The sensitivity of moiré materials to crystal quality and twist angle can make experimental results difficult to replicate. Reproducibility should therefore be built into material growth, device fabrication and experimental design from the outset.

Two-dimensional (2D) moiré superlattices are formed by stacking two or more atomically thin crystalline layers with a small relative twist angle or a mismatch between their lattice constants. These materials can host narrow flat bands near the Fermi energy, where the kinetic energy of electrons is strongly suppressed and electron–electron interactions become dominant. Owing to this property, 2D moiré materials have become a platform for exploring strongly correlated phenomena, complementing established systems such as cuprate superconductors.

Credit: Reproduced from B. Urbaszek et al. Nature 567, 39–40 (2019), Springer Nature Ltd

An example is magic-angle twisted bilayer graphene, in which unconventional superconductivity emerges at a twist angle of approximately 1.1° (ref. 1). More recently, fractionally quantized Hall conductance states at zero external magnetic field, known as the fractional quantum anomalous Hall effect, have been demonstrated in twisted bilayer MoTe2, establishing this material as a fractional Chern insulator2,3. These states host fractionally charged excitations that may obey anyonic statistics, a crucial ingredient for topological quantum computation.

However, these phenomena are sensitive to various factors. For example, the presence of defects can alter the quality of the bulk crystals that are used as sources for 2D moiré materials. Crystals grown under the same nominal conditions can exhibit sample-to-sample variation in quality, which may prevent the reproducible observation of specific effects. Therefore, addressing reproducibility at its source — namely, during material growth — is the first step for achieving reproducible experimental results. A Perspective by Lucy Clark and J. Ross Stewart — published in this issue of Nature Physics — discusses the challenges of the reproducible synthesis of quantum materials and proposes potential solutions. In particular, they argue that understanding the nature of disorder and its influence on material properties is just as important as establishing the growth of a disorder-free material. Insights into the role of disorder can provide guidance for the construction of more realistic models and for the development of theoretical frameworks that accurately capture the behaviour of quantum materials.

Additionally, small variations in the twist angle can modify the moiré pattern and, consequently, the resulting electronic structure. Such variations are likely introduced during the layer-by-layer transfer and stacking processes used to assemble moiré hetero- and homo-structures. Thermal fluctuations in the transfer set-up, unintended strain applied during the pick-up and stacking procedures, or washing away the polycarbonate film used in the transfer process can all modify the relative orientation between layers, leading to changes in the twist angle. In some cases, inconsistencies may arise within the same device, where signals from different electrode pairs vary. Consequently, the observed strongly correlated phenomena are often difficult to reproduce in samples or devices designed to have the same nominal twist angle.

The lack of reproducibility has become one of the most persistent challenges in the moiré field4,5, attracting the attention of researchers, reviewers and journal editors. Overcoming this challenge will be crucial for translating strongly correlated moiré phenomena from laboratory discoveries into practical applications.

One possible way forward is to prepare more samples or devices and repeat the measurements. However, doing so is time-consuming and may delay the release of the results. In a highly competitive field, such delay may lead to additional pressures: PhD students have to publish papers to graduate, principal investigators need publications for funding or promotion and researchers are understandably eager to share exciting results with the community.

Nonetheless, it is crucial to make reproducibility studies an integral part of research projects. From the very beginning, when preparing samples or devices, researchers could produce multiple nominally identical samples or devices rather than relying on a single sample or device.

When the same phenomena are observed consistently across multiple samples or devices, one naturally obtains sufficient data for statistical analysis — a quantitative evaluation of device-to-device variability and the reproducibility of the observed phenomena. Questions about reproducibility raised during peer review can then be addressed confidently, without the need to spend additional time fabricating new samples and repeating measurements under the pressure of an ongoing review process. Conversely, if superconductivity or another phenomenon appears only in one of several samples, the results could be an artifact, and additional experiments may be needed to rule out alternative explanations.

Therefore, disclosing to readers how many devices were fabricated and how many of them reproduced the reported results would be useful to judge the reproducibility of observations more clearly. Such repeated measurements may also have other benefits. Researchers may find other interesting phenomena that were initially unexpected, which may lead to future research directions.

The effort to establish reproducibility as an integral part of research rather than as a separate consideration will require a sustained and coordinated effort from all stakeholders, including funding agencies, publishers, university administrators, principal investigators, researchers at all career stages and students. Achieving reproducibility is a long journey, but as long as we keep making steady progress, we will get closer and closer to it.