thentheresultingsignalfromthebalancedphotodetectorwillbezero. 8{15 assumingidealoperations,anexperimentalvericationwasdoneinordertodeterminetheexactrelationbetweentheoutputvoltageofthebalancedphotodiodeandthedierentialchangeinphase.Todothis,a10mber-EOM(modelPM1060HFfromJenoptik)wasputinserieswiththeberalreadyintheinterferometerandavoltagewasappliedtotheEOMtoproduceatoneat10MHz.ThistonewasthendetectedafterHWP3byahighbandwidthphotodetectormadebytheEOTechcompanysincethebalancedphotodetectordoesnothaveabandwidthhighenoughtodetecta10MHzsignal.RotatingHWP3willcausetheamplitudeofthetonetochange.Fromthissignal,theoverallcalibration 203

PAGE 204

Todeterminethecalibrationfactor,a10MHz,3Vpeak-to-peaksignalwasappliedtotheEOM.TheoutputontheEOTechphotodetectorwasmeasuredwithaspectrumanalyzerandfoundtohaveapeakheightof70mV.Usingtheimpedanceofthespectrumanalyzer(50)andtheeciencyofthephotodetector(0.75A/W),theRMSpowerincidentonthephotodetectorwasfoundtobe0:070V 0:75A=W=1:87mW: ThephaseinducedfromapplyingthevoltagetotheEOMwasfoundbymultiplyingaconversionfactorbytheappliedvoltage.TheconversionfactorwastakenfromexperimentsdoneattheAEIinHannoverandwasfoundtobe7.2rad/VfortheparticularEOMused[ 99 ].Usingthisfactorandtheappliedvoltage,resultsinatotalphasechangeof22rad,or3.5cycles.Usingtheincidentpower,theamountofphasechangefromtheEOM,andtheresponsivityofthebalancedphotodetector(1.28V/mW),theoverallconversionfactorcanbefoundusingcal=Pactual 204

PAGE 205

8-5 andappearstobeveryat. Oncethenoiseooroftheinterferometerwasknown,aberwasplacedintheopticalpath,andthedierentialphasenoisewasmeasured.Twodierentopticalbersweretestedandbothproducedsimilarresults.Therstberis10minlengthwithFC/APCpolishedendsandwasboughtfromtheNuferncompany.ThesecondberwasboughtfromtheSchafterandKirchhocompanyandis10mlongwithFC/APCpolishedends.TheNufernberhasasignicantlythickeroutercladdingthantheotherber,butthisdidn'tappeartoeectthemeasurements.Theendsoftheberwerepolishedatanangletohelpavoidback-scatteringoftheincidentlight.Bothbersaresingle-modepolarization-maintainingbersandhaveanumericalaperture(NA)of0.12.Lightwasfocusedintothebersthroughtheuseoffree-spacecollimatinglenseswithanNAof0.15. 205

PAGE 206

Noisespectrumofthedetectorcomponentsoftheinterferometer.ThiswasfoundbyplacingamirrorinfrontofPBS1andaligningthebackreectedlight. Withproperalignment,mostofthetime70-80%ofthelightwouldgothroughthebereachwayusingthecollimatinglenses.AllopticspertainingtotheinterferometerwereplacedonaSuperInvarbreadboardboughtfromNewport(representedbythesquareinFigure 8-4 ).Typicalinputpowersintotheinterferometerwerebetween10to30mWandproducedlessthanafewmilliWattsofpoweronthedetectorphotodiodes. Theinterferometerwasplacedonavacuumtankange(representedbythecircleinFigure 8-4 ).Thetanklidcouldthenbeloweredontopoftheangethroughtheuseofapulley.Severalportswereavailableforlighttogoinandoutofthetank.Theinitialalignmentoftheinterferometerwasdonewiththelidoandthetopwasthenloweredafterthealignmentwasdone.Thissometimescausedasmallchangeinalignment.Whenthetankwasonandnotundervacuum,oneoftheportswasleftopensothelightdidn'tpassthroughaglassplate.Aluminumfoilwasplacedovertheporttohelprestrictaircurrentsandprovideasmallamountoftemperaturestabilityinsidethetank.Asmallholewaspunchedinthealuminumfoiltoallowthebeamtopassthrough.Anoticeableeect 206

PAGE 207

Therestoftheopticsoutsidethevacuumtankwereshieldedusingcardboard.Thenoiselevelsdidn'tnoticeablychangebecauseofit.Thereasonforthelowernoiselevelswiththetanklidonisunknown,butitisassumedthatthetemperatureuctuationsarelesswhenthetanklidison. Totryandimprovethestability,theportonthetankthathadbeencoveredwithaluminumfoilwasreplacedwithaglassangeandthetankwaspumpeddownusingaroughingpump.Unfortunately,asthepressuredropped,thealignmentintotheberbecameworseandeventuallynolightwasgoingthroughtheber.Themisalignmentasthepressuredropsmostlikelyoccursduetostressesthatareinducedintheopticalcomponentsthatchangethebeamtrajectory.Attemptstorealignthebeamintotheberthroughmirrorsoutsideofthetankwereunsuccessful. Afterseveraltestswiththeberintheinterferometerandthetankplacedontop,itwasfoundthataparasiticinterferometerwascausingthephotodiodesignaltobesignicantlymorenoise[ 100 ].Thesolutionwastomodulatethelaserfrequencybyapplyingatrianglewavewithamplitudeof3Vpeak-to-peakandfrequencyof5kHz.ThenoisedroppedbyalmostanorderofmagnitudeasshowninFigure 8-6 .Modulatingatthisfrequencychangedthelaserfrequencyby75GHz/s.Bymodulatingthelaserfrequencyatthisrate,theeldsreachingthedetectorhaveafrequencydierentfromthemaininputbeam[ 101 ].Inthismanner,thenoisefromtheparasiticinterferometerisnowwelloutsideofthemeasurementband. Afterthisproblemhadbeensolved,itwasthennoticedthatoverthecourseofafewdays,theamountoflightgoingthroughtheberdroppedsignicantly.Oftentheberwouldbealignedsuchthat80%ofthelightwouldbetransmittedandafewdayslaterthiswoulddropto50%orlower.TypicallythelighttransmittedfromPBS2was 207

PAGE 208

Plotofthetimeseriesbeforeandaftermodulatingthefrequencyofthelaser.Itisclearthatmodulatingthelaserfrequencyproducessignicantlylessnoise. aectedmorethanthereectedlight.Todeterminethecauseofthis,glassplateswereputinateachendoftheberandwerealignedsuchthattheanglethesurfaceoftheglassplatemadewiththeoutputoftheberwasclosetotheBrewsterangle.Thepowersofbothglassplatereectionsweremeasuredusingaphotodiode.Inaddition,anotherglassplatewasplacedafterHWP1inthesamemannertodetermineiftheinputpowerwaschangingaswell.TheresultsofallthreemeasurementstakenatthesametimeareshowninFigure 8-7 .Itiseasytoseethattheinputpowerisstayingalmostconstantwhiletheoutputpowersfromtheberarechangingovertime.Theexactcauseofthiswasneverdetermined,butitisspeculatedthatitcouldbefrompointingissuesintothebers. Whilethepowersmeasurementswerebeingtaken,theoutputofthebalancedphotodetectorwasmonitoredaswellasshowninFigure 8-8 .Theexactrelationshipbetweenthechangingberoutputpowersandthechangeinvoltageofthebalancedphotodiodeisdiculttodetermine.Asthepowerthroughtheberchanged,itwasalsodeterminedthatthemodeoftheberafterHWP3changedaswell.Whenthe 208

PAGE 209

Timeseriesoftheinputpowertotheinterferometer(left)andthetimeseriesofthepowersduringthesametimeinboththeclockwiseandcounterclockwisebeams(right). Figure8-8. Timeseriesofthebalancedphotodetectoratthesametimethepowersofthecounterpropagatingbeamsaremeasured. 209

PAGE 210

ThebestresultsforthedierentialphasenoisearepresentedinFigure 8-9 .Athigherfrequencies,thenoiseisonthe105level,butatlowerfrequencies,thenoiseisabouttwoordersofmagnitudehigher.Ifthedierentialphasenoisescaleswiththelengthoftheber,thenthelevelswouldbe10to20timeslowersincetheLISAback-linkberwillonlybe0.5to1.0minlength.ThiswouldalmostmeettheLISArequirementathighfrequencies,butasignicantamountofworkwillneedtobedoneinordertodeterminethecauseofthelowfrequencynoise. Figure8-9. Noisespectrumoftheopticalberwithcounter-propagatingbeams. AlthoughtheLISArequirementwasnotmetusingtheSagnacinterferometer,asignicantamountofinformationwaslearnedaboutthestabilityoftheopticalberand 210

PAGE 211

102 ]andJPLwereabletoachieveusingamoreLISA-likeexperiment.Inboththeseexperiments,thenoiseappearstobelimitedbytheback-scatterofthebeamintotheber.Informationaboutpointingandback-scatterfromtheberwillbeusedtodesignanotherexperimentthatwillhopefullybesensitiveenoughtoreachthenecessaryrequirements. 211

PAGE 212

Itwasfoundthatthereasontheglasswasbreakinginthersttestswereduetothewireusedinthelever-armapparatus,andnotfromthewaterinthesolutionweakeningtheglass.Fromthis,abetterwaytotestthematerialswasfoundandtheMCSTwasdeveloped.Usingthisdevice,existingglasstoglassresultswereexpandedonbyvaryingtheamountofsolutionused,varyingtheconcentrationofthesolution,andtestingbondsmadewithroughsurfaces.Itwasfoundthatthereisnosignicantdependanceinshearstrengthofthebondbasedontheamountofsolutionused,whethera1:4or1:6concentrationisused,orifaroughsurfaceisusedinsteadofapolishedsurface.Typicalshearbreakingstrengthswere2to5MPa. ItwasalsofoundthathydroxidebondingcouldbeusedonSuperInvar,whichisduetoitseaseofoxidizing.SuperInvartoBK7bondsprovedtobesuccessful,andvaryingtheamountofsolutionusedshowednosignicantdependanceontheshearbreakingstrength.Theaveragebreakingstrengthwasfoundtobe5MPa. AlthoughbondingSiCtoSiChadalreadybeendone,thebondingprocessrequiredthematerialstobeheatedtoover1000Candthebondswerenotreliablymadewithsignicantstrength.AmorereliableandsimplerwaytobondSiCwasfound,andproducedtypicalbreakingstrengthsof3to5MPa.Thesurfaceproleofthematerials,theamountofsolutionused,andthedilutionfactorofthesodiumsilicatesolutionwere 212

PAGE 213

Theversatilityofthehydroxidebondingprocessdemonstratesthatitisapromisingbondingtechniquethatcanbeusedformanyspace-basedmissions.Althoughpromisingresultshavebeenreported,furtherinvestigationsareneededtodeterminehowtoachievetheoptimumbondstrengthinarepeatablemanner,andunderwhatconditionsthebondwillsurvive.Byfurtherunderstandingthisbondingtechnique,thelimitstowhichitcanbeusedwillbefound. ZerodurwithbothopticallycontactedandhydroxidebondedmirrorsshowedalmostidenticalnoiselevelsandwerewellbelowtheLISApre-stabilizationrequirement.Althoughthehydroxidebondsmayproduceahighernoisespectrum,itwasnotdetectablewithoursystem. BothSiCandCFRPwerefoundtomeettheLISAtelescoperequirement.WhilethiswouldappeartoindicatethatbothmaterialscouldbeusedfortheLISAtelescopesupportstructure,CFRPneedstobestudiedingreaterdetailbeforeadenitiveconclusioncanbereached.ThisisduetothefactthattheCFRPcavityoutgassedandtheoverallshapeisnotsimilartowhatwouldappearontheLISAmission.AlthoughitwasseenthattheCFRPcavityinitiallyoutgassed,boththeoveralllengthcontractionandoutgassingconstituentsneedtobestudiedfurther.ItwouldalsobeusefultostudyacavitythathasashapetowhatwouldbeexpectedontheLISAmission,andnotacavitythatwasmade 213

PAGE 214

Figure9-1. ComparisonofthenoisespectrumsoftheZerodurcavitieswithbothopticallycontactedandhydroxidebondedmirrors,SiC,CFRP,andSuperInvaralongwiththeLISAtelescoperequirement. AlthoughSuperInvarwasshowntomeettheLISAtelescoperequirement,itwillnotbeusedontheLISAmissionduetoitmagneticproperties.Despitethis,itseaseofmachinability,highstrength,highstability,andlowCTEmakeitamaterialthatmaybesuitedforothermissionsthatrequireultrastablematerials.InadditiontondingthestabilityofSuperInvar,itwasalsoshownthatusingasodiumsilicatesolutiontobondthemirrorsappearedtoproducenoadditionalnoise.ThisissignicantsincethebondsproducedusingthepotassiumhydroxidesolutionwhenbondingthemirrorsfortheZerodurcavityareaboutafactoroffourthinnerthanthosewhensodiumsilicatesolution 214

PAGE 215

DespitethePZTcavityproducingnoiselevelsslightlyabovetheLISApre-stabilizationrequirement,itwasshownthatthroughtheuseofanadditionalfeedbackcontrolthatisfedtothePZT,thenoisespectrumofthecavitycanbepushedordersofmagnitudebelowthepre-stabilizationrequirement.Itwasalsoshownthatavoltageofupto200VcouldbeappliedtothePZTwithoutanoticeableincreaseinthenoisespectrum.Thisisoveranorderofmagnitudelargerthanwhatwouldbeneededforarm-locking.Thecombinationofitslownoiselevelswithanappliedvoltage,itsabilitytobeusedforarm-locking,andthefactthatitsurvivedvibrationalconditionssimilartowhatwouldbeexpectedduringlaunchmakethePZTcavityviableforuseontheLISAmission. EventhoughthePZTcavityisviablefortheLISAmission,byusingthesamePZTmaterialandthicknessandincreasingthelengthoftheZerodurspacer,thenoiseofthecavityshoulddecrease.ItshouldalsobenotedthattheexpectedtemperaturestabilityontheLISAopticalbenchisapproximatelyanorderofmagnitudebetterthanthetemperaturestabilityofthetankthePZTcavitywastestedin.AlthoughabeatnotedriftisnoticeablewhenavoltageisappliedtothePZT,aslowlychangingappliedvoltageshouldnotproduceanysignicantdrifts. Fromthepresentedresults,allmaterialsexceptthePZTandCFRPcavityreachedacommonnoiseoorabove0.01Hz(seeFigure 9-1 ),whichiswellbelowtheLISApre-stabilizationrequirement.Thiswouldimplyacommonnoiselimitofthemeasurementsystemhasbeenreachedandfurtherimprovementswillbeneededinordertopushthenoiseoordown.Atfrequenciesbelow1mHz,thenoiseislimitedbytemperatureuctuationsinthelabthatpropagatethroughthethermalshields.Althoughthethermalshieldsprovidesignicantattenuationathighfrequencies,atlowenoughfrequenciesthethermalshieldshavealmostnoattenuationascanbeseeninthebeatnotefrequency 215

PAGE 216


PAGE 217


PAGE 218


PAGE 219

TemperatureuctuationinsidethevacuumtankwillcausethecavitytoexpandandcontractbasedonitsCTE.Thepurposeofthethermalshieldswillbetoactasapassivelowpassltertoattenuatethetemperatureuctuationsinthelab.AtypicaltemperaturenoisespectrumfortheouterwallofTank2isshowninFigure A-1 .Usingtheequations FigureA-1. Noisespectrumoftheoutsidetanktemperature. =L LL=LTcav whereistheLSDofthefrequency,isthefrequencyofthelaser,ListheLSDofthelengthofthecavity,Listhelengthofthecavity,istheCTEofthecavity,andTcavistheLSDofthetemperatureofthecavity,thecorrespondingfrequencyuctuationscanbefoundvia 219

PAGE 220

A-1 ,severalordersofmagnitudeoftemperaturesuppressionwillbeneededinorderforthefrequencyuctuationstobebelow1000Hz/p A-2 Thethermalshieldsconsistof5cylindricallayersofanaluminizedPETwhosebaseisasheetofaluminum.EachsuccessivelysmallercylinderisthermallyinsulatedfromthenextusingMacorspacers.Inthecenterofthethermalshieldsisalongstandthatelevatesthecavitytoaheightsuchthatthelasercanpassthroughthecavity.Thealuminized TableA-1. Diametersandheightsofthethermalshieldcylindricalshells. Shell1Shell2Shell3Shell4Shell5 diameter(m) PEThasalowemissivityandactstoreecttheradiationfromitssurrounding.Thelargeheatcapacityofthealuminumallowsthethermalshieldtoabsorblargeamountsofenergywhilechangingitstemperatureverylittle.AlthoughthealuminizedPEThasahigherheatcapacity,itsthicknessisonly0.12mmthickandhasanegligiblemasswhencomparedtothemassofthealuminum.ThelowthermalconductivityoftheMacorspacersimpedestheheatowfromonecylindertothenext.AsummaryofpropertiesforallthreematerialsislistedisTable A-2 .Todeterminetheattenuationthethermal TableA-2. MaterialpropertiesofMacor,aluminizedPET,andaluminum. MaterialThermalConductivityEmissivityDensityHeatCapacity(W/mK)(kg/m3)(J/kgC) AluminizedPET{0.0313801190Aluminum2500.152700902Macor1.46{2520790 220

PAGE 221

Pictureofthebuiltthermalshields(left)andamodelrepresentationofthem(right).Notethemodelisnotdrawntoanyscale. shieldswillprovide,athermallumpedmodelwillbeused.Onlyconductionandradiativeheattransferwillbeconsideredsincetheconvectiveprocesswillbenegligibleduetothehighvacuumwithinthetank.AsimpliedlumpedthermalmodelofthethermalshieldsisshowninFigure A-3 .Thethermalresistanceforbothconductionandradiationcanbe FigureA-3. Lumpedthermalmodelofthethermalshields. foundusing Acond 221

PAGE 222

Thetotalthermalresistanceandcapacitanceofeachstageisgivenby wherecistheheatcapacitance,isthedensity,andVisthevolumeofthealuminumbaseplates.Thetimeconstantofthelowpasslter,,foreachcylindricalshellis Therstordercutofrequencyforthelteristhen 2RC:(A{8) ThesevaluesforeachstageofthethermalshieldsaregiveninTable A-3 .Theexpected TableA-3. Resistance,capacitance,timeconstant,andcutofrequencyforeachstageofthethermalshields. StageRradRcondRCfc(KW1)(KW1)(KW1)Jkg1(s)(Hz)

PAGE 223

A-3 from 2iif+1(A{9) whereiarethevaluesofatstagei.PlottingjH(f)jasafunctionoffrequencywillresultintheamountofsuppressionthethermalshieldswillprovideatthatfrequency.ThisisshowninFigure A-4 .Theattenuationisnotshownbelow105inordertogainan FigureA-4. TransferfunctionofthePETthermalshields. understandingofthetransferfunctionatlowfrequencies. Tocomparethiseectivenessofthismodel,thetransferfunctionofthethermalshieldscanbeusedwiththetanktemperaturedatashowninFigure A-1 todeterminewhattheexpectedbeatnoteuctuationsshouldbeduetothetemperatureuctuationsofthetank.Unfortunately,theexpectedfrequencyuctuationsdonotmatchupwellwiththemeasuredbeatnoteuctuationsandareapproximatelyanorderofmagnitudelargerthanwhatisactuallymeasuredatlowfrequencies.Thisismostlikelyduetothelimitationsofusingalumpedthermalmodelandsimplicationsofthetankdesignthatwereused.Inthelumpedthermalmode,itwasassumedthatthereisonlyalarge 223

PAGE 224


PAGE 225

Whencleaningopticalcomponentsforbonding,greatcaremustbetakentoensurethesurfacesareascleanaspossible.Usingacleanroomisoptimalasthisallowsasmallerchanceofparticlestogetonthesurfaces.Althoughacleanroomispreferred,opticalandhydroxidebondinghavebeendoneusingareasthathavebeencleanedthoroughly.Ageneralguidelineispresentedbelowtocleantheopticalsurfacesbeforebonding. 1. Inspectsurfacestobebonded.Ifthereareanylargeparticulatesonthesurface,washowithmethanolandcleanroomwipe. 2. Foldamicroberclothinhalfandplaceatthebottomofthesonicbath.Placepiecestobebondedontopoftheclothsothatthesurfacestobebondedarefacingup. 3. Fillsonicbathwithdeionized(DI)water. 4. Runsonicbathfor15-30minutes.Thedirtiertheopticalcomponents,thelongerthesonicbathshouldrun. 5. Afterthesonicbathhasstoppedrunning,draintheDIwater. 6. Wipethecomponentstobebondedusingacleanroomwipewettedwithmethanol.Ifpossible,usecompressednitrogentoblowoanyexcesswater,methanol,ordirtstillonthecomponents.Ifanystreaksorparticulatescanbeseenonthesurfacestobebonded,repeatthisstepuntiltheyhavebeenremoved. Ifasonicbathisnotavailable,analternatecleaningmethodcanbeusedtocleanthesurfacestobebonded.Thismethodisdescribedbelow. 1. WetamicroberclothwithDIwater. 2. Dabthedamppartoftheclothintosomepowderedceriumoxide. 3. Rubonthesurfacetobebondedinacircularmotion.Dothisforapproximately5minutes.Thedirtierthesurfaces,thelongerthisstepshouldbedone.Youmayneedtodabtheclothintotheceriumoxideagain(ormanytimes)duringthistimetoensureampleamountofceriumoxideisonthesurface. 4. Rinseoexcessceriumoxideusingmethanol. 225

PAGE 226

WetanotherpartofthemicroberclothwithDIwateragainanddabwithsodiumcarbonate.Youmayneedtodabtheclothintothesodiumcarbonatemanytimesduringthisstep.Makesuretouseplentyofsodiumcarbonatetoensureanyceriumoxideresidueiscleanedo. 6. Rubinacircularmotionforapproximately5minutes. 7. Rinsecomponentsthoroughlyusingmethanol.Ifalmcanbeseenonthesurface,thenanopticalwipewettedwithmethanolcanbeusedtoremovedtheresidue. Anotherpossiblemethodtocleanthebondingsurfacesistoinitiallycleanthesurfacesusingceriumoxideandthenplacethepiecesinasonicbathasdescribedabove. Althoughusingtheceriumoxidewillhelptocleanandpolishthesurfaces,a\pitting"eecthasbeennoticedwhenthisisdone.Todeterminetheeectsofusingceriumoxide,aWYKOprolometerwasusedtodeterminethesurfaceproleonseveralspotsofaSiCtile.AtypicalsurfaceprolebeforeusingceriumoxideisshowninFigure B-1 .TheSiCtilewasthenpolishedandcleanedasdescribedaboveandthesurfaceprolewasdeterminedusingtheWYKOonceagain.TheresultingsurfaceproleisshowninFigure B-2 .Althoughthermsroughnessofthetileseemstohavebecomesmoother,therearenoticeablespotswherethesurfacehasdeeppits.Thecauseofthiseectisunknown. FigureB-1. TypicalsurfaceproleofaSiCsurfacebeforecleaningwithceriumoxide. 226

PAGE 227

TypicalsurfaceproleofaSiCsurfaceaftercleaningwithceriumoxide. 227

PAGE 228

Thelinearspectraldensities(LSD)computedforthespectrumsinChapters5-8weredoneusingaMatlabscript.Thescriptcomputesthepowerspectraldensity(PSD)usingWelch'smethodofoverlappingsegmentedaveragesofmodiedperiodogramsandisbasedontechniquesdescribedin"SpectrumandspectraldensityestimationbytheDiscreteFourierTransform",G.Heinzel,et.al.[ 103 ]. Atimeserieswithaknownsamplingfrequency(fs)isloadedintoMatlab.ThetimeseriesisthendividedintosectionsofNdatapointsasdenedbytheuser.Itshouldbenotedthatthebinspacingwillbeequaltothelowestfrequencyandisequaltofs/N.Alinearregressionisttothetimesegmentsandsubtractedout.ThisresidualisthenmultipliedbyaHanningwindow(nottobeconfusedwithaHammingwindow)andadiscreteFouriertransformisperformedonthewindowedresidual.Thepowerspectraldensityfortheithsegmentisthencalculatedandanaverageistakenoverallsegmentstogetthepowerspectraldensityfortheentireset.Thelinearspectraldensityisthencomputedbytakingthesquarerootofthepowerspectraldensity. Thetimeseriesmeasurementsofthefrequencycountercanbeconvertedtoalengthspectrumusing =l l; 228

PAGE 229

[1] B.Tapley,S.Bettadpur,M.Watkins,andCh.Reigber.\TheGravityRecoveryandClimateExperiment:MissionOverviewandEarlyResults".AmericanGeophysicalUnion,(2004). [2] J.Davis,P.Elosequi,J.Mitrovica,andM.Tamisiea.\Climate-drivendeformationofthesolidEarthfromGRACEandGPS".Geophys.Res.Let.,31(L24605). [3] G.Ramilliena,F.Frapparta,A.Cazenavea,andAGountner.\Timevariationsoflandwaterstoragefromaninversionof2yearsofGRACEgeoids".EarthandPlanetaryScienceLetters,235:283{301,(2005). [4] M.Dehne,F.GuzmanCervantes,B.Sheard,G.Heinzel,andK.Danzmann.\LaserinterferometerforspacebornemappingoftheEarth'sgravityeld".JournalofPhysics:ConferenceSeries,154,(2009). [5] J.TaylorandJ.Weisberg.\FurtherexperimentaltestsofrelatavisticgravityusingthebinarypulsarPSR1913+16".Astrophys.J.,345:434{450,(1989). [6] M.Valtonenetal.\Amassivebinaryblack-holesysteminOJ287andatestofgeneralrelativity".Natureletters,452:851{853,(2008). [7] NASA.\LISALaserInterferometerSpaceAntenna".formoreinformation,goto [8] M.Andoetal.\DECIGOPathnder".JournalofPhysics:ConferenceSeries,120,(2008). [9] LISAStudyTeam.\LaserInterferometerSpaceAntennaforthedetectionandobservationofgravitationalwaves".Pre-phaseAreport,July1998. [10] S.Hughes.\LISAsourcesandscience".arXiv:0711.0188v1[gr-qc],1Nov2007. [11] B.Allen.\relativisticgravitationandgravitationalradiation".IneditedbyJ.A.MarckandJ.P.Lasota,editors,ProceedingoftheLesHouchesSchoolofPhysics,page373.CambridgeUniversityPress,1997. [12] A.Kosowsky,A.Mack,andT.Kahniashvili.Phys.Rev.D,55:7368,1997. [13] C.Hogan.\GravitationalWaveSourcesfromNewPhysics".IneditedbyS.M.MerkowitzandJ.C.Livas,editors,LaserInterferometerSpaceAntenna,page397.AIPConferenceProceedings823,Mellville,NewYork,2006. [14] M.BenacquistaandHolley-Bockelmann.Astrophys.J.,645:589,2006. [15] A.StroeerandA.Vecchio.\theLISAvericationbinaries".Class.Quant.Grav.,23:S809{S818,2006. [16] M.Haehnelt.page45.AIPConferenceProceedings456,1998. 229

PAGE 230

C.Hopman.\Astrophysicsofextrememassratioinspiralsources".IneditedbyS.M.MerkowitzandJ.C.Livas,editors,LaserInterferometerSpaceAntenna,page397.AIPConferenceProceedings823,Mellville,NewYork,2006. [18] LaserInterferometerSpaceAntenna(LISA)RequirementsFlowdownGuide,February2007. [19] DRSITAT.\LISADRSAccelerationNoiseBudget".LISAProjectinternalreport,2005. [20] A.Cavalleri,G.Ciani,R.Dolesi,A.Heptonstall,M.Hueller,D.Nicolodi,S.Rowan,D.Tombolato,S.Vitale,P.Wass,andW.Weber.\Anewtorsionpendulumfortestingthelimitsoffree-fallforLISAtestmasses".Class.Quant.Grav.,26,(2009). [21] S.Anza.\theLTPexperimentonthelisapathndermission".Class.Quant.Grav.,22:S125{S138,2005. [22] J.Armstrong,F.Estabrook,andM.Tinto.\Timedelayinterferometry".Class.Quant.Grav.,20:S283{S289,(2003). [23] M.Sallusti,P.Gath,D.Weise,M.Berger,andHSchulte.\LISAsystemdesignhighlights".Class.Quant.Grav.,26,(2009). [24] S.Turyshev,M.Shao,andK.NordtvedtJr.\Science,Technology,andMissionDesignfortheLaserAstrometricTestofRelativity".AstrophysicsandSpaceScienceLibrary,349:473{543. [25] S.Turyshev,M.Shao,andK.NordtvedtJr.\ExperimentalDesignfortheLATORMission".arXiv:gr-qc/0410044v1,8Oct2004. [26] J.Yu.\LaserAstrometricTestofrelativity(LATOR)mission".InEditedbyM.Shao,editor,JPLTechnicalMemorandum,volume2200.SPIE,1994. [27] TIE-31:Mechanicalandthermalpropertiesofopticalglass.July2004,Version2. [28] TIE-37:ThermalexpansionofZERODUR,August2006. [29] J.BaerandW.Lotz.\Figuretestingof300mmZERODURmirrorsatcryogenictemperatures".InCryogenicOpticalSystemsandInstrumentsIX,SPIEProc.,volume4822,(2002). [30] D.deChambure,R.Laine,K.vanKatwijk,andP.Kletzkine.\XMM'sX-RayTelescopes".XMMprojectbulletin100,December1999. [31] W.Sokolowski,S.Jacobs,M.Lartc,I.O`Donnell,andC.Hsieh.\DimensionalStabilityofHighPurityInvar36".MechanicalSystemsEngineeringandResearchDivision,JetPropulsionsLaboratory. [32] S.Jacobs,S.Johnston,andD.Schwab.\DimensionalinstabilityofInvars".AppliedOptics,lettertotheeditor,(1984). 230

PAGE 231

M.vanVeggel.TheBasicAngleMonitoringsystem:picometerstabilitywithSiliconCarbideoptics.PhDthesis,TechnischeUniversiteitEindhoven,2007. [34] REACTIONBONDEDSILICONCARBIDE:SFF,PROCESSREFINEMENTANDAPPLICATIONS.DepartmentofMechanicalEngineering,TheUniversityofTexasatAustin,Austin,TX78712. [35] POCOGrahpite.\SUPERSiCmaterialproperties".see [36] SaintGobain.\HexoloySASiliconCarbideTechnicalData".see [37] CoorsTek.\TechnicalCeramics".see [38] S.Suyama,T.Kameda,andY.Itoh.\Developmentofhigh-strengthreaction-sinteredsiliconcarbide".DiamondandRelatedMaterials,12:1201{1204,(2003). [39] OKubaschewskiandCAlcock.MetallurgicalThermochemistry.PergamonPress,NewYork,5theditionedition,1979. [40] S.Black.GettingtoKnow\BlackAluminum".Technicalreport,2008. [41] Electroceramics.Technicalreport,Morganelectroceramics. [42] C.PoizatandA.Benjeddou.\Onanalyticalandniteelementmodellingofpiezoelectricextensionandshearbimorphs".ComputersandStructures,84:1426{1437,(2006). [43] PiezoTechnologies.\K-180LeadZirconateTitanatePiezoelectricCeramicwithHighShearCouplingCoecient,LowPermittivity,andHighTemperatureStability".formoreinformation,goto [44] V.Greco,F.Marchesini,andG.Molesini.\OpticalcontactandvanderWaalsinteractions:theroleofthesurfacetopographyindeterminingthebondingstrengthofthickglassplates".JournalofOpticsA:PureandApliedOptics,3:85{88,(2000). [45] J.-J.Ferme.\OpticalContacting".InD.RimmerG.Roland,R.GeylandL.Wang,editors,SPIEProc.,volume5252ofOpticalFabrication,Testing,andMetrology,(2004). [46] J.vanElp,P.Giesen,andJ.vanderVelde.\AnodicbondingusingthelowexpansionglassceramicZerodur".AmericanVacuumSociety,(2005). 231

PAGE 232

M.Powers.\HermeticGlass-to-AluminumSealsforRFPackagingApplications:AUniversity-IndustryDesignProjectCollaboration".Ineditors:JohnStephensandK.ScottWeil,editors,BrazingandSoldering,Proceedingsofthe3rdInternationalBrazingandSolderingConference,(2006). [48] A.Belohlav.UnderstandingBrazingFundamentals.Technicalreport,TheAmericanWelder,2000. [49] D.Bath,D.Spain,E.Ness,S.Williams,andM.Bougoin.\Evaluationofsegmentedandbrazedmirrorassemblies".Technicalreport,collaborationbetweenCoorsTekandBoostekcompanies.technicalnote. [50] S.Gibb.TwoPartEpoxyAdhesives.Technicalreport,2006.ATechnicalMemoPreparedforOpti521IntroductiontoOpto-MechanicalEngineering. [51] J.Daly.\StructuralAdhesivesforOpticalBonding".SPIEshortcoursenotes,(2001). [52] MasterBondInc.\MasterBondPolymerSystemEP21TCHT-1ProductDescription".Formoreinformationsee [53] E.Ellie,J.Bogenstahl,A.Deshpande,J.Hough,C.Killow,S.Reid,D.Robertson,S.Rowan,H.Ward,andG.Cagnoli.\Hydroxide-CatalysisBondingforStableOpticalSystemsforSpace".Class.Quant.Grav.,22:S257{S267,(2005). [54] D.Gwo.\UltraprecisionBondingforCryogenicFused-SilicaOptics".InSPIEProc.,volume136of3435,(1998). [55] A.vanVeggel,D.vandenEndeb,J.Bogenstahl,S.Rowanc,W.Cunninghamc,G.Gubbels,andH.Nijmeijer.\Hydroxidecatalysisbondingofsiliconcarbide".JournaloftheEuropeanCeramicSociety,28:303{310,(2008). [56] G.Somerjai.\ChemistryinTwoDimensions:Surfaces".CornellUniversityPress,Ithaca,N.Y.,1981. [57] A.PrestonandG.Mueller.\BondingSiCtoSiCusingasodiumsilicatesolution".Inpreparation,(2010). [58] B.LaukeK.SchneiderandW.Beckert.\CompressionShearTest(CST)AConvenientApparatusfortheEstimationofApparentShearStrengthofCompositeMaterials".AppliedCompositeMaterials,8:43{62,(2001). [59] A.Preston,B.Balaban,andG.Mueller.\Hydroxide-BondingStrengthMeasurementsforSpace-BasedMissions".Int.J.Appl.Ceram.Technol.,5(4):365{372,(2008). 232

PAGE 233

A.vanVeggel,D.vandenEnde,J.Bogenstahl,S.Rowan,W.Cunningham,G.Gubbels,andH.Nijmeijer.\HydroxideCatalysisBondingofSiliconCarbide".JournaloftheEuropeanCeramicSociety,28:303{310,(2007). [61] S.Rowan,J.Hough,andE.Ellie.\SiliconCarbideBonding".PleasecontactMr.D.Whitefordforfurtherinformation:D.Whiteford@admin.gla.ac.uk. [62] A.Preston,B.Balaban,G.Boothe,andG.Mueller.\StableMaterialsandBondingTechniquesforSpace-BasedOpticalSystems".InNASAScienceandTechnologyConference,pageSessionC1P3,(2007). [63] J.BertholdIII.DimensionalStabilityofLowExpansivityMaterialsTimeDependentChangesinOpticalInterfacesandPhaseShiftsonReectionfromMultilayerDielectrics.PhDthesis,UniversityofArizona,1976. [64] G.WallisandD.Pomerantz.\FieldAssistedGlass-MetalSealing".JournalofAppliedPhysics,40:3946{3949,(1969). [65] J.BertholdIII,S.Jacobs,andM.Norton.\DimensionalStabilityofFusedSilica,Invar,andSeveralUltra-LowThermalExpansionMaterials".Metrologia,13:9{16,(1977). [66] D.Gwo.\Hydroxide-CatalyzedBonding".UnitedStatesPatentNo.U.S.6548176B1,2003. [67] PersonalcommunicationswithSuvrathMahadevenattheUniversityofFloridadepartmentofAstronomy. [68] A.Preston.\StableMaterialsforLISA".TalkpresentedattheFrontiersinOpticsconference,October2009. [69] B.Young,F.Cruz,W.Itano,andJ.Bergquist.\VisibleLaserswithSubhertzLinewidths".PhysicalReviewLetters,82(3799),(1999). [70] H.KogelnikandT.Li.\LaserBeamsandResonators".AppliedOptics,5(10),(1966). [71] A.Seigman.Lasers.StanfordScienceBooks,1986. [72] R.Cruz.DEVELOPMENTOFTHEUFLISABENCHTOPSIMULATORFORTIMEDELAYINTERFEROMETRY.PhDthesis,UniversityofFlorida,2006. [73] PersonalcommunicationswiththeSpaceandGravityResearchgroupattheUniversityofBirmingham. [74] J.Livas,J.Thorpe,K.Numata,S.Mitryk,G.Mueller,andV.Wand.\Frequency-tunablepre-stabilizedlasersforLISAviasidebandlocking".Class.Quant.Grav.,26,(2009). 233

PAGE 234

E.Black.\AnintroductiontoPoundDreverHalllaserfrequencystabilization".AmericanJournalofPhysics,69(1):79{87,(2001). [76] D.Shaddocketal.\LISAfrequencycontrolwhitepaper".Technicalreport,LISAprojecttechnicalnoteLISA-JPL-TN-823,2009. [77] S.Webster,M.Oxborrow,andP.Gill.\Vibrationinsensitiveopticalcavity".Phys.Rev.A,75:011801(R),(2007). [78] A.Preston,R.Cruz,J.Thorpe,G.Mueller,G.Boothe,R.Delgadillo,andS.Guntaka.\DimensionalStabilityofHexoloySASiliconCarbideandZerodurMaterialsfortheLISAMission. [79] A.Preston,R.Cruz,I.Thorpe,G.Mueller,andR.Delgadillo.\DimensionalStabilityofHexoloySASiliconCarbideandZerodurGlassUsingHydroxide-CatalysisBondingforOpticalSystemsinSpace".InJosephAntebiEditedbyEliAtad-EttedguiandDietrichLemke,editors,ProceedingsoftheSPIE:OptomechanicalTechnologiesforAstronomy,volume6273,(2006). [80] L.Conti,M.DeRosa,andF.Marin.\High-spectral-puritylasersystemfortheAURIGAdetectoropticalreadout".JournaloftheOpticalSocietyofAmericaB,20(3):462{468,(2003). [81] A.PrestonandG.Mueller.\PZT-actuatedtunablecavityforuseintheLISAmission".Inpreparation,(2010). [82] A.Stumner.\1-kVPiezoAmplierKeepsCost,NoiseLow".publishedon [83] S.Lin.\Studyontheequivalentcircuitandcoupledvibrationforthelongitudinallypolarizedpiezoelectricceramichollowcylinder".JournalofSoundandVibration,275:859{875,(2004). [84] B.ChengandM.Reese.\Stressrelaxationandestimationofactivationvolumeinacommercialhardpztpiezoelectricceramic".Bull.Mater.Sci.,24(2):165{167,April(2001). [85] NewFocusCorporation.\FMSpectroscopyWithTunableDiodeLasers,Applicationnote7".see [86] O.OrlovandV.Ustyugov.\Molecularcesiumfrequencyreferenceforfrequencystabilizationofa1.06mnd:yaglaser".Sov.Tech.Phys.Lett.,12:120{121,1986. [87] P.FritschelandR.Weiss.\Frequencymatchofthend:yaglaserat1.064mwithalineinco2".Appl.Opt.,31:1910{1912,1992. 234

PAGE 235

R.Benedict,D.Drummond,andL.Schlie.\AbsorptionspectraoftheCs2molecule".J.Chem.Phys.,66:4600{4607,1977. [89] A.Mak,O.Muravitsky,O.Orlov,andV.Ustyugov.\Newlaserforinterferometrywithlong-termfrequencystabilizationat1.06montomolecularcesiumstandard".InEditedbyZ.JaroszewiczandMPluta,editors,inInterferometry'89,volume1121,pages478{484.SPIE,1989. [90] E.Inbar,V.Mahal,andA.Arie.\Frequencystabilizationofnd:yaglasersto133Csssub-dopplerlinesnear1064nm".J.Opt.Soc.Am.B,13(7):4600{4607,July1996. [91] R.Schilling,P.Bender,J.Livas,T.Pedersen,A.Martino,P.McNamara,O.Jennrich,A.Preston,andG.Mueller.NotestakenfromtheLISAtelescopesubgroupteleconferences.internaldocument,February2006. [92] D.RobertsonandJ.Hough.InterferometryforLISA.Class.QuantumGrav.,13:A271{A277,1996. [93] et.al.RFleddermann.\Measurementofthenon-reciprocalphasenoiseofapolarizationmaintainingsingle-modeopticalber".JournalofPhysics:ConferenceSeries,154,(2009). [94] K.McKenzie.\Fibernoisemeasurementsandphotorecieverdevelopment".Pre-sentedatthe7thInternationalLISASymposium,(2009). [95] R.Bergh,H.Lefevre,andH.Shaw.\AnOverviewofFiber-OpticGyroscopes".JournalofLightwaveTechnology,LT-2(2),(1984). [96] E.Post.\Sagnaceect".Rev.Mod.Phys.,39:475{494,(1967). [97] H.LefevreandH.Arditty.\TheoreticalbasisofSagnaceectinbergyroscopes".Fiber-OpticRotationSensorsandRelatedTechnologies,32(8),(1982). [98] J.Nodaandet.al.\Polarization-maintainingbersandtheirapplications".J.LightwaveTechnol.,4(8):1071,(1986). [99] S.Barke.\PhaseCharacterizationofEOMSidebandswellwithintheMissionRequirementsoftheLaserInterferometerSpaceAntenna".TechnicalReportPresentationgivenatthe7thannualLISAsymposium,AEIinHannover,2009. [100] A.Preston.\Back-linkFiberStability".Talkpresentedatthe7thInernationalLISASymposium,2008. [101] P.Beyersdorf,R.Byer,andM.Fejer.\ResultsfromtheStanford10mSagnacinterferometer".ClaasicalandQuantumGravity,19(7):1585{1589,(2002). [102] R.Fleddermannetal.\Measurementofthenon-reciprocalphasenoiseofapolarizationmaintainingsingle-modeopticalber".InJournalofPhysics:Confer-enceSeries,volume154,(2009). 235

PAGE 236

G.Heinzel,A.Rudiger,andR.Schilling.\SpectrumandspectraldensityestimationbytheDiscreteFourierTransform(DFT),includingacomprehensivelistofwindowfunctionsandsomenewat-topwindows".Technicalreport,MaxPlankInstituteforGraviationalPhysics,HannoverGermany,February152002. 236

PAGE 237

AlixPrestonwasborninBartlesville,Oklahoma,butmovedtoFortCollins,Coloradowhenhewasve.Hespenthischildhoodandadolescencefascinatedwithmathandscience.HetookseveralAdvancedPlacementcoursesinHighSchoolandgraduatedasclassValedictorian.AlixattendedtheColoradoSchoolofMinesinGolden,Coloradoandgraduatedinthetop20%ofhisclasswithaBachelorofScienceinengineeringphysicsinadditiontotakingseveralgraduatelevelcoursesinbothmathandphysics.IntheFallof2004AlixbeganhisstudiesattheUniversityofFloridaandsoonbeganworkingwithDr.GuidoMuellerondeterminingthedimensionalstabilityofmaterialsandbondingtechniquesfortheLaserInterferometerSpaceAntenna(LISA).ThisworkcarriedintoresearchingseveralotheraspectsofLISA,andisthebulkofhisdissertation. 237