Skip to main content

Full text of "Optical Conductivity and Electronic Structure of CeRu4Sb12 under High Pressure"

See other formats


Typeset with jpsj2.cls <ver.l.2> 



Full Paper 



Optical Conductivity and Electronic Structure of CeRu4Sbi2 under High 

Pressure 

Hidekazu Okamura*, Ryosuke KiTAMURA, Masaharu M ATSUN AMI , Hitoshi SuGAWARA, 
Hisatomo Harima, Hideyuki Sato^, Taro MORIWAKI^, Yuka Ikemoto^, and Takao Nanba 

cn" 

I Department of Physics, Graduate School of Science, Kobe University, Kobe 657-8501 

• ^Graduate School of Science, Tokyo Metropolitan University, Tokyo 192-0397 

C . Japan Synchrotron Radiation Research Institute (JASRI) and SPring-8, Sayo 679-5198. 

(Received January 13, 2013) 

m ■ 

Optical conductivity [o'(w)] of Ce-filled skutterudite CeRu4Sbi2 has been measured at 
high pressure to 8 GPa and at low temperature, to probe the pressure evolution of its 
electronic structures. At ambient pressure, a mid-infrared peak at 0.1 eV was formed in 
(7{lo) at low temperature, and the spectral weight below 0.1 eV was strongly suppressed, 
due to a hybridization of the / electron and conduction electron states. With increasing 
external pressure, the mid-infrared peak shifts to higher energy, and the spectral weight 

below the peak was further depleted. The obtained spectral data are analyzed in comparison 

'. . . 

pH ^ with band calculation result and other reported physical properties. It is shown that the 

o 

m 
> 



o 



electronic structure of CeRu4Sbi2 becomes similar to that of a narrow-gap semiconductor 
under external pressure. 

KEYWORDS: Heavy fermion, filled skutterudite, optical conductivity, high pressure 



OO 

^f) [ 1. Introduction 



Physical properties of materials under external pressure have attracted much interest re- 
cently.^) By applying a hydrostatic pressure, one may reduce the interatomic distance in a 
material, and therefore can tune its physical properties in a continuous manner. In addition, 
the pressure technique does not introduce any disorder into the crystal lattice, unlike the 
case of chemical alloying. Various novel properties such as a superconductivity,^-* a transi- 
tion/crossover between localized and delocalized states,^) and quantum critical transitions^^ 
have been explored under external pressure. It is interesting to experimentally examine the 
electronic structures associated with these pressure-induced phenomena. Infrared (IR) spec- 
troscopy technique is very useful in this regard, since it can give the optical conductivity [cr(a;)] 
of a sample loaded in a pressure generating cell. cr{uj) contains much information about the 
microscopic electronic structures near the Fermi level. Note that the other common spectro- 
scopic techniques such as photoemission and tunneling spectroscopies are technically difficult 



*E-mail: okamura(a)kobe-u. ac.jp 

^Present address: UVSOR Facility, Institute for Molecular Science, Okazaki 444-8585 



1/20 



J. Phys. Soc. Jpn. Full Paper 

to perform with a pressure cell. In this work, we apply the high pressure IR technique to 
CeRu4Sbi2. 

CcRu4Sbi2 is one of the compounds with the "filled skutterudite" crystal structure, which 
have attracted a great amount of attention due to their interesting physical properties.'''^ In 
the case of Ce- filled skutterudites CeM4Xi2, the reported physical properties are widely varied 
depending on the constituent atoms M and X. For X=P and As, the compounds studied so 
far (M=Fc, Ru, and Os) all show semiconductor-like properties with the estimated energy 
gaps ranging from 5-10 meV for As compounds®' to 40-130 mcV for P compounds. ^"^''^ 
For X=Sb, in contrast, physical properties for the three compounds with M=Fe, Ru, and 
Os show more metallic characteristics. In particular, CeRu4Sbi2 shows some anomalous and 
interesting properties. The electrical resistivity (p) of CeRu4Sbi2 rapidly decreases at 
low temperature (T) below 100 K, and its electronic specific heat coefficient is moderately 
enhanced, 7 80 mJ/K^mol. These properties are characteristic of intermediate valence Ce 
compounds. In addition, pronounced non-Fermi liquid properties were observed below 4 K.^^^ 
Under external pressure, p of CeRu4Sbi2 below 100 K was found to markedly increase. At 
8 GPa, the p{T) data suggested an energy gap of a few meV,"*^*^' and it showed further 
increases at 10 GPa."'^^-' The optical conductivity (j{u}) of CeRu4Sbi2 in the IR range has also 
been studied at ambient pressure. ""^^"^^^ ^{^) showed a strong suppression below 0.1 eV at low 
j'^ie, 17) ^j^}^ ^}^g ig^[\ Qf narrow Drude response due to heavy electron state. "^^^ 

In this work, in order to probe the pressure evolution of electronic structures in CeRu4Sbi2, 
wc have derived its cr{u)) at pressures up to 8 GPa. The infrared data arc analyzed in com- 
parison with the band calculation result and other published physical properties. It is shown 
that the electronic structures of CeRu4Sbi2 indeed becomes similar to that of a semiconductor 
under high pressure. 

2. Experimental 

The samples of CeRu4Sbi2 used in this work were single crystals grown with self-flux 
method. cr{uj) spectra in vacuum were obtained from the measured optical reflectance spec- 
tra [i?(ci;)]. R{uj) in vacuum were measured under near-normal incidence configuration between 
7 meV and 30 eV, as previously discussed in detail. ^^'^^^ The Kramers-Kronig (KK) analysis 
was used to derive cr{co) from the measured R{ijo) data, where the low-energy range below 
the measurement limit was extrapolated with a Hagen-Rubens function.^^'^^) For high pres- 
sure measurements, an external pressure was applied with a diamond anvil cell (DAC), as 
previously described.^' In a DAC, a mechanically polished surface of a sample with approx- 
imately 200 X 200 X 30 /xm^ dimensions was closely attached to the culet face of a diamond 
anvil (See also Fig. 8(a) in the Appendix.) The diameter of the culet face was 800 fim. The 
diamond anvils were of type Ila with low density of impurities, which is crucial for infrared 
studies. The pressure transmitting medium used was glycerin, which has been shown to 



2/20 



J. Phys. Soc. Jpn. Full Paper 

have good characteristics as a pressure transmitting medium. ^"^^ The R{oj) of the sample was 
measured relative to a gold film mounted together with the sample in DAC [see also Fig. 8(a)]. 
To measure R{uj) of a small sample under such a restricted condition, synchrotron radiation 
(SR) was used as a bright IR source at the beam line BL43IR of SPring-8.^^) For high pres- 
sure data, Drude-Lorentz fitting analyses were used to derive (t{uj) from measured R{uj), as 
discussed later. More technical details of the high pressure R{u!) measurement with DAC and 
IR SR can be found elsewhere. 

3. Results and Discussion 

For technical reasons, high pressure measurements of R{uj) with DAC were done with 
mechanically polished surfaces, rather than with as-grown surfaces. In the course of this study, 
we have realized that R{ui) measured on a polished surface of CeRu4Sbi2 exhibit noticeable 
differences from previous data measured on as-grown surfaces. ^^'^^^ Therefore, we will first 
describe R{oo) and (7(0;) data measured on both polished and as-grown surfaces, then will 
discuss the results of high pressure experiment. 

3.1 R{uj) and (t{oj) measured in vacuum with polished and as-grown surfaces 

Figures 1(a) and 1(b) compare the R{uj) and (7(0;) spectra of CeRu4Sbi2 measured on 
mechanically polished and as-grown surfaces, respectively, in vacuum. The cr(a;) spectra were 
derived from the measured R{ijo) using the KK analysis. The strong T dependences of R{uj) 
and o-{uj) have been already discussed in detail for as-grown cases. -"^^'^^^ They result from the 
formation of a conduction (c)-/ electron hybridized state. Here, important features are the 
formation of a mid-IR peak in a{ui) at 0.1 eV and the depletion of (t{uj) below the mid-IR 
peak energy. The latter feature makes cr(a;) of CeRu4Sbi2 appear like that of an insulator. 
However, CeRu4Sbi2 shows metallic characteristics in its transport and magnetic properties 
at low T. There is a narrow, (5-function-like Drude component due to heavy-mass carriers 
below the measurement range of this study. -^^^ Such a it{uj) spectrum consisting of a narrow 
Drude component, a depletion of spectral weight (pseudogap), and a mid-IR peak have been 
observed for many intermediate- valence Ce and Yb compounds. ^^'^^"^^^ It has been shown 
that the mid-IR peak energies observed for various Ce compounds are roughly proportional to 
their c-f hybridization energies estimated with specific heat data and the single-ion Anderson 
model.-*^^'^^'^^) The spectral shapes of our a{uj) data for the as-grown case in Fig. 1(a) agree 
well with those reported by Dordevic et al.}^'^^^ 

The spectra of the polished surface in Fig. 1(b) are basically similar to those of the as- 
grown one in Fig. 1(a), but there are some noticeable differences. Most significant is that the 
optical phonon peaks for the as-grown case, seen below 40 meV, are much sharper and better 
resolved than the polished case. In the polished case, there is a broad band, indicated by 
the gray area in Fig. 1(b), superimposed on the narrower phonon lines. The peak positions 



3/20 



J. Phys. Soc. Jpn. 



Full Paper 



1.0 



0.9 



0.8 



G 

O 2 







(a) CeRu4Sb., 
^s grown 




295 K 

160 K 




(b) CeRu4Sbi 
Polished 




0.05 0.10 0.15 0.05 0.10 0.15 0.20 
Photon Energy (eV) 



Fig. 1. (Color online) Reflectance (R) and optical conductivity (a) of CeRu4Sbi2 in vacuum measured 
on (a) an as-grown surface and (b) a mechanically polished surface. The gray areas in (b) indicate 
the broadening discussed in the text. The broken curves show the extrapolated portion of the 
spectra, and the vertical dotted lines indicate the position of the mid-IR peak at 8 K. The data 
in (a) are the same as those previously reported. 



of the narrow phonon lines, however, have only small differences (less than 2 cm^^) between 
the two cases. The most likely origin for the broad band is a slight disorder in the crystal 
lattice induced by polishing. The disorder may have relaxed the momentum conservation 
rule for a phonon creation by a photon, causing the phonon band in Fig. 1(b) due to many 
phonon modes that are forbidden without disorder. Nevertheless, the cr{u}) spectra at low T 
of polished case well preserves the main features in cr{uj) of the as-grown sample, namely the 
appearance of the mid-IR peak with the spectral depletion with cooling. Therefore the data 
obtained from a polished sample should be valid for the discussion of electronic structures in 
CeRu4Sbi2 under pressure. All the results presented hereafter are based on polished samples. 

3.2 R{u}) measured under high pressure with DAC 

In a DAC, R{(jj) is measured at a sample/diamond interface, rather than the usual sam- 
ple/vacuum interface. It should be noted that the refractive index of diamond, 2.4, is much 
larger than that of vacuum. According to Fresnel's formula, R{uo) at an interface between a 
material and a transparent medium is expressed as:^^'^^) 

(n - npf + e 

^^"^^ - (n + no)2 + fc2 (1) 



4/20 



J. Phys. Soc. Jpn. 



Full Paper 





01 0.8 ■ 



Measured 
0.6 - (3) in vacuum .. 



80 K ■■ 
8 K 




Expected Measured 
(b) inDAC - (C) inDAC 







0.05 



0.10 



0.05 



0.10 



0.05 



0.10 



Photon Energy (eV) 



Fig. 2. (Color online) (a) Low energy Ro{uj) spectra measured at low T normalized by that at 295 K. 
(b) i?d('^) expected from the Ro{uj) spectra in (a), calculated using Eq. (1), rto=2.4 of diamond, 
and KK analysis as discussed in the text, (c) i?d(w) spectra actually measured in DAC at a low 
pressure of 0.2 GPa. Note that the low-energy limit of measurement is different between (b) and 



Here, n and k are the real and imaginary parts of the complex refractive index of the material, 
respectively, and no is the (real) refractive index of the medium, no=2.4 for diamond and 1.0 
for vacuum. Hereafter, we denote R{i^) relative to vacuum as Ro{uj), and that relative to 
diamond as Rd{uj). Figures 2(a)-2(c) compare measured Ro{uj), -Rd(^) expected from Rq{uj), 
and -Rd(w) actually measured in DAC at a low pressure of 0.2 GPa, respectively. To calculate 
Rdioj) of Fig. 2(b), n{uj) and k{uj) were first derived from the KK analysis of a Ro{uj) spectrum, 
and were substituted into Eq. (1) with no=2.4 to obtain R^ico). This was repeated for Ro{u) 
spectra measured at different T's. To highlight their T dependences, the spectra in Figs. 2(a)- 
2(c) have been normalized by those at 295 K. It is clear that the corresponding spectra in 
Figs. 2(b) and 2(c) agree with each other reasonably well. 

Figure 3 shows -Rd(w) spectra measured at external pressure of 0, 4, and 8 GPa. In the 
present study, Rdi^^) was measured with varying T at fixed pressures of 4 and 8 GPa, rather 
than with varying pressure at fixed T, since it was technically difficult with our gas-driven 
DAC to vary the pressure up to 8 GPa at low T. Rdi^jj) spectra at GPa in Fig. 3(a) are 
derived from Ro{u}) via KK analysis, as already discussed for the spectra in Figs. 2(a) and 2(b). 
The spectra at 4 and 8 GPa in Figs. 3(b) and 3(c) were obtained as follows. The reflectance 
spectra above 70 meV were actually measured in DAC. Those below 70 meV, i.e., the far 
IR range, were obtained by multiplying the spectra at GPa in Fig. 3(a) by the relative 
changes of i?d(^) with pressure and temperature measured in DAC. This procedure for the 
far IR range was taken because it was technically difficult to accurately determine the absolute 
value of the refiectance in DAC, due to diffraction effects of long wavelength far-IR range. 
The uncertainty in the magnitude of the overall i?d(w) under high pressure is estimated to 



(c). 



5/20 



J. Phys. Soc. Jpn. 



Full Paper 



1.0 




I I I 1 I I I I I 

0.05 0.10 0.15 0.20 

Photon Energy (eV) 

Fig. 3. (Color online) i?d('^) spectra of CeRu4Sbi2 at various pressures and temperatures. The spec- 
tra at GPa in (a) were derived via KK analysis of Rq{uj) as discussed in conjunction with Fig. 2, 
and those at 4 and 8 GPa in (b) and (c) were actually measured in DAG. 

be 2 %. Figure 3 clearly shows that, with increasing pressure, the depletion of Rdi^) with 
cooling becomes larger, and occurs over a wider photon energy. 

3.3 (j{uj) derived from measured i?d('^) o-i high pressures 

To obtain the optical conductivity cr{uj) from the Rd{^^) spectra measured in DAC, we 
have used Drude-Lorentz (DL) spectral fitting of i?d(w) rather than the KK analysis. The DL 
fitting was used rather than the KK analysis because the present study with DAC is performed 
only over a limited energy range (below 1.1 eV), although a KK analysis generally requires a 
wider spectral range. In addition, it is difficult with KK analysis to take into account multiple 
reflections within a thin layer of pressure medium, that was present between the sample and 
diamond. All the details of the DL fitting procedure and the analysis of multiple reflections 
are described in the Appendix, with a list of the obtained fitting parameters and examples of 
actual fittings. Here we present the obtained cr{u}) spectra only. 



6/20 



J. Phys. Soc. Jpn. 



Full Paper 




0.1 0.2 0.3 

Photon Energy (eV) 



Fig. 4. (Color online) (a)-(c) show optical conductivity spectra (a) at external pressures of 0, 4, and 
8 GPa, respectively. The vertical lines indicate the shift of mid-IR peak with pressure. The broken 
curves in (a) indicate the extrapolated portion of the spectra, and those in (b) and (c) indicate 
the range below the low-energy limit of the measurement. 



Figure 4 shows the obtained cr{uj) spectra of CeRu4Sbi2. The spectra at GPa in (a) 
were obtained from Ro{uj) with K-K analysis, and those at 4 and 8 GPa in (b) and (c) were 
obtained from measured Rdi^^) with the DL fitting. From the data, the T evolution of cr(u) 
at high pressure appears qualitatively similar to those at zero pressure. Namely, a mid-IR 
peak develops with cooling, and the spectral weight at the lower-energy side of the peak is 
progressively depleted with cooling. However, under pressure, the development of the mid-IR 
peak starts at higher T. For example, at GPa there is no mid-IR peak at 200 K yet, but 
at 8 GPa, it is already observed at 200 K. In addition, the mid-IR peak apparently shifts to 
higher energy with pressure, as indicated by the vertical broken lines in Fig. 4, from about 
0.1 eV at GPa to 0.13 eV at 8 GPa. Figure 5 plots the mid-IR peak energies versus T, 
which confirms the above observations. Furthermore, Fig. 4 shows that the spectral depletion 
below the mid-IR peak energy becomes much more pronounced at higher pressure. Combined 



7/20 



J. Phys. Soc. Jpn. 



Full Paper 



140 

> 

E^120 

^100 
LU 

CD 

^ 80 
1 60 



' 1 ' 1 














^^ \ 8GPa 


OGPa 




1 1 


4 GPa 

I.I.I. 



50 100 150 200 250 300 
Temperature (K) 



Fig. 5. (Color online) Mid-IR peak position in a{uj) as a function of T. Here, the peak position was 
determined as the approximate center position of the peak, rather than the center of a particular 
oscillator used in the fitting (see Appendix). The error bars indicate estimated uncertainty arising 
from the fitting. 



with our previous finding that the mid-IR peak energy is scaled with the c-f hybridization 
energy,^®^ the present result indicates that the c-f hybridization energy actually increases with 
external pressure. Although this property in Ce compounds has been widely assumed on the 
basis of transport and magnetic properties measured under external pressure, this is the first, 
direct spectroscopic evidence for the property. 

In addition to the mid-IR peak, a shoulder at 25 meV has been observed in (t{uj) of 
CeRu4Sbi2.^^^ Namely, cr{Lj) below 25 meV decreased rapidly with cooling below 80 K, in 
addition to the overall decrease of (t{uj) below the mid-IR peak. This shoulder is also seen 
in Fig. 1(a), although it is not clear due to the overlapping phonon peaks. Similar shoulder 
has been observed in cr{u}) of many other Ce and Yb compounds,^*^'^*^'^^^ and its relation to 
c-f hybridization state has been discussed. Although the pressure evolution of this shoulder 
would be quite interesting, its energy (25 meV) was close to the low-energy limit of our study 
(20 meV), and it was difficult to follow the pressure evolution. 

3.4 Analysis of the absorption edge 

The cr(a;) spectrum at 40 K and 8 GPa shows a strong depletion of spectral weight below 
the mid-IR peak energy. It is indeed very similar to that of a narrow-gap semiconductor having 
a small energy gap and a low density of free carriers. However, unlike the case of an ideal 
band semiconductor,^^' '^^^ (^{^) in Fig- 4 does not show a clear onset, and the energy gap 
magnitude is unclear. The lack of a clear onset may be partly due to the nature of Lorentz 
function itself, since, by definition, it goes through the origin. Therefore, to examine the onset 
of absorption, here we employ a technique originally used for indirect-gap semiconductors such 



8/20 



J. Phys. Soc. Jpn. 



Full Paper 




Photon Energy (eV) 

Fig. 6. (Color online) (a) Square root of the absorption coefficient (a) and (b) the optical conductivity 
(cr) of CeRu4Sbi2 at 40 K and external pressures of 0, 4, and 8 GPa. The broken lines are guide to 
the eye indicating the linear dependence discussed in the text. The upturns due to Drude response 
are not shown here for clarity. 

as Ge,^^'^^^ and also for Kondo semiconductors.^^' Near the fundamental absorption edge 
at an indirect energy gap, it has been shown that the square root of the absorption coefficient 
is proportional to the photon energy relative to energy gap, namely \J a{uS) oc (hio — Eg)?"^^ 
This dependence is due to indirect (phonon-assisted) excitation of electrons by photons. 
Figure 6(a) plots \J aiiS) of CeRu4Sbi2 at 40 K and different pressures, and Figure 6(b) plots 
the corresponding oiuS) for comparison. Here, a(a;) at GPa was obtained from the KK 
analysis of R^iuS)^ while those at 4 and 8 GPa were derived from the results of DL fitting. 
First, note that at 40 K and GPa, \J aiuS) shows a linear energy dependence below 80 meV 
as indicated by the broken line. As mentioned above, this linear dependence may suggest light 
absorption due to indirect process. The intercept of the plot with the energy axis has a large 
negative value, which should correspond to the fact that CeRu4Sbi2 at GPa is a metal, 
without an energy gap. (Since the original model for indirect semiconductors'^^' assumed a 
positive intercept with the presence of a gap, the magnitude of this negative intercept should 
not have a physical significance.) Next, consider the high pressure data at 4 and 8 GPa. 
Since the nature of Lorentz function leads to a linear dependence in a \J aiuS) plot, the linear 
dependences alone cannot prove the presence of indirect absorption. However, since such a 



9/20 



J. Phys. Soc. Jpn. Full Paper 

behavior is already observed at GPa without a fitting, it is quite reasonable to interpret 
the data as showing pressure-induced shifts of the linear portion already observed at GPa. 
Then, note that the intercept of the linear portion (dotted line) is near the origin for both 
4 and 8 GPa. This result may indicate that the overlap between the valence band (VB) and 
conduction band (CB) decreases with pressure, and that CeRu4Sbi2 is close to a crossover 
to a semiconductor. But since the intercept is still near zero energy, there is no clear energy 
gap yet on the basis of the optical spectra. This appears consistent with the reported result 
of electrical resistivity. Namely, a thermally activated dependence was observed at 8 GPa, 
but only with a small activation energy (~ 2 meV) and only over a narrow T range (20-40 K). 
Both the very small activation energy and its narrow T range indicate that the energy gap has 
not fully opened yet. A pressure higher than 10 GPa may be needed to open a clear energy 
gap in CeRu4Sbi2. The cr{uj) spectra in Fig. 6(b) also show a nearly linear dependence below 
the mid-IR peak energy, as indicated by the broken lines. The crossing of the linear portion 
with horizontal axis is located near the origin at GPa, but it shifts to about 20 nicV at 4 
and 8 GPa. This result also suggests pressure shifts of absorption edge, although the linear 
dependence does not distinguish between indirect and direct gaps. 

In contrast to the clear shift of mid-IR peak with pressure from 4 to 8 GPa in Fig. 4, 
the absorption edge seen in Fig. 6 shows almost no shift over the same pressure range. This 
results from a broadening of the mid-IR peak with pressure from 4 to 8 GPa. Although the 
significance of this result is unclear, the broadening probably corresponds to an increase of 
the bandwidth with pressure for the hybridized states near Ep. 

3.5 Electronic structures of CeRuiShi-2 with and without external pressure 

At ambient pressure, CeRu4Sbi2 shows typical properties of Ce-based intermediate- valence 
metals, with a rapid decrease of resistivity below 80 K and an electronic specific heat coefficient 
of about 80 mJ/K^mol.^^'-'^^) A Hall effect study showed that the majority carriers were 
the holes,^^) and a clear de Haas-van Alphen (dHvA) signal showed a Fermi surface with a 
cyclotron mass of about 5 vtiq?^^ These works have suggested that CeRu4Sbi2 should be a 
semimetal with at least one hole pocket in the Fermi surface. 

Figure 7(a) shows the band structure of CeRu4Sbi2 calculated with local density ap- 
proximation (LDA) and full-potential linearized augmented plane wave (FLAPW) method.^^^ 
Only the vicinity of Fermi level {E-p) is shown, along the P-F-H line in the Brillouin zone. 
According to the calculation, E-p [the broken line in Fig. 7(a)] does not cross any band, and 
hence CeRu4Sbi2 should be an insulator with an energy gap of about 0.1 eV in the total 
DOS, between the top of VB at F point and the bottom of CB at H point. In the CB, three 
flat bands are seen around P and H points, which originate from the Ce 4/ J=5/2 levels. 
The dome-shaped portion around F of the top-most band is due to hybridized Sb p state. 
The VB is mainly derived from the Sb p, but the top of VB at F has a strong mixing of Ce 



10/20 



J. Phys. Soc. Jpn. 



Full Paper 



> 

CD 
C 
LU 



0.2- 




(b) 









H 



H 



Fig. 7. (Color online) (a) Band structure of CeRu4Sbi2 in the vicinity of Fermi level {Ep, dotted line) 
and along the P-P-H line at ambient pressure, calculated with LDA and FLAPW. (b) A band 
structure suggested on the basis of the calculated result of (a), the present optical data, and the 
reported metallic characteristics. In both (a) and (b), the vertical (red) and oblique (blue) arrows 
indicate examples of direct and indirect optical excitations, respectively, which may occur in these 
band structures and may contribute to the mid-IR peak in cr{uj). The short vertical (green) arrow 
in (b) shows a "dark" transition discussed in the text. 



4/ state. In this situation, the energy gap of 0.1 eV in the total DOS should lead to a clear 
gap in (7{uj) with an onset at 0.1 eV, rather than the mid-IR peak at 0.1 eV as observed. 
[In the case of CeX4Pi2 {X=Fe, Ru, Os),^^^ an energy gap with an onset at 0.2-0.3 eV was 
actually observed ] To explain the above result and the metallic characteristics observed in 
other physical properties, we suggest that the actual band structure should be similar to that 
described in Fig. 7(b). In this suggested band structure, the flat /-derived CB states have 
been moved closer to the VB states. Note that, although Fig. 7(b) is drawn as compensated 
semimetal, it is only intended as an illustration, assuming a stoichiometric composition. In 
experiments, only one Fermi pocket in the dHvA data^^^ and a positive Hall coefiicient^^^ 
have been observed. These results may indicate a single hole pocket in an uncompensated 
semimetal. This may possibly result from a slight deviation of Ce filling from 100 %. But it 
has also been discussed that the electron pocket at H point, even if present, may be very small 
and may have a very large effective mass, resulting in the lack of corresponding dHvA signal 
and a positive Hall coefficient.^^' 

In Figs. 7(a) and 7(b), the vertical (red) and oblique (blue) arrows indicate the direct 
(momentum conserving) and indirect (phonon-assisted) optical transitions, respectively, which 
may occur in this band structure and may contribute to the mid-IR peak in a{u}). In particular, 
the indirect ones at low energy range suggested by the data in Fig. 6(a) may have actually 
resulted from such transitions as in Fig. 7(b). In this case, however, direct transitions from the 
top of VB to the low- lying, /-derived CB around F point, shown by the short vertical (green) 



11/20 



J. Phys. Soc. Jpn. Full Paper 

arrow in Fig. 7(b), do not seem to have contributed to the observed a{Lo). This is because, 
if they strongly contribute, cr{uj) would have much stronger spectral weight below 0.1 eV. 
One possible reason for such "dark" transitions is that the states near the top of VB should 
have strong mixture of 4/ component, and the transitions such as that indicated by the short 
vertical (green) arrow in Fig. 7(b) have a strong /-/ character: The Bloch function of a given 
state should be a superposition of 4/-derived component, |/), and other components, |else). 
Then the optical transition matrix element should consist of the terms {f\d\f), (else else), 
and {e\se\d\f), where d is a dipole operator. Since the /-/ term is zero due to the parity 
selection rulc,'^^'''^'^' and since the transition shown by the short vertical (green) arrow in (b) 
should have a large fraction of such /-/ transition, the transition probability may be also small. 
An effective band model based on a tight binding approximation^^' was also employed to 
calculate (t(uj) of CeRu4Sbi2 including electron correlation effects. By introducing a dispersion 
(broadening) of 4/ electron level, a compensated semimetallic band structure was obtained.^'') 
This model did not include the 4/ degeneracy, so that the obtained band structure did not 
have the low- lying, flat CB around F of Fig. 7(a). This model nevertheless reproduced the 
observed cr{Ljj) qualitatively well, which may also suggest that the 4/-derived flat bands near 
Ep do not strongly contribute to a{uj). 

Under pressure, it is theoretically expected in a Ce compound that the / electron levels 
are shifted upward relative to Ep. The CB states shown in Fig. 7 are under strong influence 
of hybridization between Ce 4/ and Sb 5p states, and therefore it is expected that the CB 
states arc also pushed upward by the pressure shift of 4/ state. This may also reduce the 
overlap between VB and CB. The experimentally observed semiconductor-like characteristics 
of CeRu4Sbi2 under pressure, namely the increase of resistivity, is most likely the result of 
this / electron energy shift. This interpretation is also consistent with the results of Fig. 6, 
i.e., an increase in the onset energy of absorption and the upward shift of mid-IR peak in 
a(uj). The upward shift of 4/ levels with pressure would also reduce the density of thermally 
populated carriers at low T, which is consistent with the observed increase of resistivity under 
pressure. If the pressure is further increased, the 4/ level shift would eventually lead to the 
opening of a true energy gap in the total DOS, similar to the situation in Fig. 7(a). Note that 
in CeM4Pi2 and CeM4Asi2, whose lattice constants are smaller than that of CeM4Sbi2, an 
energy gap is in fact observed.^"-'^^'^^) 

4. Conclusion 

The optical conductivity cr{oj) of Ce-filled skutterudite compound CeRu4Sbi2 was mea- 
sured at low temperatures under high pressure up to 8 GPa. With increasing pressure, the 
characteristic mid-IR peak in cr(a;) shifted toward higher energy. In addition, the depletion 
of o-{uj) below the mid-IR peak energy became more significant and well-developed. These 
results were discussed in comparison with other experimental results and band calculations. 



12/20 



J. Phys. Soc. Jpn. Full Paper 

The observed evolution of (t{uj) with pressure indicates that the energy separation between the 
4f-derived states above and below the Fermi level increases with pressure. The cj(i^) spectrum 
at 8 GPa is indeed similar to that of a narrow-gap semiconductor with a small density of 
residual carriers, which is consistent with the semiconductor-like behavior of resistivity with 
a small activation energy over a narrow T range. However, on the basis of our optical data, an 
energy gap is not fully open yet at 8 GPa. A pressure greater than 10 GPa may be therefore 
needed to observe a clear gap opening. Such a higher pressure study is desired in future to 
further probe the electronic structures of CeRu4Sbi2 under pressure. 

5. Acknowledment 

H. O. would like to thank Dr. H. Yamawaki for providing the gasket material used in the 
DAC. This work has been supported by the following Grants-In-Aid for Scientific Research 
from Ministry of Culture, Education, Science, Sports and Technology of Japan: Innovative 
Area "Heavy Electron" (21102512-AOl) and Scientific Research B (17340096). Experiments 
at SPring-8 were performed under the approval by JASRI (2009B0089, 2009A0089, 2008B1070, 
2008A1239). 

6. Appendix 

This Appendix describes the Drudc-Lorentz fitting procedures used for deriving the opti- 
cal conductivity cr{Ljj) from the i?d('^) spectra measured in DAC. Figure 8(a) illustrates the 
experimental condition in the present IR study. The fitting procedure takes into account ef- 
fects of diamond refractive index and a thin layer of pressure medium which is present between 
the diamond and sample. These are described in detail below. 

6. 1 Drude-Lorentz spectral fitting 

The Lorentz oscillator (classical damped harmonic oscillator) model describes the dynam- 
ical response of bound electrons with a natural (resonance) frequency of to an electromag- 
netic wave of frequency oj. According to this model, the real (ei) and imaginary (€2) parts of 
the complex dielectric function are^^'^^^ 

°°+^(a;2_^2)2+^2^2' 

2 7^ /ON 

Here, ojp is the plasma frequency, 7 is the damping (scattering) frequency, and Coo is a constant 
representing the polarizability of higher energy electronic transitions. The Drude model for 
free carriers can be obtained by setting wo=0. Complex refractive indices n and k can be 
calculated from the relations ei = — k"^ and £2 = 2nk, and Rd{oj) is calculated with Eq. (1) 
and no=2.4 for diamond. Then (t{uj) is obtained as (t{(jj) = a;e2(w)/(47r). In the fitting, the 
parameters Up, coq and 7 were varied to minimize the difference between the calculated and 



13/20 



J. Phys. Soc. Jpn. 

measured (lo) spectra on the basis of least-squares fitting scheme. 



Full Paper 



6.2 Effects of interference due to a thin gap between the sample and diamond 

In our experimental condition, as shown in Fig. 8(a), a very thin layer of pressure medium 
was present between the diamond face and the sample surface. This occurred since the surface 
of the sample was not completely flat while the diamond surface was almost perfectly flat.^^^ 
With such a layer, as shown by red arrows in Fig. 8(a), multiple reflections of light and 
their interference are expected. In fact, the visible image of a sample loaded into DAC with 
pressure medium often showed interference fringes (Newton's rings). This did not seem to 
affect the measured i?d('J-') in the far IR range, as demonstrated by the good agreement between 
the spectra in Fig. 2(b) and 2(c). In the mid-IR and ncar-IR ranges where the wavelength 
was shorter, however, the "expected" Rd{uj), calculated from the measured Rq{uj), did show 
noticeable difference from the measured i?(j(a;), as shown in Fig. 8(b). This shows that the 
thickness of the medium layer was not negligible compared with the wavelength of light in the 
mid-IR range. Note that the gold film used as the reference of rcfiectance did not have such a 
layer, since it was directly pressed onto the diamond by the gasket, as sketched in Fig. 8(a). 
The refiectance under this situation has been theoretically analyzed in detail, *^^^ and may be 
expressed, omitting the uj dependence, as^^^ 

_ (no - n'){n' + h)e'^ + (np + n'){n' - h)e-'^ ^ 
(no + n'){n' -\- n)e^^ + (no — n'){n' — n)e~'^^ ' 

where 

5 = ^n't. (5) 
A 

Here, no=2.4 is the refractive index of diamond, n' and t are the refractive index and layer 
thickness of the medium, respectively, n = n + ik is the complex refractive index of the 
sample, and A is the wavelength of light in vacuum. For simplicity, we assume that n' is a 
real constant (no absorption and dispersion in the medium), and that the layer thickness is 
constant, although in reality it may vary over the sample area. Then using n{u) and k{uj) 
of the sample obtained with the KK analysis of Ro{oj), we calculated Rd{(^) with various 
values of n't. In Fig. 9, Rd{(^) spectra calculated with n't=0.22 (iii) and 0.42 /xm (iv) are 
shown, which agree well with the measured spectra (v) and (vi), respectively. The magnitude 
of the obtained n't is quite reasonable since the visible image of the sample under microscope 
exhibited no or only one interference fringe when these spectra were measured. This means 
that 2n't should be smaller than or of the same order with the visible wavelength, which 
is consistent with the obtained n't values. Note also that both the measured and calculated 
spectra indicate that the interference effect becomes small below about 0.2 eV. 

For each high pressure run, Rd{oj) was measured in DAC before applying high pressure, 
and the obtained value of n't was used to analyze the subsequently obtained high pressure 



14/20 



J. Phys. Soc. Jpn. 



Full Paper 




Fig. 8. (Color online) (a) Schematic drawing (not in scale) for the experimental condition of om" high 
pressure IR study with a DAC. (b) Analysis of interference effects on the reflectance spectra, (i): 
i?o(w) measured in vacuum at 295 K (without using a DAC). (ii): i?d(i^) expected from Rq{uj). 
i?d(aj) was calculated with KK analysis of Ro{uj), Eq. (1), and nQ—2A for diamond, (iii) and (iv) 
(broken curves): Spectra calculated similarly to (ii), but including the interference effects with 
n't=0.22 and 0.42 ^m, respectively, (v) and (vi): Two i?d(w) spectra independently measured 
in the DAC. Note that the measured Rd{^) spectra are not shown over a narrow range around 
0.25 eV, due to strong absorption by diamond. 



data. In reality, n't may also change with pressure, but it was difficult to correctly measure 
or estimate it. Therefore we assumed the initial value of n't for the analysis of high pressure 
data. 

6.3 Actual fitting model and examples of fitted spectra 

Once the value of n't was determined, -Rd(w) was calculated with a set of parameters, 
taking into account no=2.4 through Eq. (1) and the interference effect through Eq. (4). Then 
the parameters were adjusted to minimize the deviation of the calculated -Rd('^) from the 
measured one. The fitting was done with "RefFIT" , a dedicated software for spectral fittings 
developed by Alexey Kuzmenko.^^-* Figure 9 shows examples of actual fitting for (4 GPa, 
40 K) and (8 GPa, 40 K) data. Note that the overall spectral shape of Rd{^j^) is basically 



15/20 



J. Phys. Soc. Jpn. 



Full Paper 



(a) 



1.0 





2.0 



§ 1.5 



1.0 



o 
0.5 



(1) Measured in DAC 

(2) Fit with n't=0.22 \irr\ 

(3) Expected with n't=0 

(4) Expected in vacuum 



CeRu4Sbi2 
4 GPa, 40 K 



Total fit 




0.02 0.05 0.1 0.2 0.5 
Photon Energy (eV) 



(b) 

1.0 

O 

ro 
o 

a 0.5 




2.0 

i 1.5 

a 

^ 1.0 
o 

^ 0.5 





(1) IVIeasured in DAC _ 

(2) Fit with n't=0.22 ^im _ 

(3) Expected with n't=0 . 

(4) Expected in vacuum. 



H 1 — I I I I 



H 1 — M ill 



CeRu4Sb.,2 
8 GPa, 40 K 



Total fit 




Phonon 



0.02 0.05 0.1 0.2 0.5 
Photon Energy (eV) 



Fig. 9. (Color online) Examples of fitting for i?d(<^) data measured in DAC and the resulting optical 
conductivity (cr) at (a) 4 GPa and 40 K and (b) 8 GPa and 40 K. (1) Measured i?d(w) spectrum. 
(2) Fitted spectrum with n't=0.22 /im. (3) Spectrum expected in the absence of a medium layer, 
obtained by setting t=0 in spectrum (2). (4) Spectrum expected in vacuum, obtained by setting 
no=l and t—0 in spectrum (2). The total fit (t(w) is shown both with (thin red curve) and without 
(thick red curve) the LI component used to cover the hump of R{io) at 60 meV, which is an 
instrumental artifact. ''^^ Rdi^) data between 0.236 and 0.285 eV, which could not be measured 
due to strong absorption in diamond, was interpolated by a straight line. 



16/20 



J. Phys. Soc. Jpn. Full Paper 

reproduced by one Drude and three Lorentz (L2, L4 and L5) oseillators. Two additional 
Lorentz oscillators (LI and L3) were used to cover a small hump of Rdi^j) near 60 meV and a 
shoulder of i?d(^) near 0.1 eV (0.17 eV) at 4 GPa (8 GPa), respectively. The hump at 60 meV 
is an instrumental artifact, '^^^ and therefore the (j{u}) and spectra shown in Figs. 4 

and 6 were calculated without LI. The shoulder became stronger at low T, and therefore the 
L3 oscillator was necessary for a satisfactory fit. The obtained fitting parameters are listed 
in Table I. For the Drude component, ojp and 7 were varied to reproduce R(i{oj), and then 
cjopt, namely the dc conductivity given by these parameters, was calculated and compared 
with the experimentally measured cr^c-^^'^^^ These values are also listed in Table I. Note that 
(Topt is higher than cxdc for most of the data, by a factor of up to 2.1. This is not surprising 
since a similar result is already seen in the data at 295 K and GPa, where the measured 
crdc=2600 il^^cm^^ (Ref. 13) is more than twice smaller than that given by the Hagen-Rubens 
extrapolation (Fig. 1). Similar result was also observed previously.^®) The reason for this is 
unclear, but there seems to be some scattering mechanism that affects Udc stronger than the 
high frequency one. 

6.4 Uncertainty in the magnitude of obtained conductivity 

Compared with the case of KK analysis on a reflectance measured in vacuum, there are 
many factors that increase the uncertainty in cr{uj) obtained with the above procedures. The 
largest source of uncertainty is the arbitrariness among the parameters in the DL fitting of 
-^d(^) performed in a limited energy range (below 1.1 eV). The overall uncertainty in the 
magnitude of resulting a{oj) for the high pressure case is estimated to be it 10 %. This is 
indeed much larger than that in the KK-derived cr(a;) in vacuum. However, since we are not 
making any discussion based on the absolute magnitude of cr{uj), such as spectral weight 
transfer or the optical sum rule, the uncertainty should not affect our conclusions. 



17/20 



J. Phys. Soc. Jpn. 



Full Paper 



Table I. Parameters from the Drude-Lorentz fitting of the i?d('^) data, displayed in units of meV 
unless noted. L1-L5 denote five oscillators as indicated in Fig. 9. fco=4.5 was used for all the 
fitting, and luq of L5 was fixed at 1630 meV. (Topt indicates the dc conductivity resulting from the 
fitted Drude component, and a^c indicates the actually measured values.^''' The parameters 
shown have been rounded to three figures, but this does not necessarily mean being accurate to 
three digits. A blank oscillator means that it was not needed to have a good fit. 



4GPa 




Drade 


LI 


L2 


L3 


L4 


L5 


CJ (n'cm"') 


T{K) 


COp 


y 


cOo 




Y 


(Do 


(Bp 


y 


(Do 


(Bp 


y 




COp 


y 


Wo 


(Dp 


y 


^opt 


Ode 


295 


779 


19.8 








79.6 


2030 


355 














1630 


6500 


536 


4100 


2860 


200 


740 


15.5 








62.4 


1280 


152 








247 


1500 


475 


1630 


6310 


650 


4750 


2860 


160 


687 


17.4 








74.4 


1330 


140 








290 


1470 


479 


1630 


6310 


650 


3650 


2740 


120 


599 


13.9 








87.9 


1390 


151 








323 


1310 


425 


1630 


6120 


736 


3437 


2560 


80 


440 


8.95 








106 


1440 


163 


107 


222 


34.2 


355 


1330 


456 


1630 


6120 


735 


2901 


2630 


40 


149 


0.72 


59.2 


68.4 


11.2 


118 


1450 


161 


107 


318 


51.1 


345 


1440 


469 


1630 


6070 


763 


4100 


7140 



8GPa 




Drade 


LI 


L2 


L3 


L4 


L5 


a (n 'cm"') 


r(K) 


COp 


y 


COo 


COp 


y 


Wo 


Wp 


Y 


Wo 


Wp 


y 


Wo 


Wp 


y 


COo 


Wp 


y 




Ode 


295 


788 


24.2 








89.4 


1800 


408 














1630 


5750 


924 


3400 


2380 


200 


711 


22.9 








106 


1220 


165 


162 


186 


55.8 


366 


1290 


484 


1630 


5530 


859 


2980 


2080 


160 


711 


22.9 


50.5 


64.0 


10.4 


105 


1260 


169 


159 


156 


43.4 


373 


1290 


484 


1630 


5530 


859 


2950 


1890 


120 


532 


11.8 


47.9 


82.4 


17.1 


103 


1240 


169 


159 


278 


79.0 


369 


1150 


421 


1630 


5560 


939 


3240 


1490 


80 


359 


5.26 


56.0 


92.5 


15.4 


115 


1380 


181 


158 


250 


54.9 


412 


1040 


397 


1630 


5320 


852 


1900 


970 


40 


156 


2.19 


56.0 


77.0 


13.0 


115 


1330 


163 


153 


254 


57.7 


407 


1240 


496 


1630 


5190 


828 


1500 


910 



18/20 



J. Phys. Soc. Jpn. Full Paper 

References 

1) Sec, for example, the papers in the Proeeedings of the International Conference on High Pressure 
Science and Technology (Tokyo, 2009), J. Phys. Conf. Ser. 215 (2010). 

2) See, for example, H. Kotegawa, H. Sugawara and H. Tou, J. Phys. Soc. Jpn 78 (2009) 013709. 
i) M. Matsunami, H. Okamura, A. Ochiai, and T. Nanba, Phys. Rev. Lett. 103 (2009) 237202. 
i) See, for example, J. L. Sarrao and J. Thompson, J. Phys. Soc. Jpn. 76 (2007) 051013. 
5) See, for example, Proceedings of the International Conference on New Quantum Phenomena in 

Skutterudite and Related Systems, J. Phys. Soc. Jpn. 77 (2008) Suppl. A. 

3) C. Sckine, N. Hoshi, K. Takeda, T. Yoshida, I. Shirotani, K. Matsuhira, M. Wakeshima, and Y. 
Hinatsu: J. Magn. Magn. Mater. 310 (2007) 260. 

7) C. Sckine, R. Abe, K. Takeda, K. Matsuhira and M. Wakeshima: Physica B 403 (2008) 856. 
i) H. Sato, Y. Abe, H. Okada, T. D. Matsuda, K. Abe, H. Sugawara, and Y. Aoki: Phys. Rev. B 62 

(2000) 15125. 

3) I. Shirotani, T. Uchiumi, C. Sekine, M. Hori, S. Kimura, and N. Hayama: J. Solid State Chem. 
142 (1999) 146. 

D) C. Sekine, K. Akita, N. Yanase, I. Shirotani, I. Inagawa, and C. Lee: Jpn. J. Appl. Phys. 40 (2001) 
3326. 

1) N. Takeda and Y. Ishikawa: J. Phys. Soc. Jpn. 69 (2000) 868. 

2) E. D. Bauer, A. Slebarski, R. P. Dickey, E. J. Freeman, C. Sirvent, V. S. Zapf, N. R. Dilley, and 
M. B. Maple: J. Phys.: Condens. Matter 13 (2001) 5183. 

3) M. Kobayashi, S. R. Saha, H. Sugawara, T. Namiki, Y. Aoki, and H. Sato: J. Magn. Magn. Mater. 
329-333 (2003) 605. 

i) N. Kurita, M. Hedo, Y. Uwatoko, M. Kobayashi, H. Sugawara, H. Sato, and N. Mori: Physica B 

272-276 Suppl. 1 (2004) E81. 
5) N. Kurita, H. Hedo, M. Koeda, M. Kobayashi, H. Sato, H. Sugawara, and Y. Uwatoko: Phys. Rev. 
B 79 (2009) 014441. 

3) S. V. Dordevic, D. N. Basov, N. R. Dilley, E. D. Bauer, and M. B. Maple: Phys. Rev. Lett. 86 

(2001) 684. 

7) M. Matsunami, H. Okamura, T. Nanba, H. Sugawara, and H. Sato: J. Magn. Magn. Mater. 272-276 
(2004) e41. 

i) S. V. Dordevic, K. S. D. Beach, N. Takeda, Y. J. Wang, M. B. Maple, and D. N. Basov: Phys. Rev. 
Lett. 96 (2006) 017403. 

3) M. Matsunami, H. Okamura, T. Nanba, H. Sugawara, and H. Sato: J. Phys. Soc. Jpn. 72, 2722 
(2003). 

3) H. Okamura, T. Michizawa, T. Nanba and T. Ebihara: J. Phys. Soc. Jpn. 73 (2004) 2045. 

1) G. Burns: Solid State Physics, Chapter 13 (Adademic Press). 

2) M. Dressel and G. Gruner: Electrodynamics in Solids (Cambridge Press). 

3) H. Okamura, M. Matsunami, R. Kitamura, S. Ishida, A. Ochiai, and T. Nanba: J. Phys.: Conf. 
Ser. 215 (2010) 012051. 

i) N. Tateiwa and Y. Haga: Rev. Sci. Instr. 80 (2009) 123901. 
5) T. Moriwaki and Y. Ikemoto: Infrared Phys. Tech. 51 (2008) 400. 
3) B. C. Webb, A. J. Sievers, and T. Mihalisin: Phys. Rev. Lett. 57 (1986) 1951. 



19/20 



J. Phys. Soc. Jpn. 



Full Paper 



27) J. N. Hancock, T. McKnew, Z. Schlesinger, J. L. Sarrao, and Z. Fisk: Phys. Rev. Lett. 92 (2004) 
186405. 

28) H. Okamura, T. Watanabe, M. Matsunami, T. Nishihara, N. Tsujii, T. Ebihara, H. Sugawara, H. 
Sato, Y. Onuki, Y. Isikawa, T. Takabatake, and T. Nanba: J. Phys. Soc. Jpn. 76 (2007) 023703. 

29) S. Kimura, T. lizuka, and Y.-S. Kwon: J. Phys. Soc. Jpn. 78 (2009) 013710. 

30) S. Kimura, Y. Muro, and T. Takabatake: J. Phys. Soc. Jpn. 80 (2011) 033702. 

31) H. Okamura, T. Michizawa, T. Nanba, S. Kimura, F. Iga, and T. Takabatake: J. Phys. Soc. Jpn. 
74 (2005) 1954. 

32) P. Y. Yu and M. Cardona: Fundamentals of Semiconductors (Springer, Berlin, 2001) 3rd ed.. 
Chapter 6.2. 

33) G. G. MacFarlane, T. P. McLean, J. E. Quarrington and V. Roberts: Phys. Rev. 108 (1957) 1377. 

34) B. Bucher, Z. Schlesinger, P. C. Canfield, and Z. Fisk: Phys. Rev. Lett. 72 (1994) 522. 

35) K. Abe, H. Sato, T. D. Matsuda, T. Namiki, H. Sugawara, and Y. Aoki: J. Phys. Condens. Mat. 
14 (2002) 11757. 

36) H. Sugawara, K. Abe, T. D. Matsuda, Y. Aoki, H. Sato, R. Settai, and Y. Onuki: Physica B 312-313 
(2002) 264. 

37) H. Harima and K. Takagahara: J. Phys.: Condons. Matter 15 (2003) S2081. 

38) M. Matsunami, H. Okamura, K. Scnoo, S. Nagano, C. Sckine, I. Shirotani, H. Sugawara, H. Sato, 
T. Nanba: J. Phys. Soc. Jpn. 77 (2008) Suppl. A, pp. 315. 

39) T. Mutou and T. Saso: J. Phys. Soc. Jpn. 73 (2004) 2900. 

40) T. Saso, J. Phys. Soc. Jpn. 75 (2006) 043705. 

41) It is possible to avoid such a layer of medium between the diamond and sample by using a solid 
pressure medium such as KBr and NaCl. At pressures below 10 GPa, however, liquid pressure 
medium such as glycerin can provide much more hydrostatic pressure than KBr.^^^ In view of the 
importance of having a more hydrostatic pressure for studying electronic striicturcs of strongly 
correlated materials, we chose glycerin over solid medium despite the complications in data analysis 
discussed here. 

42) O. S. Heavens, Optical Properties of Thin Solid Films (Dover, New York, 1991). 

43) A. B. Kuzmenko: RefFIT, available at http://optics.unige.ch/alexey/reffit.html. 

44) This hump of a{u)) results from a small hump of Rd{i^) at 60 meV seen in Figs. 3(b) and 3(c) 
at a few temperatures. To fit this hump, an additional oscillator LI was used as shown in Fig. 9 
(Appendix). The hump in R{u)) is also present in the low pressure (0.2 GPa) DAG data of Fig. 2(c), 
but not in the zero pressure data of Figs 1, 2(a) and 2(b). We conclude that this hump is an 
instrumental artifact created by the spectroscopy apparatus used for the high pressure experiment, 
presumably due to a structure in the spectral response of the beam splitter in the IR interferometer. 



20/20