PHYSICAL REVIEW B 98, 161303(R) (2018)

Rapid Communications

Viscous transport and Hall viscosity in a two-dimensional electron system

G. M. Gusev,1 A. D. Levin,1 E. V. Levinson,1 and A. K. Bakarov2,3

1
Instituto de Física da Universidade de São Paulo, 135960-170 São Paulo, SP, Brazil
2
Institute of Semiconductor Physics, Novosibirsk 630090, Russia
3
Novosibirsk State University, Novosibirsk 630090, Russia

(Received 5 December 2017; revised manuscript received 10 August 2018; published 15 October 2018)

Hall viscosity is a nondissipative response function describing momentum transport in two-dimensional (2D)
systems with broken time-reversal symmetry. In the classical regime, Hall viscosity contributes to the viscous
flow of 2D electrons in the presence of a magnetic field. We observe a pronounced, negative Hall resistivity at
low magnetic field in a mesoscopic size, two-dimensional electron system, which is attributed to Hall viscosity
in the inhomogeneous charge flow. Experimental results supported by a theoretical analysis confirm that the
conditions for the observation of Hall viscosity are correlated with predictions.

DOI: 10.1103/PhysRevB.98.161303

Considerable progress has been made recently in the the importance of Hall viscosity in the context of condensed
nonperturbative understanding of the interaction effects in matter physics [19], it has been demonstrated that Hall viscos-
the electronic transport properties of metals within a hy- ity arises in many different and seemingly unconnected fields
drodynamic framework [1]. A hydrodynamic description is such as hydrodynamics, plasma, and liquid crystals [30]. It
valid when the electron-electron scattering time is much has been shown that classical Hall viscosity can be extracted
shorter than the electron-impurity or electron-phonon scat- from transport measurements in the emergent magnetohydro-
tering times. The theory of the hydrodynamic regime, where dynamic regime in 2D electron systems [31–33]. Note that
transport is dominated by a viscous effect, has been developed such a possibility has been questioned in a paper [13], where
in many theoretical studies [2–8]. It has been shown that the just the conventional Hall effect was found. However, one
shear viscosity contribution can be especially enhanced in the must take into account the higher-order terms in the expansion
case where the mean free path due to the electron-electron of the electron distribution function by the angular harmonics
interaction lee is much less than the sample width W , and the of the electron velocity (related to inhomogeneities of a flow)
transport mean free path l is in the order of or greater than [34]. Therefore the experimental study of the Hall resistivity
the width, l  W . In such a hydrodynamic regime, resistivity in a viscous system may provide a useful platform for future
is proportional to the electron shear viscosity η = 41 vF2 τee , theoretical developments in Hall viscosity.
where vF is the Fermi velocity and τee is the electron-electron In the present Rapid Communication, we have gathered
scattering time τee = lee /vF [2]. It has been predicted that all requirements for the observation of the hydrodynamic
resistance decreases with the square of temperature, ρ ∼ η ∼ effect and Hall viscosity in a 2D electron system and present
τee ∼ T −2 , and with the square of the sample width ρ ∼ W −2 experimental results accompanied by a quantitative analysis.
[2–8]. For this purpose, we chose GaAs mesoscopic samples with
Works demonstrating a feasible way to realize a hydrody- high-mobility 2D electrons. We employ commonly used lon-
namic regime, so far, have been achieved in experiments with gitudinal resistance, magnetoresistance, and the Hall effect to
electrostatically defined GaAs wires [9,10] and graphene [11]. characterize electron shear viscosity, electron-electron scat-
Until very recently, experimental studies have been carried out tering time, and reexamine electron transport over a certain
in zero external magnetic field. In order to describe the large temperature range, 1.5–40 K. We observe negative correc-
negative magnetoresistance in GaAs with high-mobility elec- tions to the Hall effect near zero magnetic field, which we
trons [12], the theoretical approach has been extended to in- attribute to classical Hall viscosity.
clude the magnetohydrodynamic behavior of two-dimensional Our samples are high-quality GaAs quantum wells with a
(2D) systems [13]. Similar magnetoresistance has been ob- width of 14 nm and electron density n  9.1 × 1011 cm−2 at
served in previous studies [14–16], which could be interpreted T = 1.4 K. Parameters characterizing the electron system are
as a manifestation of the viscosity effects. Recently, it has given in Table I. The Hall bar is designed for multiterminal
been demonstrated that palladium cobaltate wires [17] and measurements. The sample consists of three 5-μm-wide con-
mesoscopic GaAs structures [18] allow for the study of the secutive segments of different lengths (10, 20, and 10 μm),
underlying physical principles of the viscous system in a and eight voltage probes. The measurements were carried out
magnetic field and the carrying out of experiments to confirm in a VTI cryostat, using a conventional lock-in technique to
theoretical predictions [13]. measure the longitudinal ρxx and Hall ρxy resistivities with
One interesting property of a 2D fluid is Hall viscosity, an ac current of 0.1–1 μA through the sample, which is
which describes a nondissipative response function to an ex- sufficiently low to avoid overheating effects. We also compare
ternal magnetic field [12,14–29]. It is remarkable that, besides our results with the transport properties of 2D electrons in a

2469-9950/2018/98(16)/161303(5) 161303-1 ©2018 American Physical Society

GUSEV, LEVIN, LEVINSON, AND BAKAROV PHYSICAL REVIEW B 98, 161303(R) (2018)

TABLE I. Parameters of the electron system in a mesoscopic

sample at T = 1.4 K. Parameters are defined in the text.

ns μ vF EF l l2 η
(cm−2 ) (cm2 /V s) (cm/s) (meV) (μm) (μm) (m2 /s)
9.1 × 1011 2.5 × 106 4.1 × 107 32.5 40 2.8 0.3

macroscopic sample [34]. Three mesoscopic Hall bars from

the same wafer were studied.
Figure 1 shows deviations from conventional Hall resistiv-
ity ρxy (T ) = ρxy (T ) − ρxy
bulk
(referred to as the H1 config-
uration) as a function of temperature. In order to determine
bulk
the bulk Hall resistivity ρxy , we measured the Hall effect
in mesoscopic samples in a wider interval of the magnetic
field (−0.2 T < B < 0.2 T) and high T ∼ 40 K temperature.
Indeed, we found ρxy bulk
= −B/ens , where e is the electron
charge. Figure 1(b) shows the ratio ρxy (T )/ρxy bulk
for dif-
ferent temperatures. One can see a strong (∼10%–20%) de-
viation from the linear slope. The slope is opposite to the
bulk Hall slope at low fields and has the same sign (neg-
ative for electrons) at large positive magnetic field and low
temperatures. Before analyzing the Hall effect quantitatively

FIG. 2. (a) Temperature-dependent magnetoresistance of a

mesoscopic GaAs quantum well. Thick curves are examples illustrat-
ing magnetoresistance calculated from Eqs. (1) and (2) for different
temperatures: (a) 2.3 K (red), 21.1 K (blue), and 40 K (magenta);
(b) 4.2 K (red), 19.2 K (blue), and 37.1 K (magenta). (b) Comparison
of the magnetoresistance for different configurations. The schematics
show how the current source and the voltmeter are connected for
measurements.

and in order to make this analysis more complete, we also

measured the longitudinal magnetoresistivity ρxx (B ) in the
conventional configuration (referred to as R1). Note that the
longitudinal magnetoresistance has been studied previously
for different configurations of the current and voltage probes
[18]. Figure 2(a) shows ρxx (B ) as a function of magnetic
field and temperature. One can see two characteristic features:
a giant negative magnetoresistance (∼400%–600%) with a
Lorentzian-like shape (except for the small feature near the
zero field) and a pronounced temperature dependence of the
zero-field resistance. In general, we expect that the character
of the viscous flow strongly depends on the geometry and
probe configurations [11]. Figure 2(b) shows a comparison of
the magnetoresistance measurements in two configurations:
a conventional R1 configuration, and when the current is
injected between probes 9 and 7 and the voltage is measured
FIG. 1. Top: Sketch of the velocity profile for viscous flow in between probes 4 and 5 (referred to as the R2 configuration).
the experimental setup used in this study. (a) Temperature-dependent Strikingly, the resistance at zero magnetic field increases in
deviations from the conventional Hall resistivity ρxy (T ) of a amplitude and the width of the Lorentzian magnetoresistance
mesoscopic GaAs well. (b) The ratio ρxy (T )/ρxy bulk
for different is slightly reduced. The features near zero magnetic field are
temperatures. Dashes: Theory with parameters described in the main also smeared out. Surprisingly, we found that the resistance at
text. B = 0 is independent of temperature for the R2 configuration

161303-2
VISCOUS TRANSPORT AND HALL VISCOSITY IN A … PHYSICAL REVIEW B 98, 161303(R) (2018)

tribution from electron-electron scattering and temperature-

independent scattering by disorder [13]. It has been shown
that conductivity obeys the additive relation and is determined
by two independent parallel channels: The first is due to the
momentum relaxation time and the second is due to viscosity
[13,31]. This approach allows for the introduction of the
magnetic-field-dependent viscosity tensor and the derivation
of the magnetoresistivity tensor [13,31–33],
 
τ 1
ρxx = 1+ ∗
ρ0bulk , (1)
τ 1 + (2ωc τ2 )2
 
2τ2 1
ρxy = ρxy
bulk
1 − rH ∗ , (2)
τ 1 + (2ωc τ2 )2
+6ls )
where ρ0bulk = m/ne2 τ , τ ∗ = W (W12η , viscosity η = 41 vF2 τ2 ,
and rH is the numerical coefficient in the order of 1 [13]. At
the limit of zero magnetic field (B → 0), one obtains negative
corrections to Hall resistivity due to Hall viscosity in the limit
of small ls , ρxy = ρxy bulk
[1 − 6rH (l2 /W )2 ].
It is instructive to collect the equations for relaxation rates
T2
separately, τ2 (T1
)
= AFL
ee [ln(EF /T )]2 + τ2,0 , and τ (T ) = Aph T +
1 1

1
τ0
, where EF is the Fermi energy, and the coefficient AFL ee can
be expressed via the Landau interaction parameter, however,
it is difficult to calculate quantitatively (see the discussion
in Ref. [13]). The term Aph is due to scattering electrons
by acoustic phonons [36,37], and τ10 is the scattering rate
due to static disorder. Note that the effective relaxation time
FIG. 3. (a) Hall effect for two configurations, T = 4.2 K. τ ∗ is proportional to the rate τ12 (not time). We represent
(b) The ratio ρxy (T )/ρxy
bulk
for different configurations. Dashes the evolution of ρxx at B = 0 with temperature in Fig. 4(a)
(magenta) present calculations from ballistic+hydrodynamic theory for configurations R1 and R2. We fit the magnetoresistance
with parameters described in the main text. curves in Fig. 2 and the resistance in zero magnetic field
shown in Fig. 4(a) with the three fitting parameters τ (T ),
[35]. We attribute these results to the enhancement of the τ ∗ (T ), and τ2 (T ). Comparing the temperature dependencies,
viscous contribution, and further, we prove it by a quantitative we extract the following parameters, τ2,0 = 0.8 × 10−11 s,
9 −1 −2
comparison with theory. Furthermore, we check the Hall resis- AFLee = 0.9 × 10 s K , ls = 3.2 μm, Aph = 109 s−1 K−1 ,
tance in a modified probe configuration [35]. Figure 3 shows and τ0 = 5 × 10−10 s for configuration R1. For configuration
a comparison of the Hall effect in the H1 configuration with R2 all parameters are the same, except for ls = 2.8 μm.
the H2 configuration, where the current is injected between Assuming that the viscous effect is small in a macroscopic
probes 9 and 7 and the voltage is measured between probes 4 sample, we attempt to reduce the number of independent
and 8. One can see that ρxy at low magnetic field is wider in parameters by measuring ρ0 (T ) ∼ 1/τ (T ) and extracting
the H2 configuration, and, therefore, the ratio ρxy (T )/ρxy bulk
Aph independently [35]. However, we find a parameter in
exhibits a wider negative peak near zero B. the macroscopic sample Amacr ph = 1.3 × 109 s−1 K−1 , which is
Classical transport can be characterized on different length slightly higher than in the mesoscopic sample [35]. Table I
scales: the ohmic case (l  W ), ballistic regime (W  l, lee ), shows the mean free paths l = vF τ , l2 = vF τ2 , and viscosity,
and the hydrodynamic regime (lee  W  l). In real sam- calculated with the parameters, which we extracted from the
ples, electrons are scattered by static defects, phonons, and fit with experimental data. Figure 4(b) shows the dependen-
the sample edge. All these processes can be expressed in cies of 1/τ2 (T ) and τ ∗ (T ) extracted from the comparison with
terms of the scattering relaxation time τ and the boundary theory. Note that τ ∗ (T ) depends on the boundary conditions,
slip length ls . Boundary no-slip conditions correspond to the and the difference in its behavior for configurations R1 and R2
ideal hydrodynamic case of diffusive boundaries with ls = 0, could be explained by the difference in the parameter ls . More
while the opposite limit (free-surface boundary conditions) diffusive boundary conditions (smaller value of ls ) correspond
corresponds to the ideal ballistic case with ls = ∞. to stronger hydrodynamic effects.
In the hydrodynamic approach, the semiclassical treat- Now we return to the issue of Hall viscosity. Figure 3(b)
ment of the electron transport describes the motion of car- shows the dependence ρxy /ρxy bulk
at B → 0 as a function
riers, when the higher-order moments of the distribution of temperature for configurations H1 and H2 with calcula-
function are taken into account. The momentum relaxation tions obtained independently from magnetoresistance mea-
rate 1/τ is determined by an electron interaction with surements. From comparison with the experiment, we find
phonons and static defects (boundary). The second moment the adjustable parameter rH = 0.4. This value agrees with
relaxation rate 1/τ2 leads to the viscosity and contains the con- numerical calculations performed in the model [31], where

161303-3
GUSEV, LEVIN, LEVINSON, AND BAKAROV PHYSICAL REVIEW B 98, 161303(R) (2018)

It is important to note that, in the ballistic regime, ρxx and

ρxy strongly depend on the magnetic field due to the size
effects [38–41]. Unfortunately the changing B scale is almost
the same ∼W/RL (RL = mVF /eB is the Larmor radius)
for both contributions [31], and ballistic and hydrodynamic
effects can obscure each other. The magnitude of the ballistic
contribution depends on the ratio W/ l. In addition, the relative
ballistic contribution ρxx ball
/ρ0bulk exhibits a strong variation
with W/RL because the resistivity directly depends on the
relaxation time τ through the boundary scattering, while the
ball bulk
relative contribution to the Hall effect ρxy /ρxy is almost
independent of W/RL , since the Hall effect does not depend
on the relaxation time (but rather the size effect) [37–39].
Note that the sign of the effects is the same: The ballistic
contribution leads to an increase in boundary scattering, an
increase of ρxx , amplification of the classical Hall slope at
W/RL = 0.55, and quenching of the Hall effect near B = 0
[39,40]. From comparison with theory, at low temperatures,
ball
we found that ρxx < ρ0 [see Fig. 2(a)]. We attempted to
fit the magnetoresistance curves with a smaller Lorentzian
amplitude, considering the features near W/RL = 0.55 due
to the ballistic contribution, and found the fitting parameters
τ
τ∗
only 10% smaller. Note also that since the ballistic and
hydrodynamic contributions have the same sign, the B scale of
ball
the magnetoresistance is almost the same, when ρxx is added
to the magnetoresistance. However, for the same parameters,
ball
ρxy is comparable with the hydrodynamic contribution and
the ballistic corrections tend to counteract the hydrodynamic
corrections in the Hall effect. The ballistic model predicts the
quenching of ρxx ball
near B = 0 [40,41], therefore, ρxy /ρxy bulk

is not affected by the ballistic effect in very close proximity

FIG. 4. (a) Temperature-dependent resistivity and the Hall effect
of a GaAs quantum well at (B → 0) for different configurations. The
to zero field. However, the ballistic contribution leads to a
ball
solid lines and dashes show calculations based on theoretical Eqs. (1) decrease in the B scale of the ρxy (B ), when ρxx is added
and (2) with numerical parameters described in the main text. (b) to the Hall effect. We performed a calculation of the ballistic
Relaxation time τ2 as a function of temperature obtained by fitting transport in our sample geometry [35]. We confirmed that
the theory with experimental results. The solid line is the theory. the billiard model reproduces earlier numerical calculations.
(c) Relaxation time τ ∗ as a function of temperature. The solid lines Figure 3(b) shows our numerical results together with the
are the theory with parameters presented in the main text. hydrodynamic model. Indeed, the ballistic contribution results
in a decrease of the width of the negative peak near B = 0.
the parameter rH ≈ 0.35 was obtained. The existence of the One can see that, for the H2 configuration with stronger
parameter rH < 1 simply reflects the fact that the viscous Hall hydrodynamic effects (smaller ls ), the calculated curve could
correction in Eq. (2) never exceeds 100%, which one expects be brought in better agreement with the measurements, indi-
even for a small ratio l2 /W [see, for example, l2 /L = 0.04 cating the relevance of this explanation.
and W/ l = 0.1, considered in Fig. 2(b) of Ref. [31]]. In conclusion, we have measured the evolution of the longi-
Figure 1(b) shows the Hall curve as a function of B tudinal and Hall resistivities with temperature in high-quality
calculated from Eq. (2). Note that the theory predicts a broad GaAs quantum wells. Our observations are correlated with the
Loreantzian-like peak, while a rapid change of the sign is predictions of classical Hall viscosity for electron flow.
observed near B ≈ 0.01 T. The discrepancy could be related
to the higher-order expansion terms of the angular velocity We thank P. S. Alekseev and Z. D. Kvon for helpful
harmonics of the electron velocity, which are not considered discussions. The financial support of this work by FAPESP,
for longitudinal magnetoresistivity [13]. CNPq and CAPES (Brazilian agencies) is acknowledged.

[1] A. V. Andreev, S. A. Kivelson, and B. Spivak, Phys. Rev. Lett. [3] M. Dyakonov and M. Shur, Phys. Rev. Lett. 71, 2465 (1993).
106, 256804 (2011). [4] M. I. Dyakonov and M. S. Shur, Phys. Rev. B 51, 14341
[2] R. N. Gurzhi, Sov. Phys. Usp. 11, 255 (1968); R. N. Gurzhi, (1995).
A. N. Kalinenko, and A. I. Kopeliovich, Phys. Rev. Lett. 74, [5] M. Dyakonov and M. Shur, IEEE Trans. Electron Devices 43,
3872 (1995). 380 (1996).

161303-4
VISCOUS TRANSPORT AND HALL VISCOSITY IN A … PHYSICAL REVIEW B 98, 161303(R) (2018)

[6] A. O. Govorov and J. J. Heremans, Phys. Rev. Lett. 92, 026803 [23] G. Y. Cho, Y. You, and E. Fradkin, Phys. Rev. B 90, 115139
(2004). (2014).
[7] R. Bistritzer and A. H. MacDonald, Phys. Rev. B 80, 085109 [24] T. L. Hughes, R. G. Leigh, and E. Fradkin, Phys. Rev. Lett. 107,
(2009). 075502 (2011).
[8] B. N. Narozhny, I. V. Gornyi, M. Titov, M. Schutt, and A. D. [25] N. Read and E. H. Rezayi, Phys. Rev. B 84, 085316 (2011).
Mirlin, Phys. Rev. B 91, 035414 (2015). [26] F. D. M. Haldane, Phys. Rev. Lett. 107, 116801 (2011).
[9] L. W. Molenkamp and M. J. M. de Jong, Solid-State Electron. [27] C. Hoyos and D. T. Son, Phys. Rev. Lett. 108, 066805 (2012).
37, 551 (1994). [28] B. Bradlyn, M. Goldstein, and N. Read, Phys. Rev. B 86,
[10] M. J. M. de Jong and L. W. Molenkamp, Phys. Rev. B 51, 13389 245309 (2012).
(1995). [29] M. Sherafati, A. Principi, and G. Vignale, Phys. Rev. B 94,
[11] D. A. Bandurin, I. Torre, R. Krishna Kumar, M. Ben Shalom, 125427 (2016).
A. Tomadin, A. Principi, G. H. Auton, E. Khestanova, K. S. [30] M. Lingam, Phys. Lett. A 379, 1425 (2015).
Novoselov, I. V. Grigorieva, L. A. Ponomarenko, A. K. Geim, [31] T. Scaffidi, N. Nandi, B. Schmidt, A. P. Mackenzie, and J. E.
and M. Polini, Science 351, 1055 (2016). Moore, Phys. Rev. Lett. 118, 226601 (2017).
[12] Q. Shi, P. D. Martin, Q. A. Ebner, M. A. Zudov, L. [32] L. V. Delacretaz and A. Gromov, Phys. Rev. Lett. 119, 226602
N. Pfeiffer, and K. W. West, Phys. Rev. B 89, 201301 (2017).
(2014). [33] F. M. D. Pellegrino, I. Torre, and M. Polini, Phys. Rev. B 96,
[13] P. S. Alekseev, Phys. Rev. Lett. 117, 166601 (2016). 195401 (2017).
[14] L. Bockhorn, P. Barthold, D. Schuh, W. Wegscheider, and R. J. [34] P. S. Alekseev (private communication).
Haug, Phys. Rev. B 83, 113301 (2011). [35] See Supplemental Material at http://link.aps.org/supplemental/
[15] A. T. Hatke, M. A. Zudov, J. L. Reno, L. N. Pfeiffer, and K. W. 10.1103/PhysRevB.98.161303 for details in mesoscopic and
West, Phys. Rev. B 85, 081304 (2012). macroscopic samples measurements, and calculation of the Hall
[16] R. G. Mani, A. Kriisa, and W. Wegscheider, Sci. Rep. 3, 2747 effect within a semiclassical billiard model, which includes
(2013). Refs. [12,13,36,37].
[17] P. J. W. Moll, P. Kushwaha, N. Nandi, B. Schmidt, and A. P. [36] J. J. Harris, C. T. Foxon, D. Hilton, J. Hewett, C. Roberts, and
Mackenzie, Science 351, 1061 (2016). S. Auzox, Surf. Sci. 229, 113 (1990).
[18] G. M. Gusev, A. D. Levin, E. V. Levinson, and A. K. Bakarov, [37] T. Kawamura and S. Das Sarma, Phys. Rev. B 45, 3612 (1992).
AIP Adv. 8, 025318 (2018). [38] M. L. Roukes, A. Scherer, S. J. Allen, Jr., H. G. Craighead,
[19] J. E. Avron, R. Seiler, and P. G. Zograf, Phys. Rev. Lett. 75, 697 R. M. Ruthen, E. D. Beebe, and J. P. Harbison, Phys. Rev. Lett.
(1995). 59, 3011 (1987).
[20] I. V. Tokatly and G. Vignale, J. Phys.: Condens. Matter 21, [39] C. W. J. Beenakker and H. van Houten, Phys. Rev. Lett. 63,
275603 (2009). 1857 (1989).
[21] N. Read, Phys. Rev. B 79, 045308 (2009). [40] Z. Qian and G. Vignale, Phys. Rev. B 71, 075112 (2005).
[22] S.-S. Lee, S. Ryu, C. Nayak, and M. P. A. Fisher, Phys. Rev. [41] T. Geisel, R. Ketzmerick, and O. Schedletzky, Phys. Rev. Lett.
Lett. 99, 236807 (2007). 69, 1680 (1992).

161303-5