Moving crystal phases of a quantum Wigner solid in an ultra-high-quality 2D electron system

Abstract

In low-disorder, two-dimensional electron systems (2DESs), the fractional quantum Hall states at very small Landau level fillings ($\nu$) terminate in a Wigner solid (WS) phase, where electrons arrange themselves in a periodic array. The WS is typically pinned by the residual disorder sites and manifests an insulating behavior, with non-linear current-voltage (I-V) and noise characteristics. We report here, measurements on an ultra-low-disorder, dilute 2DES, confined to a GaAs quantum well. In the $\nu<1/5$ range, superimposed on a highly-insulating longitudinal resistance, the 2DES exhibits a developing fractional quantum Hall state at $\nu=1/7$, attesting to its exceptional high quality, and dominance of electron-electron interaction in the low filling regime. In the nearby insulating phases, we observe remarkable non-linear I-V and noise characteristics as a function of increasing current, with current thresholds delineating three distinct phases of the WS: a pinned phase (P1) with very small noise, a second phase (P2) in which $dV/dI$ fluctuates between positive and negative values and is accompanied by very high noise, and a third phase (P3) where $dV/dI$ is nearly constant and small, and noise is about an order of magnitude lower than in P2. In the depinned (P2 and P3) phases, the noise spectrum also reveals well-defined peaks at frequencies that vary linearly with the applied current, suggestive of washboard frequencies. We discuss the data in light of a recent theory that proposes different dynamic phases for a driven WS.

Two-dimensional electron systems (2DESs) in a sufficiently strong, perpendicular magnetic field ($B$) at low temperatures and very small Landau level filling factors ($\nu<1/5$), form a many-body, ordered array of electrons, known as the quantum Wigner solid (WS) wigner1934interaction ; lam1984liquid ; levesque1984crystallization ; yi1998laughlin . The electrons, having their kinetic energy quenched significantly, succumb to the repulsive Coulomb interaction, and form a triangular lattice, maximizing the distance between each pair. In the presence of disorder, the WS breaks into domains and gets pinned to the local disorder sites which significantly alter the bulk transport properties cha1994orientational ; zhu1994sliding ; fisher1979defects ; ruzin1992pinning ; shayegan1997perspectives ; chitra1998dynamical . The longitudinal resistance ($R_{xx}$) of a pinned WS exhibits an insulating, typically activated behavior, $R_{xx}\propto e^{E_{A}/2k_{B}T}$ jiang1991magnetotransport , where $E_{A}$ is commonly associated with the WS defect formation energy narevich2001hamiltonian ; archer2014quantum ; chung2022correlated ; madathilunpublished . The pinned WS displays microwave or radio-frequency resonances that can be understood as collective excitations caused by the oscillations of the WS domains within the pinning potential andrei1988observation ; chen2006melting . It can also show non-linear current-voltage (I-V) and noise characteristics, revealing the complex dynamics of a moving WS shayegan1997perspectives ; willett1989current ; goldman1990evidence ; williams1991conduction ; li1991low ; jiang1991magnetotransport ; li1995rf ; li1996inductive ; csathy2007astability .

While the presence of disorder is a crucial ingredient in the transport properties of a pinned WS, less disordered 2DESs exhibit additional remarkable structure in transport measurements. Recent magnetoresistance measurements in ultra-high-quality 2DESs chung2021ultra indeed show clear signatures of fragile fractional quantum Hall states (FQHSs) at $\nu=1/7$ and other fillings embedded deep within the insulating phase chung2022correlated , clearly signaling the presence of interaction-driven phenomena at very small fillings. One would expect the quality to be reflected in the WS phase as well, and yield novel transport signatures. Our Letter reports striking non-linear I-V and noise characteristics of the driven WS in this extremely high-quality regime.

The ultra-high-quality 2DES in our study is realized by confining electrons to an 89-nm-wide GaAs quantum well. The sample was grown using molecular beam epitaxy, following an extensive campaign to improve the crystal purity via an optimization of the growth environment and sample structure chung2021ultra ; chung2022understanding . The quantum well has flanking Al${}_{x}$Ga${}_{1-x}$As barriers with stepped values of $x$, and Si dopants inside a doping-well structure chung2021ultra ; chung2022correlated ; chung2020working ; chung2022understanding . The very dilute 2DES has an areal density of $\simeq 2.83\times 10^{10}$ cm${}^{-2}$ and a record-high mobility of $\simeq 15\times 10^{6}$ cm${}^{2}$/Vs at such low density. We performed electrical transport measurements on a $4\times 4$ mm${}^{2}$ van der Pauw geometry, with alloyed In:Sn contacts at its corners and edge midpoints. The sample was cooled in a dilution refrigerator. For details of our measurement protocols, see Supplemental Material SupplementalMaterial .

Figure 1: Longitudinal resistance ($R_{xx}$) vs. magnetic field ($B$) for a dilute 2DES confined to an 89-nm-wide GaAs quantum well at $T\simeq$ 72 mK in the small filling range. The position of an emerging FQHS at $\nu=1/7$ is marked. Inset: $R_{xx}$ at $T\simeq$ 30 mK in the higher filling range.

Figure 1 presents $R_{xx}$ vs. $B$ for our 2DES. The sample’s exceptionally high quality is evinced from the well-developed FQHSs up to $\nu=8/17$ and $\nu=8/15$ on the flanks of $\nu=1/2$ (Fig. 1 inset), as well as an emerging FQHS between $\nu=1/3$ and 2/5, at $\nu=4/11$, which can be understood as the FQHS of interacting composite fermions wojs2000quasiparticle ; wojs2004fractional ; mukherjee2014enigmatic . The $\nu=4/11$ FQHS is extremely fragile and has been reported before only in the highest quality 2DESs and at much higher densities pan2015fractional ; samkharadze2015observation ; chung2021ultra ; chung2022correlated . Its presence in our dilute 2DES is a testament to its unprecedented quality. The data in the high-field range (Fig. 1 main) show that the resistance is approximately four orders of magnitude larger than in the low-field trace, indicating a highly insulating behavior when the filling factor is small ($\nu<1/5$). Nevertheless, we can see a clear $R_{xx}$ minimum at $\nu=1/7$, strongly corroborating the high quality of this dilute 2DES.

Figure 2(a) displays the non-linear transport traces at four different temperatures below the melting temperature of the WS which we deduce to be $T_{m}\simeq 120$ mK SupplementalMaterial ; footnote.melting ; deng2019probing . We measure the differential resistance ($dV/dI$) as a function of the driving dc current ($I$). A small sinusoidal ($ac$) excitation current of 0.05 nA is superimposed on $I$, and the ratio of the measured differential voltage to the small signal excitation gives the value of dV/dI. The frequency for the $ac$ current is kept small (0.1 Hz) to maximize the in-phase (resistive) component. The data are obtained at $B$ = 8.40 T ($\nu=0.139$), past the $\nu=$1/7 FQHS.

At the lowest $T$ [80 mK, top trace in Fig. 2(a)], $dV/dI$ vs. $I$ exhibits a strong non-linear behavior which can be broken into three distinct regimes, separated by two threshold currents. At very small $I$, $dV/dI$ drops fairly sharply, by a factor of about 5 within $\simeq$1 nA, and then shows a very abrupt drop at a first threshold current ($I_{th1}$) to negative values. We associate $I_{th1}$ with the depinning of the pinned WS. For $I>I_{th1}$, $dV/dI$ becomes non-monotonic and fluctuates around zero. Beyond a second threshold current ($I_{th2}$), the amplitude of the fluctuations significantly diminishes, and $dV/dI$ attains a small, relatively constant value, with little dependence on current. We associate the three different regimes separated by $I_{th1}$ and $I_{th2}$ with three different dynamic phases of the WS; we refer to these as P1, P2, and P3, as a function of increasing $I$, and show later in the manuscript that they each possess a distinct noise spectrum.

Figure 2: (a) $dV/dI$ vs. $I$ at $B$ = 8.40 T ($\nu=0.139$). The different traces, corresponding to different temperatures $T\simeq 80,85,90,$ and 96 mK, are offset vertically by 12, 8, 4, and 0 M$\Omega$, respectively. (b) Color-scale plot summarizing $dV/dI$ vs. $I$ curves for all temperatures. The dashed white line follows the first minimum corresponding to the negative differential resistance seen just beyond $I_{th1}$. (c) The integrated I-V characteristics at different temperatures.

Figure 3: Noise data at $B$ = 7.50 T ($\nu$ = 0.155). (a,b) $dV/dI$ (blue trace) and noise power obtained by averaging over a frequency window of $\pm$10 Hz centered at 40 Hz, defined as $S_{40}$ (red trace), vs. $I$ at $T$ $\simeq$ 61 and $\simeq$ 78 mK, respectively. (c) $S_{40}$ vs. $I$ at different temperatures. The scale for $S_{40}$ is logarithmic and spans over four decades. (d,e) Color-scale plots summarizing the noise power as a function of $f$ and $I$ at $T$ $\simeq$ 61 and $\simeq$ 78 mK, respectively. (f,g) Noise power vs. $f$ for different values of $I$ as indicated.

The traces shown in Fig. 2(a) reveal that the main peak becomes broader and $I_{th1}$ shifts towards higher values as $T$ increases. In addition, the local minima past $I_{th1}$, corresponding to negative $dV/dI$, become less negative and the amplitude of the fluctuations diminishes until it fully disappears at $T$ = 96 mK. Figure 2(b) is a color-scale plot that summarizes $dV/dI$ vs. $I$ curves at all temperatures. Figure 2(c) shows the I-V characteristics obtained by integrating the $dV/dI$ vs. $I$ curves. At the lowest temperature, I-V is step-like, with a large voltage developed for a small amount of current. This sharp jump can be understood as the voltage, or electric field, required to overcome the pinning potential and cause motion of the WS. Beyond $I_{th1}$, the voltage fluctuates about the threshold voltage value, and increases very gradually with $I$, indicating that once depinned, a negligible amount of extra electric field is required to sustain a current. Increasing temperature makes the step-like transitions more smooth, and the fluctuations become weaker, eventually vanishing at the highest temperatures vth.footnote .

To further investigate the driven phases of the WS, we measured the noise spectrum in response to a driving $dc$ current, concomitantly with $dV/dI$. Figure 3 summarizes the noise data at $B$ = 7.50 T ($\nu$ = 0.155). In Fig. 3(a), we show the non-linear $dV/dI$ (blue trace) together with the noise power averaged between 30 and 50 Hz, defined as $S_{40}$ (red trace) vs. $I$ at $T\simeq$ 61 mK. Qualitatively similar to Fig. 2(a) data, the $dV/dI$ vs. $I$ data in Fig. 3(a) exhibit two current thresholds that delineate three phases. (Note that the data in Figs. 2(a) and 3(a) were taken at different fillings and temperatures.) There is a remarkable correlation between the behaviors of $S_{40}$ and $dV/dI$ with $I$ in Fig. 3(a) in the three phases. At low drives ($I<I_{th1}\simeq 0.4$ nA), the noise is low ($S_{40}<10^{-7}$ V${}^{2}/$Hz). We identify this pinned, low-noise phase as P1. For currents between $I_{th1}$ and $I_{th2}$ ($0.4\lesssim I\lesssim 1.2$ nA), where $dV/dI$ fluctuates, $S_{40}$ rises dramatically, by about four orders of magnitude, and reaches a peak value of $\simeq$ $10^{-4}$ V${}^{2}/$Hz. This is the second phase, P2. Past P2, in phase P3, $dV/dI$ is very low and constant, noise decreases and, for large driving currents ($I\gtrsim 2$ nA), it becomes roughly constant and settles at a value of $\simeq$ $3\times 10^{-6}$ V${}^{2}/$Hz.

In Fig. 3(b) we present non-linear $dV/dI$ and $S_{40}$ data at a higher temperature of $T\simeq$ 78 mK. The $dV/dI$ trace indicates larger threshold currents, with suppressed fluctuating $dV/dI$ features in the P2 phase. The latter is also reflected in $S_{40}$ data, where the peak in $S_{40}$ in phase P2 is strongly suppressed compared to Fig. 3(a) data. Data shown in Fig. 3(c) at intermediate temperatures display the evolution of $S_{40}$ with temperature.

Besides $S_{40}$, we also measured the noise power spectrum between 5 Hz and 1 kHz as a function of frequency ($f$) at different driving currents. Figure 3(f) contains such data at $T\simeq$ 61 mK, for $I$ = 0, 1, and 3 nA. A summary of the full noise response vs. $f$ and $I$ at $T\simeq$ 61 mK is presented as a color-scale plot in Fig. 3(d). The noise spectrum reveals a distinct peak, accompanied by additional harmonics, that shift to higher frequencies with increasing $I$ in the P2 and P3 phases. The peak frequencies are linear in $I$, consistent with a signal that is of the washboard type gruner1988dynamics . However, the frequency values we observe are $\sim$$10^{3}$ times smaller than the expected washboard frequencies estimated from the expression $f=J/nea_{0}$ where $J$ is the current density, $n$ is the electron density, $e$ is the electron charge and $a_{0}$ is the WS lattice constant li1995rf ; gruner1988dynamics . Figure 3(g) contains the noise response vs. $f$ at a higher temperature of $T\simeq$ 78 mK, for $I$ = 0, 4, and 6 nA. A summarized color-scale plot is presented in Fig. 3(e). At this higher temperature, only a broad peak is observed for $I$ = 6 nA. In the Supplemental Material SupplementalMaterial , we report the temperature dependence of the slope of the $f$ vs. $I$ plots [see Figs. 3(d) and (e)], and of the power associated with the fundamental frequency of the narrow-band noise spectra.

Figure 4: Phase diagram for the WS as a function of $T$ and $I$ at $B$ = 7.50 T ($\nu$ = 0.155).

Our experimental observations can be summarized in a phase diagram for the WS as presented in Fig. 4. The data are taken at $B$ = 7.50 T ($\nu$ = 0.155), and $I$ is presented in a log-scale to visualize the width of the P2 region clearly. The black data points are obtained from the two thresholds, $I_{th1}$ and $I_{th2}$, extracted from $dV/dI$ vs. $I$ data. The extent of the P2 phase in $I$ is widest at the lowest temperatures. It narrows as $T$ increases and completely disappears above $T\simeq$ 85 mK (see Supplemental Material SupplementalMaterial for $dV/dI$ and $S_{40}$ data at $T$ $\simeq 85$ mK).

Our observations of noise and the $dV/dI$ data follow closely the numerical calculations of Reichhardt and Reichhardt reichhardt2001moving ; reichhardt2022nonlinear ; reichhardt2023noise , which suggest the presence of a pinned phase, followed by a moving WS glass phase, and then a moving smectic phase, with increasing drive. At low temperatures, below the melting temperature of the WS, the disordered WS exhibits channel flow just above the first threshold. In this phase, the conduction is filamentary, and can show jumps in $dV/dI$ which can be negative reichhardt2001moving ; reichhardt2022nonlinear ; reichhardt2023noise . Our $dV/dI$ data also show fluctuations, with negative $dV/dI$ just above depinning which disappears at $T\simeq$ 85 mK. Note that $T_{m}\simeq$ 120 mK for the WS in our sample SupplementalMaterial , and thus the P2 phase is lost before the WS melts. Our experimental observation of fluctuating and negative $dV/dI$ in phase P2 is highly suggestive of the filamentary phase reported in Refs. reichhardt2022nonlinear ; reichhardt2023noise .

Reichhardt and Reichhardt reichhardt2022nonlinear ; reichhardt2023noise also report on noise power (which we associate with $S_{40}$ in our work) as a function of the drive. Their calculations show suppressed noise in the pinned phase, high noise in the moving glass phase, and low noise in the moving smectic phase as the drive increases. Our $S_{40}$ data also reveal a similar qualitative picture, with $S_{40}$ being very low in phase P1, high in phase P2, and suppressed again in phase P3, with increasing current.

Despite the above qualitative agreements, there are some notable departures. While theory reichhardt2022nonlinear ; reichhardt2023noise predicts the emergence of the washboard signal only in the P3 (moving smectic) phase, we observe the noise peaks in phases P2 and P3. Also, although the linear dependence of the peak frequency on $I$ that we observe [Fig. 3(d)] is consistent with the washboard frequency, the three orders of magnitude discrepancy between the experimental and estimated frequencies is puzzling and leaves room for speculation. One possibility is that the origin of the noise signal is the motion of WS domains and its interaction with the disorder potential. If we assume that the domains have a reasonably uniform size, and since the domain size is much larger than the WS lattice constant, the resulting peak frequencies can be much smaller than the washboard frequency monceau2012electronic . Based on our observed $E_{th}\simeq$ 1 V/m, and the expression $L_{0}^{2}\simeq$ $0.02e/4\pi\epsilon\epsilon_{0}E_{th}$ fukuyama1978dynamics , we obtain a rough estimate of $nL_{0}^{2}\simeq$ 600 electrons per domain footnote.domain ; lili2023dynamic ; matthew2023characterization ; footnote.gaps . It is worth mentioning that narrow-band-noise peaks have been reported in the reentrant integer quantum Hall states, and there too, the observed frequencies are much smaller than the expected washboard value cooper2003observation ; sun2022dynamic . In the Supplemental Material SupplementalMaterial , we also report voltage ($V$) vs. time ($t$) traces of the noise measured with an oscilloscope. The real-time waveforms are saw-tooth like, suggesting a stick-slip motion of the domains as they encounter disorder sites gruner1994density ; sekine2004sliding .

There have been previous experiments reporting the non-linear I-V and noise characteristics of the high-field ($\nu<$1/5) WSs csathy2007astability ; li1991low ; williams1991conduction ; goldman1990evidence ; jiang1991magnetotransport ; willett1989current . While the data presented in these studies have distinct dissimilarities, a common conclusion has been the existence of two phases, a pinned phase at low drives, followed by a sliding phase at high drives. In our ultra-high-quality samples, we report an additional phase where $dV/dI$ is negative and fluctuating, and is accompanied by a large, low-frequency noise [Fig. 3(c)]. This is different from the data in previous work; so is the presence of a current-dependent peak in the noise spectrum in the depinned phases [Figs. 3(d) and (e)] Footnote.Csathy .

We also note that a rich collection of experiments on the negative $dV/dI$ exists in different platforms, suggesting a variety of possible mechanisms. In Refs. zou2021dynamical ; rees2016stick ; glasson2001observation , e.g., transport measurements of classical WSs formed on the surface of superfluid helium at zero magnetic field are reported. Reference brussarski2018transport reports two-threshold I-V characteristics past the metal-insulator transition in Si-MOSFETs (metal-oxide-semiconductor field-effect transistors). While the phenomena reported in Refs. zou2021dynamical ; rees2016stick ; glasson2001observation ; brussarski2018transport happen in the insulating phases, a more surprising observation is the similarity of our I-V features to those reported in Ref. bykov2007zero in a GaAs 2DES, but in vastly different parameter ranges: $\sim$ 30 times larger density, $\sim$ 100 times higher temperatures, $\sim$ 10,000 times larger currents, and $\sim$ 10 times lower magnetic fields ($B\sim$ 0.8 T, $\nu\sim$ 50)! The I-V characteristics were attributed to a redistribution of electrons in energy space, induced by the current bykov2007zero . These observations highlight the complexities and yet certain qualitative similarities of the electrical transport properties in remarkably different platforms and circumstances. Finally, it has been theoretically proposed that composite fermion correlations present in the FQHSs persist in the WS phase, and that composite fermion WSs are more stable than electron WSs at very low fillings archer2013competing ; archer2014quantum . Our data presented here should prove useful in testing future theories dealing with the transport characteristics of composite fermion WSs.

We acknowledge support by the National Science Foundation (NSF) Grant No. DMR 2104771 for measurements. For sample characterization, we acknowledge support by the U.S. Department of Energy Basic Energy Office of Science, Basic Energy Sciences (Grant No. DEFG02-00-ER45841) and, for sample synthesis, NSF Grants No. ECCS 1906253 and the Gordon and Betty Moore Foundation’s EPiQS Initiative (Grant No. GBMF9615 to L.N.P.). We thank S. A. Lyon for loaning us the HP-3562A spectrum analyzer, and D. A. Huse and J. K. Jain for illuminating discussions.