Ultrahigh electron mobility in suspended graphene

This paper asks whether suspending single-layer graphene from its substrate can reveal graphene’s intrinsic (nearly defect-free) transport by eliminating dominant extrinsic scattering sources. The motivation is that, despite graphene’s structural perfection, experiments typically show limited electron mean free paths (less than a micron) and mobilities far below what would be expected for an ideal crystal. The authors argue that many of these limitations arise from the substrate environment—charge traps, interfacial phonons, substrate-induced ripples, and fabrication residues—rather than from intrinsic lattice defects. Therefore, removing the substrate and improving cleanliness should increase mobility and sharpen signatures of charge neutrality.

The central contribution is experimental: the authors fabricate electrically contacted suspended graphene devices (six-probe Hall bar geometries) and demonstrate an order-of-magnitude mobility enhancement after an in situ “current annealing” cleaning step. They report mobilities exceeding 200,000 cm a2V−1s−1 at electron densities around 2×10^11 cm−2, and they connect this improvement to reduced charge inhomogeneity (narrower Dirac peaks) and earlier onset of Shubnikov–de Haas oscillations (indicating longer scattering times).

Methodologically, the study is a set of transport experiments on a small number of devices (one four-probe and two six-probe devices; device #1 is emphasized). Fabrication begins with mechanically exfoliated graphene on 300 nm SiO2/Si. Contacts are patterned by electron-beam lithography and metal deposition (3 nm Cr / 100 nm Au). Instead of oxygen plasma etching of graphene (which can introduce edge/bulk defects), the authors suspend the graphene by etching the underlying SiO2 using buffered oxide etch (BOE). They report an etching anisotropy: BOE removes approximately 150 nm of SiO2 uniformly below the flake while leaving SiO2 masked by the gold electrodes, thereby maintaining a parallel-plate capacitor geometry between gate and graphene. The device is then transferred to ethanol and dried via critical-point drying to prevent collapse of the suspended sheet.

Device integrity and geometry are verified by scanning electron microscopy and atomic force microscopy (AFM). AFM before measurements shows graphene suspended about 150 nm above the SiO2, and AFM after electrical measurements (after removing graphene with a short oxygen plasma etch) shows the remaining substrate bowing with height variation less than 20 nm. These checks support the claim that the devices are truly suspended and mechanically stable.

Transport measurements are performed in a vacuum cryostat at base temperature around 5 K (pressure < 5×10−5 mtorr). Before cooling, the devices are thermally annealed in situ to 400 K to reduce spurious doping. Electrical characterization uses low-frequency lock-in techniques with excitation current less than 100 nA. The authors measure longitudinal resistivity ρxx and Hall resistance Rxy using four-probe and six-probe configurations. They convert resistance to sheet resistivity using estimated geometric factors (width W and length L) from microscopy; they acknowledge that uncertainty in current/voltage distribution leads to less than ~30% error in ρxx.

A key experimental constraint is mechanical collapse under electrostatic gating. They limit gate voltage to ±5 V because collapse occurs in similar samples at about 20 V. They estimate the electrostatic force at ±5 V and the resulting strain in graphene (using a Young’s modulus of 1 TPa and graphene thickness 0.34 nm), concluding that the strain level (~5×10−4) should not significantly affect electronic transport.

The analysis of carrier density uses Hall measurements: n(Vg)=B/(eρxy(Vg,B)). Gate capacitance is estimated from Cg=n e/(Vg−VD) and is compared to a serial-capacitor model for graphene suspended above residual SiO2. This provides an independent verification of suspension: they find Cg≈60 aF/μm^2, close to the expected 47±5 aF/μm^2 for graphene suspended 150±20 nm above 150±20 nm residual SiO2.

The paper’s main quantitative results come from comparing transport before and after current annealing. Current annealing is implemented by repeatedly ramping current through the device to a setpoint, holding for several minutes, reducing current to zero, and remeasuring; the process is repeated until gate-response changes appear only at very large current densities (~2×10^8 A/cm^2, estimated using graphene thickness 0.34 nm). The authors interpret this as local heating sufficient to desorb fabrication residues.

Before current annealing, the suspended device already shows respectable mobility: for device #1, the mobility is about 28,000 cm2V−1s−1 at n=2×10^11 cm−2. This is comparable to the best reported unsuspended values at similar densities, implying that simply removing the substrate does not automatically eliminate the dominant scattering at this stage. After current annealing, mobility increases dramatically. For device #1, resistance decreases by more than a factor of 8 away from the Dirac point, while the Dirac peak width shrinks by about a factor of 20 with little change in the maximum resistivity. Mobility is reported to rise to 230,000 cm2V−1s−1 at n=2×10^11 cm−2. Across devices, every suspended sample exhibits mobilities higher than 60,000 cm2V−1s−1 after annealing. The authors emphasize that the peak mobility of 230,000 cm2V−1s−1 is an improvement of about a factor of 10 over the best values previously reported for traditional substrate-supported devices.

To connect mobility enhancement to sample homogeneity, the authors quantify the Dirac peak width via ΔWDirac, defined as twice the carrier density at which resistivity falls to half its maximum value. For device #1, after annealing the Dirac peak narrows to about 2×10^10 cm−2, which they state is more than a 10× improvement compared to the same device before annealing and compared to typical high-mobility unsuspended devices. They also show a correlation between reduced charge inhomogeneity and increased mobility across devices.

A third quality metric is the onset of Shubnikov–de Haas (SdH) oscillations. In a semiclassical picture, SdH oscillations begin when ωcτ≈1. They use a graphene semiclassical relation ωc=evFBSdH/(ℏ(πn)^{1/2}) with vF=10^6 m/s to infer a scattering time estimate. For the best device after annealing, SdH oscillations are observed as low as BSdH≈250 mT, whereas no SdH oscillations are observed before annealing. Across other suspended devices, BSdH ranges from 250 to 600 mT. Using their model, they estimate τ≈2×10−13 s for the best device at n=2×10^11 cm−2. In contrast, in unsuspended devices at the same density, SdH oscillations appear only for BSdH≳700 mT, corresponding to τ≈7×10−14 s. Thus, the suspended-and-annealed devices show an earlier SdH onset (about a factor of 3 improvement in onset field, as they summarize), consistent with reduced scattering.

The authors acknowledge limitations. First, the geometric conversion from measured resistances to resistivity carries up to ~30% uncertainty due to unknown current/voltage distributions. Second, the SdH onset provides only a qualitative measure of quantum scattering time τq; the authors explicitly note that other factors such as density inhomogeneity can affect the onset, so τq cannot be directly deduced. Third, they cannot fully rule out residual extrinsic scattering: even after annealing, the mean free path approaches device dimensions, and scattering from device edges and electrodes may contribute. They also speculate that larger-area devices might yield even higher mobilities.

In terms of practical implications, the results matter for both fundamental and applied directions. Fundamentally, the work supports the view that graphene’s intrinsic transport can be approached by eliminating substrate-related disorder and removing residues from both sides of the graphene sheet. This enables more direct study of Dirac fermion physics and quantum Hall phenomena with reduced disorder broadening. Practically, it provides a fabrication and cleaning strategy—suspension via BOE undercut with critical-point drying, plus in situ current annealing—that device engineers can adopt to improve performance in graphene-based high-mobility electronics and sensors. Researchers studying quantum transport, metrology, and low-disorder graphene devices should care because the paper demonstrates measurable improvements in mobility, charge neutrality homogeneity, and quantum oscillation visibility.

Overall, the study’s core conclusion is that suspended graphene, when cleaned by current annealing, reaches ultrahigh mobilities (up to 230,000 cm2V−1s−1) and exhibits substantially reduced disorder signatures (Dirac peak narrowing to ~2×10^10 cm−2 and SdH onset down to ~250 mT), demonstrating that extrinsic scattering dominates in conventional substrate-supported graphene devices. The work thereby establishes a clear experimental pathway toward accessing graphene’s intrinsic transport properties.