Electron viscosity and device-dependent variability in four-probe electrical transport in ultra-clean graphene field-effect transistors

This study investigates device-dependent variability in ultra-clean graphene field-effect transistors, attributing observed resistance fluctuations to competing scattering mechanisms and contact coupling, while proposing a phenomenological analysis method to effectively extract viscous electronic contributions in high-mobility graphene devices.

The Gist

Imagine a crowded dance floor where the dancers are electrons. Usually, in a standard metal wire, these dancers bump into walls, furniture (impurities), and each other in a chaotic, messy way. They lose their momentum quickly, like people trying to run through a crowded hallway while constantly tripping over chairs. This is called "diffusive" transport, and it creates electrical resistance (heat).

But in this paper, the researchers are looking at a very special, ultra-clean dance floor made of graphene (a single layer of carbon atoms). Because the floor is so clean and smooth, the dancers (electrons) rarely bump into the walls or furniture. Instead, they mostly bump into each other. When this happens, they start moving together like a fluid, similar to water flowing through a pipe. This is called electron hydrodynamics.

Here is a simple breakdown of what the paper found, using everyday analogies:

1. The Goal: Finding the "Perfect Flow"

The scientists wanted to prove that electrons in graphene can act like a thick, sticky fluid (viscous fluid) rather than individual particles. To do this, they built simple, rectangular "pipes" (devices) with four electrical contacts, like four people standing around a table to measure how much "traffic" is flowing.

2. The Problem: The "Device Lottery"

The researchers expected that if they built these pipes perfectly, they would all show the same "viscous" behavior. However, they found something confusing: identical-looking devices behaved completely differently.

• Device A acted like a super-fluid, showing "negative resistance." Imagine pushing a car, and instead of slowing down, it suddenly speeds up and pushes back against you.
• Device B acted somewhat normally but still showed weird fluid-like traits.
• Device C acted like a standard resistor, with no weird fluid behavior at all.

It was as if three people built the exact same model car, but one drove like a race car, one drove like a boat, and one just sat still. The paper asks: Why do these identical-looking devices act so differently?

3. The Investigation: Checking the "Edges"

The team realized that even though the graphene was incredibly clean, the edges of the device (where the metal wires touch the graphene) were the problem.

Think of the graphene channel as a river.

• In a perfect river, the water slides smoothly along the banks (no-slip condition), creating a beautiful, parabolic flow in the middle (Poiseuille flow).
• In their devices, the "banks" were slightly rough or had tiny defects. This changed how the water (electrons) interacted with the edges.

Some devices had edges that acted like a slippery ice rink (allowing the fluid to slide easily), while others acted like rough sandpaper (stopping the fluid). This difference in "edge friction" caused the same material to act like a fluid in one device and a solid in another.

4. The Evidence: How They Knew It Was Fluid

Even with the confusing results, they found strong proof that the electrons were acting like a fluid in many cases:

• The "Heat vs. Electricity" Test: In normal materials, heat and electricity travel together like two friends holding hands. In these graphene devices, they got separated. The "friendship" broke, which is a classic sign of a fluid-like electron state.
• The "Width" Test: If you make a pipe wider, a normal wire conducts electricity linearly (twice the width = twice the flow). But a fluid pipe conducts much better than that (flow increases with the square of the width). They saw this "super-conducting" behavior, confirming the fluid nature.
• The "Push Back" Effect: In some devices, when they pushed harder (increased current), the resistance actually dropped. It's like if you tried to push a heavy box, and the harder you pushed, the easier it became to move. This is a signature of electrons helping each other move.

5. The Solution: A New Way to Measure

Since the devices were so sensitive to tiny differences in their edges, the researchers couldn't just look at the raw numbers. They created a mathematical "recipe" (a phenomenological model).

Think of this recipe as a way to separate the "good fluid flow" from the "bad edge friction."

• They treated the device as a mix of two things: the viscous fluid in the middle and the messy contact points at the edges.
• By adjusting the variables in their recipe, they could mathematically "peel back" the messy edge effects to reveal the true viscosity of the electron fluid underneath.

The Bottom Line

This paper doesn't just say "electrons act like water." It says: "Electrons act like water, but only if the edges of the container are perfect. If the edges are even slightly rough, the whole experiment changes."

They showed that even in the cleanest materials, the specific way you build the device (the "architecture") dictates whether you see this amazing fluid behavior or just normal electricity. They provided a new tool (the mathematical model) to help other scientists figure out exactly how "sticky" their electron fluids are, regardless of how messy their device edges might be.

Technical Summary

Technical Summary: Electron Viscosity and Device-Dependent Variability in Four-Probe Electrical Transport in Ultra-Clean Graphene Field-Effect Transistors

Problem Statement
The hydrodynamic flow of electrons in two-dimensional mesoscopic channels offers a solid-state analogue to relativistic quantum fluids, exhibiting signatures such as Poiseuille flow, negative local resistance, and violations of the Wiedemann-Franz (WF) law. While high-mobility graphene is a premier platform for observing these phenomena due to its low intrinsic charge inhomogeneity and linear band structure, experimental observations remain inconsistent. Previous studies utilizing complex device geometries (e.g., Corbino disks, constrictions, vicinity geometries) have yielded conflicting results regarding the existence of negative resistance and other hydrodynamic signatures. These discrepancies are often attributed to the sensitivity of transport measurements to specific device fabrication processes and architectures. A unified framework to account for device-to-device variability in simple, rectangular four-terminal configurations is currently lacking, complicating the distinction between genuine hydrodynamic signatures and artifacts induced by complex geometries.

Methodology
The authors investigated a series of ultra-clean, hexagonal boron nitride (hBN)-encapsulated monolayer graphene field-effect transistors (FETs) fabricated using a simple rectangular four-terminal architecture. The devices featured channel lengths ($L$) of approximately 1–1.5 $\mu$m and widths ($W$) varying from 1 to 10 $\mu$m, with current and voltage contacts separated by a small distance $X$ (0.3–1.0 $\mu$m) to minimize non-local classical contributions.

The study employed the following experimental and analytical approaches:

1. Electrical Transport Characterization: Four-probe resistance ($R_{4p}$) measurements were conducted as a function of carrier density ($n$) and temperature ($T$) across multiple devices.
2. Viscous Flow Signatures: The authors verified hydrodynamic behavior through three key signatures:
   • Violation of the WF law via Johnson-Nyquist noise thermometry.
   • Non-monotonic variation of differential resistance ($dV/dI$) with injected current density ($J$), indicating a crossover from Knudsen to Gurzhi regimes.
   • Quadratic dependence of electrical conductivity ($\sigma$) on channel width ($W$), indicative of Poiseuille flow.
3. Phenomenological Modeling: To address the variability in $R_{4p}$, the authors developed a heuristic formalism based on the Landauer-Büttiker framework. The model treats the channel as a network of conductances comprising:
   • A viscous conductance ($G_V$) representing the hydrodynamic bulk.
   • Empirical coupling conductances ($G_a$ and $\tilde{G}_a$) accounting for non-local coupling between current and voltage contacts and potential contact asymmetries.
   • The model derives an expression for $R_{4p}$ that allows for negative values depending on the competition between these conductances.

Key Results

1. Device-Dependent Variability: Despite the use of identical simple architectures and ultra-clean materials, the devices exhibited three distinct classes of transport behavior in the doped regime ($n \gtrsim 10^{11}$ cm$^{-2}$):

   • Class I: $R_{4p}$ remains negative across the entire temperature range (100–260 K).
   • Class II: $R_{4p}$ is negative at low temperatures but transitions to positive above a characteristic temperature.
   • Class III: $R_{4p}$ remains positive across all measured temperatures.
     The sign and magnitude of $R_{4p}$ were found to correlate strongly with the intrinsic charge inhomogeneity ($n_{min}(0)$) and the specific contact configuration.

2. Confirmation of Hydrodynamics: The study confirmed the presence of viscous electron flow through:

   • A giant violation of the WF law, with the effective Lorenz number overshooting the semiclassical value by nearly two orders of magnitude.
   • A current-induced decrease in differential resistance, consistent with momentum-conserving electron-electron collisions.
   • Conductivity scaling as $\sigma \propto W^{1.8 \pm 0.1}$, closely matching the theoretical quadratic dependence expected for Poiseuille flow.

3. Viscosity Extraction and Scaling: Using the phenomenological model, the authors extracted the shear viscosity ($\eta$) and momentum relaxation length ($l_{mr}$).

   • Temperature Dependence: $\eta$ exhibited a $1/T$ dependence over the measured range ($T \ge 150$ K), contrasting with the $1/T^2$ dependence predicted by standard Fermi liquid theory. This is attributed to Fermi surface smearing at higher temperatures.
   • Density Dependence: The density dependence of $\eta$ varied by device class. Class II and III devices showed a quadratic dependence ($\eta \propto n^2$) consistent with Fermi liquid theory, while Class I devices showed linear ($\eta \propto n$) or square-root ($\eta \propto n^{0.5}$) dependencies.
   • Scattering Mechanisms: The results indicate that even in ultra-clean samples, electron-electron interactions compete with electron-phonon and electron-impurity scattering. The authors note that the scattering rates do not follow Matthiessen's rule but rather an "anti-Matthiessen's rule" where conductances add, consistent with hydrodynamic transport.

Significance and Claims
The paper claims to provide a simple, efficient experimental strategy and a phenomenological analysis tool for extracting the viscous electronic contribution in state-of-the-art high-mobility graphene FETs. By utilizing a simple rectangular four-terminal geometry, the authors demonstrate that device-to-device variability is an intrinsic feature arising from the competition between momentum-conserving (hydrodynamic) and momentum-relaxing scattering mechanisms, as well as coupling to contacts.

The authors assert that their model successfully unifies the observation of both positive and negative resistance regimes within a single framework, resolving the ambiguity often found in complex device geometries. They conclude that their findings support the simultaneous existence of multiple electron flow patterns (ranging from Poiseuille to tomographic flow) within the same device architecture. The work suggests that the observed $1/T$ viscosity and non-standard density scaling may require theoretical developments beyond conventional Fermi liquid hydrodynamics, potentially involving the "tomographic regime" of electronic transport, though they acknowledge that further theoretical investigation is needed for a deeper understanding.