You are on page 1of 10

Ann. Phys. (Leipzig) 15, No. 7 – 8, 535 – 544 (2006) / DOI 10.1002/andp.

200510198

Verifying the Drude response


Martin Dressel∗ and Marc Scheffler
1. Physikalisches Institut, Universtät Stuttgart, Pfaffenwaldring 57, 70550 Stuttgart, Germany

Received 10 October 2005, accepted 1 November 2005


Published online 26 May 2006

Key words Electronic transport, Drude model, heavy fermions.


PACS 71.10.-w, 71.27.+a, 72.10.-d
In commemoration of Paul Drude (1863–1906)
By now more than 100 years passed since Paul Drude suggested his highly acclaimed model of electronic
transport. In the form advanced by A. Sommerfeld, the Drude model is still heavily utilized to describe, for
instance, the optical properties of metals. Surprisingly, the key prediction of the Drude model, the frequency
dependence of the complex conductivity with its pronounced roll-off was directly observed in an experiment
only very recently. Here we present direct and independent measurements of both the real and imaginary
parts of the metallic conductivity in a large range of frequency around the scattering rate.

c 2006 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim

1 Introduction
At the end of the 19th century Paul Drude performed intensive studies on the optical properties of a large
number of materials. By combining these data with the theory of electrons which was quickly established
after J. J. Thomson’s experiment in 1897, Drude could explain the electrical conductivity, the thermal
conductivity, and the optical properties of metals. The model he proposed in a series of papers published
in 1900 regards metals as a classical gas of electrons executing a diffusive motion [1, 2]. Accordingly
electronic interaction is neglected, and scattering occurs only on obstacles like ions, impurities etc. The
central assumption of the model is the existence of an average relaxation time τ which solely governs the
relaxation of the system to equilibrium, i.e. the state with zero average momentum p = 0, after an external
field E is removed. The rate equation dp/dt = −p/τ expresses the hypothesis that the probability,
between two scattering events elapses the time t, decays exponentially with t; and after the period τ the
fraction 1/e ≈ 37% of the electrons underwent a scattering event at which their momentum is completely
lost. In the presence of an external electric field E, the equation of motion becomes

d p
p = − − eE . (1)
dt τ
Typically only the current density J = −N ep/m is experimentally accessible; with N the density
of charge carriers, m the carrier mass, and −e the electronic charge. For constant fields, the condition
dp/dt = 0 leads to a dc conductivity σdc = J/E = N e2 τ /m. Upon the application of an alternating
field of the form E(t) = E0 exp{−iωt} (harmonic wave), the solution of the equation of motion

d2 r m dr
m + = −eE(t) (2)
dt2 τ dt
∗ Corresponding author E-mail: dressel@pi1.physik.uni-stuttgart.de


c 2006 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
536 M. Dressel and M. Scheffler: Drude response

with p = mdr/dt gives a complex frequency-dependent conductivity [3]

N e2 τ 1
σ̂(ω) = = σ1 (ω) + iσ2 (ω) (3)
m 1 − iωτ
where the components are

N e2 τ 1 N e2 τ ωτ
σ1 (ω) = and σ2 (ω) = . (4)
m 1 + ω2 τ 2 m 1 + ω2 τ 2
There is an average distance travelled by the electrons between two collisions, called the mean free path .
Within the framework of the original Drude model,  = vth τ , where vth is the average thermal velocity
of classical particles.
Although the underlying picture is fundamentally different for electrons obeying quantum statistics, the
equation of motion for the average momentum is the same [4]. If a field E is applied, the Fermi sphere which
in the absence of an electric field is centered around zero momentum [5], is displaced by −eEτ /. Again
with J = −N ek/m and wavevector k = p/ the above Eq. (3) is recovered. However, the scattering
processes which establish the equilibrium in the presence of the electric field involve only electrons near
the Fermi surface; states deep within the Fermi sea are not influenced by the electric field. Consequently
the expression for the mean free path is  = vF τ , with vF the Fermi velocity, and differs dramatically from
the mean free path given by the original Drude model.
Advances in quantum mechanics led to a more rigorous treatment of the transport phenomena, for instance
in terms of the current-current correlation function [3, 6]:

1  ∞
σ̂(q, ω) = dt s|J(q, 0)J∗ (q, t)|s exp{−iωt} , (5)
ω s 0

where q indicates the wavevector of the radiation, the sum considers all states s and J is the operator of the
electrical current density. This so-called Kubo formula yields the Drude response (3) if two approximations
are fulfilled. First, the current-current correlation function exponentially decays in time (the relaxation time
approximation): J(t) = J(0) exp{−t/τ } for t > 0; this means that only a single, energy-independent
relaxation time exists. Second, the considerations were restricted to zero wavevector (the so-called local
limit).
The complex conductivity describes the response of the electrons as an ensemble to an external electric
field, with the attenuation expressed by the real part σ1 (ω) and the phase shift given by the imaginary part
σ2 (ω). For low frequencies ω  1/τ the response can easily follow the alternating field: σ1 (ω) ≈ σdc and
σ2 (ω) ≈ 0. This is the so-called Hagen-Rubens or diffusive regime. If ωτ  1, in the ballistic regime, the
electrons do not follow any more and the optical conductivity vanishes with σ1 (ω) ∝ ω −2 . While these
general considerations hold for any metal, we now consider different experiments concerning verification
of the Drude response.

2 Experimental verification of the Drude response


2.1 Microwave and infrared reflection measurements on metals
For typical metals like gold, silver, or aluminum the time τ between two scattering events is a few picoseconds
at room temperature, corresponding to a scattering rate 1/(2πcτ ) in the order of a few hundred cm−1 . In
this far-infrared range of the spectrum, commonly reflection measurements are conducted to determine the
electrodynamic properties of the material. In Fig. 1 the precision measurement of Bennett and Bennett [7]
is reproduced which has served as a reference for the last forty years. The presence of a plasma edge around
2 × 104 cm−1 , which in the case of gold means a drop of reflectivity not below 40% due to interband
transitions, is certainly an outflow of the Drude model. However, the plasma frequency ωp = 4πN e2 /m


c 2006 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim www.ann-phys.org
Ann. Phys. (Leipzig) 15, No. 7 – 8 (2006) 537

Energy hw (eV)
10- 1 100 101 102
1.0
s1 (10 4 W - 1 cm - 1) Reflectivity R

Au

0.5

0.0 Fig. 1 Frequency dependence of the optical properties of gold at room


10 temperature. The upper panel shows the reflectivity data R(ω) obtained
by [7, 8]. Above the plasma edge (at 2 eV) electronic transitions between
5 different bands dominate which are not included in the Drude model. In the
lower panel the conductivity σ1 (ω) is displayed as obtained by Kramers-
Kronig analysis of the reflectivity. The directly measured dc conductivity is
0 4.9 · 105 (Ωcm)−1 .
10 2 10 3 10 4 10 5 10 6
Frequency n (cm - 1)

yields information only on the carrier density and mass, but not on the characteristic dynamics. Due to the
−1
 300 cm [8]. The well-known Hagen-
high reflectivity close to unity, reliable data are not available below
Rubens behavior of the low-frequency reflectivity R(ω) = 1 − 2ω/πσdc was in fact only experimentally
observed in reflection measurements of so-called bad metals, like stainless steel (Fig. 2); i.e. materials for
which σdc is small, 1/τ is large and consequently the reflectivity deviates well from unity even in this
regime.
In order to obtain the optical conductivity from reflectivity data, the Kramers-Kronig analysis has to
be applied which requires an extrapolation of the data to zero frequency and infinity [3]. There seems
to be no way to have experimental access to the real and imaginary parts of the dynamical response
in this range of frequency. Already Paul Drude was aware of this severe problem and thus developed a
polarization-dependent reflection method, nowadays known as ellipsometry [10,11]; unfortunately reports

Energy hω (meV)
10 0 10 1 10 2
1.00
Stainless steel

0.95

Hagen-Rubens
Reflectivity R

0.90
1.00
Fig. 2 Reflectivity of stainless steel versus fre-
0.85 0.99 quency obtained at T = 300 K. The infrared data
were obtained with a Fourier transform spectrom-
0.98
eter; in the THz range the reflectivity was mea-
0.97 sured using single-frequency radiation sources; at
0.80 2 cm−1 cavity perturbation technique was used. The
0.96
solid line is a fit by the Hagen-Rubens formula with
0.95
0 50 100 150 200
σdc = 1.38 × 104 (Ωcm)−1 (after [9]). The inset
0.75 shows the same data in a linear scale.
10 0 10 1 10 2 10 3
Frequency ν (cm −1)

www.ann-phys.org 
c 2006 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
538 M. Dressel and M. Scheffler: Drude response

Frequency f (THz)
0 0.5 1.0 1.5 2.0

Si
0.12
σ 1 (Ω −1 cm −1)

0.08

0.04

(a)
0.0
(b)
0.06
σ 2 (Ω −1 cm −1)

0.04

Fig. 3 Frequency dependence of the (a) real and (b) imaginary


parts of the optical conductivity of silicon weakly doped by holes
0.02 (1.1 × 1015 cm−3 , open circles) or electrons (4.2 × 1014 cm−3 ,
solid circles). The experiments were performed at T = 300 K in
the time domain, the full lines correspond to σ1 (ω) and σ2 (ω) as
0.0 obtained by the Drude model (after [12]).
0 20 40 60
Frequency ν (cm −1)

on ellipsometric measurements in the far-infrared spectral range are extremely sparse. As a result, there are
no data available which give a real verification of the Drude response in this regard.

2.2 THz time and frequency domain transmission measurements on semiconductors


The dilemma we face is that only in the spectral range of microwaves and low THz frequencies it is possible
to measure both components of the electrodynamic response independently, but for conventional metals 1/τ
is well above these frequencies. The problem can be circumvented by looking at semiconductors which are
moderately doped. Grischkowsky and coworkers [12] investigated the electrodynamics of p and n doped
silicon using THz time domain spectroscopy. For a donor concentration of 4.2 · 1014 cm−3 the plasma
frequency ωp is 360 GHz, for instance. The optical experiments can be performed in transmission to obtain
a higher sensitivity. The THz radiation is created by the sudden discharge between two electrons triggered
by a short laser pulse. With the help of the Laplace transformation, the change in the pulse shape yields the
complex optical response of the sample. The real and imaginary parts of the conductivity of n-Si, displayed
in Fig. 3 can be nicely fitted by a simple Drude model with a scattering rate of 590 GHz. Similar values are
obtained for hole doped Si, and also for GaAs. The scattering rate does not depend strongly on the carrier
density or sign of the carriers; but cooling down to T = 80 K reduces 1/τ by about one order of magnitude.
More recently it was suggested [13] that small deviations from the Drude behavior may be fit with a
Cole-Davidson function σ̂(ω) = (N e2 τ /m)(1 − iωτ )−β which is common to describe disordered systems.
This implies that instead of a simple exponential decay, the initial drop of the time response is faster and
approaches the Drude behavior only asymptotically for large times. Nevertheless, the scattering of the data
increases at frequencies below 200 GHz and above 1.5 THz.


c 2006 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim www.ann-phys.org
Ann. Phys. (Leipzig) 15, No. 7 – 8 (2006) 539

Energy hω (meV)
0 2 4 6

1.20
N S = 15.3 × 10 11 cm −2
0.80 τ = 4.8 × 10 −13 s

0.40
(a)
0
0.60
σ 1 (10 −3 Ω −1)

N S = 8.3 × 10 11 cm −2
τ = 4.5 × 10 −13 s
0.40

0.20
(b)
0
0.24 Fig. 4 Conductivity σ1 (ω) of a two-dimensional sili-
N S = 3.6 × 10 11 cm −2 con inversion layer measured versus frequency at 1.2 K.
τ = 4 × 10 −13 s The open circles are data taken by an infrared spec-
0.16
trometer, the closed circles are from microwave mea-
surements. The lines correspond to the Drude behavior
0.08 predicted from the dc conductivity at the indicated value
(c) of the carrier concentration NS for two dimensions and
Eg
0 of the scattering time τ (after [14]).
0 10 20 30 40 50
Frequency ν (cm −1)

Obviously, any more sophisticated analysis requires high-quality data in a much larger frequency range,
at least two orders of magnitude which is possible only in the microwave regime. There are two possibilities
to tackle the problem of how to shift the scattering rate to much lower frequencies. First, in two-dimensional
electron gases generated at semiconductor interfaces, the mobility and thus the electronic scattering time
can be extremely large at low temperatures. Some experiments in this direction have been conducted [14],
however, the reduced dimensionality seems to become increasingly important. In Fig. 4 the conductivity
σ1 (ω) of a two-dimensional silicon inversion layer is plotted. As the carrier concentration per area NS is
reduced, localization sets in due to disorder; seen by a pseudogap in the optical spectra, indicated by Eg in
Fig. 4c.

3 Broadband microwave spectroscopy on heavy fermions


3.1 Dynamics of correlated electron systems
The approach we have chosen here to shift the scattering rate to lower frequencies utilizes electronic
correlations [15]. The above mentioned models of the charge transport did not take the interaction of
the underlying lattice or electron-electron interaction into account. Broadly speaking these effects can be
summarized assuming an effective mass m∗ and also an effective scattering time [16]. The concept to treat
electronic correlations as a renormalization goes back to L. D. Landau [17] and is know as Fermi-liquid
theory. It is most successfully applied to so-called heavy fermions which are intermetallic compounds whose
electronic structure is determined by the interaction of nearly localized f -shell electrons with quasi-free

www.ann-phys.org 
c 2006 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
540 M. Dressel and M. Scheffler: Drude response

conduction-band electrons. This leads to an effective mass of the quasiparticle which is orders of magnitude
larger than the bare electron mass [18, 19]. Accordingly, the spectral weight [3]

π N e2
σ1 (ω)dω = (6)
2 m∗

and the scattering rate 1/τ ∗ = (m/m∗ )(1/τ ) are significantly reduced [20, 21]. With other words, the
charge carriers are extremely slow due to the electron-electron interaction. A number of heavy-fermion
compounds seem to confirm this picture since they indicate scattering rates shifted well into the GHz
range [22]. However, the analysis is in general heavily based on the Hagen-Rubens extrapolation and the
dc conductivity. In no case spectral data were obtained in the actual range of the reduced scattering rate.
Thus the actual low-frequency behavior of the complex conductivity remains an open question.
Electron-electron interaction leads to an enhanced thermodynamic mass, as observed in the electronic
contribution to the specific heat or the magnetic susceptibility, but it also adds a T 2 contribution to the
resistivity. More general, Fermi-liquid theory predicts a renormalized frequency-dependent scattering rate
[3, 16, 23]:

1/τ ∗ (ω, T ) = A(kB T )2 + B(ω)2 (7)

with the prefactors increasing as the square of the effective mass [24, 25], and A/B = (2π)2 . Although
various attempts have been made over the years to extract a frequency-dependent scattering rate [26–29],
the results were never really convincing due to the limited range of frequency and quality of data.
In order to address the question of the Drude roll-off of correlated electron systems, we have investigated
the electrodynamic properties of the heavy-fermion compound UPd2Al3 in the very broad frequency range
from 50 MHz to 40 GHz (0.002 cm−1 to 1 cm−1 ) at temperatures 1.6 K < T < 300 K. During the last
years UPd2Al3 matured as a model compound [30] which exhibits typical heavy-fermion behavior below
the coherence temperature T ∗ of approximately 50 K. Because of its two-component electronic character,
UPd2Al3 exhibits both pronounced local magnetic moment and heavy-fermion itinerant behavior [30–32].
The 5f -shell has an average occupation of slightly less than three, with two electrons localized in the U4+
state and the remaining 5f -electrons considered to be itinerant due to their large hybridization with the
conduction electrons [33,34]. This coexistence is seen in the large magnetic moment (0.85 µB ) well below
the antiferromagnetic ordering temperature TN ≈ 14 K which is almost atomic-like, while at the same time
the itinerant electrons display a large coefficient of the electronic specific heat γ = 140 mJ mol−1 K−2 which
indicates m∗ /m ≈ 50. The magnetic ordering is of local origin and the formation of a spin density wave
can be ruled out [35,36]. The heavy quasiparticles condense in the superconducting state below Tc ≈ 2 K.
In order to reconcile magnetic order and superconductivity it was first suggested that different parts of the
Fermi surface lead to antiferromagnetism and to the heavy-fermion ground state. Tunneling spectroscopy
and inelastic neutron scattering evidence, however, that magnetic excitons produce the effective coupling
between the itinerant electrons and are responsible for superconductivity in UPd2Al3 [37, 38]. Previous
investigations of the optical reflectivity [36] indicated that the spectral weight is significantly reduced in
the far infrared, but even extensive measurements in the THz range could not determine low-frequency
dynamics of the itinerant quasi-particles [28].

3.2 Experimental methods and results


To avoid the extremely high reflectivity of bulk samples, we studied 150 nm thick epitaxial UPd2Al3 -films
deposited on LaAlO3 substrates by molecular beam epitaxy [39]. The physical properties of these thin films
have been characterized carefully and match those of bulk single crystals. The microwave conductivity
was determined with a Corbino spectrometer [40]: the reflection coefficient of the flat sample terminating
a coaxial cable reveals the sample impedance. For cryogenic measurements (base temperature: 1.6 K) we
have performed a full three-standards calibration that could be improved by using the superconducting state


c 2006 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim www.ann-phys.org
Ann. Phys. (Leipzig) 15, No. 7 – 8 (2006) 541

Fig. 5 Real part σ1 (ω, T ) of the complex con-


ductivity of UPd2Al3 as a function of frequency
(45 MHz to 20 GHz) and temperature (2 K to 300 K).
The temperature dependence additionally plotted on
the left wall (white line) represents the dc conductiv-
ity as determined with a two-lead measurement si-
multaneously with the microwave experiments. The
coincidence with dc conductivity indicates that the
low-frequency microwave data is obtained at fre-
quencies ω  1/τ . The strong high-frequency roll-
off in σ1 (ω) for the lowest temperatures on the other
hand occurs for ω > 1/τ . These two regimes also
lead to the characteristic, non-monotonic tempera-
ture dependence at high frequencies with a maxi-
mum in σ1 at the transition (ω = 1/τ ) between
these regimes.

of the UPd2Al3 sample (Tc = 2 K) as a short standard. This also makes it possible to extend the upper
frequency limit to 40 GHz. Data analysis [40] from complex reflection coefficient via sample impedance to
complex conductivity is performed for each frequency separately and requires no additional assumptions
than that the sample is thin compared to the skin depth (which we have checked for the complete frequency
and temperature range). The material does not enter the regime of the anomalous skin effect because the
Fermi velocity is very low, leading to a mean free path much shorter than the skin depth even for the
extremely low scattering rate found here.
In Fig. 5 the optical conductivity spectrum σ1 (ω, T ) of UPd2Al3 in the frequency range from 45 MHz
to 20 GHz is plotted as a function of temperature. At low frequencies as well as at dc the conductivity
resembles the dc behavior of bulk samples [30], which proves the high quality of the films. Clearly seen
is the minimum around 80 K, i.e. close to the coherence temperature T ∗ . At TN = 14 K a kink in the
temperature-dependent conductivity evidences the freeze out of magnetic scattering. At low temperatures
the resistivity follows a quadratic behavior in accord with the Fermi-liquid theory: ρ(T ) = ρ0 +A T 2 . As far
as the frequency dependence is concerned, at high temperatures the optical conductivity is constant according
to the Drude behavior of metals; infrared reflection experiments indicate a room-temperature scattering rate
around 8500 cm−1 [36]. With the temperature decreasing below T ∗ , the many-body ground state builds-up
and confines the charge excitation spectrum to very low frequencies, i.e. electronic correlations shift the
effective scattering rate 1/(2πτ ∗ ) to values as low as 3 GHz at T = 2.5 K.

3.3 Discussion
For a closer inspection, the real and imaginary parts of the low-temperature conductivity spectrum of
UPd2Al3 are displayed in Fig. 6. The optical conductivity σ1 (ω) is constant until the frequency approaches
the GHz-range where it slowly decreases and vanishes to zero as σ1 (ω) ∝ ω −2 . In this regime the imaginary
part of the conductivity σ2 (ω) exhibits a distinct maximum at 3 GHz. As demonstrated by the solid line,
the complete behavior can be perfectly fitted by the Drude model (3), using the dc value σdc = 1.05 ·
105 (Ωcm)−1 and the effective scattering rate 1/(2πτ ∗ ) = 3 GHz.
Our system therefore represents a perfect mélange to exhibit the Drude response: strong electronic
interactions lead to an extremely slow relaxation suitable for direct spectroscopic detection, and the influence
of impurities is strong enough to let the simple Drude description with a single relaxation rate be valid, but
weak enough to let the relaxation rate remain in the observable low-frequency range. In fact, concerning the
complex conductivity of any metal this is the first report that continuously covers such a broad frequency
range of almost three orders of magnitude right at the scattering rate.

www.ann-phys.org 
c 2006 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
542 M. Dressel and M. Scheffler: Drude response

Fig. 6 Optical conductivity spectrum (real and


imaginary parts) of UPd2Al3 at temperature T =
2.75 K. The fit by σ1 +iσ2 = σdc (1−iωτ )−1 , with
σdc = 1.05 · 105 (Ωcm)−1 and τ = 4.8 · 10−11 s,
documents the excellent agreement of experimental
data and the Drude prediction. The characteristic
relaxation rate 1/(2πτ ) is marked by the decrease
in σ1 and the maximum in σ2 around 3 GHz.

Fig. 7 Optical conductivity spectrum of


UPd2Al3 at different temperatures as indicated. The
relaxation rate 1/τ shifts to higher frequencies with
increasing the temperature. The results are in per-
fect agreement with the temperature dependence of
the dc conductivity as demonstrated in the inset.

Fig. 7 shows the optical conductivity spectra at different temperatures: the Drude roll-off described by
the scattering rate 1/τ ∗ (T ) shifts to lower frequencies as the temperature decreases. The behavior can be
compared to the scattering rate obtained by the dc resisitivity ρdc (T ) = m/[ne2 τ (T )]. As seen in the inset,
the values obtained from transport and from electrodynamic studies perfectly agree over a considerable
temperature range, indicating the consistency of the description. Our low-temperature findings can be
perfectly described by a Drude model with a renormalized, frequency-independent scattering rate 1/τ ∗ .
It is worth while to deliberate upon observing Fermi-liquid behavior in these strongly correlated electron
systems. Electron-electron interaction leads to an enhancement of the effective mass below the coherence
temperature T ∗ , i.e. typically starting somewhere below 50 K. Accordingly the Fermi velocity is slowed
down and the scattering rate reduced which shifts the interesting electrodynamics down into the microwave
range where it eventually can be directly observed. The energy scale of the Fermi liquid on the other hand can
be quite different. The dc resistivity of heavy-fermion compounds often displays a T 2 -behavior only below
1 K; and this is the temperature range where frequency-dependent experiments have to be performed to find
fingerprints of Fermi-liquid behavior. The probing frequency has to correspond to equally low energies,
i.e. microwave frequencies are appropriate. To make sure that electron-electron scattering dominates over
impurity scattering, the samples have to be extremely clean. Furthermore the effective mass should be large
because the prefactors of the Fermi-liquid prediction, Eq. (7), increase proportional to (m∗ )2 [24].


c 2006 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim www.ann-phys.org
Ann. Phys. (Leipzig) 15, No. 7 – 8 (2006) 543

4 Summary
The experimental verification of the Drude response in the electrical conductivity turns out to be a challenge.
For conventional metals, the roll-off is somewhere in the very far infrared and basically impossible to
access by standard optical reflection techniques which always require a low-frequency extrapolation for the
Kramers-Kronig analysis. Impurity conduction in semiconductors allows for a reduction of charge carriers;
however, the mean free path is affected by disorder. An extremely high mobility, which shifts the scattering
rate into the GHz range, is achieved in two-dimensional inversion layers, but for lower carrier concentrations
localization effect become effective. In heavy fermions electronic correlations lead to a significant reduction
of the scattering rate as the effective mass increases. Eventually this allows us to nicely observe the Drude
roll-off in the complex conductivity of UPd2Al3 . Over about three orders of magnitude in frequency the
dynamical response perfectly follows the behavior of the simple Drude model.

Acknowledgements We thank Martin Jourdan and Hermann Adrian for providing the UPd2Al3 samples. The work
was supported by the Deutsche Forschungsgemeinschaft (DFG).

References
[1] P. Drude, Physikalische Zeitschrift 1, 161 (1900).
[2] P. Drude, Ann. Phys. (Leipzig) 1, 566 (1900); 3, 369 (1900).
[3] M. Dressel and G. Grüner, Electrodynamics of Solids: Optical Properties of Electrons in Matter (Cambridge
University Press, Cambridge, 2002).
[4] A. Sommerfeld and H. Bethe, Elektronentheorie der Metalle, in: Handbuch der Physik, Vol. 24/2, edited by H.
Geiger and K. Scheel (Springer-Verlag, Berlin, 1933).
[5] J. M. Ziman, Principles of the Theory of Solids, 2nd edition (Cambridge University Press, London, 1972).
[6] J. Callaway, Quantum Theory of the Solids State, 2nd edition (Academic Press, Boston, 1991).
[7] H. E. Bennett and J. M. Bennett, Validity of the Drude Theory for Silver, Gold and Aluminum in the Infrared, in:
Optical Properties and Electronic Structure of Metals andAlloys, edited by F.Abelès (North Holland,Amsterdam,
1966).
[8] D.W. Lynch and W. R. Hunter, Comments on the Optical Constants of Metals and an Introduction to the Data
for Several Metals, in: Handbook of Optical Constants of Solids, Vol. I, edited by E. D. Palik, (Academic Press,
Orlando, 1985), p. 275.
[9] M. Dressel, O. Klein, S. Donovan, and G. Grüner, Ferroelectrics 176, 285 (1996).
[10] P. Drude, Ann. Physik und Chemie 36, 532 and 865 (1889).
[11] A. Röseler, Infrared and Spectroscopic Ellipsometry (Akademie-Verlag, Berlin 1990); R. M.A. Azzam and N. M.
Bashara, Ellipsometry and Polarized Light (Elsevier, Amsterdam, 1999); Handbook of Ellipsometry, edited by
H. G. Tompkins and E. Irene (Springer-Verlag, Berlin, 2005).
[12] M. van Exter and D. Grischkowsky, Phys. Rev. B 41, 12140 (1990); N. Katzenellenbogen and D. Grischkowsky,
Appl. Phys. Lett. 61, 840 (1992).
[13] T.-I. Jeon and D. Grischkowsky, Phys. Rev. Lett. 78, 1106 (1997); Appl. Phys. Lett. 72, 2259 (1998).
[14] S. J. Allen, D. C. Tsui, and F. DeRosa, Phys. Rev. Lett. 35, 1359 (1975).
[15] M. Scheffler, M. Dressel, M. Jourdan, and H. Adrian, Nature 438, 1135 (2005).
[16] D. Pines and P. Nozières, The Theory of Quantum Liquids, Vol. 1 (Addison-Wesley, Reading, 1966).
[17] L. D. Landau, Sov. Phys. JETP 3, 920 (1957).
[18] Z. Fisk, D.W. Hess, C. J. Pethick, et al., Science 239, 33 (1988).
[19] N. Grewe and F. Steglich, Heavy fermions. in: Handbook on the physics and chemistry of rare earths, Vol. 14,
ed. by K.A. Gscheidner Jr. and L. Eyring (Elsevier, Amsterdam, 1991), p. 343.
[20] C. M. Varma, Phys. Rev. Lett. 55, 2723 (1985); C. M. Varma, in: Proceedings of the 8th Taniguchi Symposium
on the Theory of the Valence Fluctuating State, Vol. 62 of Springer Series in Solid State Science, edited by T.
Kasuya and T. Saso (Springer-Verlag, New York, 1985).
[21] A. J. Millis and P.A. Lee, Phys. Rev. B 35, 3394 (1987).
[22] L. Degiorgi, Rev. Mod. Phys. 71, 687 (1999).

www.ann-phys.org 
c 2006 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
544 M. Dressel and M. Scheffler: Drude response

[23] A.A. Abrikosov, L. P. Gor’kov, and I.Y. Dsyaloshinskii, Methods of Quantum Field Theory in Statistical Physics
(Prentice-Hall, Englewood, 1963).
[24] K. Kadowaki and S. B. Woods, Solid State Commun. 58, 507 (1986).
[25] M. Dressel, G. Grüner, J. E. Eldridge, and J. M. Williams, Synth. Met. 85, 1503 (1997).
[26] P. E. Sulewski, A. J. Sievers, M. B. Maple, et al., Phys. Rev. B 38, 5338 (1988).
[27] A. M. Awasthi, L. Degiorgi, G. Grüner, et al., Phys. Rev. B 48, 10692 (1993).
[28] M. Dressel, N.V. Kasper, K. Petukhov, et al., Phys. Rev. Lett. 88, 186404 (2002); M. Dressel, N.V. Kasper, B.
Gorshunov, et al., Phys. Rev. B 66, 035110 (2002).
[29] P. Tran, S. Donovan, and G. Grüner, Phys. Rev. B 65, 205102 (2002).
[30] C. Geibel, C. Schank, S. Thies, et al., Z. Phys. B 84, 1 (1991).
[31] R. Caspary, P. Hellmann, M. Keller, et al., Phys. Rev. Lett. 71, 2146 (1993).
[32] R. Feyerherm, A. Amato, F. N. Gygax, et al., Phys. Rev. Lett. 73, 1849 (1994).
[33] K. Knöpfle, A. Mavromaras, L. M. Sandratskii, and J. Kübler, J. Phys.: Condens. Matter 8, 901 (1996).
[34] G. Zwicknagl and P. Fulde, J. Phys.: Condens. Matter 15, S1911 (2003); G. Zwicknagl, A. Yaresko, and P. Fulde,
Phys. Rev. B 68, 052508 (2003).
[35] A. Krimmel, P. Fischer, B. Roessli, et al., Z. Phys. B 86, 161 (1992); A. Krimmel, A. Loidl, R. Eccleston, et al.,
J. Phys.: Condens. Matter 8, 1677 (1996).
[36] L. Degiorgi, H. R. Ott, G. Grüner, et al., in: Strongly Correlated Electronic Materials, edited by K.S. Bedell et
al. (Addison Wesley, Reading, 1994), p. 96; L. Degiorgi, S. Thieme, H. R. Ott, et al., Z. Phys. B 102, 367 (1997).
[37] M. Jourdan, M. Huth, and H. Adrian, Nature 398, 47 (1999).
[38] N. K. Sato, N. Aso, K. Miyake, et al., Nature 410, 340 (2001).
[39] M. Huth, A. Kaldowski, J. Hessert, et al., Solid State Commun. 87, 1133 (1993).
[40] M. Scheffler and M. Dressel, Rev. Sci. Instrum. 76, 074702 (2005).


c 2006 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim www.ann-phys.org

You might also like