Optical waveguide chemical sensors


Material Information

Optical waveguide chemical sensors
Physical Description:
ix, 161 leaves : ill. ; 29 cm.
Weiss, Martin Neil
Publication Date:


Subjects / Keywords:
Optical wave guides   ( lcsh )
Chemical detectors   ( lcsh )
Electrical and Computer Engineering thesis, Ph. D   ( lcsh )
Dissertations, Academic -- Electrical and Computer Engineering -- UF   ( lcsh )
bibliography   ( marcgt )
non-fiction   ( marcgt )


Thesis (Ph. D.)--University of Florida, 1996.
Includes bibliographical references (leaves 150-160).
Statement of Responsibility:
by Martin Neil Weiss.
General Note:
General Note:

Record Information

Source Institution:
University of Florida
Rights Management:
All applicable rights reserved by the source institution and holding location.
Resource Identifier:
aleph - 002283410
oclc - 38254005
notis - ALN6532
System ID:

Full Text







This work is dedicated to my parents,
Edward and Catherine


John Donne once wrote, "no man is an island, entire of itself". This dissertation

stands as a testament to those words. As I look back on the various experiments,

processing techniques (read "tricks"), and theoretical models devised herein, I cannot help

but recall the vast myriad of people who aided me in the completion of this study. Most of

all I would like to thank Dr. Ramakant Srivastava, my advisor. In addition to providing me

with the wide range of resources and training I needed for my research, Dr. Srivastava was

a constant source of encouragement and support. Without his leadership and guidance, I

doubt I would have been able to accomplish as much in this work. I would also like to

thank Dr. Peter Zory, the cochairman of my Ph.D. committee, and Dr. James Winefordner,

for giving me access to their laboratories for materials characterization. Thanks also go to

Dr. Toshikazu Nishida, Dr. Stephan Schulman, and Dr. Ewen Thomson for serving on my

Ph.D. committee. I would like to thank my friends Howard Groger and Dr. Peter Lo of the

American Research Corporation of Virginia who were in fact the people who first sparked

my interest in the chemical sensor field. I am indebted to James Chamblee, Tim Vaught,

Steve Shine, and Frank Tavano for maintaining the Electrical Engineering department's

cleanroom where I spent many an hour fabricating my various sensor devices. I am also

grateful to Dr. Ben Smith for help in characterizing the fluorescence spectra of dye-doped

polymers and Dr. Sheng Li for the HeCd laser used in the later chapters. Some of the more
obscure tips on material processing were provided by Dr. Drew Roza of OCG

Microelectronics Materials, Inc., Dr. Charles Sullivan of Sandia National Laboratories,

and Richard Steppel of Exciton.

A portion of my dissertation work was performed during a brief visit to the Federal
University of Pernambuco in Recife, Brazil. I would like to thank Dr. Cid de Araujo, and

his colleagues Dr. Ricardo Correia, Dr. Anderson Gomes, Dr. J. F. Martins-Filho, and

Breno Neri for making my wonderful stay very productive.

I am grateful to the Microfabritech program at the University of Florida, the

National Science Foundation, and the American Research Corporation of Virginia for

funding this work.

Lastly, I am deeply grateful to my beloved fiancee Vicki and my dearest parents,

Edward and Catherine, for their unending love and support over these long years of

research during my career at the University of Florida.


ACKNOW LEDGM ENTS .................................................................................. iii

A B STR A C T .................................................................. ........................................viii


1.1 Overview ............................................................ ................ 1
1.2 Organization ...................................... ................. .......... ...... 2
1.3 Sensor Evaluation Criteria ................................... ............. 2
1.3.1 Sensitivity ................................... .......... .............. 3
1.3.2 Selectivity ................................... .......... .............. 3
1.3.3 R eversibility ......................................... .............. 3
1.3.4 Cost Effectiveness ................................... ............ 4
2.1 Review of Waveguide Theory .............................................. 5
2.2 Modelling Graded Index Waveguides: The Transfer Matrix ....8
2.3 Numerical Solution of the Transfer Matrix .............................12
2.4 Summ ary ......................................... .........................................13
3. SURFACE PLASMON WAVEGUIDE SENSORS ...............................14
3.1 Overview of Surface Plasmon Sensors .................................... 14
3.2 Theoretical Formulation of Surface Plasmon Resonance ..........15
3.2.1 Modelling of Surface Plasmon Waveguides ............... 18
3.2.2 Integrated-Optic Surface Plasmon Waveguide
Structures ................................... ...........................24
3.3 Experimental Investigation of Surface Plasmon Devices ..........29
3.3.1 Device Fabrication ..............................................29
3.3.2 Experimental Measurement of Refractive Index ........32
3.3.3 Humidity Measurement ............................................40
3.4 Proposed Surface Plasmon Structures With Improved
Perform ance ................................... .....................................43
3.5 Application of Surface Plasmon Resonance to Monolayer
D election ............................................................................49
3.6 Conclusion ................................................ ......................... 51

WAVEGUIDES ...........................................................................52
4.1 Advantages of Polymer Waveguides ........................................52
4.2 Fabrication of Polyimide Waveguides .....................................53
4.2.1 Substrate Preparation ....................................... ...54
4.2.2 W afer Priming ................................................ ..... 54
4.2.3 Polyimide Deposition ............................................56
4.2.4 Photolithography ................................... ............ 57
4.2.5 Curing of Polyimide Films .......................................58
4.2.6 Doping (Optional) ...............................................58
4.2.7 Endfacet Preparation ........................................ ...59
4.3 Characterization of Polyimide Films .......................................60
4.4 Summary ....................................... ....................................66
WAVEGUIDES ...........................................................................68
5.1 The Evanescent-Wave Absorption Sensor ..............................68
5.2 Detection of Aqueous Ammonia .............................................74
5.2.1 Choice of Sensing Layer Materials ...........................74
5.2.2 Fabrication of Oxazine-Doped Nafion .....................77
5.2.3 Characterization of Oxazine-Doped Nafion ...............79
5.2.4 Bulk Ammonia Sensor Response .............................81
5.2.5 Demonstration of Reversibility .................................87
5.2.6 Selectivity of the Nafion Response ...........................89
5.2.7 Waveguide Issues in Evanescent Wave Sensor
D esign ....................................................... ............91
5.2.8 Evanescent Wave Sensor Fabrication .......................92
5.2.9 Performance of Evanescent Wave Absorption
Ammonia Sensors .............................................92
5.3 Summ ary ....................................................................................100
SENSORS ........................................ ................. ............................ 101
6.1 Introduction ................................... ........................................101
6.2 Principle of Operation ..............................................................102
6.3 Theoretical Formulation of Fluorescence Capture by Guided
M odes ...................................................................................... 102
6.4 Sensor Fabrication ................................................................... 108
6.4.1 M materials ................................................................... 108
6.4.2 Waveguide Fabrication ............................................116
6.5 Sensor Characterization .......................................................116
6.5.1 Laser Pum ping .......................................................... 116
6.5.2 LED Pumping ........................... ......................117
6.5.3 Conventional Evanescent Wave Measurement ........... 121
6.6 Limitations of Fluorescence-Excited Waveguide Sensors ........122
6.7 Summary ........................................................... ................. 122

7.1 Introduction ................................... ..........................................125
7.2 Optical Amplification ........................................................... 127
7.3 Characterization of Optical Gain in Dye-Doped Polyimide
Waveguides ..........................................................................128
7.3.1 Active Waveguide Fabrication ..................................130
7.3.2 Experimental Set-up ...........................................130
7.3.3 Measurement of Optical Gain ...................................132
7.4 Optical Amplifiers as Chemical Sensors .................................137
7.5 Conclusion .............................................................................138
8. CONCLUSIONS AND FUTURE WORK ............................................139
8.1 Summary ............................................................................... 139
8.2 Future Work ............................................................................... 140
8.2.1 Improved Fluorescence-Excited Evanescent
Waveguide Absorption Sensors ..............................141
8.2.2 Active Waveguides for Chemical Sensing .............. 145
8.2.3 Polyimide Waveguides as Selective Chemical
Recognition Elements ..........................................147

LIST OF ACRONYMS........................................................................148

REFERENCES ..........................................................................................150

BIOGRAPHICAL SKETCH .................................................................. 161

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



Martin Neil Weiss

December 1996

Chairman: Ramakant Srivastava
Major Department: Electrical and Computer Engineering

Optical sensors convert information about the surrounding environment such as

temperature and chemical composition into changes in light intensity and phase. Compact,

lightweight waveguide and fiber-based sensors offer a high level of detection capability

with an inherent immunity to electromagnetic interference. Several types of waveguide

sensors for chemical detection have been explored, both theoretically and experimentally.

Primary emphasis has been placed on optimizing sensitivity through the design of the

waveguide structure, rather than the sensing material.

Using surface plasmon resonance, sensors comprised of metal-clad dielectric

waveguides have been designed with enough sensitivity to measure refractive index

variations on the order of 10-5 and adsorbed film thicknesses of less than 1 nm. The basic

surface plasmon resonance (SPR) waveguide structure has been modified to include a

dielectric tuning layer, which simplifies the design process and reduces the number of

materials needed in sensor fabrication. A humidity sensor was created by coating an SPR

waveguide with a thin layer ofNafion, an ion-exchanging fluoropolymer whose refractive

index varies with atmospheric moisture content.

Conventional endfire-excited evanescent wave absorption (EWA) sensors were

produced by coating waveguides with materials whose optical absorption changes upon
analyte exposure. Ammonia in aqueous solution has been detected at sub-part-per-million
levels using polyimide waveguides clad with Nile-blue doped Nafion. A novel variant of

the EWA sensor, which uses fluorescence generated inside a waveguide to probe the

cladding absorption, was also studied. This latter device, known as the fluorescence

excited evanescent wave absorption sensor provides a higher level of sensitivity and is

virtually free from the stringent alignment tolerances that often plague other EWA sensors.

Lastly, a sensing technique based on measuring analyte-induced perturbations in
the optical gain of a waveguide amplifier or laser was proposed. An optical amplifier, with

a gain of 14.4 dB at 670 nm, was achieved in a cresyl violet-doped polyimide waveguide.

Unfortunately, fluorophores were found to exhibit minimal analyte sensitivity when

immobilized in the polyimide matrix.


1.1 Overview

Modem society's recent trend toward increased environmental awareness has led

to a need for development of advanced chemical detection systems. Sensors which utilize

optical detection techniques, such as fluorescence excitation, have proven to be highly

effective in this regard. Highly sensitive, compact, and lightweight optical sensors have

been demonstrated for a large number of chemicals.

Optical chemical sensors employ either bulk or integrated-optical (IO)
components, such as fibers and waveguides, as sensing elements. Analyte interactions

with the sensing element are converted into optical information, such as light intensity or

phase, through a number of transduction mechanisms. A wide variety of optical

phenomena can be used for analyte detection, including surface plasmon resonance,

fluorescence excitation, optical absorption measurement, and refractive index

perturbation. The most common examples of optical chemical sensors are fiber evanescent

wave devices and bulk surface plasmon resonance devices.

Unlike their electronic counterparts, optical sensors neither produce nor are

affected by electromagnetic interference. In many cases, optical sensors can respond more

rapidly and with a higher sensitivity than electronic sensors. Although optical device

fabrication is in general not as well-established as semiconductor processing, IO sensors
still benefit greatly from optoelectronic technology developed for the optical

telecommunications industry.

Of the various types of optical sensors, waveguide-based devices are in general

more versatile than bulk ones. This is particularly true of integrated-optic sensors, where

multiple sensing and reference channels can be built on the same substrate. Processing
functions needed for conditioning of the sensor output signal, such as filtering,

polarization splitting, and wavelength demultiplexing can also be performed on-chip by

additional waveguide structures. In addition, integrated-optic chemical sensors are smaller
and lighter than bulk ones, and can be easily deployed in remote areas via optical fiber

delivery. Despite the promising outlook though, commercial development of integrated

optic sensors has proceeded rather slowly, due to concerns about durability in harsh

environments and high packaging costs.

1.2 Organization

We begin with a brief review of waveguide theory in chapter 2 to establish basic

concepts and terminology. A numerical simulator for calculating waveguide mode field

distributions and propagation constants is also introduced. In chapters 3 through 7, four

optical waveguide sensors are developed: the surface plasmon resonance waveguide, the

evanescent wave absorption sensor, the fluorescence-excited evanescent wave absorption
sensor, and the chemically-sensitive amplifier. The waveguide principles underlying the

operation of each device are presented, along with modelling predictions. Considerable

effort has been devoted to the sensor fabrication and experimental characterization. In this

work, the primary emphasis will be placed on the advantages and disadvantages

associated with each structure, rather than the actual sensing chemistry.

1.3 Sensor Evaluation Criteria

Each of the sensors presented in chapters 3 through 7 will be evaluated with regard

to the criteria of sensitivity, selectivity, reversibility, and cost effectiveness.

1.3.1 Sensitivity

The sensitivity of a device with respect to a given analyte is defined as the

derivative of the sensor output with respect to the concentration of that analyte.1 The

minimum amount of analyte which produces a measurable response (i. e. above the value

determined by the signal-to-noise ratio) is known as the lower detection limit (LDL). The

range of analyte concentration over which the sensitivity is non-zero is known as the

dynamic range. Sensitivity can be improved through optimization of waveguide

properties. However, as will be seen in chapters 5 and 6, certain waveguide structures are

inherently more sensitive than others.

1.3.2 Selectivity

Selectivity is the ability of a sensor to differentiate between multiple analytes.1

Devices with low selectivity are prone to mistake one chemical species for another.

Sensors which contain analyte-specific receptors, such as proteins or antibodies, offer

highly selective responses. As will be seen in chapters 4 and 5, polymer-based sensors can

also have some inherent level of selectivity when matrices with analyte-specific diffusion

properties are employed. In addition, individual sensor elements with generalized (i. e.

nonselective) but differing analyte responses can often be combined into an array, forming

a highly selective sensor system.

1.3.3 Reversibility

Reversibility is the extent to which a sensor returns to its initial state after exposure

to an analyte.2 We use the term reversible to specifically describe systems which

automatically return to their pre-analyte state when placed in an analyte-free environment.

In some cases, irreversible sensors can be restored to their initial state by application of an

outside influence, such as a rinse chemical. We refer to this latter type of sensor as

resettable. Only sensors which can be fully restored to their pre-analyte condition can be

reused effectively. Devices which are neither reversible or resettable are limited to a single


1.3.4 Cost Effectiveness

Optical waveguide devices are notorious for having high packaging costs. This

single aspect, perhaps more than any other, has hindered widespread commercialization of

integrated-optic devices. Not surprisingly, disposable waveguide sensors that are

discarded after only one use are not economically viable. Waveguide-based sensors must

either be fully reversible (or resettable), thereby ensuring long operating lifetimes, or

utilize inexpensive packaging schemes in order to be cost effective. Thus, packaging

issues will play an important part in the discussions of each of the sensors presented



2.1 Review of Waveguide Theory

We shall begin with a brief review of basic waveguide concepts and establish a few
definitions and terminology in order to facilitate our later discussion of guided wave

chemical sensors. When light travelling in a dielectric medium of refractive index n2 is

incident on a second medium with index n, < n2 as shown in figure 2.1(a), total internal

reflection (TIR) occurs when the angle of incidence, 0, exceeds the critical angle, defined


0e = asin (nl/n2) (2.1)

By bringing two such interfaces into close proximity, as shown in figure 2.1(b), light can

be constrained to propagate only in the direction tangential to the interfaces. This structure

represents a one-dimensional waveguide. The higher-index central region is referred to as

the core, while the lower-index surrounding regions are known as the claddings. In many

waveguide sensor applications, it is the top layer of index n1 which is exposed to the

analyte. This layer is called the superstrate or the sensing layer. It is also useful to

characterize the refractive index mismatch between the top and bottom claddings in terms

of the normalized waveguide asymmetry parameter, defined as3

2 2 2 2
aE = (n- n / (n2 n3) (2.2)

For energy to propagate a significant distance in the waveguide, two criteria must

be satisfied: total internal reflection (at both interfaces) and constructive interference at all

points along the ray path. As a result of the latter requirement, only a discrete set of rays,


\ 0 0

ni (cladding)

n2> nl, n3

Figure 2.1 Light incident on dielectric media. n2 > nl
(a) single interface
(b) two interfaces waveguidee)


d -

corresponding to waveguide modes, can be guided. For the guided light, two polarization

states are possible. When the electric field, E, is perpendicular to the plane defined by the
direction of propagation and the interface normal vector, the field is said to be TE-
polarized. Conversely, when E is parallel to this plane, the field is TM-polarized.4 For the
TE guided modes, transverse field distribution can be written as

Ae3, y < 0 (2.3)
E(y) = Be + Ce- ,0
De-, (y-d) > d

i= ko n2-N2

Y1 = k -n2

3 = ko N2 n
ko = 2n/A0 (2.4)

where A, B, C, and D are constants, Xo is the vacuum wavelength, and N is called the

mode index. The total field associated with the ith TE waveguide mode is

E1 (y,z) = RE E(y)e-ji2 (2.5)

where ki = (27x/X0)Ni is called the propagation constant. In lossy waveguides, ki becomes
complex-valued, with an imaginary term which characterizes mode attenuation:

ki = kreal jkma (2.6)

The mode indices and field distributions correspond to the eigenvalues and eigenfunctions
of the waveguide structure and are obtained by solving the eigenvalue (characteristic)
equation. TM mode fields have forms analogous to those presented above for the TE case,
with some minor modifications.

As we will see later, the performance of a waveguide sensor depends critically on

the amount of power propagating in the sensing region. For the ith mode, the fractional
power propagating in the region yl < y < Y2 is defined as5

Y2 0
ri = (Ei x H) dy/ (E x Hi) dy (2.7)
YI -00

and simplifies to

TE = E2 (y)dy/E2(y) dy (2.8)
Y1 -00

for the TE polarization.

2.2 Modelling Graded Index Waveguides: The Transfer Matrix

We have used the well-known transfer matrix technique for solving the waveguide

eigenvalue problem for each structure. In this approach, a continuous refractive index

profile, n(y), is quantized into a series of slabs, each with a constant index value, ni, as

shown in figure 2.2(a). Application of boundary conditions at the interfaces between

adjacent slabs allows the mode indices and transverse field distributions of the structure to

be determined.

Starting with the TE polarization, with Ey = Ez = Hx = 0, it is evident from figure

2.2(b), that the total electric field in the ith layer, Exi, is the sum of the components
transmitted through the i-1h interface and reflected off the ith interface. Defining Ei' and

Ei- respectively as the components travelling toward and away from the ith interface and

the total depth d'i as

n (2.9)
d'i = ,di






E2+ E2-





Figure 2.2 Determination of waveguide mode indices and field distributions by the
transfer matrix method.
a) discretized refractive index profile, n(yi)
b) multilayer stack formulation of the waveguiding problem

where n is the total number of layers. The total field confined in layer i may be expressed


E = + -jkiyY y (2.10)
EXi~=Ee +E e d'i-i

where k = jn -2. From equation (2.10), it can be seen that real values of ky (ni > N)

yield sinusoidal solutions while imaginary values of kyi (ni < N) result in exponentially

growing or decaying fields. We expect the field in the core to be characterized by an

oscillatory solution while fields in the surrounding media decay. To solve for the actual

mode indices and fields, the tangential electric fields as well as their derivatives along the

y-direction are forced to be equal at each interface. Thus,

Exi = Ex(i-_) y=di_ (2.11)

dExi dEx(i 1) (2.12)
S =YY d'(2.12)
dy dy A

Noting that aExi/oy = kyExi and inserting (2.10) into (2.11) and (2.12) yields

E+ e-jkyid'i-I jkyid-i- (2.13)
y = d'i-I
= E 1e + E kiie I Y = d' I

+- i e-jky(i-)d'i+ k E jky~ U d
-k(i 1) i-1e y(i-1) i-i
= kyiE e-jk d.i-I + kyiE-ejkid'i -y = d'

Finally, through equations (2.13) and (2.14) the fields in the top layer can be expressed in

terms of those of the bottom layer as

E = M En (2.15)

M= mi (2.16)

1+k2ky e(ky(i-)-kyi)d'i- 2kyi j(ky(- + kyi) d'i-I

= y(i 1) y ky(i- 1) (2.17)
I Ji ( +k )ky)d'i- 1 + P2ki j (ky(i-1) kyi)d'i-
ky (i-i) ) ky(i -1)

In (2.17), P is a constant which equals unity for the TE polarization. When radiation

couples into one of the guided modes of a waveguide consisting of n layers, El" and En+
must be zero. Thus, determining the mode indices of the structure is equivalent to solving
for the roots of the equation M,1 (N) = 0. Inserting the values of mode index into
equation (2.15) yields the corresponding field distribution. Under this algorithm, the
integrated intensity distribution of each mode has an arbitrary value:

JEi (y)12dy = Ki (2.18)

where the subscript designates the mode order. For convenience, we define the normalized
mode fields as

ei (y) -E(y) (2.19)

so that

Plei(y)12dy= 1 (2.20)

The derivation of the transfer matrix for the TM polarization is analogous to the
TE analysis presented above. In the TM case, the fields are expressed in terms of Hx,
rather than Ex, and an identical transfer matrix is obtained, with the exception that in
(2.17), P = ni_ /ni.

2.3 Numerical Solution of the Transfer Matrix

In order to solve for the roots of the transfer matrix, the Newton-Rahpson
algorithm is employed6'7. In this method, the function Mil(N) is expanded in a Taylor
series about an initial point as8

M, (N+8) = M, (N) +M', (N)8+M", (N) 82/2+... (2.21)

where 8 is the difference between the estimated and the actual values of the root. Setting

M11(N+8) = 0 and retaining only the linear terms in (2.21) yields

8 = -M, (N)/M'1, (N) (2.22)

This correction is then added to the current estimate of the root and the process is repeated
iteratively until 8 becomes sufficiently small. Starting values for each of the roots of

M11(N) are found from a course sampling of the function, during which the structure is
assumed to be lossless.9 When multiple roots exist, the algorithm generally converges to
the one which is nearest to the initial estimate.
The Newton-Raphson technique is attractive for root finding because it converges
very quickly. In fact, it can be shown that the magnitude of the error associated with the
difference of the actual value of a root versus its value after a given iteration decreases
quadratically.7 In practice however, stability issues, such as the presence of local minima
in the function of interest, often make it necessary to allow the algorithm to update the

current estimated value of the root with only a fraction of the full Newton step. Since the

Newton-Rahpson technique is easily extended to solving multidimensional roots, it is

useful for determining the roots of lossy structures, such as metal-clad waveguides. We
have been able to duplicate theoretical results given by Harris and Wilkinson'o in the

modelling of a surface plasmon waveguide structure, thus confirming the accuracy of our


2.4 Summary

A basic introduction of waveguide concepts and terminology has been presented.

We have introduced the concept of fractional power which will figure prominently in the

remainder of this work. A numerical approach to solving the waveguide eigenvalue

problem has also been described. In the following chapters, the simulator will be used

extensively for modelling and optimizing a number of waveguide-based chemical sensors.


3.1 Overview of Surface Plasmon Sensors

Optical surface plasmon resonance (SPR) devices provide a highly sensitive
means for detecting environmental changes involving small perturbations in refractive

index. Sensors of this type have been employed in many diverse fields, ranging from

pollution monitoring11 and humidity measurement12'13 to immunoassay14 and molecular

self-assembly studies.15,16

SPR sensors operate by measuring changes in the refractive index of their

surrounding media.17 As such, they are inherently generic devices with regard to what is
actually detected. An SPR sensor can be made to detect a specific environmental

characteristic by coating the device with a layer of material whose optical properties are

changed by the occurrence of a particular event, such as exposure to a certain chemical.

Such a medium is often referred to as a transducing layer. In immunoassay studies, for

example, this is commonly done by taking advantage of the natural affinity of

complementary protein-ligand pairs.15 An SPR device coated with one material from the
pair becomes a sensor for the other. Antibody-antigen complexation may be monitored in
the same fashion.

Most commonly used SPR sensors consist of a dielectric prism with a metal
cladding of typically either gold or silver on one face.14'15'16 This arrangement is known

as the Kretschmann configuration. In this setup, a collimated beam of light shines into the
prism through one of the clear facets, reflects off the metal film and exits through the
remaining clear facet.18 The incident beam must be TM-polarized with respect to the

metal film. Surface plasmon excitation can be observed by monitoring the reflected power

as a function of the angle the beam makes on the front of the prism. At a particular angle

of incidence, Osp the reflectivity drops to nearly zero. Osp is highly dependent on the
refractive index of the surrounding media. However, these systems tend to be somewhat

cumbersome and are better suited to laboratory settings than field conditions.

A considerable improvement in the functionality of such sensors can be achieved

by taking an integrated optic (10) approach and using SPR waveguides. The 10 format

enables the realization of devices which are compact and lightweight and, furthermore,

allows the possibility of additional signal processing, such as polarization and wavelength
filtering, to be performed on-chip. In addition, multiple sensing elements and reference

channels can be incorporated into a single device. Guided-wave SPR sensors have been

explored previously using a number of geometries, including D-fiber,17 tapered fiber,19

side-polished fiber,20'21 and ion-exchanged GRIN waveguides.11,13,22,23

3.2 Theoretical Formulation of Surface Plasmon Resonance

A surface plasmon is a lossy TM-polarized wave supported by a metal-dielectric

interface.4'24 Physically, the plasmon wave is an optically excited electron plasma

oscillation in the metal. When the metal is a film of finite thickness, individual surface

plasmon waves are supported on both metal-dielectric interfaces. Furthermore, if the

metal film is sufficiently thin (on the order of the penetration depth of the optical wave),

these two waves will couple to form the so-called symmetric and antisymmetric bound

and leaky surface plasmon modes.20

The simplest integrated optic SPR waveguide sensor consists of a dielectric single

mode waveguide overlaid with a thin layer of metal. A dispersion plot of the uncoupled

waveguide TM mode and the plasmon modes is shown schematically in figure 3.1. When

the propagation constants of the waveguide mode and an SP mode are equal, the two
couple to form a lossy normal mode whose attenuation is proportional to the fractional





Figure 3.1. Qualitative dispersion of the individual (uncoupled) TM propagation
constants of the waveguide, kWG, symmetric bound SP mode, k"B, and
antisymmetric bound SP mode, kAB. The normal TM mode of the composite
(coupled) structure is shown by the dashed lines. Field distributions for the
individual plasmon modes are shown in the inset.

symmetric antisymmetric
bound plasmon bound plasmon




power propagating in the metal layer. When this condition, also known as phase-matching,
is not satisfied, propagation loss is considerably lower. SPR excitation in these devices is

thus strongly wavelength dependent. Wavelengths less than XSPR cause the TM

waveguide mode to couple primarily to the symmetric bound plasmon mode, while

wavelengths greater than this mainly excite the antisymmetric bound plasmon mode. TE-

polarized waveguide modes do not interact with the surface plasmons and experience a

small, relatively wavelength-independent loss due to the presence of the metal layer.

The response of the IO SPR waveguide is characterized in terms of the

polarization extinction ratio (PER), defined at a particular wavelength as the ratio of the

propagation losses of TM to TE-polarized normal modes, or

PER = AT- ATE (3.1)

where ATM,TE are the propagation losses of the TE and TM modes in dB/cm. The losses

are related to the normal mode propagation constants as

A = 10log{exp(-2ki z)}/z (3.2)

where ki = (2r/A0) Ni, X0 is the vacuum wavelength, z is the propagation length, and Ni is

the imaginary part of the mode index. Since plasmon resonance induces large losses in the

TM mode, while leaving the TE one relatively unaffected, PER generally takes on
negative values. For single-mode waveguides, the PER varies linearly with device length.

In multimode devices, it is necessary to account for interference between the normal

modes when computing the extinction ratio.25 By exciting SPR waveguides with

circularly polarized light, the TE signal can be used as an internal reference for signal

normalization since its loss is relatively independent of the excitation wavelength. We

define the peak resonance wavelength, XSPR, as the wavelength at which the magnitude of

the polarization extinction ratio achieves its largest negative value. Perturbations to the

refractive index of the region immediately surrounding the sensor, also known as the

superstrate, affect XSPR by producing unequal changes in the propagation constants of the

uncoupled SP and waveguide modes. To simplify further discussions, we will refer to the
largest negative value of the polarization extinction ratio as PERma.

superstrate adsorbedd film)

high index tuning layer overlay
t. low index buffer layer
waveguide --
input signal substrate output signal

Figure 3.2. Schematic representation of the basic surface plasmon waveguide sensor.

3.2.1 Modelling of Surface Plasmon Waveguides

SPR waveguides have been modelled extensively, using the previously described
numerical simulation. The first device considered here is based on a K+-Na+ ion-
exchanged planar waveguide in BK7 glass, as shown in figure 3.2. In order to excite
resonance, several dielectric films a thin metal layer are deposited on the top surface of the
waveguide. We refer to these films, including the metal one, as the plasmon overlay. For
reasons that will be discussed later in section 3.2.2, the plasmon overlay in this case is
comprised of thin films of SiO2 (buffer), TiO2 (tuning), and gold, at thicknesses of 500
nm, 55 nm, and 30 nm respectively. The K+-Na+ ion-exchange in BK7 glass is well
characterized and produces high quality waveguides with a graded refractive index
(GRIN) profile given by26

n (x) = AnTE, TMerfc ( (x t) /dx), x to


where dx is the waveguide depth, to is the total thickness of the plasmon overlay, and
AnTETM are the respective surface index changes for the TE and TM polarizations
respectively. In order to ensure single mode operation in the red to infrared portion of the

spectrum, dx was chosen to be 3 gim. Although the K+-Na+ ion-exchange process in BK7
glass is known to produce birefringent waveguides, with AnTE = 0.008 and
AnTM =0.0092, only the latter value was used for modelling purposes since the

attenuation experienced by the TE mode is small (< 1 dB/cm) and relatively independent

of An. Refractive index dispersion data for the materials comprising the structure in figure
3.2 are taken from references 27-30.

In order to best illustrate the function of a basic SPR waveguide as a sensor, we

first treat the superstrate region as a thin film of index 1.415 adsorbed onto the surface of

the device and calculate PER as a function of wavelength. By 'thin,' we mean that the

thickness of the adsorbed film is less than the penetration depth of the evanescent wave at

the surface. The medium surrounding the device is assumed to be air (n = 1). As shown in

figure 3.3, prior to film adsorption, the calculated PER spectrum is essentially flat, with

the exception of a small positive-valued peak in the green region. This positive peak is not
related to the surface plasmon effect but rather arises from coupling between leaky TE
modes in the tuning layer and the guided TE mode. Upon the adsorption of a thin film,

surface plasmon resonance occurs and is evident as a large, negative-valued dip in the
PER. In this case, the full-width half-maximum (FWHM) of the resonance is about 30 nm.
XSPR increases as the adsorbed layer becomes thicker. At the same time, the magnitude of

pERmax increases since the optical waveguide mode becomes less tightly confined at the

longer wavelengths and interacts more strongly with the metal film. This effect continues

until the thickness of the film exceeds the penetration depth of the evanescent wave at the

surface, at which point XSPR becomes constant. Thus, film thickness can be measured
simply by monitoring the extinction ratio at a wavelength in the vicinity of SPR

E -20
S" '*\ \" /
0 200 nm \\ *
S-40- --r v^
o -40 I
235 nm
o 1
300 nm \ i
-80 "
1000 nm

0.5 0.55 0.6 0.65 0.7 0.75
wavelength (lim)

Figure 3.3. Simulated device extinction ratio as a function of wavelength with a thin
adsorbed film of refractive index 1.415 as the superstrate. Adsorbate
thicknesses are indicated next to the respective curves.

Intensity distributions corresponding to the cases from figure 3.3 are shown for the

TM and TE modes at 670 nm in figures 3.4(a) and (b), respectively. Plasmon excitation,

evidenced as an enhancement of the TM field at the air/metal interface (x = 0 in

figure 3.2), becomes stronger at 670 nm as the adsorbed film thickness approaches

1000 nm. As expected, the increase in the fractional power propagating in the metal layer

results in higher loss for the TM mode. In contrast, the TE mode shows no surface field


Extending this analysis to the case of an infinitely thick superstrate, theoretical

plots of PER as a function of superstrate refractive index are presented in figure 3.5, using

several different excitation wavelengths. For each wavelength, there exists a unique value

of superstrate refractive index which maximizes the polarization extinction ratio. This

variation of PER with superstrate index can be used to measure the latter. We define the

minimum detectable change in the superstrate refractive index as

Anin Mx PER (n)
min -
=LPER (?, n)
anc (3.4)
n =n

where M is the signal-to-noise ratio (as a percentage) and no is the nominal superstrate

index. If we assume a superstrate index of 1.42 and a signal-to-noise ratio of 20 dB

(1% measurement precision), using a 670 nm excitation source yields Anmin 7x10-5

As is usually the case, there is a trade-off between sensitivity and dynamic range. In

regions where the derivative term in equation (3.4) is large, and the sensitivity is high,

refractive index can only be monitored over a small interval. Conversely, when the

derivative term is small, refractive index can be measured over a wider range of values,

but Anmi is larger. Measurement of the actual refractive index of the superstrate requires a

knowledge of the extinction ratio at two or more wavelengths.

0.40 -

0 0.5 1 1.5 2 2.5 3 3.5 4
position (um)


X-- = 670 nm

-------- Onm
300 nm -
6------ 1000 nm

0.5 0 0.5 1 1.5 2 2.5 3 3.5 A
position (um)


Figure 3.4. Intensity profiles of the SPR waveguide at 670 nm with various thickness
superstrates of index 1.415.
(a) TM mode fields
(b) TE mode fields

-2 0.25
' 0.20
- 0.15

594 nm
-40 611 nm -- -

U \ "' I
-60 633 nm

c -80

-100 -670 nm I


705 nm I
-140 i
1.34 1.35 1.36 1.37 1.38 1.39 1.40 1.41 1.42 1.43 1.44
superstrate refractive Index

Figure 3.5. Response of the surface plasmon sensor to changes in superstrate refractive
index. The superstrate is assumed to be infinitely thick. Excitation wavelengths
are shown next to their respective curves.

From a comparison of figures 3.3 and 3.5, it is apparent that in order to maximize

the derivative term in equation (3.4), the spectral width of the plasmon resonance must be

as narrow as possible. The resonance width is proportional to i/E2r, where Er and E, are the

real and imaginary parts of the permittivity of the metal layer, and in general decreases at

longer wavelengths.31 As we will see later, the resonance wavelength can be readily

controlled by the tuning layer. Thus, operating in the infrared region would improve

sensitivity. In addition, other metals, notably silver,32 support narrower plasmon

resonances than gold. Optimization of the sensor design parameters should allow Anmin to

be reduced to the order of 10.6 or less.

3.2.2 Integrated-Optic Surface Plasmon Waveguide Structures

The key to designing an SPR waveguide is to build the device in such a way that

resonance occurs at a convenient wavelength. Additionally, it is necessary to control the

PERmax, so that the TM output signal is measurable. The structures we have investigated,

shown in figure 3.2, address both of these issues.

As noted earlier, surface plasmon resonance is excited by placing a thin metal film

in close proximity to an optical waveguide, so that the TM-polarized waveguide and

plasmon modes are coupled. The thickness of the metal film needs to be on the order of

the penetration depth of the optical wave into the metal, typically about 20 to 30 nm, to

ensure maximum sensitivity to refractive index perturbations in the surrounding

environment. As seen in figure 3.6, changing the thickness of the gold layer used in the SP

waveguide design has a significant impact on both XSPR and PERmaX. As the metal

thickness is increased, XSPR moves to longer wavelengths, while PERmax decreases. At

the same time, the FWHM of the resonance broadens, resulting in decreased refractive

index sensing capability. From these calculations, it is evident that 20 to 40 nm of gold are

required for a useful device. Furthermore, it is evident that tight fabrication tolerances are

required to build SP waveguide devices with the desired resonance characteristics.








0.6 0.62 0.64 0.66 0.68 0.7


wavelength (m)

Figure 3.6. Extinction ratio as a function of wavelength for several different metal (gold)
layer thicknesses. The device is coated with a 1 pm thick adsorbed layer of
refractive index nc = 1.415.

In this configuration, propagation losses are controlled by the insertion of a low
index buffer layer (figure 3.2), placed between the waveguide and the metal film, as
demonstrated in figure 3.7. Since the ion-exchanged waveguide is slightly asymmetric, the
evanescent wave in the silica buffer layer decays rather rapidly and as a result, only a
relatively thin buffer layer of less than 1 pgm is required to achieve an acceptable PER.
Decreasing the thickness of the buffer layer from 750 nm to 400 nm increases PERm" by

an order of magnitude, from about -20 dB/cm to -200 dB/cm at 670 nm. In the quest to
minimize Anmin, losses need to be maintained at levels which are measurable. Metal-
induced losses to the TE mode, which are nearly 2 dB/cm when the buffer thickness is 400
nm also need to be minimized. An important consequence of this design is that XSPR is
virtually unaffected by buffer thickness.
The primary feature that sets these designs apart from others17'20,22,25 is the

inclusion of a high index dielectric tuning layer deposited either directly above or below
(figure 3.2) the metal layer, which provides a large degree of flexibility in choosing the
resonance wavelength. The function of this layer can be explained by a simple coupled-

mode argument: by virtue of proximity, the propagation constants of the SP modes are
strongly dependent on the thickness of the high index layer, while those of the waveguide

modes are not. Thus, in accordance with figure 3.1, changing the thickness of the tuning

layer shifts the resonant wavelength. In figure 3.8, the variation in XSPR is shown against
tuning layer thickness. The peak resonance wavelength can be easily tuned over a 130 nm
range, while maintaining a reasonable level of TE loss. )SPR is very sensitive, changing

by over 2 nm per nanometer of tuning layer thickness, which again underscores the need
for close tolerances during fabrication.
Conventional SPR waveguides lack the tuning layer and require that XSPR and the
PERmax must be set by simultaneously adjusting the metal and buffer layer

thickness.17'20,22'25 Waveguide parameters, such as maximum index change and depth
which determine mode index, are also important in determining XSPR in conventional SPR

50 -



-50 -

o-100 -

-150 -

-200 -

14 2

10 -
8 5

6 u

0.6 0.62 0.64 0.66 0.68

wavelength (pm)

Figure 3.7. Extinction ratio as a function of wavelength for several different buffer (SiO2)
layer thicknesses. The device is coated with a 1 ugm thick adsorbed layer of
refractive index nc = 1.415.



E 670

1 650



1 590

* 570



0 10 20 30 40 50 60 70 80
tuning layer thickness (nm)

Figure 3.8. Peak resonance wavelength and TE insertion loss at resonance as functions of
tuning layer thickness. The device is coated with a 1 pIm thick adsorbed layer
of refractive index nc = 1.415.

devices. In contrast, the tuning layer design allows PERma" and XSPR to be set relatively

independently of one another. Furthermore, the large flexibility in choosing XSPR in tuning
layer-equipped devices allows the waveguide characteristics and metal film to be chosen
fairly independently.
An alternative approach to tuning the resonance wavelength is to use electro-optic
(EO) materials, such as liquid crystal, in the design of the SPR device.33 In this case, an
EO material is positioned next to the metal layer of the plasmon device. Applying a
voltage across the EO material changes its refractive index which consequently alters
XSPR. Thus, incorporation of EO materials in SPR devices allows active control of the

plasmon response. Shifts in resonance wavelength of almost 200 nm have been achieved

by applying 30 volts across a liquid crystal layer in a bulk SPR device.33

3.3 Experimental Investigation of Surface Plasmon Devices

3.3.1 Device Fabrication

The SPR sensors were built on planar K+-Na+ ion-exchanged waveguides. BK7

glass substrates were cleaned by sequential immersion in trichloroethane, acetone, and
methanol for 10 minutes each. The substrates were then rinsed with deionized water,
blown dry with a filtered nitrogen gun, and baked in an oven at 90 C for 1 hour. The ion-

exchange was performed by placing the substrates in a bath of molten potassium nitrate
housed in an aluminum vessel at 375 C for 3.5 hours, as shown in figure 3.9. This

produces a birefringent waveguide with AnTE = 0.008 and AnTM = 0.0092, and dx = 3 pm.

For wavelengths above 600 nm, the waveguide is single-moded. A thermocouple and
temperature controller (Omega CN8500) were used to maintain a constant bath
temperature. The samples were then removed from the melt and cooled slowly to avoid

thermal stress-induced cracking and rinsed in deionized water. Next, thin films of SiO2,
TiO2, and gold were deposited sequentially on top of the waveguide in an electron-beam



Figure 3.9. Fabrication apparatus for constructing potassium-sodium ion-exchanged

evaporator, at thicknesses of 500 nm, 55 nm, and 30 nm, respectively. Lastly, the substrate

was cut into 1 cm squares, and endfacets were polished to permit endfire excitation.

In the polishing process, the sample was first waxed to a dummy piece of glass

with a low temperature wax to prevent chipping of the edges. The sample/dummy

composite was then mounted vertically with wax in a vise-like holder, with the endfacet

held parallel to the plane of polishing. A paste consisting of polishing powder (SiC or

A1203 from Buehler) and water was prepared on a clean, flat sheet of glass and the sample

was moved across this paste (by hand) in a figure-eight pattern for a few minutes. It was

polished sequentially with 400 grit until the edge was flat (typically less than 5 minutes),

20 gpm size for 5 minutes, and 5 pm size for 10 minutes. Between steps, the sample was

rinsed thoroughly with water. Next, the sample was placed on a mechanical polishing

wheel and ground on a 3 pm pad wetted with 1 p.m alumina paste for about 60 minutes. At

this point, the endfacet was generally free of major scratches. The sample received its final

polish on a soft TEXMET pad (Buehler), wetted with MASTERPOLISH 2 (Buehler), a

colloidal suspension of 0.06 p.m silica in a high pH solution. MASTERPOLISH 2

provides a chemical/mechanical polishing action, for optimum results. The final polish

typically took about an hour, after which the endfacet appeared completely free of

scratches under a 20X microscope lens. The sample was typically about 9 mm long after


It was found that at first, the polishing process tended to cause extensive scratches

to the exposed and highly fragile gold film. To alleviate this problem, a 75 nm layer of

SiO2 was evaporated over the gold prior to polishing. The superior abrasion resistance of
this top SiO2 layer, referred to as the cap, allowed the devices to be polished with minimal

damage to the gold layer. The modelling shows that the addition of the cap will shift the

plasmon response to slightly longer wavelengths.

3.3.2 Experimental Measurement of Refractive Index

Figure 3.10 shows the experimental arrangement for the device characterization.

Light from a monochromator (Digikrom 240, dispersion 3.2 nm/mm) was collimated with

a 25 mm focal length lens, passed through a chopper (Rolfin-Sinar), and fed into a 100 p.m

core multimode fiber through a 20X microscope objective. The fiber was then butt-

coupled directly to the SPR waveguide. Light leaving the waveguide was collected by a

combination of a 40X microscope objective and a cylindrical lens, passed through a high

extinction (30 dB) Glan-Thompson polarizer to resolve the individual polarization

components and ultimately detected with a silicon photodiode. A slit aperture was placed

on the front of the photodiode to serve as a spatial filter and block light launched into

substrate radiation modes. A lock-in amplifier was used to improve the signal-to-noise

ratio. With the monochromator output slit at 2 mm, corresponding to a spectral resolution

of 6.4 nm, typical output powers collected from the waveguide were on the order of a few

tens of nanowatts at wavelengths off resonance.

The polarization extinction ratio of the SPR waveguide was measured by coating

the device with different index matching oils.34 The stated accuracy of the refractive

indices of these fluids is +/- 0.0002. Transmission spectra for the TE and TM modes were

measured separately by adjusting the output polarizer. In order to compute the PER, the

loss terms from equation (3.1), ATE and ATM, are expressed as dB per unit distance as

TE,TM (1 pTE,TM pTE,TM (3.5)
A = 10- log (Pin / out )

where L is the length of the waveguide and Pin and Pout are the power coupled into and out

of the waveguide. In this experimental configuration, only the output power, Pout, can be

readily measured. In our launching scheme, using incoherent light and a multimode fiber,

it is reasonable to assume that the fractional powers launched into the TE and TM modes


lens (f= 25 mm)

SPR waveguide
I Szzzvgwe



20X lens

multimode fiber

40X lens

cylindrical lens


slit (spatial filter)


Figure 3.10. Experimental setup for characterizing the SPR waveguide sensor.


are approximately equal. This allows equation (3.1) to be rewritten as
PER= TMiO log TE
PER = -10 log (Pout/P out)
L )(3.6)

In figures 3.11 (a) (d), the TE and TM transmission spectra and corresponding PER are
shown, for superstrate refractive indices of 1 (air), 1.340, 1.380, and 1.410, respectively.
Plasmon resonance, evidenced by a superstrate-dependant decrease in the TM

transmission over a narrow wavelength interval, is clearly observed. In accordance with
theoretical predictions, the resonance shifts to longer wavelengths as the superstrate index
increases. As expected, the TE transmission spectra are relatively unaffected by the

superstrate refractive index.
The wavelength dependance of the polarization extinction ratio is summarized and
compared against the theoretical model (updated to include the SiO2 cap) in figures 3.12
(a) and (b), with slightly better PER than shown in figure 3.11. This improvement was
realized by using a narrower slit over the detector. Experimental PER values are about an
order of magnitude less than those predicted by the model. In addition, the peak resonance

wavelengths are about 30 nm greater than expected. The large discrepancy in PERma
between theory and experiment is most likely due to imperfect spatial filtering of light
launched into substrate radiation modes. PER is computed from the loss of the guided

modes, in particular, from the guided TM mode which excites plasmon resonance.
Radiation modes do not significantly interact with the surface plasmon modes supported
by the gold layer, and as such, experience minimal polarization-dependant loss. TM-

polarized radiation modes which reach the detector can easily overwhelm the strongly-
attenuated signal from the guided TM mode, resulting in the appearance of a larger TM
signal at the detector and causing the TM loss term in equation (3.1) to seem considerably
smaller. While this problem could in principle be circumvented to some extent by using a
single-mode fiber to excite the waveguide, it was found to be extremely difficult to couple


-10- 0

0 .0 -10
500 550 600 650 700 750 800
wavelength (nm)


80 -- 15


70 ---- -------
0 / 10

o0 ,. .,-- .

S0 -0 550 600 650 700 750 8050

30 (b) = 1.340

'" PER

600 550 600 650 700 750 800
wavelength (nm)


Figure 3.11. Transmission and polarization extinction ratio spectra for the capped SPR
waveguide sensor for differing superstrate indices
(a) nsup = 1 (air)
(b)nsup= 1.340
(c) nsp = 1.380
(d) nup = 1.410

60 -15

0 /\ .TM --10

0 0
70 -- -5_

60 -A- .--



90 T. M.\... -8

10 0---- 1 -- ---- --

500 550 600 650 700 750 800

wavelength (nm)


Figure 3.11. Transmission and polarization extinction ratio spectra for the capped SPR
waveguide sensor for differing superstrate indices
(a) nsup = 1 (air)
(b)np= 1.340

(c) nup = 1.380
(d) np = 1.410


II 0
0 \
---- n=1------


-60 .
n 1.310 :
-80 n 1.335
.. .----- n 1.380
-100 i .
50 50 600 0 60 650 700 750 800
wavelength (nm)


Sn= = 1.000
10.00 --- n1.310
n n1.335
S5.00 -........ n -1.380

0.00- -\--

500.00 550.00 600.00 650.00 700.00 750.00 800.00
wavelength (nm)


Figure 3.12. Polarization extinction ratio for the capped SPR waveguide sensor
(a) theoretical prediction
(b) experimentally measured response

a detectable amount of power into such a fiber, using the available equipment. In
particular, a tunable laser would be highly desirable.
To a lesser extent, defects in the gold film may have also contributed to some

degradation of the PER.21 From figure 3.6, it is clear that variations in the thickness of the
gold film significantly affect resonance characteristics. Metal films with thicknesses on
the order of a few tens of nanometers tend to be somewhat porous and exhibit grain

boundary-related roughness, which contributes to light scattering and possibly
polarization conversion. Some evidence of metal roughness is seen in the width of the

plasmon resonances shown in figures 3.12(a) and (b). The measured resonance width is

about 60 nm, roughly twice that predicted by the simulation. In addition, the refractive
index of metal films varies with both the film thickness and the deposition technique.
Refractive index data used in modelling were taken for a 50 nm thick gold film,24

compared to the 30 nm thickness in our experiment. This may account for the difference in
resonance wavelengths between theory and experiment. High temperature annealing could

have been employed to improve the quality of the gold layer, but would have altered the

waveguide characteristics, complicating the overall design of the device. From figures 3.6
and 3.8, it is also apparent that errors in the thickness of the gold layer, as well as that of

the TiO2 tuning layer would result in a shift in XSPR

To calibrate the SP waveguide for refractive index measurements, the polarization
extinction ratio was measured at 658 nm and 708 nm, as a function of superstrate index.

This data is shown in figure 3.13. This device is clearly useful in measuring refractive

index over a wide range, depending on the choice of excitation wavelength. Using
equation (3.4), a sensitivity, or minimum detectable index change of -7x105 had been
predicted for this device, at a signal-to-noise ratio of 13 dB. However, as noted earlier,

experimentally measured PER values were found to be about an order of magnitude less

than predicted. Consequently, the minimum index change which was experimentally
resolvable was only about 5x10 .


-C 0
l O _\ _


1.3 1.32 1.34 1.36 1.38 1.4 1.42
Superstrate Refractive Index

0 658 nm, measured
] 708 nm, measured
658 nm, theoretical, +-10
Q -12 -------------

-708 nm, theoretical, +10

Figure 3.13. Theoretically predicted (lines) and experimentally measured (symbols)
polarization extinction ratio against superstrate refractive index for excitation
wavelengths of 658 and 708 nm. Theoretical curves have been scaled to be a
factor of 10 smaller for comparison purposes.

3.3.3 Humidity Measurement

Many materials, in particular organic polymers, exhibit a humidity-dependant

refractive index. By coating a thin layer of such a material onto the surface of an SP

waveguide, it is possible to create a device in which the PER at a given wavelength varies

with atmospheric moisture content. In this manner, the SP waveguide can be used as a

humidity sensor. It was observed that Nafion fluoropolymer exhibits a tendency to swell to

the point of cracking when immersed in water. As such, this material was deemed to be

appropriate for use as a humidity transducing layer.35 Nafion fluoropolymer was obtained

from Aldrich as a 5% polymer solution in a mixture of lower aliphatic alcohols and water

and diluted to 1% by the addition of methanol. A few drops of the dilute solution were

deposited onto the SP waveguide and allowed to dry at room temperature. This produced a

layer of Nafion on the order of 5 ntm thick, which covered the entire surface of the device.

Variations in the thickness of the Nafion layer do not significantly affect device

performance, since the film thickness is always greater than the penetration depth of the

evanescent tail of the plasmon into the polymer. The humidity-induced variation in the

PER of the Nafion-coated SP waveguide is shown in figure 3.14 over a range of 20% to

50% relative humidity (RH). Over this range, the sensor exhibits a reasonably linear

response, changing by 0.030 dB/%RH-cm for 658 nm excitation and -0.073 dB/%RH-cm

for 708 nm excitation. Humidity-dependent changes in the index of Nafion occur rather

rapidly, on the order of tens of seconds, and appear to be fully reversible. No attempt was

made to control the temperature, which varied from 220 C to 260 C over the course of

these measurements. Comparing figures 3.13 and 3.14, we find that the refractive index of

Nafion changes from 1.3460-0.0005 at 20% relative humidity to 1.3580.002 at 50%

relative humidity. For comparison, Fan and Harrison36 have measured the refractive index

of Nafion as 1.320.03 at 632.8 nm, using ellipsometry.

At first glance, it appears counterintuitive that the refractive index of a material

like Nafion which swells in water should increase for higher humidities. In the following,






0 .



20 25 30 35 40 45 51
Relative Humidity (%)

S= 658 nm
E = 708 nm

Figure 3.14. Humidity response of the capped surface plasmon waveguide when coated
with a thin film ofNafion fluoropolymer.

we analyze this result theoretically. From the Lorentz-Lorenz relation, the refractive index
of a material may be related to its density, p, as37

(n2 1)
S Kp (3.7)
(n + 2)
where K is a constant. At zero humidity, the density of the film is

p = mF/VF (3.8)

where mF and VF are the mass and volume of the dry film, respectively. Thus, in principle,
humidity-induced swelling should reduce the density of the Nafion film, resulting in a
consequent decrease in index. However, in addition to swelling, Nafion also absorbs large
quantities of water from the surrounding atmosphere, resulting in a significant mass
increase. Sadaoka et al.38 find that the water-content of Nafion films can be as high as
-110 mg/g at 80% relative humidity, depending on the film processing conditions. To
examine the simultaneous effects of swelling and water gain, we take the derivative of

(3.7), yielding

2nAn[ 21 n2 = KAmF mFAV (3.9)
n 2+2 (n2+2)2. VF V
where AmF and AVF are the humidity-induced changes in film mass and volume
respectively, and An is the consequent change in refractive index. Rearranging terms in
(3.9) to solve for the index change gives

(n2- 1) (n2 +2) AmF AV,
=n 6n m I (3.10)
6n mF VF
Clearly, water absorption and swelling compete in the overall change in refractive index.
When (AmF/mF) < (AVF/VF), swelling dominates, and refractive index decreases with
increasing humidity. Conversely, and evidently in the case for Nafion in the humidity
range studied, mass increase due to moisture uptake dominates volume increase, and
refractive index increases with higher humidity, when (Amp/mF) > (AVF/VF).

The refractive index of polyimide, another material with a well-known moisture-

dependant refractive index, behaves similarly in that the index increases with humidity.

Moisture uptake in polyimide is on the order of 2%. Franke et al.9 report that the

refractive index of the polyimide SIXEF 33 changes from 1.5512 at 52% RH to 1.5525 at

96% RH, which is about an order of magnitude less than for Nafion.

3.4 Proposed Surface Plasmon Structures With Improved Performance

The SPR sensor discussed to this point offers an excellent means for detecting

extremely small index changes arising from changes in the environment. This high

sensitivity, in conjunction with appropriate transducing layers, makes these devices

appropriate for use in a wide range of applications. However, from a "real world"

viewpoint, large volume manufacture of these devices would probably not be possible.

The unfortunate drawback of integrated optics is that in general, packaging issues, namely

endfacet polishing and fiber coupling, may account for as much as 90% of the final device

cost. Thus, even though the ion-exchanged SPR waveguides are simple in design, it is

unlikely that a low-cost commercial product will be realizable through this route. Because

of this, we have also explored designs for SPR sensors based on SiO2/Si waveguides.

Silicon is perhaps the most mature processing technology, due to the large

competition in the semiconductor sector. Building devices on silicon wafers offers the

inherent advantage of batch processing, allowing the potential for high volume

manufacture. Silicon has a high quality natural oxide, SiO2, which can be grown either by

high temperature oxidation or a number of other means. The single-crystal nature of

silicon allows waveguide endfacets to be prepared simply by cleaving the substrate along

the appropriate crystal plane. Furthermore, through anisotropic etching with solutions of

either potassium hydroxide/isopropanol/water40 or tetramethylammonium hydroxide/

water,41 V-groove structures to facilitate passive fiber coupling can be built into the

substrate. Thus, packaging of silicon-based waveguides is considerably easier and less

expensive than that of glass ones.

Waveguides are deposited on silicon substrates by chemical vapor deposition

(CVD), flame hydrolysis, or plasma enhanced CVD (PECVD). These techniques allow

layers of silica to be deposited at micron thicknesses. Unlike thermal oxidation, these

techniques allow the addition of dopants to the oxide during deposition. Typically, the

oxide is doped with boron and/or phosphorous to reduce stress and the resultant glass is

known as a boron-phosphorous silicate (BPSG). Silicon oxynitride, SiOxNi.x, can also

be deposited. Both SiOxNi._ and BPSG have higher refractive indices than silica, and

through tailoring of the exact composition, substantial refractive index variations are

possible.42'43 As an example, a typical SiO2/Si waveguide could consist of an SiO2

cladding sandwiched between a BPSG core and a silicon substrate. The quality of CVD

deposited oxides is not as high as thermally grown ones. As such, the starting point for

SiO2/Si waveguides is generally a thermally oxidized silicon wafer.
We have designed and modelled an SPR waveguide based on the SiO2/Si

technology. Shown in figure 3.15(a), the structure uses an oxidized silicon wafer for the

substrate. The waveguide core, either SiOxNi.x or BPSG, is deposited next and has a

step-index profile. A second layer of silica, serving as the buffer layer is deposited on top

of the core, followed by a 35 nm layer of silver and an 86 nm Si02 tuning layer. The core

and buffer layers are 1.5 pm and 2.5 pim thick respectively. The refractive index of the

waveguide core is chosen to be 0.01 higher than that of pure silica, which is compatible

with existing deposition processes. The dispersion of the core is assumed to be equal to

that of pure silica. In principle, the waveguide core and buffer layers can be made during a
single run, simply by changing the gas chemistry. Figure 3.15(b) shows the proposed SPR

device on a substrate with an integrated V-groove structure for fiber coupling. In this case,

a cut from a wafer saw would be used to prepare a flat waveguide endfacet at the fiber

pigtail, rather than cleaving.


input signal

superstrate adsorbedd film), ns

tuning layer (SiO2, ZrO2)

SiO2 (top buffer)
doped SiO2 core (BPSG, SiOxNi.x)
SiO2 (isolation buffer)
silicon substrate

output signal

End View





silicon substrate

Figure 3.15. The SiO2/Si surface plasmon waveguide structure in the protected metal
(a) device structure
(b) device with integrated V-grooves for fiber coupling

Note that several modifications have been made to the original SPR waveguide

design (figure 3.2). Silver has been used instead of gold as the metal layer, which, as will
be seen shortly, improves sensitivity. Since the waveguide is symmetric, a much thicker
buffer layer is required to achieve a reasonable PER. An advantage of using the thicker

buffer layer however, is that metal-induced TE propagation losses are reduced to less than
0.01 dB/cm, a considerable improvement over the previous GRIN waveguide designs

which had TE losses on the order of 1 dB/cm! In addition, the tuning layer has been

repositioned to be above the metal, which both protects the metal layer and increases the

range over which XSPR can be tuned. Furthermore, the accessibility of the tuning layer

makes possible post-deposition trimming to correct errors in the resonance wavelength

which may arise from fabrication tolerances in the thickness and index of the various

constituent layers in the plasmon overlay. This design offers a degree of durability which

is generally lacking in SPR devices and we aptly refer to it as the "protected metal


As noted earlier, the advantage of substituting silver for gold in the SPR sensor lies

in the width of the spectral response of the PER. In figure 3.16, PER for the SiO2/Si SPR

waveguide is plotted against wavelength for various adsorbed films with thicknesses
ranging from 0 to 300 nm and an index of 1.415. The medium above the adsorbed film is

water (n = 1.33). The observed plasmon resonance has a FWHM of about 11 nm, roughly

one-third of that of the gold-based GRIN SPR device. Additionally, the shift in XSPR

produced by adsorption of the thin films is much larger than in the previous device.

Adsorption of a 100 nm thick film changes XSPR by 90 nm. An interesting consequence of

positioning the tuning layer above the metal film is that the positive-valued PER peak at

shorter wavelengths seen in figure 3.3 vanishes. In fact, the plasmon resonance dip is

observed even prior to film adsorption.

Extending this analysis to the case of infinitely thick superstrates, we next present

the PER as a function of superstrate index in figure 3.17. Excitation wavelengths of 633

0 -100



B -


650 700 750 800

wavelength (nm)

Figure 3.16. Spectral response of the SiO2/Si silver-based SPR waveguide (protected
metal configuration) during the adsorption of thin films of varied thickness.
The adsorbed material has a refractive index of 1.415.



2 -30-


-60 -


-80 -

1.32 1.33 1.34

1.35 1.36 1.37 1.38

superstrate Index

Figure 3.17. Refractive index dependance of the polarization extinction ratio for the SiO2/
Si silver-based SPR waveguide (protected metal configuration). The
superstrate is infinitely thick. The excitation wavelength is 633 nm.

nm and 670 nm are used. Again, the range of index values over which resonance is excited

is much smaller than the previous device. Using equation (3.4), at a superstrate index of

1.334, Anmin 4x10 assuming a signal-to-noise ratio of 20 dB. This represents a 40%

improvement in sensitivity over the gold-based sensor.

3.5 Application of Surface Plasmon Resonance to Monolayer Detection

An important area of interest for surface plasmon sensors is immunoassay, in

which extremely thin films must be detected.14'15'16 In such applications, the SPR device

is coated with a protein or antibody monolayer film, on the order of 1 to 2 nanometers

thick. This film serves as a transducing layer to catalyze various types of chemical

interactions, such as complexation and desorption, with analytes in the surrounding

environment. These reactions are monitored through changes in the plasmon resonance

characteristics.22 For instance, a common immunoassay study is molecular self-assembly,

which involves the accumulation of several monolayers of different materials. During the

process, the addition of each monolayer produces a discrete shift in the plasmon

resonance. Morgan et al.15 have used this technique to monitor the sequential deposition

of monolayer films of biotin, avidin, and bisbiotin onto a gold surface. The highly

selective nature of protein-ligand and antibody-antigen binding offers the potential for

realization of analyte-specific sensors.

In order to examine the applicability of the SiO2/Si SPR sensor to immunological

studies, we have calculated PER as a function of adsorbed film thickness, using a fixed

excitation wavelength of 633 nm. The adsorption of films of index 1.40 and 1.45 is

analyzed and the medium surrounding the device is water (n = 1.33). As shown in figure

3.18, the device is quite sensitive and is in fact able to detect films thinner than 1 nm!

Adsorption of a 2 nm film causes the PER to change from -40 dB/cm initially to

-33 dB/cm and -27 dB/cm for films of index 1.40 and 1.45 respectively. The plasmon

response is inherently nonlinear and the sensitivity, defined as the derivative of PER with

I 633 nm n = 1.45

S-10 /

%-15 .*/
S2 /0 / nc= 1.40

1 /


-4 5 .. .. . .
0 2 4 6 8 10 12 14
superstrate thickness (nm)

Figure 3.18. Response of the SiO2/Si silver-based SPR waveguide (protected metal
configuration) to ultra-thin adsorbed films. The excitation wavelength is 633
nm. The tuning layer thickness is chosen so that XSPR initially (prior to film
adsorption) coincides with the excitation wavelength.

respect to adsorbed film thickness, decreases as the film becomes thicker. Nevertheless,
films of up to about 10 to 15 nm can easily be measured with this device, which is

adequate for detecting monolayers. Sensitivity is also clearly larger for high index films.
The improvement in the sensitivity of this device as compared to the design discussed
previously results from narrowing the spectral width of the plasmon resonance.

3.6 Conclusion

Surface plasmon resonance provides a highly sensitive means for detecting small

perturbations in the refractive index of the surrounding environment. Integrated-optic SPR
sensors, consisting of waveguides coated with thin metal and dielectric layers have been

modelled extensively. A dielectric tuning layer which simplifies the design process was

added to the basic SPR waveguide structure. Designs based on both GRIN and SiO2/Si
waveguides have been examined. A GRIN SPR sensor have been fabricated by depositing

a plasmon overlay with a thin layer of gold onto an ion-exchanged waveguide.

Measurements of the polarization extinction ratio of the GRIN SPR waveguide show

reasonable agreement with theoretical predications and a humidity sensor was produced

by coating the device with a thin film of the moisture-sorbing polymer Nafion.
Simulations of an SiO2/Si SPR waveguide which uses silver instead of gold in the design
show the potential for a 40% improvement in refractive index sensing capability over the

GRIN device. Furthermore, calculations show the SiO2/Si SPR sensor to be capable of

monitoring the adsorption of films thinner than 1 nm, making this device attractive in
immunological studies such as molecule self-assembly and protein-ligand binding.


4.1 Advantages of Polymer Waveguides

Waveguiding in optically transparent polymers has been studied since the early
1970s. Today, polymers with a wide range of optical, chemical and mechanical properties

are available, offering numerous possibilities for novel optical devices. Extremely low

propagation losses have been achieved, using polymethyl methacrylate (0.12 dB/cm at
633 nm),44 polycarbonate (0.19 dB/cm at 830 nm),45 optical grade epoxy (0.3 dB/cm at
633 nm),46 deuterated fluoromethacrylate (0.1 dB/cm at 1300 nm),47 and others. A
number of companies, including DuPont, Amoco, Allied Signal, and others are presently
developing materials with even better performance. From an economic viewpoint,

polymer waveguide fabrication requires little specialized equipment and, within the
constraint of reliability issues, has the potential to be considerably less expensive than
other guided-wave technologies such as Ti:LiNbO3 and SiO2/Si.
The application which perhaps receives the largest benefit from the diverse nature

of polymers is the sensor field. Material properties, such as preferential adsorption of
specific chemicals and environmentally-induced swelling, can be exploited in polymer
waveguide sensors, allowing unparalleled levels of performance and versatility. A
significant advantage which polymer waveguides have over conventional integrated-optic
technologies such as lithium niobate and ion-exchanged glass is the relative ease with
which a chemically-sensitive dopant can be incorporated into the polymer matrix.46'48'49
This offers possibilities not readily achievable otherwise and allows the detection of a
wide range of analytes. Several integrated-optic chemical sensors based on waveguides

fabricated from a material known as polyimide will be presented in chapters 5 through 7.
However, in order to facilitate a better understanding of those devices, we will first
describe the polyimide waveguide fabrication process.

4.2 Fabrication of Polyimide Waveguides

Polyimides are a broad class of polymers which possess excellent chemical,

mechanical, and thermal stability. They are used extensively in the semiconductor industry
as dielectric materials in integrated circuits50 and are also used in planarization51 and

micromachining52 applications. Additionally, many polyimides may be
photolithographically patterned through commonly used photoresist processing
techniques. Recently, a number of companies, including DuPont, Amoco, OCG

Microelectronics Materials, Hitachi, and Hoest-Celanese have developed optically
transparent polyimides, laying the foundation for the use of polyimide in waveguide
applications. Propagation losses in planar polyimide waveguides as low as 0.2 dB/cm at
800 nm53 and 0.3 dB/cm at 1300 nm54 have been reported. Unlike many other types of
polymers waveguide materials, the glass transition temperature of these materials is very
high, often well in excess of 300 oC.55,56,57 Thus, polyimide waveguides can survive
elevated temperature environments. In contrast, the glass transition temperature of
polymethyl methacrylate is only 85 oC .44
We have experimented with the Probimide 400 series of photosensitive polyimide

from OCG Microelectronics Materials*. This material is obtained as a solution of fully
imidized benzophenone tetracarboxylic dianhyride-alkylated diamine (BTDA) polyimide,
dissolved in y-butyrolactone (GBL) solvent and is slightly amber in color. Probimide 400
may be deposited on various substrates by spin coating and behaves as a negative resist for
photolithographic purposes. Three products are available in this series: Probimide 408

OCG Microelectronics Materials, Inc.
200 Massasoit Ave., East Providence, RI 02914

(8.5% solids, 580 cs viscosity), which can be deposited from a thickness of 0.5 Ipm to 4.0

glm, Probimide 412 (12.5% solids, 3500 cs viscosity), which can be deposited from 3.0

pim to 12.0 im, and Probimide 410 (8.5% solids, 8200 cs viscosity), which can be
deposited from 5.0 pim to 20.0 pm.58 The glass transition temperature for this material is

356 oC. The procedure employed to produce polyimide waveguides is shown in figure 4.1.

In the following, we describe each step.

4.2.1 Substrate Preparation

Thermally oxidized silicon wafers are attractive as substrates for spin-cast polymer

waveguide, both because of the high optical quality of the SiO2 layer and the simple fact

that spin-deposition processes work best on round substrates. As will be seen shortly, the

refractive index difference between polyimide and SiO2 is rather large. Therefore, as seen

from equation 2.4, 73 is large and the field decays rapidly into the Si02 layer. As such,

only relatively thin buffer layers of the oxide are required to reduce the absorption due to

the underlying silicon. Oxidation is performed by placing clean, 2" diameter, <100>

oriented silicon wafers into a tube furnace at 1050 OC. Steam is pumped into one end of

the tube, creating a "wet oxygen atmosphere", which greatly enhances the rate of

oxidation. Under these conditions, the growth of a 2 pim thick layer of SiO2 takes 10

4.2.2 Wafer Priming

Although Probimide adheres reasonably well to glass and silicon, it is generally

necessary to treat substrates with a silane-based adhesion promoter for optimum results,

particularly with respect to endfacet preparation (section 4.2.7). The adhesion promoter,

available from OCG as QZ 3289, is diluted at a ratio of 1 to 9 with a solution of 90/10 v/v

ethanol/water (QZ3290). The diluted adhesion promoter solution is stirred thoroughly and

allowed to sit for at least one hour prior to use to ensure proper mixing. During this time,

wafers are baked at 120 OC to remove moisture from the surface of the oxide layer. After

1. Oxidize 2" silicon wafer
10 hours at 1050 C under wet 02

2. Apply -0.8 mL 1:9 diluted
silane adhesion promoter,
spin 4000 rpm/20 sec,
bake on hotplate at 100 oC/20 sec

3. Dispense 0.6 mL P412
(large diameter syringe)

4. Spin wafer: 400 rpm/20 sec,
1000 rpm/10 sec, 2000 rpm/30 sec
Wait 3-5 minutes,
soft-bake on hotplate at 100 oC/15 min

5. Photolithographic patterning:

Sio2 I





UV light


UV exposure

cross-linked region





6. High temperature cure

7. Dope with organic dye (optional)

(Same structure)

dye-diffused polyimidi



Figure 4.1. Schematic representation of the polyimide waveguide fabrication process.


the dehydration bake, the wafer to be primed is placed on the spinner and the dilute silane

adhesion promoter is applied in sufficient quantity to cover the entire wafer. For a 2"

wafer, typically about 0.8 mL of solution is required. The sample is then spun at 4000 rpm

for 20 seconds. This leaves a thin layer of silane on the wafer surface. The presence of

moisture on the wafer can interfere with uniform dispersal of the silane film and should be


4.2.3 Polyimide Deposition

Deposition of a uniform polyimide film is the key to realizing high quality

waveguides. However, spin-coating of polyimide is somewhat more complicated than that

of conventional photoresists. Specifically, in order to prevent the inclusion of air bubbles

in polyimide films during deposition, a syringe with a large diameter tip needs to be used

to transfer polyimide solution onto the wafers. This is accomplished by cutting off the end

of a 3 mL plastic Luer-Lok tipped syringe (Becton-Dickinson), producing a tube with a

diameter of 8 mm. Butyrolactone was used to wipe off the ink markings on the bottom of

the syringe, so as not to contaminate the stock polyimide solution. The syringe plunger is

pushed down until there is no air trapped in the tube. Approximately 0.6 mL of Probimide

412 (12.5% solids, viscosity 3500 cps) is drawn into this modified syringe and held

vertically to maintain vacuum. The polyimide is then dispensed onto the center of the

wafer into a puddle approximately 2 cm in diameter. Any air bubbles which are visible at

this time must be removed, either by piercing them with a sharp, clean object, or simply by

letting them rise to the surface and break of their own accord. A three-step spin process is

then employed to distribute the polyimide solution. The spinner is set for 60 seconds at
400 rpm and activated. After 20 seconds, while the wafer is in motion, the spin speed is

increased to 1000 rpm. This rate is held for 10 seconds and then increased to 2000 rpm for

the remaining 30 seconds. This procedure is well-suited to uniform dispersal of the high

viscosity polyimide across the wafer. The wet film is left to sit for 3 minutes to settle and

allow any air bubbles trapped at the polyimide/SiO2 interface to migrate out of the film.

The polyimide-coated wafer is then placed on a hotplate at 100 OC and baked for 15

minutes. During this soft-bake, the silane forms a strong chemical bond with the

polyimide film, ensuring excellent adhesion to the oxidized silicon substrate.

It is important that the soft-bake step be performed on a hotplate, rather than a

conventional box-type oven. When polyimide films are soft-baked in an oven, the air-

exposed surface of the film dries first, forming a skin which impedes solvent removal. As

the solvent evaporates, this outer skin tends to crack, leading to large surface roughness.

At the same time, microvoids are formed inside the film. Both effects cause strong

scattering and increase propagation loss by about an order of magnitude. When dried on a

hot-plate however, polyimide dries at the substrate interface first and the solvent is

efficiently removed without formation of defects.

4.2.4 Photolithography Planar Waveguides

While planar waveguides do not require photopatterning, a moderate UV dose

(365 nm) is nevertheless required to cross-link the polyimide chains and make the films

insoluble in organic solvents. The required UV dose varies with thickness. Typically, for

the process described above, a dose of about 0.8 J/cm2 is necessary. Moreover, cross-link

density can also be used to control the diffusion of an organic dye into the polyimide

matrix48 (section 4.2.6), and in some cases, UV doses of 2 to 3 J/cm2 may be required to

optimize concentrations. This topic will be dealt with more in chapter 6. In general, the

polyimide takes on a darker shade of amber when exposed to UV radiation, due to

increased absorption at shorter wavelengths. Channel Waveguides

Photomasks with various channel waveguide patterns are placed between the

polyimide-coated silicon wafer and the UV light source, so that only the exposed regions

are cross-linked, much like a negative photoresist. The masks were designed with the

MAGIC and CADENCE software packages and fabricated on a model GCA MANN 3600

pattern generator. A UV dose of 0.6 to 0.8 J/cm2 is used. The film is then placed in a

developer solution (50/50 wt. butyrolactone/xylene) housed in an ultrasonic bath for 6

minutes. During this time, unexposed regions are dissolved. Ultrasonic agitation during

the developing stage improves pattern contrast.51 After developing, the sample is placed

sequentially in two identical solutions (50/50 wt. developer solution/xylene), referred to as

crossover baths, for 20 seconds each. Lastly, the sample is rinsed in xylene for 30 seconds

and blown dry with nitrogen. The crossover solutions are necessary to avoid precipitation

of the polyimide when going from the highly polar butyrolactone-based developer to the

non-polar xylene rinse. Precipitation causes the polyimide to turn opaque white, rendering

it useless for waveguiding purposes.

4.2.5 Curing of Polyimide Films

A high temperature cure is required to achieve good mechanical properties in the

polyimide. This needs to be performed under a nitrogen atmosphere, as the presence of

oxygen during curing degrades the mechanical performance and increases coloration in

these materials. Samples are placed in a vacuum oven with a nitrogen purge and ramped

up to 280 oC at a rate of 5 oC/minute. Peak temperature is held for 1 hour, after which the

samples are cooled back to room temperature at about 1 OC/minute. After the cure cycle,

planar polyimide films and wide (> 50 ptm) channel waveguides are approximately 5 pm

thick. Narrow channel waveguides tend to be up to 1 pIm thicker, depending on the

channel width.50

4.2.6 Doping (Optional)

A wide variety of dyes can be used to dope the polyimide, using a simple diffusion

process. Cross-linked and partially cured polyimide waveguides are soaked for 10 to 15

minutes in 1x10-4 M (or less) butyrolactone solutions of the desired dye. Butyrolactone

causes cross-linked polyimide to swell, allowing dye molecules to easily penetrate the

matrix. After the desired diffusion time, samples are quickly rinsed with methanol to

remove residual dye from the surface, immersed in water, blown dry with nitrogen, and

baked on hotplate at 100 OC for 30 minutes to remove residual solvent. After solvent

evaporation, the polyimide matrix contracts, effectively trapping the dye and preventing

clustering and migration. The peak dye concentration introduced into the polyimide films

by this process varies inversely with the degree of UV-induced cross-linking.48 The

uniformity of the doping mimics the quality of the polyimide waveguide. The laser dyes

cresyl violet 670, oxazine 720, nile blue 690, oxazine 725, oxazine 750, LDS 698, and

LDS 751 has been successfully introduced into the polyimide matrix using this technique.

Dyes can also be introduced into polyimide from methanol solutions, but this requires

longer diffusion times and results in lower peak dye concentrations than when

butyrolactone is used. It appears that the effectiveness of various solvents in transporting

dye into the polyimide matrix is related to the amount of swelling induced.

4.2.7 Endfacet Preparation

One of the main advantages of depositing polymer waveguides on oxidized silicon

wafers is that endfacets can be prepared by cleaving along the crystal planes of silicon,

provided the polymer exhibits sufficient adhesion. In this case, the <100> oriented wafers

are cleaved along <110> directions, producing rectangular samples. A small scratch is

made on the wafer in the desired cleave direction. Applying a small pressure to the wafer

so that it bends along the direction of the scratch will cause the wafer to cleave along that

plane. With practice, this process takes only a matter of seconds and produces quality

endfacets. In contrast, the preparation of good endfacets on glass ion-exchanged

waveguides by the polishing process described in chapter 3 takes several hours.

4.3 Characterization of Polyimide Films

The nominal value of the refractive index of Probimide 400, measured by prism

coupling, is found to be about 1.626 at 633 nm, for the TE polarization. Process
parameters, primarily the peak bake temperature and the UV dosage affect this value to
some extent. By measuring the index over the wavelength range of 594 nm to 780 nm, we
have fitted the TE refractive index to the second-order Sellmeier equation,

2 1.1413212 0.2113021 X2
nTE = 1 +----- + -- (4.1)
2 + 0.3670977 2 0.1570585
where X is in micrometers. TE refractive indices are shown in figure 4.2. The
birefringence, defined as

n = nTE -nTM (4.2)

was found to be very small (3x10 ) in Probimide 400 films. This is consistent with the
amorphous nature of the BTDA polyimide, which is comprised of short, flexible polymer

chains.60 As such, this material lends itself to use in devices such as splitters and switches,
where polarization-insensitivity is desirable. In contrast, the polyimides SIXEF 33 (Hoest-
Celanese) and Ultradel 9000 (Amoco), which are based on longer, stiffer polymeric
chains, are more crystalline in nature and have birefringences of 2x10-3 and 3.3x102,
The propagation losses of polyimide waveguides have been measured in the

visible and near infrared spectrum, using a tunable Helium-Neon laser (PMS Electro-
Optics, LSTP-0010) and semiconductor diode lasers operating at 677 nm and 780 nm.
Light is coupled into a guided mode using the prism coupler and an optical fiber is
scanned along the length of the guided streak to collect scattered light, as shown in figure
4.3. The scattered light is proportional to the power propagating in the mode and decays
exponentially along the length of the waveguide as given by

Is, i (z) = KIi (z) = KiIi (0) exp (-2k magz) (4.3)

1.640 -


1.630 -



600 650 700 750

wavelength (nm)

Figure 4.2. TE refractive index of the Probimide 412 polyimide, fitted to a second-order
Sellmeier equation.



optical fiber

input beam coupling
prism scan
p uscane

s Z


Figure 4.3. Experimental set-up for measuring propagation loss in waveguides.

where Ki, Ii(O), and kimag are a constant, the initial power launched into the ih mode, and

the field attenuation coefficient of the ih mode, respectively. To a rough approximation, Ki

is proportional to the intensity of the mode field at the air/polyimide interface. A typical

measurement of scattered intensity as a function of position along the guided streak is

shown in figure 4.4, along with the values of the loss coefficients obtained by fitting data
to equation (4.3). Unfortunately, the technique tends to overestimate losses when used to

evaluate polymer waveguides. The pressure applied to the substrate to hold the waveguide

against the prism tends to physically deform the polymer film, causing light to be

launched simultaneously into several modes at once, instead of just one. Thus, the

detected scatter signal is actually the sum of the individual scatter signals from each

excited mode. Higher order modes scatter more strongly than lower order ones (a

consequence of having a larger intensity at the air/polyimide interface), and tend to be

more lossy. Thus, excitation of higher order modes while attempting to measure the

attenuation of a low order mode can make the loss appear to be larger. Signal-to-noise

constraints generally restrict the use of this method to modes with losses of greater than

1 dB/cm.

Propagation losses are shown in figure 4.5 for the TEo mode of a planar

waveguide. At 780 nm, the loss is less than 1 dB/cm. At shorter wavelengths, losses are

considerably higher due to increased absorption. In fact, a weak orange fluorescence is
observed in the material when 543.5 nm radiation is launched into the films. Loss was not

observed to depend significantly on the film thickness, which indicates that scattering

losses arise primarily from roughness at the air/polyimide interface rather than from

scattering sites distributed within the bulk of the polyimide film itself.

Propagation losses can be improved by optimizing the cure process. During

curing, the polyimide becomes darker in color and exhibits increased absorption. This

effect is more pronounced at higher cure temperatures.50 The use of a lower peak cure

temperature sacrifices some of the polyimide's chemical and mechanical resistance, but

0 .5 .. . . .
0 1000 2000 3000 4000 5000 6000 7000 8000
position (urn)

Figure 4.4. Variation in scattered power along the length of the waveguide for the TEo
mode at 670 nm. R is the correlation coefficient, which describes the degree to
which experimental data fall along the exponential fitting curves.





I- 0--

590 640 690 740

wavelength (nm)

Figure 4.5. Propagation losses for the TE0 mode of a 3.6 pm thick polyimide waveguide
on a soda-lime glass substrate.


appears to be necessary to attain good optical properties.62 Slowing the heating rate during
the cure cycle to -1 OC/min also reduces loss.56,* Using a modified cure cycle, we have

been able to reduce propagation losses considerably. Under the new cure schedule,

samples are heated to 150 OC, at a rate of 5 C/min. After 15 minutes at 150 C, the
temperature is ramped to a final temperature of 250 OC at the same heating rate. Peak

temperature is maintained for 30 minutes, after which samples are cooled slowly to room

temperature over several hours. Using this modification, we have achieved propagation

losses of 3.1 dB/cm at 633 nm and 1.5 dB/cm 670 nm in the TMo mode of a 5 pm thick
planar polyimide waveguide.

Finally, it should be noted that the coloration of polyimide solutions becomes more

pronounced with age.** The quoted shelf-life of the Probimide 400 series is 1 year.
However, this specification is based largely on changes in photospeed, rather than

absorption. Most of the waveguides in chapters 5 through 7 were deposited from
solutions which were 2 to 3 years old and still showed good performance.

4.4 Summary

Polymer waveguide fabrication is less expensive than other guided-wave
technologies, such as Ti:LiNbO3 and Si02/Si, but requires careful attention to process

conditions in order to produce devices which are competitive with respect to loss. Using
the photosensitive polyimide Probimide 400, we have obtained propagation losses on the
order of 1 to 2 dB/cm at 670 nm in multimode waveguides deposited on oxidized silicon

substrates. This material is easily doped with a variety of organic dyes using a simple
diffusion process, making it particularly attractive for sensor applications. The refractive
index has been measured over a wide wavelength range and fitted to a second order

C. T. Sullivan, Sandia National Laboratories, private communication
C. T. Sullivan, Sandia National Laboratories, private communication
D. Roza, OCG Microelectronics Materials Inc., private communication


Sellmeier equation. We shall now present several chemical sensor designs which utilize

polyimide waveguides.


5.1 The Evanescent-Wave Absorption Sensor

The most common types of optical sensors are those which employ changes in the

optical absorption characteristics of an indicator material, such as a pH sensitive dye, as a
means of analyte detection. In sensors based on optical waveguides, the indicator is

immobilized in an analyte-permeable host material and used as either the core or the

cladding (defined in figure 2.1). The relative ease with which organic polymers can be

doped with a variety of analyte-sensitive dopants, makes this class of materials very

attractive in this application. We refer to the portion of the waveguide which houses the

indicator as the sensing region. When the sensing region is a cladding layer, as is

commonly the case due to fabrication requirements, only the evanescent tails of guided

modes interact with the indicator and the device is accordingly termed an evanescent wave

absorption (EWA) sensor.63'64 Table 5.1 lists a few examples of absorption-monitoring
waveguide sensors

Most absorption-monitoring integrated optic sensors (and in fact all of the

examples given in table 5.1) are based on multimode waveguides. Therefore, a brief
review of the formulation of wave propagation in multimode structures is in order. We

shall consider a simple two-dimensional multimode waveguide, consisting of a high index
media of index n2 bounded between two media with indices of n1 and n3 respectively as
shown in figure 5.1. For the sake of generality, we will also assume the each of the

materials forming the waveguide has a bulk absorption coefficient, Ki.

Table 5.1 Examples of Absorption-Monitoring Waveguide Sensors

Ref. core cladding indicator analyte

65 K+-Na IE PWG SiOz lutetium Chlorine
66 multimode fiber silicone analyte hydrocarbons
67 Ag+-Na IE- CWG silicone analyte trichloroethylene
68 Ag+-Na+ IE- CWG sol-gel bromcresol purple NH40H
69 glass rod poly vinyl sodium picrate sodium cyanide
70 AgClxBrl.x fiber none (*) fiber core SF6
71 sol-gel film on none (*) bromophenol blue pH

IE CWG: glass ion-exchanged channel waveguide
IE PWG: glass ion-exchanged planar waveguide
(*): analyte-sensitive material located in waveguide core

A coherent light source is to be endfire-coupled into the waveguide. At z = 0, the power

launched into the ih order mode is

Pi P e(x) einput(x) dx


where Po is the input power, ei(x) and einput(x) are the respective transverse field
distributions of the ith order mode and the excitation source at z = 0. Thus, input power is

launched unequally into the various guided modes. Inside the waveguide, the power

propagating in a given mode can be written as

Pi (z) = Piexp (-2k imagz)


SInput field distribution, Einput

Z nl, K,
SX \(sensing layer)

Light n2, K

n3, K3

field distributions, Ei

nl < n2 > n3

Figure 5.1 Launching light from a free-space beam into the guided modes of a waveguide
(endfire excitation).

where kimag is imaginary part of ki, the propagation constant of the ith mode. The total
power in the waveguide is thus

P (z) = XPiexp (-2kimaz) (5.3)

We define the transmission of a waveguide of length L as P(L)/P(0).
For a given mode, kimag depends on the fractional power, defined in equation
(2.7), travelling in each of the three regions shown in figure and the bulk absorption
coefficient of that region:

kimag = (riK + T,2iK + F3iK)/2 (5.4)

In an evanescent wave absorption sensor, losses occurring in the sensing layer (region 1)

must dominate all other forms of loss and thus, FliK1 ) r2iK2 + F3iK. Under this
assumption, inserting (5.4) into (5.3) yields

P (z) = XPiexp (-rliK1z) (5.5)
In the evanescent wave sensor, the absorption coefficient of region 1, K1, is expected to
change in the presence of analyte. Therefore, in order to maximize device sensitivity (i. e.
analyte-induced change in the transmission of the waveguide), we would like to have as
much power travelling in the sensing layer as possible.
We have used the waveguide simulator developed in chapter 2 to analyze the mode
fields of a simple three layer waveguide. In this model, we have chosen n2 = 1.625

(polyimide), n3 = 1.45 (SiO2), thickness d = 5 pim, and used a wavelength of 633 nm. In
figure 5.2(a), Fli, the fractional power carried by the ith mode in region 1, is shown for a
few values of n1. As expected, higher order modes are less tightly confined than lower
order ones and propagate a larger fractional power in region 1. This is particularly true
when (n2 nl) is small. From figure 5.2(a) and equation (5.5), we see that the launching
conditions at the input of the waveguide (z = 0) play a key role in determining the overall

propagation loss, and hence, sensitivity of a multimode waveguide absorption sensor. For

- 0.12
n 0.06

0 1 2 3 4 5 6 7 8 9
mode order


Power distribution in a three-layer waveguide. n2 = 1.625, n3 = 1.45, and the
thickness of region 2 is 5 pm. The excitation wavelength is 633 nm.
(a) Fractional power carried in region 1 by each guided mode.
(b) Total fractional power (summed over all guided modes) and waveguide
asymmetry as a function of sensing region index, nj.

Figure 5.2

example, an EWA sensor with power launched predominately into lower order modes will
be less sensitive to analyte-induced changes in cladding absorption than an identical

sensor with more power launched into higher order modes.
Continuing the numerical example, we next calculate the total fractional power

propagating in region 1, defined as

F1 = 1 (5.6)
as a function of nI. These results are presented in figure 5.2(b), along with the
corresponding waveguide asymmetry factor, aE, defined by equation (2.2). In these
calculations, each mode is assumed to carry equal power for convenience. As the core/

cladding index difference (n2 nl) decreases and the decay term in equation (2.4) becomes
smaller, the evanescent tails of the guided modes become longer and the total power in
region 1 increases. Sharp oscillations in the total power in region 1 occur as individual
modes go to cut-off. Decreasing waveguide asymmetry also results in more power in
region 1. Up to 10% of the light launched into the waveguide can propagate in region 1

when (n2 nl) is less than 0.1.

In contrast to the EWA sensor, many optical detection systems measure the
transmission of a free-space beam passing through a bulk absorption cell. In order to
establish a comparison between the EWA sensor and a bulk device, we define the effective
interaction length of light in an absorption sensor (either bulk or EWA) as

Lcf = FL (5.7)

In bulk absorption measurements, where a probe beam is passed through a dye cuvet, thin
film, etc., F1 = 1. However, as seen from figure 5.2, Ti is typically only a few percent in

EWA sensors, and is also dependant on the launching conditions at the waveguide input.

Examples of bulk and EWA sensors will be presented in the following sections 5.2.4 and
5.2.9, respectively.

Note that this analysis of absorption-based sensors is equally applicable to the

surface plasmon devices examined previously. Having established the basic operating
principles of evanescent wave absorption sensors, we shall now present a specific example

of an EWA device designed to monitor the concentration of ammonia in water.72

5.2 Detection of Aqueous Ammonia

Industrial pollution of rivers and lakes poses a serious hazard to wildlife in affected

areas. Increased algae blooms and the consequent red tides are common side-effects of

fertilizer run-off into swamplands and coastal waters. Pesticide and fertilizer run-off into

groundwater result in elevated ammonia and nitrite levels which can be hazardous to

wildlife at even very low concentrations. For example, total dissolved ammonia

concentrations (NH3 + NH4+) on the order of a few parts-per-million are harmful to

fish,73'74 on a time scale of hours.75'76 Whereas the ammonium ion is relatively

innocuous, the non-ionized form of ammonia is highly toxic to aquatic life and must not

exceed 40 to 400 ppb, depending on species, temperature, water chemistry, and pH.73,75'77

In order to monitor industrial pollution effectively, highly sensitive devices with short
response times are required. Ideally, in order to minimize cost issues, sensors also need to
be reusable. In this chapter, we focus on the development of an optical waveguide based-

sensor for the detection of aqueous ammonia which satisfies these criteria.

5.2.1 Choice of Sensing Layer Materials

Much of the efforts to date in the development of ammonia sensors have dealt with

the detection of vapor phase ammonia. Guiliani et al.78 were able to rapidly and reversibly

detect ammonia vapor concentrations of 60 ppm by monitoring the transmission of a
quartz rod coated with the pH sensitive dye oxazine 170. A similar system utilizing a

sensing layer comprised of ninhydrin immobilized in films of poly (vinyl alcohol) and

poly (vinyl pyrrolidone) exhibited a detection limit as low as 60 ppb, but required nearly
an hour to achieve full response and was not reusable.79 Oxazine 750-doped silicone and

bromcresol purple doped porous sol-gel films have been used as claddings on optical
fibers80 and ion-exchanged waveguides68 respectively, producing ammonia vapor sensors

with detection limits of less than 1 ppm with rapid, reversible responses. In most cases,
humidity has a significant effect on the performance of sensors based on pH indicators.

Unfortunately, the aqueous environment generally places more severe restrictions

on the materials used in EWA sensors. Structural degradation prohibits use of water-

soluble materials in the sensor design. Immobilization of the indicator in the sensing layer
is of paramount importance to avoid problems such as dye migration and clustering.

Aqueous-based ammonia sensors with rapid, reversible responses have been demonstrated
using bromophenol blue71 and oxazine 17081 entrapped in inorganic sol-gel matrices, but

some degree of leaching (loss of indicator) was observed. Oxazine 170-doped poly

(methylmethacrylate) films deposited on glass have also been investigated for ammonia
sensing, but were found to respond very slowly and exhibit poor substrate adhesion when
immersed in water for long periods of time.81 Thus, realization of a useful sensor for an

aqueous environment requires a careful optimization of the mechanical as well as
chemical characteristics of the constituent materials.

We have studied the performance of pH sensitive dyes in two polymers,

polymethylmethacrylate and Nafion. In the following, we describe their performance. Polymethylmethacrylate

Polymethylmethacrylate (PMMA) is an extremely transparent polymer often used

in the fabrication of plastic optical fibers. Previous studies have explored the use of
PMMA as a host material for various laser dyes, such as rhodamine 6G,82 rhodamine B,83

and DCM.84 It was assumed that the immobilization of a pH sensitive dye in a PMMA

matrix would produce a sensor useful for the detection of ammonia. Following the
procedure outlined by Chernyak et al.81 a chloroform solution of PMMA (5% wt. solids)
was doped to 1x10 4M with oxazine 170. The dip-coating technique was used to prepare
a 1 gpm thick film ofPMMA/oxazine 170 on a soda-lime microscope slide. Unfortunately,
this material failed to exhibit a significant change in absorption when exposed to ammonia
vapor. In aqueous environments, the PMMA film separated from the substrate (though
remained largely intact). These results were not sufficiently encouraging to merit further
study of this material. Nafion

We have investigated the use of DuPont's Nafion fluoropolymer, a copolymer of
polytetrafluoroethylene (PTFE) and an acid (S03) -terminated perfluorovinyl ether,85 as a
cladding material for evanescent wave sensing of aqueous ammonia. The chemical
structure of Nafion (1100 equivalent weight) is shown in figure 5.3. This material has been

[(CF2CF2)n (CF2CF)] -


Figure 5.3 Chemical structure of Nafton fluoropolymer.

investigated for this purpose previously by Churchill et al.,86 who produced a fast and
highly sensitive, albeit irreversible, ammonia vapor sensor based on Naflon films
containing the dyes oxazine 720, Nile blue 690, and bromothymol blue. Nafion has also

been used previously by Zen and Patonay87 for pH measurement and by Ballantine et al.88
for acid vapor detection.
Nafion has several properties which make it attractive for ammonia sensing. It is

expected that sulphonic acid groups of Nafion will react strongly to the presence of
ammonia, producing a large change in the pH of the polymer. Annealed Nafion films

exhibit ion-selective transport properties,89 which favor the diffusion of positively

charged species, such as NH4+, through the polymer network over that of negatively
charged ones like Cl .90 In addition, Nation swells considerably in water, allowing rapid

penetration of the analyte into the polymer matrix. We have successfully demonstrated an

EWA sensor for the detection of aqueous ammonia using Nation doped with pH-sensitive
indicator dyes from the oxazine family as a sensing layer.
5.2.2 Fabrication of Oxazine-Doped Nafion

Nation was obtained as a 5% solution of 1100 equivalent weight polymer in lower

aliphatic alcohols and water (Aldrich). Alcohol solutions of the oxazine dyes cresyl violet

670, oxazine 720, Nile blue 690, oxazine 725, and oxazine 750 were mixed with the
Nafion solution, producing mixtures that contained 1% polymer (by weight) and -0.1

mmol dye. The chemical structures of these dyes are shown in figure 5.4.91'* Dye

concentration was kept reasonably low to avoid dimerization. After thorough mixing, the
dye-doped Nation solutions were deposited on clean glass microscope slides using a
coverage of 30 pL/cm2 and allowed to dry at room temperature (24 C, 45% relative

humidity). The resultant films were reasonably uniform in color and thickness in the

*Exciton, private communication

--- N
H Ox
Cresyl Violet 670

Oxazine 725


Oxazine 720

Oxazine 750

"/ N
H5C2 leBu 0
Nile Blue 690

Figure 5.4 Chemical structure of several oxazine dyes.

H5C2 /


center of the samples, but were considerably thicker at the outer edges of the slide. The

films were on the order of 1 to 2 gm thick in the center of the slides.

During the course of this study, it was found that careful attention had to be paid to

subtle processing details to produce a high-quality film. Specifically, the choice of solvent

used for the dye, which forms 80% of the total solvent in the dye-doped polymer solution,

is critical to determining stability of the final Nation film in water.92,93 Films deposited

from methanol-based dye solutions showed poor mechanical properties, crumbling into

pieces and separating from the glass substrate when immersed in water. Baking had been

proposed as a means to improving the mechanical integrity of Nation films by inducing

some degree of crystallinity in the polymer matrix. However, baking alone was found to

be insufficient to improve the resistance of methanol-deposited films to water. The water

stability problem was finally solved by switching the dye solvent to isopropanol, which

produced films which were pliable and cohesive. When immersed in water, isopropanol-
deposited films exhibited excessive swelling and eventual substrate separation, but

remained largely intact. More importantly and unlike the case in methanol-deposited

films, it was observed that baking the isopropanol-deposited films at 120 oC for 60

minutes was sufficient to completely suppress these undesirable effects, resulting in the

realization ofNafion-coated glass slides which were stable in water.

It was also discovered that the dye-doped Nation solutions had shelf-lives of less

than one day in liquid form. Films deposited from "old" solutions showed different

absorption characteristics than ones deposited from fresh mixtures. Therefore, the dye and

Nation solutions were mixed only immediately prior to film deposition.
5.2.3 Characterization of Oxazine-Doped Nafion

Oxazine dyes like those shown in figure 5.4 respond to local pH variations by

reversible deprotonation of an amino endgroup (auxochrome).91'94 In dye-doped Nafion

films deposited by the above method, the polymer matrix is highly acidic in nature and the

dyes are fully protonated and generally lightly colored. However, when exposed to

ammonia, these films change color rapidly, corresponding to the deprotonated (basic) state

of the dye. Furthermore, the deprotonated state is retained after the ammonia source is

removed. We propose that in the presence of ammonia, hydrated ammonium ions rapidly
penetrate the Nation network and bind the polymer's acid groups by the reactions
NH3 + H20 + NH4+ + OH- (5.8)

SO3- + NH4+ -> NH4SO3 (5.9)
Based on the irreversible nature of the reaction of Nafion to ammonia, we may conclude

that the resultant ammonium salt is very stable at room temperature. With the side groups
in the salt form, the pH of the Nafion matrix is significantly higher than in its acid form,

causing deprotonation of the indicator. Sadaoka et al.95 have suggested a similar reaction
for poly (acrylic acid). Based on this model, Nation-based sensors can measure only
ammonium ion concentration. However, from this information, the non-ionized ammonia

concentration in a solution can be calculated, given a knowledge of the temperature, pH,

and other factors.96'97

Immersion of oxazine-doped Nafion films in pure water was also observed to

cause a color change corresponding to dye deprotonation, but only while the samples were

wet. When removed from water and dried, these films returned to the initial acidic state
whereas those exposed to ammonia did not. As might be expected, films in acidic state are
sensitive to atmospheric moisture and a study of the humidity dependency of the

absorption characteristics of oxazine-doped Nation has been performed separately.*
Bulk absorption spectra were measured in the range of 500 nm to 800 nm by

passing collimated light from a monochromator (Digikrom 240) through the dye-doped
Nafion-coated glass slides at normal incidence and monitoring transmission with a silicon

photodetector. The output slit on the monochromator was set to 1 mm, corresponding to a

G. A. Stewart, "Humidity dependant transmission characteristics of Nile blue
doped Nation," Independant study project, University of Florida, 1996.

spectral resolution of 3.2 nm. Light was also passed through an uncoated glass slide as a
reference. The absorption was determined as

A = 1 Tc (k) /Tr (,) (5.10)

where Tc(X) and Tr(,) are the transmissions of the Nafion-coated and reference slides,

respectively. Films were measured in the as-deposited state, exposed to fumes from an
ammonium hydroxide solution, and remeasured. We denote the films to be in the acidic
state prior to ammonia exposure and in the basic state after ammonia exposure. Figures

5.5 through 5.9 show the absorption spectra of the oxazine dyes cresyl violet 670, oxazine
720 Nile blue 690, oxazine 725, and oxazine 750, respectively, in Nafion films before and
after exposure to ammonia vapor. In all cases, the presence of ammonia causes a radical

change in the absorption spectrum. The primary concern here is the wavelength-
dependance, rather than the peak absorption value for each system, since dye
concentration and film thickness vary in each case. Interestingly, despite the relative
similarities in the structures of these dyes, two distinctly different types of absorption
response are observed. Cresyl violet 670, oxazine 720, and Nile blue 690 all exhibit weak
absorptions initially, but form strong and well-defined absorption bands in the visible
region when exposed to ammonia. Conversely, oxazine 725 and oxazine 750 show strong
initial absorption in the near-infrared and exhibit a 30 to 40 nm blue-shift when exposed to
ammonia. Nile blue was chosen as the most convenient indicator dye for the ammonia
sensing, since it showed a large ammonia-dependant change in absorption over the red
region of the spectrum where laser diodes and HeNe lasers are readily available.
5.2.4 Bulk Ammonia Sensor Response

Nile blue-doped Nation solutions were prepared by mixing 1 part Nation solution
to 4 parts of a 0.3 mmol mixture of Nile blue in isopropanol. Films were then made by
depositing 30 pL/cm2 of solution onto clean soda-lime glass microscope slides. The





0 0.400
m 0.300

0.200 --


0.000 45
400 450 500

550 600 650 700 750 800 850

wavelength (nm)

Figure 5.5 Absorption spectra of cresyl violet 670/Nafion (methanol solution) in the acid
and base forms. The film is yellow in the acid form and purple in the base




base form

Said form

0.100 ----
I "/ \ \
0.050- \

0.000 |. -
400 450 500 550 600 650 700 750 800 850
wavelength (nm)

Figure 5.6 Absorption spectra of oxazine 720/Nafion (methanol solution) in the acid and
base forms. The film is green in the acid form and blue in the base state.


0.35 -



S 0.20 -
[- -
- 0.15-


0.05 -


450 500 550 600 650 700 750 800 850

wavelength (nm)

Figure 5.7 Absorption spectra of Nile blue 690/Nafion isopropanoll solution) in the acid
and base forms. The film is yellow in the acid form and blue in the base

1.000 aj o I -
Naflon / Oxazine 725
0.900 \ \

acid form
0.700- -_-f--

S0.600 /
0.60 base form
. 0.400-
U / \ / \
0.300 -/--


0.100 --

0.000- ---- -yrr
400 450 500 550 600 650 700 750 800 850
wavelength (nm)

Figure 5.8 Absorption spectra of oxazine 725/Nafion (methanol solution) in the acid and
base forms. The film is black in the acid form and deep blue in the base state.

500 550 600 650

700 750 800 850

wavelength (nm)

Figure 5.9 Absorption spectra of oxazine 750/Nafion (methanol solution) in the acid and
base forms. The film is aqua in the acid form and blue in the base state.





. 0.500





Naflon I OxazIne 750

base form

/ ^I

acid form /

7 / I

....1-- -\



solution was allowed to dry at room temperature and was then baked at 120 OC on a

hotplate for 60 minutes. The final film was approximately 1.4 gpm thick and yellow in

color. The indicator-doped Nafion films prepared in this manner were found to be very

stable in aqueous environments. In fact, a film immersed in water for a period of 5 weeks

exhibited negligible leaching of the indicator and only minor swelling at the edges of the


Measurements of the change in the bulk transmission of Nile blue-doped Nafion

films with ammonia concentration were performed at 632.8 nm using a HeNe laser. The

coated slides were repeatedly immersed in ammonia solutions in order of increasing

concentration, up to about 3 ppm, for 1 minute, blown dry with nitrogen, and baked on a

hotplate at 50 to 70 oC for 5 minutes. In this manner, the reversible reaction of Nafion to

water could be differentiated from the irreversible reaction to ammonia. Transmission was

measured after each immersion/drying cycle.

The response of the bulk device is shown in figure 5.10. For ammonia

concentrations below 1 ppm, the Nafion film is in the acidic state (yellow) and the Nile

blue indicator absorbs only weakly. In the range of 1 to 2 ppm, localized blue spots begin

to form on the film which persist after drying. With each subsequent immersion/drying

cycle, these spots grow in size until the entire film is blue. When one of these spots

intersects the HeNe beam, in this case at 2 ppm, transmission drops sharply.

Unfortunately, the step-like response observed here results in a sensor with an exceedingly

small dynamic range. Given the film thickness and taking into account the transmission of

the glass substrate, we have calculated the absorption loss of the Nile blue-doped Nafion

to be 16,000 200 dB/cm at 633 nm when the polymer is in the base state. In the acid

state, the absorption is too low to be measured reliably by this technique.

5.2.5 Demonstration of Reversibility

In order to allow Nile blue-doped Nafion films to function as reusable ammonia

sensors, a rinse technique was developed which allows ammonia-exposed films to be reset


90 ----------- -- --------- --, ------ ..-.--

80 ----- ------ -- --"---- ---



I I i i I
I \

50 ------.---- ---- -- ------- --

40 .
0 500 1000 1500 2000 2500 3000
ammonia concentration (ppb)

Figure 5.10 Bulk transmission of a Nile blue-doped Nafion film on a microscope slide to
various aqueous ammonia concentrations. The excitation wavelength is
632.8 nm. The film is approximately 1.4 jim thick.

back to the original acidic (yellow) state. In this process, base-state (blue) sensors are

immersed in 1:20 solutions of acetic acid/deionized water for 30 seconds, blown dry with

nitrogen, and dried on a hotplate at 90 OC. After the rinse, the films are again yellow in

color. To demonstrate the viability of this technique, we have monitored the transmission

of a bulk sample alternately exposed to a high concentration ammonium hydroxide vapor

followed by the dilute acetic acid process. As shown in figure 5.11, the rinse process is

extremely effective in restoring the acidic form of Nafion, cycle after cycle. Thus, this

technology has the potential to be economically viable, as sensors can be reused many

We would like to emphasize that during the course of these experiments, no

problems with substrate adhesion were encountered.

5.2.6 Selectivity of the Nafion Response

As may be inferred from figure 5.3, the transmission-based Nafion sensor exhibits

little selectivity amongst strong bases. For example, exposure to aqueous NaOH converts

Nafion's acid groups into ionic salts in the same manner as NH4OH and produces the

same color change in the Nile blue indicator. The fact that Nafion responds irreversibly to
bases (prior to the acid rinse) but reversibly to water offers some level of discrimination.

In a sensor based instead on the diffusion rates of neutral species through a polymer

matrix, Nation's permaselective nature would provide some measure of selectivity.

However, in order to obtain a selective response to ammonia, the Nafion sensor would

have to be used in conjunction with several other sensors in a multielement array. Ideally,

each sensor element would exhibit different analyte response characteristics. The
composite array response could then be treated as a vector in a multidimensional pattern

space and analyzed by a neural network algorithm, providing both selectivity and

multianalyte detection capability.

acid acid acid acid


& 70

I -


55 ammonia ammonia ammonia

0 1 2 3 4

number of rinse cycles

Figure 5.11 Bulk transmission of a Nile blue-doped Nafion film on a microscope slide at
632.8 nm after sequential exposure to aqueous solutions of 5% acetic acid
and concentrated ammonium hydroxide vapor. The film is approximately
1 Jm thick.

5.2.7 Waveguide Issues in Evanescent Wave Sensor Design

An EWA sensor for the detection of aqueous ammonia has been developed by
coating a polyimide channel waveguide clad with a thin layer of Nile blue-doped Naflon.
An oxidized silicon wafer is used as the substrate. The entire device is shown in figure
5.12. An advantage of the choice of the materials in this structure is that refractive indices

output light

dye-doped Nafion



input light

Figure 5.12 Device structure of the evanescent wave absorption sensor.

of the SiO2 substrate (n = 1.45) and the Nafion cladding (n = 1.35, undoped35'36) are fairly
similar. This low degree of waveguide asymmetry increases the penetration depth of the
evanescent wave associated with each of the guided modes into the Nafion cladding,
thereby enhancing sensitivity (see figure 5.2(b)). Rectangular cross-sectioned ridge
waveguides were used instead of planar ones, in order to further increase the fractional
modal power propagating in the Nation cladding. Through numerical simulations, the
total fractional power travelling in the Nafion cladding of this structure is estimated to be