The Flumen of Hidden Order: Resolving Competing Phases in Moiré Heterostructures

Moiré heterostructures have become a playground for competing electronic phases. In a single device of twisted bilayer graphene, one can find superconductivity, correlated insulators, charge density waves, and even ferromagnetism—often within a few millielectronvolts of each other. The challenge is not just to observe these phases, but to resolve which one is truly dominant under a given set of conditions. For researchers who have already spent hours aligning flakes and tuning gates, the frustration of ambiguous transport data or conflicting scanning probe signals is all too familiar. This guide is for those who need a practical, step-by-step approach to untangling phase competition in moiré systems, without relying on black-box analysis or oversimplified phase diagrams.

Why Phase Competition Is the Norm, Not the Exception

In moiré heterostructures, the flat bands near the magic angle create a situation where kinetic energy is quenched, and interactions reign. This leads to a delicate balance between multiple ordered states. The same sample can host a superconducting phase at one doping and a correlated insulator at another, with the transition occurring over a gate voltage change of just a few volts. The underlying mechanism is that the Coulomb repulsion, bandwidth, and Fermi surface topology all conspire to produce nearly degenerate ground states. Small perturbations—strain, twist angle variations, dielectric environment—can tip the balance. Understanding this landscape is essential because the observed phase often depends on the measurement probe itself: transport might see a metal, while local compressibility sees a gap. The key is to recognize that competition is intrinsic, not a sign of a bad sample.

Degeneracy and the Role of Flat Bands

Flat bands enhance the density of states, making the system susceptible to various instabilities. The bandwidth in magic-angle twisted bilayer graphene is on the order of 10 meV, comparable to the Coulomb scale. This means that the system is always on the verge of ordering, and the exact ground state is determined by subtle details. For example, the presence of a hexagonal boron nitride (hBN) substrate can break the sublattice symmetry, favoring a correlated insulator over a superconductor. Similarly, alignment with the hBN can induce a moiré potential that further modifies the band structure. Researchers should always characterize their device's alignment and strain state before interpreting phase competition.

Why Different Probes See Different Phases

A common frustration is that transport measurements may show a metallic state, while scanning tunneling microscopy (STM) reveals a gap. This is not a contradiction—transport is sensitive to extended states, while STM probes local density of states. In a phase-separated system, the percolation path for transport may be metallic, even if most of the sample is insulating. Conversely, local probes might miss rare superconducting puddles. To resolve competition, one must use multiple complementary techniques on the same device. A practical rule: if two probes disagree, the phase is likely inhomogeneous or the measurement timescale matters.

Prerequisites: What You Need Before Starting

Before diving into phase mapping, ensure your experimental setup and sample quality meet certain thresholds. The most common failure mode is attributing a phase transition to physics when it is actually due to a bad contact or gate leakage. Here we outline the essential checks and calibrations.

Sample Quality and Twist Angle Homogeneity

Twist angle variations as small as 0.1° can shift the flat band energy significantly. Use Raman spectroscopy or low-energy electron diffraction to verify the twist angle across the device. If the variation exceeds 0.2°, consider the sample unsuitable for phase competition studies—the observed phases may be an average over different local angles. Also check for bubbles and cracks; they create local strain that can induce spurious insulating states. A good practice is to perform a basic transport characterization at multiple locations using a multi-terminal Hall bar geometry.

Gate Calibration and Hysteresis

Moiré devices often require dual-gate control (top and bottom) to independently tune doping and displacement field. Calibrate the gate capacitance using quantum Hall measurements at high magnetic field. Hysteresis in the gate response, especially in hBN-encapsulated devices, can shift the apparent doping by up to 10^12 cm^-2. Always sweep gates in the same direction when comparing phases, and record the sweep rate. For reliable phase boundaries, use a slow sweep rate (below 1 V/min) and wait for the current to stabilize at each point.

Cryogenic and Magnetic Field Setup

Many competing phases appear only below 1 K. Ensure your cryostat can reach base temperature below 300 mK, and that the thermometry is accurate. Use a calibrated RuO2 thermometer near the sample. For magnetic field studies, a vector magnet is ideal because it allows you to probe both in-plane and out-of-plane anisotropy—a key signature of different orders. If only a single-axis magnet is available, at least measure both field directions by rotating the sample.

Core Workflow: Building a Phase Diagram Step by Step

The central task is to construct a phase diagram as a function of doping (carrier density) and displacement field (or temperature). The following steps provide a systematic approach, from raw data to a published phase diagram.

Step 1: Measure Longitudinal and Hall Resistance

Start with a four-probe measurement of Rxx and Rxy as a function of gate voltages at base temperature. Use a low excitation current (1-10 nA) to avoid heating. Identify regions of superconductivity (zero resistance) and correlated insulators (peaks in Rxx). Plot Rxx as a color map in the (n, D) plane. This gives a first sketch of the phase boundaries. Note that the superconducting dome often appears as a narrow region near half-filling, while insulating states appear at integer fillings.

Step 2: Compressibility Measurements

Use a capacitance bridge or a single-electron transistor to measure the electronic compressibility dμ/dn. A negative compressibility indicates a correlated insulator, while a large positive value suggests a metal. Compressibility is often more sensitive than transport to small gaps. For example, a charge density wave gap may show up as a dip in compressibility even when transport still shows finite conductivity. Overlay compressibility maps on the transport phase diagram to resolve ambiguous regions.

Step 3: Local Probe Confirmation

If possible, perform scanning tunneling spectroscopy (STS) at a few key doping points. Look for gaps at the Fermi level (insulator) or coherence peaks (superconductor). STS can also reveal the spatial extent of phases—for instance, whether the insulating state is uniform or consists of puddles. For devices that cannot be scanned, use a scanning photocurrent or microwave impedance microscopy to map local conductivity.

Step 4: Temperature Dependence

Measure the evolution of Rxx and compressibility with temperature. Plot the phase boundaries as a function of T. The superconducting critical temperature Tc is defined as the point where Rxx drops to zero. For insulators, extract the activation energy from Arrhenius plots. A crossover from activated to metallic behavior with increasing T often indicates a competing phase. Use these data to refine the phase diagram and identify the nature of each phase (e.g., BCS-like vs. unconventional superconductivity).

Step 5: Magnetic Field Response

Apply a perpendicular magnetic field and measure the evolution of the phase diagram. Superconductivity is suppressed by fields above the upper critical field Hc2. Insulating states may show a magnetoresistance that is positive (for a band insulator) or negative (for a correlated insulator with spin ordering). The field response can also reveal the presence of a quantum Hall state, which competes with superconductivity at high fields. Use the field dependence to extract the Ginzburg-Landau coherence length for superconductivity and the magnetic length for insulating states.

Tools and Setup Realities

Not every lab has access to a dilution refrigerator with a vector magnet and STM. Here we discuss practical alternatives and trade-offs for common experimental constraints.

Transport-Only Approaches

If you only have transport capabilities, you can still resolve many competing phases. Use the temperature dependence of Rxx to distinguish between activated (insulator) and metallic behavior. A useful diagnostic is the derivative dRxx/dT: a positive slope indicates metallic behavior, negative slope indicates insulating. For superconductivity, look for a sharp drop to zero resistance. However, transport alone cannot distinguish between a charge density wave and a correlated insulator—both show a peak in Rxx. In that case, measure the nonlinear I-V characteristics: a charge density wave often shows a threshold voltage for depinning, while a correlated insulator shows a smooth nonlinearity.

Capacitance-Based Compressibility

A home-built capacitance bridge can be assembled with a lock-in amplifier and a low-temperature transformer. The sensitivity is sufficient to resolve the compressibility features in magic-angle graphene. The key is to minimize stray capacitance and use a high-frequency (10-100 kHz) excitation. This technique works down to 300 mK and does not require a separate probe. It is an excellent complement to transport because it is insensitive to contact resistance.

When to Use a Scanning Probe

Scanning probe methods (STM, AFM, microwave impedance) are essential when the phase is spatially inhomogeneous. If your transport data show a broad transition or hysteresis, it is likely that the sample has phase separation. A local probe can reveal the domain structure. However, these techniques are time-consuming and require ultrahigh vacuum. Reserve them for the most interesting samples where transport alone cannot resolve the competition. A practical compromise is to use a scanning photocurrent setup at room temperature to check for macroscopic inhomogeneities before cooling down.

Variations for Different Moiré Systems

The workflow above is tailored for twisted bilayer graphene, but the same principles apply to other moiré heterostructures. Here we discuss adjustments for three common variants.

Twisted Double Bilayer Graphene

In twisted double bilayer graphene (two Bernal-stacked bilayers twisted relative to each other), the flat bands are more tunable by an out-of-plane electric field. The phase diagram is richer, with multiple insulating states at different integer fillings. The key difference is that the displacement field can induce a band gap even without correlations. To separate the trivial gap from the correlated gap, measure the temperature dependence: a trivial gap is temperature-independent at low T, while a correlated gap closes at a characteristic temperature. Also, the superconducting dome is often narrower and appears only at very low temperatures (< 100 mK).

Transition Metal Dichalcogenide Moirés

In TMD moirés (e.g., WSe2/WS2), the interactions are even stronger, and the competition is between Mott insulators, charge density waves, and excitonic condensates. The main difference is that the bandwidth is smaller, and the system is more susceptible to disorder. Use photoluminescence to detect excitonic phases, which have no transport signature. For transport, focus on the Hall effect: a sign change in Rxy can indicate a transition from electron-like to hole-like carriers, often accompanying a phase transition. Also, the spin-valley locking in TMDs leads to Ising-type superconductivity with very high in-plane critical fields.

ABC Trilayer Graphene on hBN

This system has a moiré potential that creates narrow bands even without twisting. The phase competition is between a quantum anomalous Hall state, a correlated insulator, and superconductivity. The key experimental signature is the Chern number, measured via the Hall conductance. Use a combination of transport and magnetometry to identify the topological nature of the insulating state. The phase diagram is highly sensitive to the alignment angle with hBN—a misalignment of 0.5° can destroy the quantum anomalous Hall effect. Always characterize the alignment using atomic force microscopy or Raman before extensive measurements.

Pitfalls and Debugging: What to Check When Data Contradict

Even with careful planning, phase competition experiments often yield confusing results. Here are the most common issues and how to resolve them.

Hysteresis and Slow Relaxation

If the phase boundaries shift when you sweep the gate in opposite directions, the sample has trapped charges in the hBN or at the interface. This is especially problematic at low temperatures where charge relaxation is slow. To mitigate, use a light-emitting diode to illuminate the sample at base temperature—this can redistribute trapped charges and reduce hysteresis. Alternatively, apply a large gate voltage pulse (e.g., ±10 V) for a few seconds to reset the charge state. If hysteresis persists, consider the phase diagram as a function of sweep direction and report both.

Contact Resistance Artifacts

High contact resistance can mimic an insulating state. Check the two-terminal vs. four-terminal resistance: if they differ by more than a factor of two, the contacts are likely Schottky. Anneal the device at 150°C in vacuum for 12 hours to improve contacts. Another test: measure the resistance as a function of temperature—a true correlated insulator shows activated behavior (R ~ exp(Δ/2kBT)), while a bad contact shows a weaker temperature dependence (often R ~ T^{-1/2}).

Phase Separation Misinterpretation

When transport shows a gradual transition rather than a sharp phase boundary, the sample may be phase separated. In that case, the measured resistance is a percolation average. To confirm, measure the noise spectrum: phase separation often produces 1/f noise that peaks at the transition. Alternatively, use a scanning probe to directly image the domains. If you cannot do that, model the transport using effective medium theory to extract the fraction of each phase. A common mistake is to interpret a percolation transition as a continuous phase transition—they have different scaling behaviors.

Frequently Asked Questions and Practical Checks

This section addresses common questions that arise when resolving competing phases, along with quick diagnostic checks.

How do I distinguish a correlated insulator from a band insulator?

A correlated insulator shows a gap that scales with interaction strength and can be suppressed by doping away from integer filling. A band insulator has a fixed gap determined by the band structure. Experimentally, measure the gap size as a function of displacement field: a correlated gap changes with D, while a band gap is relatively constant. Also, a correlated insulator often shows a large negative magnetoresistance due to spin fluctuations, while a band insulator has positive magnetoresistance.

Why does my superconducting Tc vary between cooldowns?

This is often due to charge trapping or strain relaxation. The superconducting dome is extremely sensitive to the local doping. Between cooldowns, trapped charges can shift the effective doping by up to 10^11 cm^-2. To get reproducible Tc, always cool down with the same gate voltages applied (e.g., ground all gates) and use a thermal cycle that includes a high-temperature anneal (200 K) to reset charge traps. Also, check that the contacts are not degrading—a small change in contact resistance can shift the measured Tc.

What is the best way to publish a phase diagram?

Present the phase diagram as a color map of Rxx or compressibility in the (n, D) plane, with overlaid symbols for the phase boundaries determined from temperature and field dependence. Include error bars that reflect the hysteresis width. Always specify the sweep direction and rate. Provide raw data for key line cuts (e.g., Rxx vs. n at fixed D) in the supplementary material. For the community to trust your phase diagram, show that the same boundaries are observed with at least two different probes.

How do I know if my sample is in the clean limit?

The clean limit is reached when the mean free path exceeds the superconducting coherence length or the Fermi wavelength. In moiré systems, the mean free path is often limited by twist angle disorder. A practical check: measure the quantum oscillations (Shubnikov-de Haas) at high field. If you see well-resolved oscillations with multiple frequencies, the sample is relatively clean. If the oscillations are absent or broad, disorder is significant, and phase competition may be smeared. In that case, focus on the most robust phases (e.g., the correlated insulator at half-filling) rather than subtle competitions.

Next Moves: From Phase Diagram to Understanding

Once you have a reliable phase diagram, the next step is to connect it to microscopic theory. Compare your experimental boundaries with mean-field calculations of the Hubbard model on the moiré lattice. If the phase diagram shows a superconducting dome adjacent to an antiferromagnetic insulator, it suggests a spin-fluctuation pairing mechanism. If the superconductor appears next to a charge density wave, the pairing may be mediated by charge fluctuations. To test these ideas, measure the gap symmetry using tunneling or Andreev reflection. Another important check is the isotope effect—if available, substitute carbon with 13C to see if Tc changes, indicating phonon-mediated pairing. Finally, share your data with the community through open repositories like the Materials Data Facility to enable meta-analyses. The flumen of hidden order is not a single river but a network of competing currents—mapping it is the first step toward controlling it.