Skip to main content Skip to secondary navigation

Thin films, Interfaces and 2D Materials

Main content start

Crystal lattices with confined geometries, like thin films and interfaces, host plenty of novel phenomena and possess unique functionalities. Over the past decade, artificial nanoscale material structures, such as heterointerfaces and monolayer 2D materials, were a major driving force toward the realization of novel physics in quantum materials. We engineer a variety of crystal thin films and interfaces at atomically thin limits using molecular beam epitaxy (MBE).

In addition to pioneering novel techniques for quantum material synthesis, we investigate the properties of these materials by performing in situ ARPES and STM measurements, giving us direct feedback on material quality and novel electronic properties.

Enhanced superconductivity in monolayer FeSe
2D Materials
Strong near-neighbor attraction in 1D Cuprate chains

Enhanced superconductivity in monolayer FeSe

Interfacial Coupling

Bulk FeSe is an iron-based superconductor that has a maximum transition temperature (Tc) of 8 K. Surprisingly, the Tc of a single layer of FeSe film grown epitaxially on an SrTiO3 substrate is enhanced to over 55 K. We perform an in situ synchrotron-based MBE/ARPES experiment to understand how Tc can be enhanced from the bulk Tc of 8 K at this 1UC (one unit-cell) FeSe/STO interface. We found two reasons for its enhancement: (1) additional electron doping through interface charge transfer, and (2) interfacial mode coupling. Prominent replica bands in the photoemission spectra, separated by energies on the order of phonon modes on the STO substrate, indicate the existence of electron-phonon coupling across the interface. The similarity of replica and main bands imply that emission/absorption of phonons by electrons occurs with very little momentum transfer (nearly q = 0 scattering). This forward scattering could help boost Tc even in superconductors whose order parameter possess a sign change, as is the case in the cuprate and possibly iron-based superconductors.


Substrate Effect

STO exhibits a wide range of unique properties, such as superconductivity and incipient ferroelectricity. It is therefore challenging to identify what plays the major role in the enhancement of Tc in the 1UC (one unit-cell) FeSe/STO system. We explored the substrate effect by studying 1UC FeSe grown on a rutile TiO2 (100) surface. We found that 1UC FeSe/TiO2 and 1UC FeSe/STO have similar electronic structure, superconducting gap size, and Tc. The similarities between these two systems suggest that the dielectric constant, strain, and crystal symmetry of the substrate do not play important roles in the enhancement of Tc. Instead, it shows that interfacial e-ph coupling is most likely crucial for enhancing superconductivity in 1UC FeSe/STO systems.


Robust Doping Level

Charge transfer across an abrupt interface between two materials with different work functions is a common phenomenon in epitaxial thin films. In our recent study, monolayer FeSe is grown on top of a LaTiO3 (LTO)/STO heterostructure. Due to the lower work function of LTO, additional charge transfer is expected between 1UC FeSe and an STO substrate with more LTO layers. Surprisingly, electron density as measured by ARPES Fermi surface maps, and superconducting gaps for 1UC FeSe/LTO/STO films, remain nearly unchanged with different LTO thickness. This shows that superconductivity in 1UC FeSe thin films is robust and anchored by a 'magic' doping level.

To understand the unique phenomena of a 'magic' doping level in the context of iron-based superconductivity, consider the possible interplay of first-order phase transition and superconductivity. Above a doping level of 0.11, an insulating phase may form, resulting in phase separation between superconducting and insulating phases. As a result, only ~0.11 doping is visible in the ARPES experiment.

2D Materials

The synthesis and characterization of 2D monolayer crystals have advanced in recent years, and they have emerged as a new platform for experimental condensed matter physics.

The bulk (3D) analog of a 2D material is usually composed of atomic layers with weak out-of-plane van der Waals bonds. While 2D materials could possibly be produced by mechanical exfoliation from the bulk, few high-quality 2D materials are actually produced this way. A famous example is high mobility graphene (2D), which is made by exfoliation from graphite (3D). Other 2D materials, such as transition metal chalcogenides, cannot be easily exfoliated from bulk crystals. Therefore, developing advanced monolayer materials by vapor deposition methods, such as MBE, is crucial for producing 2D materials and understanding their underlying physics.

We study various new 2D materials in our group. They are made by layer-by-layer growth of high-quality thin films, and subsequently analyzed using ARPES. Two of our papers are summarized here: a measurement of the transition from direct to indirect bandgap from bulk to monolayer MoSe2, and an observation of the quantum spin Hall state in monolayer WTe2.

Direct observation of the transition from indirect to direct bandgap in atomically thin epitaxial MoSe2

(a) Crystal structure of MoSe2. (b) Theoretical band structure along the Gamma-K direction of the monolayer MoSe2 film. (c) Experimental band structure, shown as second derivative of the data to enhance visibility. (d) Raw experimental band structure.

In 2014, Yi Zhang et al. measured the band gap in MoSe2 using ARPES. High quality thin films of MoSe2 were grown by molecular beam epitaxy on a bilayer graphene substrate, then immediately transfered to an in situ ARPES set-up. The work provided the first direct experimental evidence of the distinct bandgap transition in thin film samples with varying thicknesses. Studying films from one to eight layers, we found that the direct-to-indirect bandgap transition occurred at one to two monolayers, which is consistent with theory.

Quantum spin Hall state in monolayer 1T’-WTe2

A quantum spin Hall (QSH) insulator is a novel two-dimensional state of matter featuring quantized Hall conductance without a magnetic field. The quantum spin Hall state arises from topologically-protected dissipationless edge states that bridge the energy gap opened by band inversion and strong spin-orbit coupling. In a study done by Shujie Tang et al., monolayer WTe2 was grown on a bilayer graphene substrate using molecular beam epitaxy. Using ARPES and scanning tunneling spectroscopy, we established that monolayer 1T’-WTe2 is a new class of QSH insulator, with nontrivial band inversion, the opening of a 55 meV bulk band gap, and a conducting edge state coexistent with an insulating bulk. This finding provides a platform for studying QSH insulators in 2D transition metal dichalcogenides, and for developing novel device applications.


Strong near-neighbor attraction in 1D Cuprate chians

While superconductivity in cuprates exists in their 2D layers, so far, there remains no exact solution of many-body Hamiltonians in two or more dimensions, making a quantitative comparison between theory and experiment a challenge. Nevertheless, this theory problem can be simplified greatly in 1D systems: the reduced dimensionality enables exact solutions of the microscopic Hubbard Hamiltonian. For years, researchers have tried to find or create such 1D systems in order to test the theoretical models. In the past 25 years, researchers have successfully managed to design 1D chains of the cuprate family, but tunning their doping levels, which is an essential aspect for theory verifications, has always been a challenge.

After two decades of trying, in 2021, Z. Chen et al. finally doped 1D cuprate chains over a wide range of hole-doping values by synthesizing Ba2-xSrxCuO3+d (BSCO) through the technique of ozone-reactive molecular beam epitaxy (MBE), which gave us the ability to study this long-running problem.

(A1 to A6) Fermi surface maps with different hole doping levels. (B1 to B6) Measured ARPES spectra taken along the cuts shown with white dashed lines in (A1) to (A6). (C1 to C6) Second derivatives of energy distribution curves and momentum distribution curves are summed with sign reversed for comparison. (D1 to D6) Spectral function simulation results based on 16-site cluster perturbation theory (CPT) with varied corresponding hole doping levels.

In a 1D system, strong correlations manifest as so-called spin-charge separation, through which a photoinduced hole will fractionalize into a “spinon” and a “holon,” carrying spin and charge, respectively, giving rise to spectral branches with different propagation velocities. In case of the cuprate chain, the ARPES spectra are mainly characterized by the main holon branch “h” crossing EF at kF, the “3kF” branch extending from h and bends back at 2π-3kF, the spinon branch “s” which flattens near k||=0, and the holon folding branch “hf” at |k|>|kF| as previously being attributed to a holon-holon interaction induced by spin superexchange. The persistence of spin-charge separation up to 40% doping and the presence of the hf branch agree with predictions using the single-band Hubbard model with fitted parameters U=8t, t= 0.6eV, and J=4t2/U=0.3 eV.

(A to C) MDCs for varied energies at 9, 14, and 33% hole doping. (D) CPT simulation based on single-band Hubbard model without Fermi-Dirac suppression above EF for the 12.5% doping case, shown in the color-scale plot. (E) Peak positions extracted from MDCs for different dopings (squares: 9%; circles: 14%; diamonds: 33%) overlaid on theoretical lines of single-band Hubbard prediction.

Despite such consistencies, we find that the simple Hubbard model is fundamentally deficient in accurately addressing additional spectral features. Specifically, the Hubbard model prediction suggests that the relative intensity of the hf feature is rather weak compared to the 3kF branch, whereas the experimental observation is the opposite. This implies that spin superexchange cannot fully describe the holon-holon interaction. With simulations, we find that an attractive nearneighbor Coulomb interaction V≈–1.0t must be introduced to the Hubbard Hamiltonian to reproduce the experimental spectra and explain the suppression of the intensity of 3kF branch. More importantly, this additional nearneighbor attraction is an order of magnitude stronger than the inherent attraction in the Hubbard model mediated by spin superexchange (which has magnitude ~J/4~0.1t). This difference implies that the single-band Hubbard model misses a sizable, attractive interaction between neighboring holes, which likely originates from electron-phonon coupling, considering the evidence of such interactions in a variety of cuprates.

(A) CPT simulation incorporating a near-neighbor interaction term V (blue curves), compared with experimental MDC at –0.7 eV below EF for 14% doping (red curve). (B) Doping-dependent experimental MDCs at energies approximately 0.6 eV above the holon bottom. (C) Folding peak intensity relative to the major peak intensity as a function of doping. (D) –V as a function of CPT simulated folding peak relative intensity f as a phenomenological measure of the net near-neighbor attraction (squares: 9%; diamonds: 12.5%).

Given that the Hubbard model with strong on-site repulsive U and strong near-neighbor attractive V explains the essential physics in 1D, it is plausible that such an augmented Hubbard model provides a holistic picture for all cuprates, including their 2D forms.