of experimental findings, the electron-doped system will be denoted by Cal_-La6B6, the
intended stoichiometric material will be labeled CaB6, and the Ca deficient compound
will be referred to as Cal-6B6.
These single crystals form in a rectangular shape with typical dimensions being
approximately 2mm x 0.5mm x 0.5mm. However, we believe that electron transport may
be confined to a fraction of the crystal smaller than the physical size would suggest. As
an adequate means of measuring these reduced dimensions is absent, an attempt has been
made to conduct all transport measurements corresponding to a particular doping
concentration on the same crystal, so that the results can be related within the data set.
4.2 Transport Measurement Techniques
The oblong crystal shape adopted by these samples suggests a natural orientation
for contact formation. The contact configuration shown in Figure 4-1 was found to be
suitable for resistivity, Hall effect, and magnetoresistance simultaneously. A significant
advantage to this permanent configuration became apparent during studies of the Ca
deficient compound, where it was discovered that repeated processing of this material
generates small cracks in the crystal, drastically altering the transport properties of the
sample. Tunneling experiments employed a tri-layer structure that will be described in
 Takashi Ichinomiya, "Impurity-induced ferromagnetism in a doped triplet excitonic
insulator," Physical Review B 63, 045113 (2001).
 B.I. Halperin and T.M. Rice in Solid State Physics, F. Seitz, D. Tumbull, and H.
Ehrenreich, eds. (Academic Press, New York) 21, 115-192 (1968).
 L.J. Sham and M. Schluter, "Density functional theory of the band gap," Physical
Review Letters 51, 1888-1891 (1983).
 J.E. Hirsch, "Metallic ferromagnetism without exchange splitting," Physical Review
B 59, 6256-6265 (1999).
 T. Jarlborg, "Ferromagnetism below the Stoner limit in La-doped SrB6," Physical
Review Letters 85, 186-189 (2000).
 R. Monnier and B. Delley, "Point defects, ferromagnetism and transport in calcium
hexaboride," Physical Review Letters 87, 157204 (2001).
 H.R. Ott, M. Chernikov, E. Felder, L. Degiorgi, E.G. Moshopoulou, J.L. Sarrao, and
Z. Fisk, "Structure and low temperature properties of SrB6," Zeitschrift Physik B
102, 337-345 (1997).
 P. Vonlanthen, E. Felder, L. Digiorgi, H.R. Ott, D.P. Young, A. D. Bianchi and Z.
Fisk, "Electronic transport and thermal and optical properties of Cal-xLaxB6,"
Physical Review B 62, 10 076-10 082 (2000).
 J.L. Gavilano, Sh. Mushkolaj, D. Rau, H.R. Ott, A. Bianchi, D.P. Young, and Z.
Fisk, "Anomalous NMR spin-lattice relaxation in SrB6 and Cal-xLaxB6," Physical
Review B 63, 140410 (2001).
 J.D. Denlinger, J.A. Clack, J.W. Allen, G.-H. Gweon, D.M. Poirier, C.G. Olson, J.L.
Sarrao, A.D. Bianchi, and Z. Fisk, "Bulk band gaps in divalent hexaborides,"
 Donovan Hall, D.P. Young, Z. Fisk, T.P. Murphy, E.C. Palm, A. Teklu, and R.G.
Goodrich, "Fermi surface measurements on the low carrier density ferromagnet Cal-
xLaxB6 and SrB6," cond-mat/0104184 (2000).
 Satoru Kunii, "Surface-layer ferromagnetism and strong surface anisotropy in Cal-
xLaxB6 (x = 0.005) evidenced by ferromagnetic resonance," Journal of the Physical
Society ofJapan 69, 3789-3791 (2000).
 Taichi Terashima, Chieko Terakura, Yuji Umeda, Noriaki Kimura, Haruyoshi Aoki,
and Satoru Kunii, "Ferromagnetism vs. paramagnetism and false quantum
oscillations in La-doped CaB6," Journal of the Physical Society ofJapan 69, 2423-
temperature during growth. That the size of the clusters probably depends on growth
conditions lends credence to a mechanism of interplay between an impurity-based
magnetic moment and critical cluster size, possibly coupled with a lower limit on the
carrier concentration in the case of a carrier-mediated magnetism, above which
ferromagnetic order is possible.
7.2 Future Directions
CaB6 may ultimately be successfully applied to the field of spintronics. There is
much work to be done, however, before the integration can be accomplished. Further
investigations into off-stoichiometry effects are imperative to the conclusive
understanding of ferromagnetism in these compounds.
Complications to experimentation in these samples are associated with a charge
current distribution within these crystals that is presently unclear and may be non-
uniform through the thickness of the sample. Thin film growth may be a method of
reducing the effects of inhomogeneous charge transport. Alternatively, efforts to
fabricate thin films of these materials are expected facilitate several technical aspects of
scientific investigations. A greater degree of dopant control may be achievable, and
quality of electrical contacts could be significantly improved, including those required for
possible future spin-polarized tunneling experiments to determine spin imbalance of
The formation of field-gated devices using CaB6 as a counter-electrode may
enable simultaneous study of the metal-insulator and ferromagnet-non-ferromagnet
transitions. From a pure physics standpoint, the ability to observe these transitions could
give rise to interesting experiments, while from a device standpoint, the manipulation of
0.007- 1 T=2K
S 0.006- o0
S 0.005- s -5 .
-6 -4 -2 0 2 4 6
T T=4.2 __K Voltage (mV) _._
S 0.004- -- T = 10 K
T =25 K
T =30 K Cal-6LagB6
-40 -20 0 20 40
Figure 5-16. Tunneling conductance versus bias voltage in Cal-6La6B6.
spectrum at low temperatures, as shown in Figure 5-17(b), but the width and energy
scales are rather large to be attributed to a zero bias phenomenon. In the bias range
shown, there appear to be features centered about 10, 20, and 30. These features are
reminiscent of a system in which some type of inelastic excitation couples to the
conduction electrons of the system. A discussion of what these excitations may represent
will be given in chapter 6.
LIST OF REFERENCES
 H.C. Longuet-Higgins and M.deV. Roberts, "The electronic structure of the higher
borides MB6," Proceedings of the Royal Society of London Series A,
Mathematical and Physical Sciences 224, 336-347 (1954).
 Robert W. Johnson and A.H. Daane, "Electron Requirements of Bonds in Metal
Borides," Journal of Chemical Physics 38, 425-432 (1963).
 J.P. Mercurio, J. Etourneau, R. Naslain, and P. Hagenmuller, "Etude compare des
proprietes electriques des hexaborures d'europium et d'ytterbium," Materials
Research Bulletin 8, 837-844 (1973).
 R.E.J. Sears, "11B chemical shifts and quadrupole coupling constants in the alkaline-
earth hexaborides," Journal of Chemical Physics 76, 5651-5652 (1982).
 A. Hasegawa and A. Yanase, "Electronic structure of CaB6," Journal of Physics C:
Solid State Physics 12, 5431-5440 (1979).
 S. Massidda, A. Continenza, T.M. de Pascale, and R. Monnier, "Electronic structure
of divalent hexaborides," Zeitschriftfur Physik B 102, 83-89 (1997).
 J. Etourneau and P. Hagenmuller, "Structure and physical features of the rare earth
borides," Philosophical Magazine B 52, 589-610 (1985) and references therein.
 P.H. Schmidt and D.C. Joy, "Low work function electron emitter hexaborides,"
Journal of Vacuum Science and Technology 15, 1809-1810 (1978).
 J. Etourneau, J.P. Mercurio, R. Naslain, and P. Hagenmuller, "Structure electronique
de quelques hexaborures de type CaB6," Journal of Solid State Chemistry 2, 333-
 B.T. Matthias, T.H. Geballe, K. Andres, E. Corenzwit, G.W. Hull, and J.P. Maita,
"Superconductivity and antiferromagnetism in boron-rich lattices," Science 159,
 R.L. Cohen, M. Eibschutz, and K. West, "Electronic and magnetic structure of
SmB6," Physical Review Letters 24, 383-386 (1970).
 A. Menth, E. Buehler, and T.H. Geballe, "Magnetic and semiconducting properties
of SmB6," Physical Review Letters 22, 295-297 (1969).
which the magnetization exhibits no hysteresis. These interpretations are consistent with
the existing data.
The recent theoretical predictions given by Monnier and Delley , discussed
above, combined with the magnetic domain structure suggested by Terashima et al. may
serve to explain the variation in moment size as a function of growth technique and even
among separate attempts of the same technique. The magnetic domains may represent
clustering of B6 vacancies or substitutions so that domain size could account for the
reported discrepancies in moment magnitude. This model is discussed relative to our
work in chapter 6.
Terashima et al. also performed resistivity measurements on these samples, giving
results shown in Figure 2-15 that appear to be uncorrelated with the crystal composition.
The crystals grown by FZ are characterized by a high resistivity. Interestingly, the La-
doped resistivity extrapolates to a higher value than that of CaB6 at zero temperature.
The Al-flux grown crystals instead show a low and metallic resistivity with an anomalous
drop off at low temperatures. This feature was attributed to the superconductivity of Al
inclusions, a conclusion supported by the magnetic field induced suppression of the
In response to the contradictory results for the same nominal doping
concentrations, Morikawa et al. undertook a systematic study of CaB6 in an attempt to
isolate the effect Ca vacancies have on the electronic properties of the material . The
crystals were prepared by reacting CaO with B at various temperatures and at various
growth rates and resulted in polycrystalline samples. By manipulating the growth
temperature and rate, the workers asserted the ability to control the Ca content of the
crystals. The assumption was that a lower growth rate enabled the escape of Ca from the
1.1.3 Valence Fluctuating SmB6 and CeB6
The rare earth compound SmB6 is not easily classified as a divalent or trivalent
hexaboride, as it is a mixed-valence system. In the mid-1970s, the considerable interest
in intermetallic compounds motivated extensive studies of SmB6, in which the ratio of
Sm3+ to Sm2+ was found to be 7:3 and roughly independent of temperature . A
simple analysis dictates that this material should be metallic, since each Sm3+ ion donates
one conduction electron to the system. Resistivity measurements have instead revealed
semiconductor-like behavior initially attributed to a small insulating gap, featuring an
activated rise at high temperatures followed by a low-temperature plateau region.
-4 -2 0 2 4 /
S1.0 Voltage (meV)
0.8 \ j 49 K
',,\ 33 K
0.7 23 K
-40 -20 0 20 40
Figure 1-2. Tunneling conductance versus bias voltage in SmB6. The arrow denotes the
location of the gap edge in energy. The depletion of the density of states near zero is
evidence for a pseudo-gap .
The form of the resistivity for SmB6 is consistent with the development of a small
gap in the density of states at low temperatures suggested by several workers [12-17].
Tunneling measurements of SmB6, shown in Figure 1-2, confirm the appearance of a
the charge carriers. A positive value corresponds to holes, and a negative value indicates
Corrections to this simple analysis are required for a compensated system, in
which both electrons and holes participate in conduction. Complications arise because
the Hall coefficient is dependent not only on the carrier concentrations for each type of
carrier, n andp, but also on the corresponding mobilities, [in and [p, as shown in Equation
(3-5). As a result, the Hall coefficient reveals the sign of the dominant carrier type, but
quantitative results are difficult to extract.
e e(p kp + n pn)2
The presence of interacting magnetic moments in a material can further
complicate the Hall effect. In addition to the externally applied magnetic field, the
magnetism of the material itself influences the trajectories of the charge carriers. Under
these conditions, the anomalous Hall effect appears. Instead of a linear dependence of
Hall voltage on magnetic field, a hysteretic component can contribute to the signal.
In a ferromagnet that displays hysteresis in magnetization versus field, the Hall
voltage will also be hysteretic at fields less than the saturation field. The origin of this
behavior is easily explained. When an external magnetic field is applied, the magnetic
moments become increasingly aligned with increasing field. The fraction of aligned
moments determines the magnitude of the internal field experienced by the charge
carriers. Since the Lorentz force is proportional to magnetic field, the internally induced
component of the Hall voltage should follow the response of the magnetization to applied
3.3.2 Z ero B ias A nom alies ................................................................................. ..... 56
4 EXPERIM ENTAL TECHNIQUES ........................................ ........................... 60
4.1 Crystal Growth and D oping.................... ................................ .......................... 60
4.2 Transport M easurem ent Techniques .................................. ............ ................. 61
4.2.1 Resistivity, Hall Effect, and Magnetoresistance ................... .................62
4.2.2 Tunneling Spectroscopy ........................................................ ............. 65
5 PRESENTATION OF RESULTS........................................................ ............ 72
5.1 Resistivity M easurem ents ...................................................................... 73
5.2 H all E effect M easurem ents .......................................................................... ... ... 79
5.3 Magnetoresistance Measurements.......................... ....................... 84
5.4 T unneling M easurem ents ........................................................................................ 86
6 INTERPRETATION AND MODELING OF THE DATA ........................... ........ 91
6 .1 Interpretation of th e D ata ......................................................................................... 9 1
6.1.1 Cal-~La6B6 .................................... ................... .......... ........ 91
6.1.2 Stoichiom etric C aB 6 .............................. .. ..................................... 102
6 .1 .3 C a l-~ B 6 ...................... .................................. ................. 1 0 6
6 .2 B and Structure M odel...................................................................... .. .......... 111
7 SU M M A R Y ................................................................................... 1 15
7.1 Synopsis of Experim ental R esults................................... ................................... 115
7.2 Future D directions ................................. ...... ........ ............... ...... .. 116
L IST O F R E F E R E N C E S ....................................................................... .................... 118
BIOGRAPHICAL SKETCH..... ....................................... 125
In this chapter, the technical aspects of this work will be discussed. First, a brief
description of the growth of the CaB6 crystals will be given, followed by sample
preparation methods prior to our transport measurements, including resistivity, Hall
effect, and magnetoresistance. An overview of electronic, magnetic field, and data
acquisition equipment will follow. In addition, a presentation of planar tunneling
techniques, as well as the low-noise electronics required for these investigations, will be
4.1 Crystal Growth and Doping
Single crystals of CaB6 were grown and provided by Z. Fisk's group at Florida
State University. The samples were grown out of a molten Al flux using high-purity
starting materials, encapsulated in an alumina crucible. Stoichiometric CaB6 required an
excess of Ca that was incorporated into the melt, due to the suggestion of a tendency of
this material to form Ca vacancies. In the initial growth process, it was believed that the
excess Ca is not incorporated into the resultant crystal. Further discussion of this
assumption is given in the conclusions presented in Chapter 6.
Electron doping was achieved by substitution of La for Ca. The data presented in
this work were obtained from samples that were La doped at a level of roughly 0.5%. A
second doping procedure involved the exclusion of a small amount of Ca from the melt,
in an effort to intentionally incorporate Ca vacancies into the material. In the discussion
Work by Vonlanthen et al. presents FIR spectra as a function of La concentration
. Results of optical reflectivity versus wavevector are shown in Figure 2-10 and have
been related to the excitonic insulator model, since a minimum in the reflectivity
spectrum indicates the presence of a gap. In semiconducting CaB6, there appear two
narrow minima close to each other in energy at frequencies 240 and 275 cm-1. These
features are shown in higher resolution in the inset and may imply the presence of gaps
for each spin species. Excitonic ordering in a material with a small degree of incomplete
nesting of the Fermi surface can produce a similar effect resulting from asymmetry-
induced band splitting. The possibility of such asymmetries in the Fermi surface cannot
be excluded for this material.
r so -
1 40 00 200 300 400
0 .. *1 2....
10 10 10 10 10
Frequency (cm1 )
Figure 2-10. Optical reflectivity versus frequency in Cao.995Lao.005B6 (dash),
Cao.99Lao.o0B6 (dot/dash), and CaB6 (solid) .
Upon examination of the reflectivity in 0.5% La doped CaB6, we find two clear
minima in the spectrum at higher energies of 700 and 1200 cm-1. With further La doping
to a level of 1%, the features broaden considerably but appear to preserve the energy
scale set by the 0.5% La-doped material. As the doped excitonic insulator model requires
different gap values for the different spin flavors, these minima may be evidence for a
0 -- Linear fit of Ca1 _gB at high T
In (p Po) = -12.48 + 3457/T
0.0036 0.0040 0.0044 0.0048
In (pip 0)
0.01 -- Linear fit of CaIgB6 at low
In (p Po) = -0.0013 + 0.5303/T
0.00 1 ,
0.02 0.04 0.06 0.08 0.10 0.12
Figure 5-4. Fit of resistivity versus temperature to activated form in Cal_-B6.
(a) High temperature limit.
(b) Low temperature limit.
these features may be a signature of ferromagnetism. A more detailed discussion of zero
bias anomalies is given in chapter 3, while comparison of this data to the results of this
work is presented in chapter 5.
I i I I I I I I I I I I I I I
-200 -100 0 100 200
Figure 2-1. Tunneling conductance versus bias voltage in EuB6 .
amplifier. The output of the lock-in, as well as the independent parameter (temperature,
magnetic field) were measured and transmitted out of the shielded room to an automated
data-acquisition LabVIEW program by fiber optic cable.
For the Hall effect and magnetoresistance measurements, an AMI
superconducting magnet was used to produce fields up to 8 Tesla. The magnet was
employed using an AMI 4Q-05100 four-quadrant power supply and an AMI 420 power
supply programmer. Data points were taken by ramping the magnet to a specific field
and recording a number of points, usually 5 points/field, to be averaged in data analysis.
This eliminated two undesirable effects: not only is there a high level of noise associated
with field ramping, but this technique also corrects for a lag in sample voltage with
respect to field due to the time interval intrinsic to the data acquisition technique.
Attempting to measure the anomalous Hall effect is an exception to this static-field
procedure. During these measurements, the magnet ramp rate was sufficiently low that
the time lapse between recording the field and recording the sample voltage was taken to
Data analysis was performed using Origin. Since magnetoresistance tends to
contain contributions from the Hall effect, and vice-versa, it was important to take
magneto-transport data at both positive and negative magnetic fields. In this way,
accurate results were obtained by eliminating any anti-symmetry seen in the
magnetoresistance and any symmetry in the high-field Hall voltage. This procedure is, in
effect, the van der Pauw method of correctly measuring said quantities .
4.2.2. Tunneling Spectroscopy
Further investigation of the electronic structure in the form of planar tunneling
spectroscopy was performed on a crystal of the La doped material. Tunneling is a
system is independent of carrier concentration and is instead related to a reduction of
symmetry in the crystal.
10-1- (2) _
I I ,
0 100 200 300
Figure 2-18. Resistivity versus temperature in polycrystalline CaB6. Samples (2)-(4) are
diamagnetic and sample (9) is ferromagnetic .
Another possibility can be derived from the results of Terashima et al. If there are
variations in magnetic domain size, then a lower growth rate might be expected to
produce larger grains and perhaps larger magnetic domains. This may justify the
appearance of ferromagnetism in only samples (5) through (9). More difficult is relating
this effect to the electrical resistivity, but the variations in Ca content suggested by these
workers is insufficient to account for the wide variations in transport data accompanying
this work. For example, it is difficult to see how samples (2), (3) and (4) exhibit such
different behaviors, and in particular, why (3) is more resistive than sample (9), if (2)
through (4) are nominally stoichiometric. It is apparent that fabricating crystals with
stoichiometric precision currently limits coherent progress in the field.
charge carriers and modulation of the magnetism in a semiconducting environment at
room temperature is a technological advance that is greatly anticipated.
I I I I
2 (1) (8)
----(4) /o "
I I ,, I
0 1000 2000 3000 4000 5000
Figure 2-16. Magnetization versus magnetic field for polycrystalline CaB6: growth
conditions. Diamagnetic samples (1)-(4) were grown at a high rate or low temperature
compared to the ferromagnetic samples (5)-(8), grown at a low rate and high temperature
Figure 2-17. Magnetization versus magnetic field in polycrystalline CaB6:
ferromagnetism and diamagnetism. (1) was synthesized at a 1200 C for 12 hours and (9)
at 1500 C for 6 hours. The inset shows ferromagnetic hysteresis in sample (9) .
'-- 2 4 6 8 10
1) ^H (kOe)
I I11 11 11 1
field. The observed Hall resistivity, the ratio of Hall voltage to longitudinal current, takes
the form of Equation (3-6):
PH = RB +Rs 4nM, (3-6)
where Ro is the conventional Hall coefficient, B is the applied magnetic field, Rs is the
anomalous Hall coefficient, and Mis the sample magnetization, which is field-dependent
until saturation of the moments.
3.3 Electron Tunneling Spectroscopy
The density of states of a material contains valuable information that can advance
the understanding of the physical properties of that material. One method by which the
density of states can be probed is electron tunneling spectroscopy, in which the quantum
mechanical transmission of electrons across a thin insulating barrier can be monitored
and used to interpret the distribution of electronic states within a solid. The technical
aspects of tunneling will be presented in chapter 4. Subsection 3.3.1 will serve to
introduce the basic theory of tunneling, while subsection 3.3.2 will discuss features
known as zero bias anomalies that can appear in tunneling spectra.
3.3.1 Theory of Tunneling
A tunnel junction consists of a sandwich structure, where a thin insulating
material separates two electrodes. Many variations of tunneling spectroscopy exist, in
which the electrodes can have very different electronic environments. Commonly, the
electrodes are normal metals, superconductors, or magnetic materials, and the type of one
electrode can differ from that of the opposite electrode. The appearance of the spectra
may differ with type of electrode, but what it represents to first order is characteristic of
the material, in the sense that the choice of barrier does not influence the form of the
deficient material that is semiconducting, and with further Ca depletion, the material
approaches a mid-gap Fermi level.
There are data that suggest that the semiconducting nature of this material is
unconventional. This is shown in Figure 6-5, which depicts Hall mobility versus
temperature. These data were calculated by taking the ratio of Hall coefficient to the
resistivity. This plot indicates that the mobility is temperature-dependent, judging by the
large deviation occurring at room temperature. In the absence of scattering, the mobility
of a simple semiconductor does not depend on temperature. To see this, it is necessary to
analyze the temperature dependence of the Hall coefficient and the resistivity. The Hall
coefficient is known to follow the form of l/n, which to first order, exhibits a temperature
dependence /1exp(-A/kT), giving RH oc exp(A kT). Meanwhile, the resistivity follows the
form p oc exp(A kT). When the ratio of the two quantities is formed, the exponentials
cancel, giving a temperature-independent mobility.
Including the effect of scattering events adds temperature dependence to the Hall
mobility. One type of temperature dependence can be associated with phonon freeze-out,
which would give an increase in mobility with decreasing temperature. Scattering
mechanisms associated with defects tend to lower the mobility with decreasing
temperature. The temperature dependence seen in Cal-5B6 may be due to a combination
of factors. At high temperatures, we find a mobility that is reduced with decreasing
temperature, indicating the presence of defects, possibly in the form of Al inclusions, B6
vacancies, or anti-site substitutions. Judging by the temperature at which the mobility is
suppressed, such a defect density would be unexpectedly high, but this remains a
possibility in light of recent proposals that defects are present. In the low temperature
Figure 4-2. Depiction of 4He cryostat with pump-out port used for magnetotransport
(a) Measurement leads.
(b) Flange to mate with top of dewar, pressure fitting to adjust height.
(c) Vacuum can with Ag solder seal (first layer of vacuum).
(d) Thermal insulation of stage from He bath.
(e) Cu sample stage with 0-rotation capability.
(f) Outer vacuum jacket with pump-out port (second layer of vacuum).
All transport measurements were performed in an RF shielded room using analog
electronics, including a Princeton Applied Research (PAR) 124A lock-in amplifier
equipped with a PAR 116A preamplifier and accompanied by low-noise Ithaco 1201
preamplifiers. The sample was driven with an AC level out of the lock-in. Constant
current was achieved by placing in series with the sample current leads a dropping
resistor of value Rdrop = 100 x Rsample or greater. The current through the sample was
measured during data acquisition to confirm its constancy. The voltage across the sample
was first fed into an AC-coupled preamplifier, whose output was followed by the lock-in
- T = 15 K
S T = 20 K
T =25 K
T =30 K
T =35 K
- T =40 K
-40 -20 0 20 40
Bias Voltage (mV)
-20 0 20
Bias Voltage (mV)
Figure 5-17. Tunneling conductance versus bias voltage in Cal-6La6B6 (a) for 15 K < T <
40 K and (b) for 2 K < T < 20 K.
-3 -2 -1 0 1 2 3
we can solve for the magnetoconductance using
0 0 + 03 i 2
o 03/2 ro3/2
9+ = SF t BH
1.41 2.00 2.45
0.000---- r -m____...
-m-- T = 5 K
--'-- T = 50 K
T = 100 K
-0.004 Ap/p = -7.7e-4 H CaB
0 1 2 3 4 5 6
Magnetic Field (T)
Figure 6-4. Magnetoresistance in CaB6 versus H with fit to linear for 2 T < H < 6 T.
Inset shows magnetoresistance versus H2 for 0 T < H < 2.5 T.
Finally, we obtain
A (sp + CH^)3/- s (p/ pH)3/2- 3/2
o 3 /2 + 3/2
= 1+ + 1- -(2
- ApIp =-2.05e-4 H2
with our experimental results, indicating that disorder may play a role in the low-
temperature properties of the La-doped material and may even indicate a degree of
disorder introduced to the system by a structural transition.
An alternate effect known as the Kondo effect can also produce a minimum and
subsequent increase in resistivity with decreasing temperature. Kondo-like behavior
occurs in the presence of a dilute concentration of magnetic impurities. The electron gas
couples to the magnetic impurities through the Ruderman-Kittel-Kasuya-Yosida (RKKY)
interaction, where Friedel-like oscillations in the type of coupling, ferromagnetic or
antiferromagnetic, form radially in the space around the ion. Through the conduction
electrons, the magnetic ions interact with each other indirectly. The resistance minimum
occurs at low temperatures due to scattering by the exchange coupling, which effectively
localizes conduction electrons around the magnetic ions .
The Kondo theory provides a prediction of the form of resistivity versus
temperature, as shown in Equation (6-2). At high temperature, the resistivity is
dominated by the normal T5 contribution. At low temperatures, a competition ensues
between the T5 dependence and the Kondo contribution, which is of the form InT for a
p = aT5 + cpo cp 1 In T, (6-2)
where c is the concentration of the magnetic impurity,
p o is a measure of the exchange scattering,
and p 1 = 3 zJ/IF (z = number of nearest neighbors and
Jis the exchange constant).
, as shown in Figure 3-1. The negative signal features a saturating effect at the
highest fields, implying an eventual turnover. The positive component at low field may
be due to magnetic hysteresis.
350 T = 4.2 K
0.0 0.5 1.0 1.5 2.0 2.5 3.0
Figure 3-1. Resistivity versus applied magnetic field in EuB6 .
Another mechanism by which a negative magnetoresistance can be attained is
observed in semiconducting systems. The resistivity of the semiconductor can be
reduced due to the application of a magnetic field and the subsequent paramagnetic
splitting of the conduction band. The resultant band structure features one spin band
closer to crossing the valence band. Since the conduction of this band is exponentially
dependent on its proximity to the valence band, it will dominate the transport and give
rise to a lower resistivity resulting from the reduced gap. This effect is clearly
represented by the high-field magnetoresistance of semiconducting SmB6 , shown in
Fermi level marginally crossing the conduction band separated in energy from the
valence band. This result contradicts theoretical expectations of a semiconductor or
compensated semimetal. The La-doped counterpart is metallic, as expected. The Ca-
deficient compound, however, retains a low electron concentration while exhibiting
semiconducting transport properties, implying that the Fermi level resides near the
bottom of the conduction band within the semiconducting gap.
Recent theoretical efforts, in addition to predictions of a polarized electron fluid
and the formation of a doped excitonic insulator, have proposed that an impurity-induced
magnetic moment is conceivable. The results of this work are consistent with the
presence of an impurity band and may indicate the validity of a theory of impurity-driven
Owing to the combination of a high Tc, low carrier density, and proximity to
semiconducting behavior, this novel ferromagnet may be applicable to the field of
spintronics, in which the exploitation of the spin degree of freedom aims for integration
of spin-based devices into the semiconductor industry. The control of electronic spin in
semiconducting devices suggests a variety of technologically important applications,
including spin-based transistors, light emitting diodes, and optical sensors.
Under certain circumstances, inelastic excitations can couple to the quasiparticles
in the electrode and appear as broad features at characteristic energies in the tunneling
conductance versus bias voltage . An example of this phenomenon can be found in
so-called strong-coupling superconductors like Pb and Hg. In these materials, electron-
phonon interactions induce a spreading of the single-particle states so that they are no
longer eigenstates of the BCS Hamiltonian. The signatures of these states appear as
deviations centered about 7 meV and 11 meV in the tunneling spectrum as shown in
Figure 3-6 for Pb.
0 4. 8.0 12.0
Figure 3-6. Tunneling conductance versus bias voltage in a Pb-I-Pb junction. The
coupling of phonons to the quasiparticle spectrum is seen as maxima in the broken data.
The solid curve represents BCS theory .
In a manner similar to the phonon structure discussed above, evidence for other
types of inelastic excitation can be seen in tunneling. For example, a ferromagnetic metal
may reveal magnons or magnetic polarons in the form of smoothly varying bumps or
wiggles . The identity of the feature depends critically on the energy scale at which
the features appear.
position of the sample within the dewar, taking advantage of the vertical temperature
It is useful to take tunneling data for a broad range of temperatures. In this way, it
is possible to investigate the progression of the density of states as the temperature is
altered, an ability especially valuable in the neighborhood of a phase transition. These
very sensitive measurements are best taken with an x-y plotter, advantageous for its high
resolution. The spectra were then converted into a set of x-y points using Un-Scan-It.
Analysis of the data was finally performed using Origin.
temperatures. Nevertheless, activated transport is clearly seen in Figure 5-4 (b), and our
results can be interpreted to indicate that inelastic effects are minimal at these
The implication of our findings is that thermal activation over a large gap governs
the high-temperature data. The low-temperature data seems to be dominated by a much
smaller band gap, where lower-lying thermal excitations remain sufficient to promote
carriers into the conducting regime. The large band gap frozen out at high temperatures
may correspond to the intrinsic separation between the conduction and valence bands,
while the small band gap could represent the presence of an impurity band residing very
close to the bottom of the conduction band.
The value of po in the relation p(T) = po exp(A/kT) can be easily obtained for the
high temperature gap by noting the limiting value of the resistivity in the high
temperature limit. By extrapolating the low temperature data to a high temperature limit,
a different value of po has been obtained. This difference in the high-temperature
resistivities can be attributed to a difference in carrier population. For a two-gap model,
such an analysis yields an effective population across the dominant gap at high
temperature that is nearly 200 times that promoted across the low-temperature gap.
If we turn our attention to the magnetoresistance exhibited by this material, we
find an effect at high field that is similar to its metallic relatives. We expect that
paramagnetic splitting of the conduction band by the external field will produce a smaller
gap for one spin band. The energy of the gap that dominates in the low temperature
region of the data has been found to be nominally 0.05 meV. If we compare this energy
to that of the field-induced paramagnetic splitting with ltBH at 1 T being 6x10-5 eV, or
accompanied by low d-band energies give large values for J, which apparently
collaborate with the small N(O) in Ag to give local moments, as observed experimentally
for spin-1 Ti and Ni. The larger N(O) in Al accommodates larger spin values and d-band
energies, giving a smaller J, to satisfy the requirement for local moment formation.
These predictions are consistent with experiment, as moments appear to be formed in the
Al electrode for a broad range of dopant spin.
Table 3-1. Dopants purposefully implanted into metallic tunnel junction electrodes and
corresponding properties .
This brief discussion of zero bias anomalies is intended to form the framework
within which comparisons can be made between the data from intentionally doped
electrodes presented above and tunneling spectra of ferromagnetic compounds such as
EuB6 (see chapter 2) and, in a preliminary way, CaB6, as will be presented in chapter 5.
Further information on zero bias anomalies in compounds less related to this work can be
found in Principles of Electron Tunneling Spectroscopy by E. L. Wolf .
Dopant n (3d) Junctions Nominal Spin
Ti 2 Y Y 1
V 3 N Y 3/2
Cr 4 N Y 2
Mn 5 N N 5/2
Fe 6 N Y 2
Co 7 N Y 3/2
Ni 8 Y Y 1
explicitly fit to the weak localization model (shown in solid red). The qualitative
agreement between experiment and theory is good at the lowest temperatures, while the
data tend to deviate from -4H as the temperature is increased. From this fit, we can
estimate the value of 0 as defined in Equation (6-3), which is unity for free electrons.
This calculation reveals the value of 0 to be roughly 5x105-, an unexpectedly low figure
that indicates either an overwhelmingly detrimental contribution from electron
correlations or the inapplicability of weak localization theory in this form to the low-
temperature behavior of the La-doped system.
--= 5 K
- T = 10 K
T =22 K
--- T = 36 K
-- ARR2 = -0.0095 H12
Figure 6-3. ARR 2 versus H1/2 in Cal-6La6B6.
A relevant mechanism for the negative magnetoresistance appearing in this
material may also lie in the proximity of its electron transport properties to the
normalized tunneling spectra but may superimpose a linear or quadratic background onto
the conductance signal.
The quantity measured in a tunneling experiment is the conductance across the
tunnel barrier, dI/dV. By applying a DC bias voltage across the junction, the chemical
potentials of the electrodes can be shifted with respect to each other, as depicted in Figure
3-4 . The tunneling conductance represents the product of the densities of states of
the two electrodes, and the bias voltage indicates the energy level with respect to the
Fermi energy that is being investigated.
Ev I E 2= 0
0 1 x
Figure 3-4. Schematic of electron tunneling. The electronic wavefunction tunnels from
electrode 1 to electrode 2 through an insulating barrier of height U. The bias voltage, V,
is applied to shift the densities of states with respect to one another .
One arrives at the conclusion that the tunneling process reveals information about
the density of states through a calculation of transition probability per unit time .
This quantity is approximately obtained by treating the tunneling Hamiltonian as a
perturbation with respect to the Hamiltonians, H1 and H2, of each electrode. There is
assumed to be no electron-electron interaction across the barrier. The conserved
semiconducting state. The overlap of the Fermi level near the minimum in the
conduction band implies that a negative contribution may result from externally induced
paramagnetic splitting of the conduction band, enhancing the conductivity for a majority
spin band and reducing the sample resistance by a small but finite amount. The ratio of
the band splitting induced by an external field to the Fermi energy can give us a
prediction of the magnetoresistance we would expect from such a paramagnetic splitting
in a semimetallic material.
A simple calculation, based on the Drude model of conductivity in a metal, can be
performed for comparison to the experimental data. This is done below (see Equations
(6-4) through (6-6) and Figure 6-4) for the other semimetallic compound CaB6, whose
magnetoresistance is not complicated by a low-temperature transition like that occurring
at 28 K in this material. The value obtained for CaB6, however, is expected to be roughly
applicable to both materials because of their similarity in electronic environments.
Limited tunneling studies, as shown previously in Figure 5-15, support the finding
that La-doped CaB6 exhibits ferromagnetism, with the appearance of a ZBA-like feature,
in analogy with EuB6, in the early tunneling spectra at low temperatures. Subsequent
tunneling results have been shown in Figures 5-16 and 5-17, with the I-V characteristic
included in the inset. The zero bias peaks seen in previous spectra are not convincingly
present in these data. Owing to the sensitivity to external influences, such as remnant
magnetic field or cooling rate, the later data may reflect dissimilarity in environment,
compared to that of the previous junction. It must also be noted that the later spectra
were taken with an x-y plotter, while the early data were recorded digitally using a
programmable voltage source and a Lab View program. The voltage resolution of the
crystal must be smooth, the Pb counter-electrode must be protected against shorting to the
electrode, and the method of defining the junction area must be able to withstand cold
temperatures without cracking.
The process that was found to be optimal for tunneling into CaB6 was preceded by
a 5-minute etch in a weak nitric acid solution. This was designed to remove residual Al
left by the growth process from the crystal surface. The next task was to define the
junction area. This consisted of the selection of a smooth portion and an increase in the
likelihood of forming a uniform barrier. The remainder of the crystal is electrically
isolated during this step by spinning a layer of 1350J photoresist at 4000 rpm for 30
seconds. After baking the photoresist layer at 75 C for at least 20 minutes, optical
lithography was used to expose a region of the surface roughly 20 Mtm2 in area. The
photoresist was then developed in AZ351, a sodium boride-based developer, for 20
seconds and rinsed in water. A thin insulating layer was then formed on the bare surface
of the crystal by a 20-minute exposure to a reactive ion etch of CC12F2 at 20 Watts. This
procedure is believed to remove boron from the surface of the crystal, leaving the Ca-rich
surface free to oxidize upon exposure to the atmosphere. Using a Cu evaporation mask to
limit coverage of the counter-electrode, the sample was loaded into a high vacuum
chamber with a base pressure of 5x10-7 Torr, and 500 nm of Pb were thermally deposited
to form the counter-electrode.
While photoresist is used to not only define the junction area but also to
electrically isolate the crystal from the Pb, cracks can easily develop along the crystal
edges at low temperatures. The material found to be well suited to withstand low
temperatures is a mixture of one part Duco cement to three parts n-Butyl Acetate. A drop
Figure 6-2 shows a plot of normalized resistivity versus InT from 5 K to
approximately 28 K. The experimental results are shown in black, while a fit to Kondo
behavior is shown in red. For the temperature range plotted, the agreement appears to be
reasonable. By calculating a rough estimate of the resistivity at 300 K to be 10 mQ-cm,
using approximate crystal dimensions, we find that pi = 0.011 10 mQ-cm = 0.11 mQ-
cm (see Kondo fit in Figure 6-2). We can infer from this modeling procedure a measure
of the exchange energy in the case that the Kondo model is relevant to this material. By
taking the Fermi energy to be 0.062 eV, the magnetic impurity concentration to be 0.005,
and the number of nearest neighbors for this cubic crystal to be 6, an approximate value
4.48 7.39 12.2 20.1 33.1
Q. 0.710 CalGLagB6
Sp (T)/p (300 K)
-- Linear fit to Kondo theory
0.705 p (T)/p (300 K) = 0.743 0.011 In(T)
1.5 2.0 2.5 3.0 3.5
Figure 6-2. Normalized resistivity versus InT in Cal-6La6B6 for 5 K < T < 28 K. A fit to
linear behavior is displayed in red and represents the behavior predicted by Kondo
strength as a product of the s-d exchange energy and density of states of the electrode at
the Fermi level.
4 (a) Ti-
C448 .44.1 1
5- 11 160 C
-- 38 _8 ___ 75 Undoped
-60 -40 -20 0 20 40 60 -60 -40 -20 0 20 40 60
Voltage (mV) Voltage (mV)
Figure 3-8. Zero bias anomalies in the tunneling conductance versus bias voltage for
magnetically doped electrodes of (a) Ag and (b) Al. The data are expressed in the same
units of conductance but shifted with respect to each other for presentation purposes. The
zero bias conductance is given as the number corresponding to each curve .
Table 3-1 lists the dopants used, the number of 3d electrons each possesses,
whether a moment exists when incorporated into the host electrode, as interpolated by the
presence or absence of a zero bias peak, and the strength of the moment. The s-d
exchange constant is inversely proportional to both the magnitude of the moment and the
energy of the d-band relative to the Fermi level. The presence of a magnetic moment is
expected to correspond to a lower limit of the coupling strength, JN(O). It is known that
Ag has a smaller density of states at the Fermi level than does Al. Since J is inversely
proportional to the product of the total spin of the impurity and the energy of the d-band
relative to the Fermi energy, a qualitative prediction can be made. Small spins
TABLE OF CONTENTS
A C K N O W LE D G M E N T S ............................ ............... ...............................................iii
LIST OF FIGU RE S ......... .... .. ...... ...... ......................... ............. ... ........ ..... ... vi
A B STR A C T ................................................................................................................ ix
1 IN TR O D U C TIO N ........................................... .......... ........ ... ... ........... .. 1
1.1 Electronic and Magnetic Properties of the Hexaboride Family of Compounds ....... 2
1.1.1 The D ivalent H exaborides ..................................................................... 3
1.1.2 The Rare Earth H exaborides ........................... ....... ................................ 5
1.1.3 Valence Fluctuating SmB6 and CeB6 ......................................... ............ 6
1.1.4 Ferrom magnetic EuB 6 ........................................................... .................. 8
1.2 D oping Studies of the H exaborides................................... ................................... 11
1.2 .1 C arb on D hoping of E uB 6.................................................................................... 11
1.2.2 Lanthanum and Ytterbium Doping of SmB6 ............................................ 13
1.2.3 Cerium and Thorium Doping of CaB6 ..................................................... 14
2 PREVIOUS WORK IN CaB6 AND RELATED COMPOUNDS ................................ 17
2.1 Electron Tunneling Spectroscopy in EuB6.......................................... ......... ..... 17
2.2 Recent Experimental and Theoretical Results in CaBI and SrB6 .......................... 20
2 .2 .1 T heoretical Studies .............. ............ .. ......... ...... ..... .. ............. ..... 20
2.2.2 Experim ental Studies ......... .................................... ................ ............... 25
3 PURPO SE OF EXPERIM EN T ............................................... ........................... 41
3.1 Technological M otivation.............. ........... .......... ....... .............................. 42
3.2 Electrical Conductivity in Metals and Semiconductors ...................... ..............44
3.2.1 Electrons in Applied Magnetic Field I: Magnetoresistance............................45
3.2.2 Electrons in Applied Magnetic Field II: Hall Effect .................. ........ 48
3.3 Electron Tunneling Spectroscopy.................................................. .... .. .............. 51
3.3.1 Theory of Tunneling ......... ................. ................... .................. .............. 51
presented in chapters 5 and 6. This chapter contains a description of the technological
motivation for our studies, followed by brief theoretical descriptions of electrical
conductivity and electron tunneling microscopy.
3.1 Technological Motivation
The quest to integrate the spin degree of freedom into the semiconductor-based
electronics industry is the driving force of the emerging field of spintronics. Through the
use of spin-polarized carriers, both the electronic and magnetic responses of
semiconductor devices can be exploited.
The materials that are applicable to the field of spintronics must satisfy a number
of criteria. The usefulness of a candidate material depends on the degree of electron spin
imbalance, the operating temperature relative to Tc, and the ability to effectively transfer
spin-polarized current into the materials used in the existing semiconductor-based
industry. Future progression of the field of spintronics relies exclusively on the ability to
develop a compound satisfying the above requirements.
Spin-polarization of electrons is associated with ferromagnetic materials. The
known ferromagnetic metals were employed as the first attempt to inject spin-polarized
current into semiconductors. While these ferromagnetic metals have high ordering
temperatures and do produce spin-polarized current, they lack the ability to effectively
transfer this current to a neighboring semiconductor. The band structures of the two
materials differ to such an extent that the wave vector mismatch across the boundary
significantly hinders the transmission of current. Efforts now focus on another approach:
to develop a ferromagnetic semiconductor.
5.1 Resistivity Measurements
Resistivity measurements are of importance in the effort to identify the electronic
environment in these materials. In the process, the consistencies of various band
structure predictions with experimental data have been inspected and will be addressed
further in the next chapter. As will be demonstrated here, however, behaviors in the
resistivity can be complex, with physical interpretations that are not immediately
Figure 5-1 shows resistivity versus temperature for the electron-doped material.
The metallic nature seen here is in agreement with that reported earlier. The resistivity,
normalized to the value at 300 K, shows a decrease of roughly 30 percent in the
temperature range studied. At the lowest temperatures, there is a small upturn that may
indicate a type of charge carrier localization. The nature of the interaction will be
revisited in the following chapter of this work.
The stoichiometric material was expected to exhibit semiconducting behavior,
according to a portion of the previously published work. Resistivity versus temperature,
as shown in Figure 5-2, instead appears to retain marginal metallicity. The temperature
dependence of the resistivities of roughly a dozen crystals grown from the same melt
were measured to verify the seemingly contradictory behavior. Small variations are
observed within the sample batch since the transport behavior of these crystals are highly
dependent on the ratio of Ca to B6, but the data shown corresponds to a representative
CaB6 crystal. At high temperature, an initial decrease in resistivity is seen, followed by
an increase, a small maximum at 50 K, and another minimum at lower temperature. In
fact, the form of the resistivity appears to indicate a competition between high and low
1 7 -v-
V- I I-
magneto-resistance Hll ct
Figure 4-1. Contact configuration for simultaneous measurement of resistivity,
magnetoresistance, and Hall effect.
4.2.1 Resistivity, Hall Effect, and Magnetoresistance
Initial efforts to establish electrical contact to the crystals entailed submersion into
a weak nitric acid solution for 5 minutes, followed by a rinse in water, and manual
application of conductive silver paint in the desired contact configuration. Upon the
observation that the contacts deteriorated with thermal cycling, it was apparent that a
more careful process would be required. Good-quality, low-resistance contacts were
formed in a high vacuum chamber with a base pressure of 5x10-7 Torr. To clean the
surface and improve sticking probability, the crystal was exposed to a high-purity Xe
plasma at a beam voltage of 450 V. This was followed by thermal evaporation of 20 nm
of Au in the pattern discussed above defined by a shadow mask. Subsequent 8- to 12-
I dedicate this dissertation to my parents for giving me the intellectual freedom to make
my own way.
A simple metal in which there is one type of carrier and has a perfectly spherical
Fermi surface exhibits no change in resistance with applied magnetic field. In this simple
metal, an electric field is set up by the charge carriers that is transverse to both the current
direction and the magnetic field and perfectly cancels the Lorentz force due to the applied
magnetic field. This transverse electric field makes Hall measurements possible, as will
be discussed below.
In the presence of anisotropy in the Fermi surface or when more than one type of
carrier is responsible for the charge transport, the resistivity of a typical metal will
become enhanced when placed in a magnetic field. This phenomenon is known as
magnetoresistance, and to first order follows a positive H2 dependence on magnetic field.
This dependence can be easily understood from symmetry arguments in the sense that the
magnetoresistance should be symmetric with respect to the sign of the magnetic field. To
complicate matters, when dealing with correlated electron systems, there can be
significant contributions from other sources.
In many ferromagnets, it is common to observe a reduction in sample resistance
with the application of a magnetic field. This occurs due to the alignment of moments
along the direction of the field. The state of lower disorder leads to reduced spin-
dependent scattering and increases the electronic mean free path, thus reducing
resistivity. At the saturation field, the moments are completely aligned, so that no further
reduction of the resistivity is attained by the application of additional field. At fields
sufficiently above the saturation field, the negative effect becomes overpowered by the
standard positive H2 behavior, and the magnetoresistance turns over and approaches
positive values. A relevant example of negative magnetoresistance is exhibited in EuB6
 Tomoaki Morikawa, Takashi Nishioka, and Noriaki K. Sato, "Ferromagnetism
induced by Ca vacancy in CaB6," Journal of the Physical Society ofJapan 70, 341-
 H. Munekata, H. Ohno, S. von Molnar, Armin Segmuller, L.L. Chang, and L. Esaki,
"Diluted magnetic III-V semiconductors," Physical Review Letters 63, 1849-1852
 N. Theodoropoulou, A.F. Hebard, M.E. Overberg, C.R. Abernathy, S.J. Pearton,
S.N.G. Chu, and R.G. Wilson, "Magnetic and structural properties of Mn-implanted
GaN," AppliedPhysics Letters 78, 3475-3477 (2001).
 I. Bat'ko, M. Bat'kova, K. Flachbart, J. Kovac, E.S. Konovalova, Yu.B. Paderno,
and S. Ramakrishnan, "High residual resistivity of carbon doped EuB6," Solid State
Communications 98, 895-898 (1996).
 A. Lacerda, D. Rickel, M.F. Hundley, P.C. Canfield, J.D. Thompson, Z. Fisk, P.
Haen, and F. Lapierre, "High field magnetoresistance in SmB6," Physica B 199 &
200, 469-470 (1994).
 E.L. Wolf, Principles of Electron Tunneling Spectroscopy, Oxford University Press,
New York, 20, 395-408 (1985).
 W.L. McMillan and J.M. Rowell in Superconductivity, R.D. Parks, ed. (Dekker,
New York) 561 (1969).
 Michael Tinkham, Introduction to Superconductivity, (McGraw-Hill, New York) 70-
 Heather D. Hudspeth, "Electron tunneling measurements on ferromagnetically
doped lanthanum manganite films," Doctoral Dissertation, University of Florida
 J.M. Rowell and Y.L.Y. Shen, Physical Review Letters 17, 15 (1966).
 F. Mezei and A. Zawadowski, Physical Review B 3, 3127 (1971).
 M.A. El-Semary, Y. Kaahwa, and J.S. Rogers, Solid State Communications 12, 593
 J.R. Cooper and A.F.G. Wyatt, Journal of Physics F3, L120 (1973).
 A.F.G. Wyatt, Journal of Physics C7, 1303 (1974).
 L.J. van der Pauw, "A method of measuring specific resistivity and Hall effect of
discs of arbitrary shape," Philips Research Reports 13, 1-9 (1958).
spectra of ferromagnetic EuB6 below its Curie temperature. Additional information
provided by these measurements has been influential in the analysis of the Cal-6La6B6
In early measurements, a sharp feature at zero bias in a tunneling spectrum of La-
doped CaB6 was observed at very low temperature, as shown in Figure 5-15. Many
external conditions can influence the appearance of these features, including stray
magnetic fields, temperature variation, and even the passage of time. Therefore, while
the detection of this feature is supporting evidence for ferromagnetism, the data cannot
stand alone as a convincing signature of magnetic order. Additionally, it is important to
20- --T = 11K 4
10- T=20K 0
T T = 25 K U
0- -T=30K __T=2K
T = 40 K -2.0 -115-11 -015 o i 05 l0 1.5 210
10IV 40 Voltage (mV)
-10 i i i i
-100 -75 -50 -25 0 25 50 75 100
Bias Voltage (mV)
Figure 5-15. Tunneling conductance versus bias voltage in Cal-6La6B6.
note that the I-V characteristic, shown as an inset, appears to be of a lower quality than is
typically desired. While it is possible that the conductance peak at zero bias may be
 D.P. Young, "Dilution studies of the Kondo insulators and alkaline earth and rare
earth hexaborides," Doctoral dissertation, Florida State University (1998).
 D.P. Young, D. Hall, M.E. Torelli, Z. Fisk, J.L. Sarrao, J.D. Thompson, H.-R. Ott,
S.B. Oseroff, R.G. Goodrich, and R. Zysler, "High-temperature weak
ferromagnetism in a low-density free-electron gas." Nature 397, 412-414 (1999).
 T. Kasuya, "Exchange-pair Jahn-Teller effects in GdB6," Journal ofMagnetism and
Magnetic Materials 174, L28-L32 (1997).
 J.L. Gavilano, B. Ambrosini, H.R. Ott, D.P. Young, and Z. Fisk, "Spin-lattice
relaxation studies of selected hexaboride compounds," Physica B 284-288, 1359-
 L.V. Keldysh and Yu.V. Kopaev, "Possible instability of the semimetallic toward
coulomb interaction," Soviet Physics -Solid State 6, 2219-2224 (1965).
 Jacques des Cloizeaux, "Exciton instability and crystallographic anomalies in
semiconductors," Journal of the Physics and Chemistry of Solids, 26,259-266
 F. Bloch, Zeitschrift Physik B 57, 549 (1929).
 D.M. Ceperley and B.J. Alder, "Ground state of the electron gas by stochastic
method," Physical Review Letters 45, 566-569 (1980).
 David Ceperley, "Condensed-Matter Physics: Return of the intinerant electron,"
Nature 397, 386-387 (1999).
 G. Ortiz, M. Harris, and P. Ballone, "Zero temperature phases of the electron gas,"
Physical Review Letters 82, 5317-5320 (1999).
 H.J. Tromp, P. van Gelderen, P.J. Kelly, G. Brocks, and P.A. Bobbert, "CaB6: a new
semiconducting material for spin electronics," Physical Review Letters 87,
 M.E. Zhitomirsky, T.M. Rice, and V.I. Anisimov, "Ferromagnetism in the
hexaborides," Nature 402, 251-253 (1999).
 Victor Barzykin and Lev P. Gor'kov, "Ferromagnetism and superstructure in Cal-
xLaxB6," Physical Review Letters 84, 2207-2210 (2000).
 Leon Balents and Chandra M. Varma, "Ferromagnetism in doped excitonic
insulators," Physical Review Letters 84, 1264-1267 (2000).
minimum, and we find that a high-temperature region appears to be discrete from a low-
Because the Fermi level lies in very close proximity to the conduction band edge
in the stoichiometric sample, it may be that subtle features of and near the conduction
band account for the complex form of the resistivity as a function of temperature. For
example, the effects of a small splitting of the conduction band or the existence of an
impurity band might significantly influence the transport properties in this material.
The carrier concentration is found to be nearly an order of magnitude lower than
what is observed in the electron-doped compound. This result is consistent with a
reduction in Fermi level accompanying the lack of trivalent La. The observation that the
carriers are electrons, accompanied by the moderate electron density, verifies that in this
stoichiometry, the material is not of a semiconducting nature.
Paramagnetic splitting of the conduction band in magnetic field may be the origin
of the negative magnetoresistance in this material. In a field of 6 T, the paramagnetic
shifts of the opposite spin components of the conduction band are expected to be small,
only marginally increasing the density of states at the Fermi level for one spin band and
reducing it equally subtly for the other.
Using the simple Drude model to estimate the change in conductivity in the limit
of zero temperature and governed by the relations
(3,X2 20 )2B
2m e2 z (6-4)
region, the mobility appears to be increasing with decreasing temperature, as shown in
the inset of Figure 6-5, indicating that phonon freeze-out may dominate in this regime.
Higher resolution data at all temperatures are desirable to determine the reproducibility of
these temperature dependence.
10 j *
S0 50 100 150 200 250
0 IIIII II I
0 50 100 150 200 250 300
Figure 6-5. Hall mobility versus temperature in Cal-1B6. The solid line is a guide for the
eye. The inset shows in detail the temperature dependence in the low mobility regime.
In a fit to a semiconducting model where p(T)=po exp(A/kBT), the two
temperature ranges suggested by the resistivity can be individually interpreted to
represent thermally activated behavior. The values of the gap can be compared, and it is
found that the high-temperature gap of roughly 350 meV is larger by orders of magnitude
than the low-temperature gap, which was found to be approximately 0.05 meV. This
unusually small value may be susceptible to a significant amount of error, due to the
comparatively large amount of thermal energy available to excite the electrons at these
-4 1: Al-flux La 0.5% #7
2: Al-flux La 0.5% #1
-8 3: FZ La 0.5%
0 T "FZ CaB,
-10 -5 0 5 10
Magnetic Field (kOe)
Figure 2-14. Magnetization versus magnetic field in Cao.995Lao.oo5B6 and CaB6 grown by
FZ and Al-flux techniques .
Other workers have recently grown single crystals of electron-doped and
stoichiometric CaB6 by two methods: out of an Al flux, the method used by Young et al.,
and a floating-zone (FZ) method . Comparisons of crystals grown by the different
methods are somewhat conflicting. A ferromagnetic signal was detected in one of the
La-doped crystals grown by the Al-flux method, as shown in Figure 2-14 (curve 1).
Another of the flux-grown La-doped crystals (curve 2) exhibited no hysteresis but finite
magnetization, as did a La-doped crystal grown by FZ (curve 3). Magnetization versus
field is also given in the lower plot for stoichiometric CaB6 grown by FZ as a reference.
Speculating on the origin of these apparently conflicting results, Terashima et al.
have suggested the presence of an inhomogeneous ferromagnetic phase, containing
magnetic domains separated by paramagnetic regions. These workers point out the
possibility that there is a lower critical limit on ferromagnetic domain size required for a
material to exhibit hysteresis. For sufficiently small domains, then, the response of the
material to an applied magnetic field may be superparamagnetic in nature, a model in
temperature data in the same temperature range, a concurrent local maximum is observed,
as shown in Figure 2-8. Similar behavior of these quantities has been observed in SrB6
Gavilano et al. performed 11B nuclear magnetic resonance studies (NMR) to
measure the spin-lattice relaxation rates in SrB6 and Cal_6La6B6 . The results were
compared to NMR spectra in LaB6. The workers discovered no discernable difference in
the spin echo intensity of all three materials at low temperatures. It is therefore apparent
that the drastically different resistivities, and in turn, the electron densities, have no direct
relation to the rate of spin relaxation. Furthermore, it was noted that the intensities are
too large to be accounted for by itinerant electrons alone.
The relaxation rates for SrB6 at various magnetic fields are plotted versus
temperature in Figure 2-9. The field dependence of T'1 was found to be of the form 1/H
(not explicitly shown). Data for LaB6 at H=0 are shown as a constant-slope solid black
line for comparison and exhibit a linear dependence on temperature, as is expected for
metallic materials. In contrast, SrB6 demonstrates what may be a linear dependence on
temperature below a few Kelvin accompanied by a crossover temperature labeled by To.
Above To, the relaxation rate appears to be independent of temperature. The constancy of
the relaxation rate suggests that phonons are not the mechanism behind the spin
Fits to various relaxation mechanisms, such as localized moments, itinerant
magnetic moments, and paramagnetic impurities, were performed and yielded
unsatisfactory results. A viable possibility involves scattering of conduction electrons
from localized electronic states that only a fraction of the material's electrons occupy.
These circumstances would produce a temperature-independent effect and would give a
significant clustering of these defects may also occur more so in the flux growth method
than in the more thermally uniform growth method of float-zone techniques. It may
follow that larger magnetic domains, that is, larger clusters of B6 defects, are formed
more often in flux-grown single crystals than in float-zone grown single crystals.
The above discussion serves as justification for the correlation between Al flux
growth and ferromagnetism, as discussed in chapter 2. The finding that La substitution
for a B6 octahedron lacks a magnetic moment serves to explain the disappearance of
ferromagnetism at high La doping levels, where B6 vacancies may begin to be filled with
substitutional La. Because ferromagnetism appears to be confined to the two
semimetallic systems, it may be that the order depends not only on the cluster size but
also on the carrier concentration, possibly indicating a carrier-mediated exchange.
Variations in size of magnetic domains with differing growth conditions, such as
temperature and technique, is a highly applicable speculation. The anomalous Hall effect
data presented in Figure 5-10 are puzzling because of the disappearance of hysteretic
behavior following the first application of magnetic field. This crystal may not be of a
ferromagnetic nature. Instead, the domains may be in the paramagnetic or
superparamagnetic regime so that full hysteresis is not seen. That the initial voltage at
zero field differs from that seen in subsequent sweeps is yet unexplained, but the effect
has been found to reappear upon thermal cycling of the sample. The apparent
contradiction with the hysteresis present in magnetization versus field may be resolved by
recognizing that, of the multiple samples used for the measurement, a fraction of the
crystals could possess large domains and exhibit ferromagnetism, while the remaining
crystals possess small domains and exhibit paramagnetism or superparamagnetism.
minimum of the resistivity just above Tc. As the fully ferromagnetic state evolves, the
moments become aligned, so that the effective masses of the conduction electrons are
drastically reduced. The absence of these polarons in the fully magnetic state serves to
explain the large resistive drop below Tc discussed above.
1.2 Doping Studies of the Hexaborides
The physical properties exhibited by the hexaborides were found to be very
sensitive to electron concentration, prompting a series of doping studies to induce
electronic and magnetic transitions in these compounds. For example, by substituting a
trivalent cation for a divalent cation, the metal-insulator transition becomes accessible.
Furthermore, the dependence of the magnetic interactions on carrier concentration can be
probed by substitutionally doping the magnetic hexaborides with cations of dissimilar
size. Descriptions of C-doped EuB6, La- and Yb-doped SmB6, and the Ce- and Th-doped
CaB6 relatives of Cal_6La6B6 are given below.
1.2.1 Carbon Doping ofEuB6
The ferromagnetism exhibited by EuB6 is unique among rare earth hexaborides.
For this reason, many studies of this compound have focused on determining the
mechanism behind its ferromagnetism and how it differs from that responsible for the
antiferromagnetism typically seen in the rare earth hexaborides.
In initial efforts, carbon doping of EuB6 has been found to suppress its
ferromagnetism [32,33] and is therefore a technique that is used to facilitate the
aforementioned investigation. Carbon is incorporated into EuB6 as a boron substitution
and acts as a single electron donor. The radius of carbon is smaller than that of boron so
-=-- before anneal
--- after anneal
0 50 10
Figure 5-5. Resistance versus temperature in Cal_6B6 prior to (black squares) and
following (red circles) 02 anneal.
U3 -" -- before anneal
Su -a- after anneal
Figure 5-6. High temperature region of Figure 5-5.
2.2 Recent Experimental and Theoretical Results in CaB6 and SrB_
In this section, a collection of results obtained for SrB6 and CaB6 are provided.
Comparison of CaB6 to SrB6 reveals that the compounds are isovalent and similar in bond
lengths. Following the discovery of ferromagnetism in CaB6 , long-range order was
inferred to exist in SrB6 . As a result, experimental data and theoretical results for
both materials are compiled here.
2.2.1 Theoretical Studies
Since the discovery of novel ferromagnetism in lightly electron-doped CaB6, there
has been a surge of experimental and theoretical work that forms the basis of current
knowledge pertaining to this system. Theoretical efforts have been undertaken by a
number of workers. While there are a few avenues that have been pursued in attempting
to describe the ferromagnetism, special attention has been given to an excitonic insulator
picture. The theoretical foundation of the excitonic insulator model was published nearly
simultaneously in 1965 by two independent efforts: Keldysh and Kopaev  and des
Cloizeaux . The model has been recently revived in applications to CaB6 and the
closely related compound SrB6.
Initial theoretical efforts focused on a prediction made by Bloch in 1929 , in
which a polarized electron fluid forms at electron densities between the upper limit of a
free electron gas and the lower limit of an insulating Wigner crystal. This intermediate
state occurs for a particular range of carrier concentrations, where electron exchange
interactions are dominant over Coulomb interactions, the bounds of which have been
calculated by several workers [43-46] with a large variation in results. There are two
characteristics of the system under scrutiny that appear to be fatal to the success of this
theory in application to CaB6. The first is that the electron densities at which
 Z. Fisk, D.C. Johnston, B. Cornut, S. von Molnar, S. Oseroff, and R. Calvo,
"Magnetic, transport, and thermal properties of ferromagnetic EuB6," Journal of
AppliedPhysics 50, 1911-1913 (1979).
 C.N. Guy, S. von Molnar, J. Etourneau, and Z. Fisk, "Charge transport and pressure
dependence of Tc of single crystal, ferromagnetic EuB6," Solid State
Communications 33, 1055-1058 (1980).
 T. Kasuya, K. Takegahara, M. Kasuya, Y. Isikawa, and T. Fujita, "Electronic
structure of EuB6, transport and magnetic properties," Journal de Physique,
Colloque C5 41, C5161-C5170 (1980).
 S. Sullow, I. Prasad, S. Bogdanovich, M.C. Aronson, J.L. Sarrao, and Z. Fisk,
"Magnetotransport in the low carrier density ferromagnet EuB6," Journal of
AppliedPhysics 87, 5591-5593 (2000).
 S. Sullow, I. Prasad, M.C. Aronson, S. Bogdanovich, J.L. Sarrao, and Z. Fisk,
"Metallization and magnetic order in EuB6," Physical Review B 62, 11626-11632
 T. Fujita, M. Suzuki, and Y. Isikawa, "Specific heat ofEuB6," Solid State
Communications. 33, 947-950 (1980).
 L. Degiorgi, E. Felder, H.R. Ott, J.L. Sarrao, and Z. Fisk, "Low-temperature
anomalies and ferromagnetism of EuB6," Physical Review Letters 79, 5134-5137
 P. Nyhus, S. Yoon, M. Kauffman, S.L. Cooper, Z. Fisk, and J.L. Sarrao,
"Spectroscopic study of bound magnetic polaron formation and the metal-
semiconductor transition in EuB6," Physical Review B 56, 2717-2721 (1997).
 M. Kasaya, J.M. Tarascon, J. Etourneau, and P. Hagenmuller, "Study of carbon-
substituted EuB6," Materials Research Bulletin 13, 751-756 (1978).
 S. von Molnar, J.M. Tarascon, and J. Etourneau, "Transport and magnetic properties
of carbon doped EuB6," Journal of Applied Physics 52, 2158-2160 (1981).
 J.M. Tarascon, J.L. Soubeyroux, J. Etourneau, and R. Georges, "Magnetic structures
determined by neutron diffraction in the EuB6-xCx system," Solid State
Communications. 37, 133-137 (1981).
 Mitsuo Kasaya, Hitoshi Kimura, Yosikazu Isikawa, Toshizo Fujita, and Tadao
Kasuya, "Valence instabilities and electrical properties of the La- and Yb-
substituted SmB6" in Valence Fluctuations in Solids, L.M. Falicov, W. Hanke,
and M.B. Maple, eds. (North-Holland Publishing Co., New York), 251-254
shown to reflect different activation energies associated with promotion from the valence
and impurity bands into the conduction band.
(a) --, ,,
(b) ---"---,---- ^ -^_---:-----
Figure 6-7. Band structure model depicting shifts in the Fermi level as a function of
(b) CaB6, and
(c) Cal_-B6 (shading indicates impurity band).
The suggestion of Monnier and Delley (see chapter 2) that there exist local
moments with B6 vacancies or anti-site substitutions is consistent with our picture. These
workers suggest that the moments are confined to the surface of the single crystals, in
moderate agreement with the ESR studies performed by Kunii, and to grain boundaries in
polycrystalline samples. We propose that the moments need not be confined to the
surface. Inhomogeneous distribution of these defects, as implied by the work of
Terashima et al., is possible because of local fluctuations in temperature during Al flux
growth. Not only could localized thermal pockets produce inhomogeneities, but
Transport and optical studies were undertaken to investigate the possible
appearance of evidence for the excitonic insulator picture [56,57]. The single crystals of
these studies were grown under the same conditions as those of Young's work, namely
by Z. Fisk's group, as were the samples studied for the purposes of this dissertation.
According to these workers, the samples called Cai+1B6 are thought to be stoichiometric,
and the crystals labeled CaB6 are thought to be hole-doped. Measurements of resistivity,
shown in Figure 2-6, asserted that stoichiometric CaB6 is semiconducting. The La-doped
and Ca-deficient compounds were proclaimed to be metallic, due to assumptions that the
Ca deficiency corresponds to hole doping. It should be emphasized that the transport
results to be presented in chapter 5 disagree with these conclusions for the stoichiometric
and Ca-deficient compounds, specifically in that the transport properties of the crystals
seem to be interchanged.
10-2 Cal_1LaSB6 _
0 100 200 300
Figure 2-6. Resistivity versus temperature in the electron-doped CalpLa5B6, Ca-deficient
CaB6 (to be denoted Cal-5B6), and stoichiometric Cai+6B6 . Note that these
identifications conflict with the results of our work for the stoichiometric and Ca-
hour annealing in an inert Ar environment at temperatures of approximately 100 C
further established the low-resistance character of the contact pads.
The Hall effect and magnetoresistance studies, for which a magnetic field is
required, were conducted in a 4He cryostat, as shown in Figure 4-2. The dewar is
equipped with a pump-out port, making possible a base temperature of 1.6 K, as well as
an American Magnetics, Inc. (AMI) superconducting magnet with a maximum field of 8
Tesla. The cryostat is equipped with temperature sensors, a small resistive heater for low
temperature regulation, and a power resistor with which temperatures near 300 K can be
attained. Two layers of vacuum provide good thermal isolation of the sample stage from
the liquid helium bath making high temperature measurements in magnetic field possible.
Temperature regulation was performed using a LR-400 resistance bridge coupled with a
LR-130 temperature controller. At low temperatures, the small 50 Q resistor provided
adequate heating. Above roughly 15 K, the 5-Watt power resistor served to bring the
equilibrium temperature up, while the small resistor functioned as a fine control.
A smaller 4He probe with one layer of vacuum isolating the sample from the
helium bath was employed for resistivity measurements. The temperature was varied in a
very simple way. Cold N2 gas was first circulated through the N2 jacket, without
collecting liquid. Liquid He was transferred into the dewar only until the sample reached
4.2 K. With the aid of automated data acquisition, which will be described below, the
measurement can be performed without supervision, as the dewar slowly warms. The
voltage output of the LR-400 is used to record the sample temperature.
pseudo-gap below approximately 40 K , where the density of states is steadily
reduced with a power-law dependence on temperature. Note the redistribution of
electronic states to energies above the gap energy, denoted by the arrow. The origin of
the gap is thought to be hybridization between the closely situated f- and d-bands, as
proposed by Mott in 1974 . Other workers have instead proposed the formation of a
Wigner crystal in which the interaction dominating the kinetic energy may not be
coulombic in nature . Later work has proposed that intra-gap impurity bands, due to
such defects as Sm vacancies, dominate the low-temperature region [20-22].
While the trivalent Sm cation carries a magnetic moment, SmB6 has not been
found to exhibit long-range order. A paramagnetic response of magnetization to the
application of magnetic field appears to persist to temperatures well below 1 K. The
most likely scenario to account for the absence of long-range magnetism is the significant
spatial separation between trivalent cations. In such a configuration and with a carrier
concentration low enough to produce semiconducting behavior, magnetic ordering is
difficult to produce by way of the RKKY interaction.
CeB6 is also classified as a valence fluctuating compound. Magnetic ordering in
CeB6 has been given special attention due to its complicated phase diagram, which
exhibits three distinct phases as the material is cooled . At high temperatures, CeB6
behaves like a typical dense Kondo system with a Kondo temperature of approximately 1
K. At temperatures between 2.4 K and 3.2 K, the material undergoes an
antiferroquadrupolar ordering. With the ground state of these moments being the F8
quartet, the Ce atoms are best described by quadrupole moments. Antiferromagnetic
ordering of these moments commensurate with the lattice is the phase in which CeB6
typically linear, with a correction term that approaches T5 at lower T, containing
information about electron-phonon scattering, which contributes a T3-dependence, and
small-angle electron-electron scattering, which contributes a factor of T2. At the lowest
temperatures, there is insufficient thermal energy to excite phonon modes, and ultimately,
the dominant scattering mechanism is that of electron-impurity scattering. In a perfect
crystal at zero temperature, the resistivity vanishes.
In contrast, perfect semiconductors exhibit infinite resistivity at zero temperature.
At finite temperature, these materials rely on thermal energy to promote valence band
electrons across an energy gap into the conduction band. The resistivity in the presence
of thermal excitation follows an exponential form as a function of temperature, as in
p(T)= pO e A/
In the following subsections, the effect of an applied magnetic field on the
electron transport in metals and semiconductors will be presented.
3.2.1. Electrons in Applied Magnetic Field I: Magnetoresistance
The magnetoresistance of a material refers to the effect of an applied magnetic
field on its electronic transport properties. There are two main configurations commonly
used to measure the magnetoresistance: longitudinal, and transverse. In the longitudinal
configuration, the current is injected, the magnetic field is applied, and the voltage is
recorded along the same crystal direction. In the transverse orientation, the field is
applied in a direction perpendicular to that of the parallel current injection and voltage
Figure 3-3. Contact configuration for performing Hall effect measurements.
The Hall effect is a manifestation of the manner in which the magnetic field exerts
forces on different charge carriers. The electrons are pushed to one side of the crystal,
and the holes are pushed to the other. The transverse voltage can be interpreted to give
the carrier concentration by Equations (3-3) and (3-4), applicable to a system with a
majority carrier type.
VH =RH IH/d (3-3)
RH = (ne)-, (3-4)
where VH is the Hall voltage, RH is the Hall coefficient, Iis the longitudinal current, His
the applied magnetic field, d is the thickness of the crystal, n is the carrier concentration,
and e is the charge of the particle. The sign of the Hall coefficient conveys the nature of
A recent theoretical result proposed recently by Monnier and Delley 
investigates a possible off-stoichiometry effect in which the single crystals contain boron
vacancies. These workers have performed calculations within the local density
approximation (LDA) of density functional theory (DFT) to produce theoretical values of
the formation energy and magnetic moment associated with plausible types of defect.
The defects studied include various substitutions of La and Al (from the flux growth
technique of Young et al.) for Ca and B6, interchange of Ca and B6, and Ca, B, and B6
vacancies. Their findings imply the presence of a magnetic moment associated with B6
vacancies and Ca substitutions on B6 sites. Ca substitution for B6 is predicted to produce
a moment roughly half the size of that associated with a B6 vacancy. It was also
predicted that the substitution of La or Al for B6 quenches the moment, an observation
that is consistent with a decrease in magnetic moment with La doping at concentrations
above 0.5% , as is described below.
Monnier and Delley suggest that the most likely source of B6 vacancies is the
 set of crystal surfaces, since the cleaving process of single crystal hexaboride
surfaces occurs through severing of inter-octahedral bonds at the surface. This region
appears to be a good approximation to the B6 vacancy density corresponding to the
experimentally determined magnitudes of magnetic moment. The viability of this
argument will be related to the results of this work in chapter 6.
2.2.2 Experimental Studies
The experimental investigations that have been performed to date on the CaB6-
like divalent hexaborides have produced a multitude of unclear results. Contradictory
effects are seen, and even the compositions of samples are often ambiguous, as a result of
disagreeing behaviors between crystals that are labeled by the same stoichiometry. For
that, with increasing carbon concentration, the lattice parameter of the compound
decreases. This doping results in two transitions, one magnetic and one electronic.
A ferromagnet to antiferromagnet transition occurs with increasing carbon
concentration, a progression that may be equivalent to introducing a smaller cation with
increased valency. Results show a change in sign of the paramagnetic Curie temperature,
Op, as the doping level is increased from x=0 to x=0.21 . Further investigations have
yielded information about the intermediate doping regime in which the transition takes
place. Neutron diffraction data suggest the coexistence of the low-doping ferromagnetic
phase and a high-doping helimagnetic phase, a spiral structure that can be formed through
antiferromagnetic interactions .
Measurements of resistivity as a function of doping concentration reveal a
reduction in resistivity with increased carbon content, consistent with the addition of
electrons to the conduction band, which seems to dominate over any decrease in mean
free path accompanying the reduction in lattice constant. The correlation between
enhanced carrier concentration and the suppressed onset of ferromagnetism and eventual
antiferromagnetic order indicates that the higher density of conduction electrons
promotes antiferromagnetic order, as seen in the trivalent rare earth hexaborides. These
observations are supportive of an RKKY-mediated ordering.
The suggested competition between the addition of charge carriers and the
ferromagnetic state is also consistent with the polaronic mechanism for ferromagnetism
advocated by Nyhus et al . High carrier concentrations increase electronic screening
and tend to inhibit magnetic polaron formation, which would similarly suppress the
ferromagnetism characteristic of stoichiometric EuB6.
of J is determined to be 70 meV. In temperature, 70 meV corresponds to roughly 700 K,
a figure not far from the reported Tc of this material.
It is important to emphasize that the Kondo effect is typically incompatible with
the presence of ferromagnetic order. This observation seems to provide two possible
scenarios for this material that are mutually exclusive. For a compound that is truly
ferromagnetically ordered, we would be obliged to abandon this model in favor of a
mechanism that can coexist with long-range order. If the La-doped material is actually
well described by the Kondo picture, the presence of disordered local magnetic moments
is implied. In conjunction with the recent theoretical prediction that magnetic moments
may be associated with B6 vacancies or substitutions with Ca, as described in chapter 2,
the suggestion of a Kondo state is a consistent one, provided they are disordered. This
second possibility may constitute further evidence for non-uniform sizes and distributions
of moment-carrying defect pockets, in which superparamagnetic or paramagnetic
behavior is possible in place of ferromagnetism, the proposals of which were discussed in
A few important caveats regarding these suggestive models should be noted.
First, it should be recognized that InTbehavior could arise from mechanisms other than
the Kondo effect. These other mechanisms are governed by electron correlation effects
and, as it is not clear that they are necessarily related to this system, have been omitted
from the discussion here. It is also essential to note that, due to the narrow range in
temperature over which these fits have been performed, the agreement to the models
discussed above only suggests a particular mechanism. The fits cannot be considered to
superconducting gap. Another test of junction quality is to determine the ratio of the
four-terminal dynamical resistance at energies less then the gap to that at energies greater
than the gap. A ratio of RA = 100 at T < Tc/2 is considered an indication of a good-
quality barrier. Additionally, phonon modes, which appear as local extrema in the
tunneling conductance, can be detected at bias voltages 7 mV and 11 mV. In the
presence of inelastic tunneling mechanisms, which are highly undesirable when present
in the barrier, this phonon structure will be undetectable.
-4 -2 0 2 4
Figure 4-4. I-V characteristic of a EuB6-I-Pb tunnel junction. The gap about zero bias is
clearly seen, as is the break in slope occurring at roughly +2 mV.
The most significant difficulty has been to fabricate a high-quality insulating
barrier. Many variables contribute to the success of a barrier. Not only must a
fabrication method be optimized, which is a challenging process, but the surface of the
structure for the doped system in the absence of excitonic order, and (b), which shows the
band splitting for one spin species in the presence of excitonic order .
There is only a narrow doping range in which the excitonic state is stable, which
is shown to be in agreement with experimental results by Young et al., presented below.
Specific to CaB6, it is not established that the valence and conduction bands are
sufficiently symmetric to each other to enable excitonic condensation. It is also not clear
whether CaB6 is a fully or partially compensated semimetal or a direct band-gap
semiconductor. The prediction of this dissertation is given in chapter 6.
An experimental signature of excitonic condensation of this type can be seen in
far infrared (FIR) spectroscopy. In the purely insulating regime, there should occur one
minimum in the reflectivity, denoting a gap that is of equal value for the spin-up and
spin-down bands. If asymmetry is introduced into the valence and conduction bands with
respect to each other, or if the system is driven from full compensation, as with doping,
for example, the gap values will become different for the two spin species. In this case,
two minima in the optical conductivity are expected to appear, which represent the
energy gaps for each spin species. These FIR studies have been performed by
experimental workers and are outlined below.
In addition to the excitonic insulator model, other theoretical proposals have been
made. In an argument made by Hirsch , the minima of two sub-bands, possibly
identifiable with opposite spin species, lie at the same energy, but the curvatures are
different. The theory relies on a broadening of one spin band with respect to the other
upon ordering. This loss of kinetic energy, Hirsch has argued, drives the ferromagnetism,
and because of the difference in curvatures of the two spin bands, one becomes a majority
crystal. The results of the study indicated that there was no correlation between the
inferred crystal stoichiometry and the resistivity of the sample.
Al-flux La 0.5% #7 /
2.0 FZ La 0.5%
t 1./ i"/ --FZ CaEB
1 .6 ----- (a) -
S 030 I I I
a 0.5 -026 -
0.4 o0 1i 2.0
Al-flux La 0.5% #1
0 50 100 150 200 250 300
Figure 2-15. Resistivity versus temperature for Ca0.995Lao.005B6 and CaB6
(a) grown by FZ technique and
(b) grown by Al-flux technique .
Magnetization versus magnetic field data is shown in Figure 2-16, where samples
(1) through (4), which were grown at either 1200 C for 12 hours or at 1500 C for one
hour, are diamagnetic, and (5) through (8), which were grown at 1500 C for longer
times, exhibit ferromagnetism. In Figure 2-17, the magnetization for a ferromagnetic
sample (9) and a diamagnetic sample (1) is shown. Comparison with the resistivity
versus temperature for diamagnetic samples (2) through (4) and ferromagnetic sample (9)
in Figure 2-18 indicates no correspondence between ferromagnetism and metallic or
semiconducting behavior. Morikawa et al. therefore assert that ferromagnetism in this
growth and preparation, low temperature methods, and measurement procedures,
followed by a thorough account of results in chapter 5. Chapter 6 will consist of an
interpretation of the work presented in this dissertation, and chapter 7 will summarize.
1.1 Electronic and Magnetic Properties of the Hexaboride Family of Compounds
Although early work on the metal borides dates back to the early 1950s, this class
of compounds first became a focus of intense research in the late 1960s. Many of the
studies have involved the rare earth and alkaline earth hexaborides. These materials exist
in the CsCl crystal structure, in which a cage of B6 octahedra surrounds the metal atom,
as shown in Figure 1-1. Divalent hexaborides were long considered semiconducting, as
will be discussed below, and the presence of trivalent (tetravalent) cations gives rise to
metallicity with an estimated one (two) conduction electron(s) per metal atom. It follows
that these compounds are excellent candidates for the study of the metal-insulator
transition. Because the hexaborides are isostructural in the presence of different cations,
doping studies have been a major component of the experimental efforts concerning these
Subsection 1.1.1 will serve to present generally obeyed properties of the divalent
hexaborides. Subsection 1.1.2 will introduce the electronic and magnetic characteristics
of the rare earth hexaborides, while subsections 1.1.3 and 1.1.4 focus on the intermediate
valence compound SmB6 and the ferromagnet EuB6, respectively.
As the field has evolved, the experimental and theoretical investigations that have
been conducted on these compounds have become the source of much dispute. In
particular, the electronic transport properties of the divalent hexaborides, including EuB6,
are not easily interpreted to be metallic or semiconducting. The origins of long-range
This summary chapter serves as the closing of this dissertation. Section 7.1 will
briefly reiterate the major results of this work, and suggestions for future experimental
work on CaB6 will be presented in section 7.2.
7.1 Synopsis of Experimental Results
The transport studies of La-doped, intended stoichiometric, and Ca-deficient CaB6
have produced a number of clarifying results for single crystals grown by the Al flux
technique. We have established that the crystals labeled CaB6 exhibit semimetallic
behavior, in contrast with expectations of intrinsic semiconductivity or perfect electron-
hole compensation. This finding may imply that the CaB6 samples are not truly
stoichiometric, and some type of impurity or defect contributes electrons to the system.
This identification has enabled a model band structure to be presented and has given rise
to an improved understanding of the electronic environment present in this system.
Through results of resistivity, Hall effect, and mobility studies as a function of
temperature and doping level, we suggest the existence of an impurity band in the
semiconducting material, giving rise to the low-temperature plateau observed to correlate
with off-stoichiometry effects. Vacancies of B6 octahedra and anti-site substitutions are
two possible defects that could produce an impurity band while also possessing a small
magnetic moment, according to a recent theory. It is speculated in this work that these
impurities are formed in clusters, in a process driven by local fluctuations in melt
ferromagnetism appears are rather high with respect to the calculated values. In addition,
the polarized electron fluid is treated for one type of carrier, and the transport properties
of the divalent hexaborides are generally governed by a combination of electrons and
holes. Because of these pathologies, the intermediate density polarized electron fluid was
somewhat neglected in favor of the excitonic insulator [47-52].
An excitonic insulator is theorized to form in compensated semimetals where
Coulomb interactions are significant in the absence of substantial screening. The band
structure corresponding to such a semimetal is illustrated in Figure 2-2 (a). The
symmetry between the electron and hole bands makes possible a pairing of electrons and
holes in the same k-state. These excitons are bosonic in nature and form what is known
as an excitonic condensate. The energy required to form an electron-hole pair is reflected
in the formation of a small energy gap in the band structure, at the middle of which lies
the Fermi level, as shown in Figure 2-2 (b). This band structure defines the excitonic
\t / 0
Figure 2-2. The evolution of a semimetallic band structure with excitonic ordering:
(a) Band structure for a compensated semi-metal.
(b) Band structure upon formation of an excitonic insulator, where the Fermi
level lies at mid-gap .
There are degenerate singlet and triplet states that correspond to a charge density
wave (CDW) and a spin density wave (SDW), respectively. This degeneracy can be
behavior, which would indicate that the Ce cations exist in a trivalent state at dilute
Resistivity versus temperature as a function of doping level for Ce concentrations
of 0.1%, 0.25%, 0.5%, and 0.75% reveals that metallic behavior appears to correlate with
the addition of low levels of Ce doping levels. In addition, measurements of
magnetization versus magnetic field yield an unexpected hysteretic behavior. This
hysteresis is clear evidence for ferromagnetic order. The assumption that the f-type
moments originating from the Ce are responsible for the ferromagnetism conflicts
significantly with the RKKY interaction thought to mediate antiferromagnetism in the
magnetically dilute regime associated with these observations.
Efforts to simplify the problem and explore the role played by the 4-f Ce moments
prompted Young to fabricate La-doped CaB6 single crystals. The addition of La
preserved the trivalent valence state while excluding the magnetic nature of the Ce. An
exceptional result was found. In the presence of small La concentrations, ferromagnetic
hysteresis in magnetization versus magnetic field was discovered in CaB6 .
The saturation magnetic moment was found to be a maximum value of 0.07 [tB/La
ion at a doping level of 0.5% and was additionally found to persist to 600 K. These data
will be presented in chapter 2. Important evidence in support of the intrinsic nature of
this ferromagnetism was obtained by subsequent Th-doping studies. If the addition of
one electron per dopant ion gives rise to long-range order, what is the effect of doping
with a tetravalent element, thus adding two electrons per dopant ion? The results are
consistent with those seen in Cal-aLaaB6. The maximum saturation magnetization occurs
lifted in two ways. In the event that electron-phonon interactions are important, the
CDW state lies lower in energy and implies that charge is the relevant degree of freedom
in the system. Conversely, short-range coulomb interactions can dominate, forming a
spin-modulated state. The latter possibility is thought to be the ground state of the
excitonic insulator in the hexaboride materials, owing to the long mean free paths and
low defect content of these compounds.
/ N\ <^
X(a) I (b)
Figure 2-3. The band structure of an excitonic system with electron doping:
(a) Semi-metallic band structure in the absence of excitonic condensation.
(b) Band structure showing the excitonic gap for one spin species .
The semiconductivity originally observed in CaB6 in the absence of La has been
thought to correspond to the insulating state described above. When carriers are then
added to the system, the excitonic condensate becomes slightly unstable. For low doping
concentrations, the electron-hole pairing can be maintained by reducing the insulating
gap. Additionally, the lowest energy state for the excitonic condensate occurs when all
added carriers are placed into one spin band. The cost in energy of accepting electrons
unpaired with holes into the system is compensated by a reduction in the excitonic gap
for that spin species. The position of the Fermi level is correspondingly raised in the
process of doping, and same-spin electrons are responsible for conduction within the
doped triplet excitonic insulator. This is shown in Figure 2-3 (a), which shows the band
In the magnetoresistance observed in the La-doped compound, the curvature
changes sign at a temperature near 28 K, suggestive of a correlation between this effect
and the minimum in resistivity versus temperature. It was seen in the results presented in
Figure 5-12 of chapter 5 that at 22 K and below, the curvature of the data is positive,
while at temperatures greater than 36 K, the curvature becomes negative in sign.
Deviations from positive H2 magnetoresistance suggest a more complex
electronic environment than a simple free-electron model. Here we consider mechanisms
that can produce a negative magnetoresistance, including the Kondo effect, weak
localization where spin-orbit scattering is negligible and an alternate mechanism in which
a decrease in resistivity can be produced due to band shifting at the Fermi level in the
presence of a magnetic field.
At temperatures below TK, at which the Kondo minimum in the resistivity occurs,
Kondo theory predicts a negative power law behavior in magnetoresistance versus
magnetic field. Monod  determined the field dependence of the resistivity to be
approximately p -H", where n z 1.7 to 1.8. This functional form, however, gives a
negative curvature for magnetoresistance versus temperature, which is contradictory to
our findings. We therefore assert that the Kondo effect is most likely not the mechanism
driving the 28 K transition in Cal_6La6B6.
In the presence of weak localization in three dimensions and where spin-orbit
scattering is sufficiently weak, we expect the quantity Ap/p 2 to behave as -4H :
Ap 0.605 e2 (eH
p2 2-2 -h -h
Experimental data are shown for Cal-5La6B6 in Figure 6-3, where the 5 K data are
at a Th level of 0.25%. This value is half the doping concentration but corresponds to the
same electronic content required in the La doping studies.
The discovery of ferromagnetism in Cal-6La6B6 directly inspired the work
presented in this dissertation.
deviation from these fits at intermediate temperatures may represent interactions poorly
described by the independent two-component model employed to investigate the
asymptotic limits of the resistivity results in Cal-sB6.
To ensure that the activated behavior in the Ca-deficient material is due to Ca
vacancies, a post-growth method was developed to further deplete the Ca concentration.
Annealing in an oxygen atmosphere exploits the low vapor pressure of CaO to remove Ca
from the surface of the crystal. This anneal was conducted at a temperature of roughly
250 "Celsius. A subsequent anneal in an inert Ar environment at a slightly higher
temperature was performed in an effort to restore uniformity to the crystal, thus
enhancing the concentration of Ca vacancies in the bulk.
The effect of the annealing procedure on the behavior of resistivity with
temperature is shown in Figure 5-5. The data in black were taken on a Ca-deficient
crystal prior to anneal. After anneal, as shown in red, the same crystal displays no
detectable change in the value of the high-temperature resistivity, as can be seen in
Figure 5-6, which is a magnified view of the high-temperature region of Figure 5-5. The
values of the high- and low-temperature gaps do exhibit a 2-3 percent enhancement after
anneal. The change in gap values implies that, by removing the electron donor cations
from the system, the Fermi level has been shifted downward in energy and apparently
deeper into a band gap. These data suggest the presence of the thermal excitation of
carriers from a source in addition to the valence band.
Normalized resistivity versus temperature in Cal-6La6B6.
0 50 100 150 200 250
Normalized resistivity versus temperature in CaB6.
0 50 100 150 0 2 250 300
electrons. For a group of six boron atoms, then, there are 18 electrons that can participate
in bonding. In preparation for using the tight-binding approximation, these workers
enumerated the orbitals available for filling for each octahedron in terms of symmetry
groups. By calculating the energy contained in each bond, bonding orbitals were
differentiated from anti-bonding orbitals. Of the bonding type, there are seven, each of
which can hold two electrons of opposite spin. Combining these 14 orbitals with those
bonding between borons and the neighboring octahedron, of which there are six, the total
electron capacity required for full covalent character is 20. The implication is that the
combination in a unit cell of a divalent metal with the 18 electrons native to the B6 group
gives a fully covalent structure. Measurements of the temperature dependence of
resistivity in single- and poly-crystals of the divalent, non-magnetic hexaborides
produced activated behaviors, enforcing the apparent correctness of the semiconducting
Heat capacity studies of the divalent hexaborides reveal a small electronic
contribution at low temperatures . The magnitude of this signal is interpreted to reflect
a small density of states located at the Fermi level, which in turn reflects a low carrier
concentration, in agreement with the semiconductor-like behavior of the resistivity.
Nuclear magnetic resonance (NMR) measurements of 11B chemical shifts in
CaB6, SrB6, and BaB6 lent further credence to the claim that the alkaline earth
hexaborides are semiconducting . The B11 chemical shifts measured in powdered
samples indicate a diamagnetic response, which the author interprets to be a signature of
3.3.2 Zero Bias Anomalies
Sharp features centered about zero voltage in the tunneling conductance known as
zero bias anomalies (ZBAs) have been associated with inelastic mechanisms present in
the electrode . The concept of a ZBA was introduced in chapter 2 during the
discussion of tunneling results in EuB6, and it was noted that the appearance of the zero
bias peak is simultaneous in temperature with ferromagnetic ordering. It is important to
mention that there is no consensus concerning the origins of these features, and the term
anomaly certainly suggests incomplete understanding.
One type of ZBA is known as the giant resistance peak and is the least understood
of all types of ZBA. This type of feature has appeared in work on Cr-CrOx-Ag tunnel
junctions , as shown in tunneling resistance (dV/dl) versus bias voltage (Figure 3-7).
Initial observations noted that there are most likely magnetic moments present in the
barrier, as CrO2 is ferromagnetic and CrO3 is antiferromagnetic. Mezei and Zawadowski
presented a Kondo-like effect as a possible mechanism leading to the increase in
resistance . This mechanism explained a logarithmic dependence of the resistance on
temperature but was inadequate to explain the magnitude of the effect, which can be two
orders of magnitude larger than the resistance at 100 meV. A guideline provided by this
model is that conductance peaks are produced by moments in the barrier and dips are a
result of moments in the electrode. Artificial implantation of magnetic Ni into Al
electrodes gave a giant resistance peak  of the correct magnitude, but no conclusions
have been reached as to the actual mechanism responsible for the effect.
PREVIOUS WORK IN CaB6 AND RELATED COMPOUNDS
The occurrence of ferromagnetism in CaB6 has no true predecessor, that is, a
material composed of non-magnetic constituents that exhibits ferromagnetic order.
Section 2.1 includes a review of electron tunneling results for EuB6, the only other
ferromagnetic hexaboride to date, which may prove helpful in determining the origin of
certain features that appear in CaB6 data.
Section 2.2 will focus on recent data obtained for CaB6 and SrB6, which possess
very similar electronic and crystal structures. In fact, some workers have recently
inferred the existence of ferromagnetism within the SrB6 compound because of its
similarity to CaB6.
2.1 Electron Tunneling Spectroscopy in EuB6
Electron tunneling spectroscopy is a method by which the electronic density of
states of a material can be qualitatively measured. Features appearing in the tunneling
spectrum of a compound can provide information on how the states evolve with changes
in the external environment, examples of which are temperature variations and the
application of a magnetic field. Features important to the understanding of the electronic
structure in the material can be resolved, including gap structures, evidence for the
presence of magnons, and substantiation for the presence of other inelastic excitations. A
more in-depth discussion of tunneling is given in chapter 3.
Specific to EuB6 is the tunneling spectrum shown in Figure 2-1 . The
counter-electrode used in these tunneling studies was Pb, and the I-V characteristic of this
of green paint was commonly added to make the adhesive visually distinguishable during
manual application. Once the contacts are confirmed to be electrically isolated from each
other and from ground, the sample is mounted onto the 4He probe. Thin Au wire is
attached to the contact pads in a four-terminal configuration such that current is injected
across the barrier, and voltage is measured independently across the barrier. This
eliminates the voltage contribution from the leads that would be present in a two-terminal
When conducting tunneling measurements, the quantities of interest are typically
very small in magnitude. It is essential, therefore, to perform these studies in a low-noise
environment. All data sets to be presented were taken in an RF shielded room with
analog electronics. Care must be taken to prevent voltage transients across the junction,
which can easily destroy the thin barrier. For this reason, shorting the junction leads is
necessary when setting up the measurement and electrically connecting the sample to the
There are two components of the voltage applied to the junction. One is a DC
level produced by a completely passive analog circuit with ramping capabilities as well
as a manual voltage adjustment. A small AC voltage at a frequency of roughly 1 kHz,
but not a harmonic of 60 Hz, originates from a PAR 124A lock-in amplifier and is added
to the larger DC signal. This circuit is shown in Figure 4-5. The DC voltage is the bias
voltage. For high-resolution spectroscopy, the AC voltage must be much smaller in
amplitude than the DC level, VAC < 0.05 VDC, since we will be using this configuration to
measure the dynamical resistance, or the inverse of the tunneling conductance, versus
bias voltage. The objective is to keep a constant current, with the use of a dropping
exists at intermediate temperatures. Below 2.4 K, CeB6 exists in a classical
1.1.4 Ferromagnetic EuB6
Until the discovery of ferromagnetism in CaB6, EuB6 was the sole ferromagnetic
exception to the generally antiferromagnetic ordering shown in the hexaborides. A
review of experimental results for EuB6 may prove helpful in determining the origin of
certain features appearing in CaB6 data.
7.0 I I I I
S5.0 / *H=
S H=15 kOe
.' 2.0 8
0.0 I I I i
0 50 100 150 200 250 300
Figure 1-3. Resistivity versus temperature in EuB6. Closed circles represent zero-field
data, while the open circles correspond to a 15 kOe applied magnetic field .
Examination of resistivity versus temperature in EuB6 reveals a sharp maximum
followed by a dramatic decline below 16 K [24,25], as shown in Figure 1-3. Since Eu
exists in a +2 valence state in this compound, semiconducting behavior was expected.
The large drop in resistivity with decreasing temperature was interpreted to indicate semi-
washing out the Pb structure, contributions due to undesirable inelastic transport
mechanisms may also be present.
It is also noticeable that the peak diminishes rather quickly with increasing
temperature and appears to be completely absent for temperatures above 25 K. Whether
this reduction in intensity indicates a transition, magnetic or otherwise, is not definite.
These results may be interpreted to be evidence for a second transition well below the Tc
of the material that has no bearing on the bulk ferromagnetism. While a double transition
is reminiscent of that seen in ferromagnetic EuB6, we hesitate to draw parallels between
the two systems based on these data.
Subsequent tunneling spectra on additional crystals are shown in Figures 5-16 and
5-17. Figure 5-16 is accompanied by the I-V characteristic in the inset for this particular
junction and exhibits a well-defined gap region about zero voltage. In the tunneling
conductance, the superconducting Pb gap appears at zero bias for temperatures below To
= 7.2 K. There is an unusual feature in the center of the gap, however. It seems that
there may exist a zero bias component to the low-temperature spectra that is unrelated to
the superconducting gap. Due to the absence of this feature in the tunneling spectra of
Figure 5-17, it is clear that more extensive studies are required to ascertain its origin.
The spectra exhibited in Figure 5-17(a) represent higher temperature behavior and
may indicate the formation of a pseudo-gap centered about 22 mV, a feature and energy
scale common to many of the previously studied hexaborides. The low-temperature data
do not display a clean zero bias peak, as seen previously. There are many maxima in the
At the mid-point of her fourth year, Stephanie began work under Professor Fred
Sharifi. Initial experimental responsibilities focused on the preparation of substrates for
applications to nanofabrication using electron-beam lithography. Before long, Stephanie
was familiar with thin film growth and electron tunneling spectroscopy techniques and
enjoyed her work to such a degree that she decided to continue her research under
Professor Sharifi as a graduate student, beginning in August of 1998.
Stephanie's graduate studies included electron transport and tunneling
measurements in the doped CaB6 system. In addition, she demonstrated viability of a
process to create evaporation masks of sub-micron features in thin Si3N4 membranes
using electron-beam lithography. Her tunneling work also included the formation and
preliminary study of tunnel junctions of UPt3, considered to be an unconventional
In the spring of 2001, Stephanie's work with Professor Sharifi was concluded, and
she proceeded to write her dissertation under the advisement of Professor Art Hebard.
She hopes to obtain future employment either in industrial research or at a national
Stephanie April Getty was born to Paul and Michelle Getty on April 12, 1977 in
Douglasville, Georgia. At the age often months, she moved to Hialeah, Florida and
remained a south Floridian until the age of seventeen. She began pre-school at Advent
Montessori, and by the time she transferred to the first grade at a public school, she had
learned the basics of cursive writing and multiplication. She credits the unconventional
teaching techniques of the Montessori program with a great deal of her scholastic success
Stephanie graduated as salutatorian from Piper High School in June 1994. She
decided to attend college at the University of Florida in Gainesville and, accompanied by
30 Advanced Placement credit hours, enrolled as a sophomore in the fall of 1994. She
began to investigate the field of experimental condensed matter physics in December of
1996, under the advisement of Professor Gary Ihas. Her duties as undergraduate research
assistant included the restoration of a 4He cryostat, assisting in low temperature
measurements of conducting polymers, and thermometry calibrations.
During the following summer, Stephanie participated in the Summer Research
Program at Bell Laboratories, Lucent Technologies under the guidance of Dr. C.H. Chen.
She learned to operate the transmission electron microscope and used bright field and
selected area diffraction modes to characterize impurities in Ni films. Upon her return to
the University of Florida, Stephanie continued work with Professor Gary Ihas until
December of 1997.
electron-doped analog. Notice, however, that the Ca-depleted compound retains
electron-like carriers. This indicates that the mid-gap position has not been crossed in
reducing the Fermi level in energy. The electron concentration is found to decrease in the
region of increasing resistivity, as to be expected. As the resistivity levels off, the
electron density demonstrates a simultaneous leveling. The concomitant behaviors of
resistivity and carrier concentration are consistent with a standard interpretation. Here
the increase in resistivity is due to a combination of phonon freeze-out and a reduction in
thermally excited electrons about the Fermi level.
SCarrier Concentration C
0 50 100 150 200 250 300
Figure 5-9. Carrier concentration (red) and normalized resistivity (black) versus
temperature in CaB6.
The stoichiometric material exhibits an interesting behavior in carrier
concentration versus temperature, as seen in red in Figure 5-9. The carrier concentration
5.2 Hall Effect Measurements
A series of Hall effect measurements was performed with the intention of
determining not only the sign of the carriers in each doping regime but the carrier
concentration as a function of doping level. We initially expected that in the La-doped
system the carriers would be electrons, that in the stoichiometric system there would be
equal numbers of electrons and holes, and in the Ca-deficient system the carriers would
be exclusively holes. The data contradict these expectations but are consistent with the
resistivity versus temperature behaviors presented above. A band structure model
incorporating this information will be fully developed in chapter 6. An alternative motive
of pursuing these measurements was to investigate the possible appearance of the
anomalous Hall effect as a signature of ferromagnetism.
1.6 8 S
C Carrier Concentration 3
0.6 -0 o Io I I o
0 50 100 150 200 250 300
Figure 5-7. Carrier concentration (red) and normalized resistivity (black) versus
temperature in Cal-6La6B6.
The Hall effect has been studied as a function of temperature for each of the three
doping concentrations. Figures 5-7 through 5-9 display the results of Hall effect
measurements in red. The resistivities versus temperature for the same samples are
included for comparison and are shown in black. The La-doped material exhibits a
carrier concentration of roughly 5x1019 electrons/cm3, as seen in Figure 5-7. This
electron concentration is approximately independent of temperature, indicating that the
conduction electrons reside in the conduction band and are not thermally excited from a
band lower than the Fermi level.
300- 2 -
Ca 1 gB M
0 Carrier Concentration -
S 50 100 150 200 250 300
Figure 5-8. Carrier concentration (red) and normalized resistivity (black) versus
temperature in Cal-6B6.
Figure 5-8 shows data for the Ca-deficient material, with the carrier concentration
shown in red and accompanied by the resistivity in black for comparison. The electron
concentration is and is found to be an order of magnitude lower than that seen in the
physical manifestation of the model. It is troubling, however, that these features are very
far from the energy scale of the undoped material. This large shift in energy is not
predicted to accompany slight doping of an excitonic insulator.
The results of band structure calculations by Massidda et al. that predict an
overlap between the conduction and valence bands and form the foundation for the
excitonic insulator seem to be contrary to the findings of several workers, the results of
which indicate a gapped band structure. Among these is angle resolved photoemission
spectroscopy data  that provides evidence for a band gap of the order of 1 eV. The
ARPES technique is a direct probe of the band structure, which is shown for CaB6 in
Figure 2-11. The band structure in solid black represents the calculations performed by
Tromp et al. that are in good agreement with the experimental findings. It is worthwhile
to note that the valence band image differs significantly from the doubly peaked band that
is predicted for an excitonic insulator, and the band gap is substantially larger than an
excitonic instability is expected to yield.
F X r
0.0 0.5 1.0 1.5
Figure 2-11. Angle resolved photoemission spectra in stoichiometric CaB6. The
theoretical GW-calculated band structure is shown in the dotted lines .
101' a.--'- -~ ,---.. Lal+6-6
_p = 8.8 + 730 TV
0 le-3 2e-3
10-1 100 101 102
Figure 2-7. Resistivity versus temperature in Cal+6B6. The log-log scale emphasizes the
low-temperature decrease in resistivity at roughly 0.2 K. The inset shows a fit to T 3 at
the lowest temperatures .
Figure 2-8. Specific heat versus temperature in Cal+6B6 and Cao.995Lao.oo5B6. Note the
local maximum at roughly 0.2 K .
Low temperature resistivity measurements by Vonlanthen et al. of the
semiconducting material show a slight decrease with decreasing temperature below a few
tenths of Kelvin. This is shown in Figure 2-7, the inset of which emphasizes a T3
dependence of low-temperature resistivity. When compared to specific heat versus
PRESENTATION OF RESULTS
In this chapter, the results of this work on CaB6 and its electron-doped and Ca-
deficient variants will be presented. The contents included here will be the foundation
for the interpretation to follow in chapter 6. Section 5.1 is devoted to a description of the
electronic resistivity as a function of temperature and electron content. In section 5.2,
Hall effect data are provided in the form of the Hall voltage versus field and carrier
concentration versus temperature and doping concentration. Magnetoresistance data are
the subject of section 5.3, and electron tunneling spectra are presented in section 5.4.
One fundamental complication with attempting to measure transport properties in
the single-crystal samples has been the variation in sample dimension and in sample
composition. The absolute values of the resistivity are difficult to determine because the
effective sample thickness, where the current is uniform, is suspected to be smaller than
the physical crystal thickness. In addition, the variations in composition occur even
within a single batch of crystals. To compensate for these sources of error, resistivity
data that is normalized to the value at 300 K is presented for the electron-doped and Ca-
deficient compounds and to the value at 50 K for the stoichiometric material. The issue
of non-uniformity within a crystal batch has been addressed by using a single sample
where possible throughout our series of measurements. In this way, the resistivity, Hall
effect, and magnetoresistance data shown below correspond to representative samples of
the same crystal stoichiometry and physical dimensions for each doping concentration.
2-14. Magnetization versus magnetic field in Ca0.995Lao.005B6 and CaB6 grown by FZ and
A l-fl u x tech n iqu es. .............................. ............. ........... ........... ... ........ ........ 3 6
2-15. Resistivity versus temperature in Ca0.995Lao.005B6 and CaB6 grown by FZ and Al-
flux techniques ............... ............. .................... .................. 38
2-16. Magnetization versus magnetic field in polycrystalline CaB6: growth conditions...... 39
2-17. Magnetization versus magnetic field in polycrystalline CaB6: ferromagnetism and
diam agnetism ......................................... ......... 39
2-18. Resistivity versus temperature in polycrystalline CaB6 ............... ..... ................ 40
3-1. Resistivity versus applied magnetic field in EuB6. ............................................ 47
3-2. Magnetoresistance versus applied magnetic field in SmB6. ........................................ 48
3-3. Contact configuration for performing Hall effect measurements .............................. 49
3-4. Schem atic of electron tunneling ........................................................................... .... 52
3-5. Normalized tunneling conductance versus positive bias voltage in a BCS
superconductor .................................................. .............. 54
3-6. Tunneling conductance versus positive bias voltage in a Pb-I-Pb tunnel junction........ 55
3-7. Zero bias anomalies in the tunneling resistance versus bias voltage for a Cr-CrOx-Ag
tunnel junction. ....................................................... .............. 57
3-8. Zero bias anomalies in the tunneling conductance versus bias voltage for
magnetically doped electrodes of (a) Ag and (b) Al. ...................................... 58
4-1. Contact configuration for simultaneous measurement of resistivity,
magnetoresistance, and Hall effect................................................ .............. 62
4-2. Depiction of 4He cryostat with pump-out port used for magnetotransport
measurements. .......................................... 64
4-3. C ross-section of tunnel junction............................................................................. .. 66
4-4. I-V characteristic of a EuB6-I-Pb tunnel junction. ............................................... 67
4-5. Circuit diagram for tunneling measurements ............................................... 70
5-1. Normalized resistivity versus temperature in CalpLa5B6........................................ 74
5-2. Normalized resistivity versus temperature in CaB6. ......................................... 74
5-3. Normalized resistivity versus temperature in CalB6................................ .............. 75
scattering. It is interesting to note that, under the conditions of weak localization,
inelastic scattering actually has the potential to increase the conductivity of a material as
it destroys the phase coherence of the carrier wavefunction.
Predictions made by Altshuler and Aronov [81,82] indicate that electron
interaction effects can induce localization in three dimensional systems with finite
disorder. At low temperatures, the conductivity is proposed to behave with a correction
term proportional to the square root of temperature, as shown in Equation (6-1):
cI(T) = co 1 kF+T[ 21
where a is a constant that can be positive or negative, kF is the Fermi wavevector, / is the
mean free path, and c is the elastic scattering time. Figure 6-1 shows normalized
conductivity versus the square root of Tfor Cal-5La6B6. The temperature range that has
been used in the fit is shown as the upper axis. The fit turns out to be reasonable. The
intercept of the linear fit is given by the value of the conductivity extrapolated to zero
temperature. From the fit and using 200 Q-1-cm-1 as an estimate of the conductivity at
300 K using typical crystal dimensions, the value of oo is roughly 135 Q-1-cm1.
Further comparison to Equation (6-1) can yield an estimate of the mean free path.
Using the approximations that a 1, l/t VF, and m ~ me, and by taking the value of the
Fermi energy to be 0.062 eV , the mean free path is found to be roughly 5 microns.
This result is consistent with predictions that the hexaborides are typically clean systems.
We verify that this second term is truly a correction in that kF / ~ 104, and consequently
much larger than unity. While Fermi liquid theory enhances the effective carrier mass,
which is proportional to the square of the mean free path but not accounted for in this
 J.C. Nickerson, R.M. White, K.N. Lee, R. Bachmann, T.H. Geballe, and G.W. Hull,
Jr., "Physical properties of SmB6," Physical Review B 3, 2030-2042 (1971).
 J.W. Allen, B. Batlogg, and P. Wachter, "Large low-temperature Hall effect and
resistivity in mixed-valent SmB6," Physical Review B 20, 4807-4813 (1979).
 N.F. Mott, "Rare-earth compounds with mixed valancies," Philosophical Magazine
30, 403-416 (1974).
 B. Batlogg, P.H. Schmidt, and J.M. Rowell, "Evidence for a small energy gap in
SmB6," in Valence Fluctuations in Solids, L.M. Falicov, W, Hanke, and M.B.
Maple, eds. (North-Holland Publishing Co., New York), 267-269 (1981).
 S. von Molnar, T. Theis, A. Benoit, A. Briggs, J. Flouquet, J. Ravex, and Z. Fisk,
"Study of the energy gap in single crystal SmB6," in Valence Instabilities, P.
Wachter and H. Boppart, eds. (North-Holland Publishing Co., New York), 389-
 B. Amsler, Z. Fisk, J.L. Sarrao, S. von Molnar, M.W. Meisel, and F. Sharifi,
"Electron tunneling studies of the hexaboride materials SmB6, EuB6, CeB6, and
SrB6," Physical Review B 57, 8747-8750 (1998).
 T. Kasuya, K. Takegahara, Y. Aoki, K. Hanzawa, M. Kasaya, S. Kunii, T. Fujita, N.
Sato, H. Kimura, T. Komatsumbara, T. Furuno, and J. Rossat-Mignod,
"Anomalous properties of valence fluctuating CeB6 and SmB6," in Valence
Fluctuations in Solids, L.M. Falicov, W. Hanke, and M.B. Maple, eds. (North-
Holland Publishing Co., New York), 215-224 (1981).
 N.F. Mott in Valence Instabilities, P. Wachter and H. Boppart, eds. (North-Holland
Publishing Co., New York.), 397-403 (1982).
 Jozef Roman, Karol Flachbart, Thomas Herrmannsdorfer, Stephan Rehmann, Frank
Pobell, Elena S. Konovalova, and Youri B. Padreno, Low temperature magnetic
properties of samarium hexaboride," Czechoslovak Journal of Physics 46, 1983-
1984 (1996), Suppl. S4.
 S. Gabani, K. Flachbart, P. Farkasovsky, V. Pavlik, I. Bat'ko, T. Herrmannsdorfer,
E. Konovalova, and Y. Paderno, "The energy gap of SmB6 at low temperatures,"
Physica B 259-261, 345-346 (1999).
 J.M. Effantin, J. Rossat-Mignod, P. Burlet, H. Bartholin, S. Kunii, and T. Kasuya,
"Magnetic phase diagram of CeB6," Journal ofMagnetism and Magnetic
Materials 47&48, 145-148 (1985).
-0.4 T=4 K
-0.54 \ I
0 10 20 30 40 50
Figure 3-2. Magnetoresistance versus applied magnetic field in SmB6 .
3.2.2. Electrons in Applied Magnetic Field II: Hall Effect
The Hall effect is a widely used method for measuring the carrier concentration in
a material. While the magnetoresistance, as discussed above, is the change in resistivity
along the direction parallel to the current, or longitudinally, in the presence of a magnetic
field, the Hall effect is the change in resistivity along the direction perpendicular to both
the current direction and the magnetic field. In Cartesian coordinates, if the current flows
in the y-direction, the Hall voltage is measured along the x-direction, and the magnetic
field is applied in the z-direction, as shown in Figure 3-3.
There are two conclusions we can draw about the hysteretic Hall voltage effect presented
in this work: the agreement in saturation field with magnetization data indicates that the
unusual first-sweep behavior appears to be related to the magnetic environment of the
sample, and this La-doped crystal is most likely non-ferromagnetic.
We therefore believe our data to be consistent with the presence of
inhomogeneous clustering of locally magnetic defects, the characteristic size of which
determines the bulk magnetism in this material. A summary of this work is given in
chapter 7 to follow.
-.- T *-
-0.005 -K 4 -
-* *"- _
-0.010 -- T=5K -., '.
--'- T = 10 K = 10.
-0.015T = 16 K E a
--4- T = 31K K .
T = 40 K
-0.020 T = 60 K
T = 75 K Cal- B6 "
-0.025 I I I
0 1 2 3 4 5 6 7
Magnetic Field (T)
Figure 5-14. Magnetoresistance versus magnetic field in Cal-6B6.
exhibits a magnetoresistance of a different functional form than that seen in the
stoichiometric compound. This observation will be addressed in chapter 6.
Data for the vacancy-doped compound is shown in Figure 5-14. These data
indicate a much larger effect than that seen in the two metallic materials. The maximum
value of the negative magnetoresistance at 5 K and 6 T is nearly 2.5%, a signal that is
approximately five times larger than that seen in the electron-doped and stoichiometric
analogs. The trend of decreasing signal with increasing temperature is preserved in this
5.4 Tunneling Measurements
Tunneling measurements were performed in advance of the transport studies
discussed above. The initial goal in undertaking tunneling experiments was to investigate
the possible appearance of zero-bias anomalies, in analogy with the features seen in the
We obtain a result that, at very small fields, depends quadratically on magnetic field.
With a Fermi level located very near the bottom of the conduction band, one band will
become dominant at higher fields, while the Fermi level overlap with the other is
suppressed, thus giving linear dependence on the magnitude of the field. At much higher
fields, the H3/2 behavior will become apparent. These are fields not attained in our
magnetoresistance studies, since for [IBWHF to be of the order of unity, H 103 T. The
results presented in this work are confined to a maximum field of 6 T and remain within
the linear regime at the highest fields, where one spin band dominates over the other.
We can calculate a zero-temperature estimate of the intermediate field
magnetoresistance in CaB6. Using the values EF = 0.062 eV and ltB = 5.8x105 eV/T, an
effect of roughly 0.75% can be expected at a field of 6 T. This result is in moderately
good agreement with the 0.4% effect seen at 5 K, as shown in Figure 5-13 of the
The magnetic field dependence described by the band shifting calculation
performed above (see Equations 6-4 through 6-6) can be tested against experimental
results. This is done in Figure 6-4, which shows magnetoresistance versus H2 for small
field in part (a) and magnetoresistance versus H for moderate field in part (b).
We next investigate the prefactors of the low-field H2 and high-field H
dependencies of magnetoresistance at low temperature. In Figure 6-4(a), the coefficient
of the H2 dependence is found to be -2.05x10-4 T-1. From the calculation, we expect this
value to approximate the quantity -(tB/EF)2 = -10-6 T-1. It appears that the calculation is
an underestimate of the experimental findings by two orders of magnitude. The origin of
metallic behavior instead and was attributed to a reduction in magnetic scattering upon
moment alignment. Negative magneto-resistance corroborates this picture [26,27], as
shown in Figure 1-4.
1500 A EuE6
S 0 OkOe
4 2.7 kOe
S13 } 5.A kOe-
500 8.1 kOe
x 10.8 kOe
I IH et 16. kOe
0 10 20 30 40 50 60 70 80 90
Figure 1-4. Resistivity versus temperature in EuB6 at various magnetic fields. At low
temperatures, the magneto-resistance is clearly negative .
Further studies of the magnetic transition in EuB6 demonstrated that there exist
two critical points associated with the onset of ferromagnetic order [28-30], one
occurring at 15.1 K and the other at 12.6 K. Figure 1-5 shows that these two transitions
appear as sharp peaks in the quantity dp/dTas a function of temperature. Subsequent
work has attempted to differentiate between the origins of the two transitions.
Accompanying this work  is a series of Arrott plots of magnetization versus magnetic
field for various temperatures. These curves indicate that EuB6 is ferromagnetic, i.e.
exhibits a positive value of magnetization at zero magnetic field, up to a temperature
these reasons, care must be taken in the interpretations applied to the results outlined
The discovery of ferromagnetism in La-doped CaB6 arose out of analogous
studies by Young  using Ce as a dopant, as described in chapter 1. In an attempt to
simplify the compound and understand the effect of a trivalent, non-magnetic dopant on
the CaB6 system, single crystals of Cal_6La6B6 were grown. Surprisingly, a small
magnetic moment of 0.07 [tB/La ion was discovered .
CaI La B
SCa 0.995 La0.005 6
Figure 2-4. Magnetization versus temperature in Ca0.995Lao.005B6 showing a Tc of
approximately 600 K .
Figure 2-4 shows the preservation of a finite magnetization in this material to a
temperature of at least 600 K, a Tc that is astounding in its own right because of the
extremely small moment. Recent workers have quoted the Tc to occur as high as 900
K.With variations in 6, it was found that there exists a narrow region in doping
I I I I ______
magnetic order in EuB6 and the topic of this work, CaB6, are still subject to much
Figure 1-1. CsCl crystal structure of the hexaborides.
1.1.1 The Divalent Hexaborides
The electronic properties of the divalent hexaborides, represented by CaB6, SrB6,
BaB6, YbB6, and EuB6, to be discussed separately below, were long considered to be well
understood. Longuet-Higgins and Roberts  performed early calculations of the
electronic configuration of the B6 octahedron. The analysis considers that each boron
donates 3 electrons to the system; one is a 2p electron, while the other two are 2s
behavior is less easily modeled, presumably because of a variety of contributions from a
complex band structure in a regime between semimetallic and semiconducting behaviors.
6.2 Band Structure Model
The interpretations given above for all three compounds can be integrated to form
a self-consistent model for this system. To begin, a candidate band structure model is
presented in Figure 6-7. This model includes a gap between the conduction and valence
bands. The conduction band may experience a splitting into spin up and spin down
components in response to local moments, and this has been tentatively incorporated into
the model as dashed curves. The position of the Fermi level depends on the doping level
of the material.
The Fermi level of the La-doped compound is depicted in position (a). This
positioning accounts for the metallic character exhibited by this material at temperatures
greater than 28 K. A Fermi level position characteristic of the stoichiometric material is
shown in position (b), though this positioning will vary with small changes in the ratio of
Ca:B6 of the crystal. A significant overlap of one band compared to the other may give
rise to the two-domain behavior of the resistivity as a function of temperature, as shown
in Figure 5-3. The Ca-deficient compound corresponds to the Fermi level in position (c),
where an impurity band located at the Fermi level lies at higher energy than mid-gap to
represent that the carriers are electrons.
The magnitude of the magnetoresistance for each sample is also consistent with
the model. The metallic overlap of the Fermi levels for both the La-doped and parent
materials imply the small magnetoresistances seen in experiment. There are two gaps
quantities in tunneling spectroscopy are taken to be the transverse momentum and total
energy. Based in these approximations, Fermi's golden rule can be used to obtain the
transition probability per unit time between electrodes 1 and 2. The transition probably
differs from the tunneling current by a prefactor, denoted as A in the following
112 =A | TI2N1E )(E)N(E (E+eV)[1-f(E+eV)] dE
1 = 1-2 2-1
=A ITI 2 N(E) N (E+eV)[f(E)-f(E+eV)]dE
The expression, as shown in Equation 3-7 , is intuitive. Ni and N2 are the
densities of electronic states in electrodes 1 and 2, respectively. An electron with energy
E is annihilated in electrode 1, leaving a vacant state. An electron with energy E+eVis
created in electrode 2, filling an empty state and conserving transverse momentum. The
product of these densities of states is multiplied by the tunneling probability 17]2 and the
corresponding Fermi distribution functions, so that the expression is valid at finite
temperature. At zero temperature, dI/dVis simply proportional to the product of the
densities of states. At finite temperature, the tunneling conductance is, to first order,
proportional to the product of the densities of states of the two electrodes, and correction
terms only become important when the thermal energy becomes comparable to the Fermi
energy, which is typically tens of thousands of Kelvin. The form of this product, and in
turn the appearance of the tunneling spectrum, depends on how the densities of states of
the electrodes vary with bias voltage.