Home  /  Research

Research

From molecules to modules: connecting the full perovskite photovoltaic chain

Perovskite solar cells now approach the efficiency of conventional silicon, but to reach industrial production several barriers must be overcome at once: material purity, interfacial chemistry, crystallization kinetics, device stability and scalable manufacturing. The group puts materials chemistry and device physics on the same bench, giving rise to five interlocking research directions.

Semi-transparent perovskite devices laid over the chemical design scheme of the interfacial molecules
Inverted (p–i–n) perovskite devices, with the chemical structures of the interfacial molecules behind them.
Spin-coating of a perovskite thin film
Wet processing of perovskite thin films — solution chemistry here directly determines film quality.

Direction 01

Novel Photoactive Materials Design

Novel photoactive materials design

The molecular structure of interfacial and transport materials often sets the ceiling for device efficiency. Taking "chemical compatibility" as its central design principle, the group systematically develops functional molecules for perovskite photovoltaics: carbazole-based phosphonic acid self-assembled monolayers (SAMs) and their derivatives, conjugated hydrazide additives, symmetrical molecules with long-range-ordered π–π stacking, and hybrid molecular systems for interfacial passivation and charge extraction.

The question we care about is not simply whether a material conducts well, but what happens between it and the precursor, the solvent and the perovskite surface under real film-forming conditions. This perspective was shown in our 2026 Science paper to be the key to understanding a device performance bottleneck.

Self-assembled monolayers (SAMs) Conjugated hydrazide additives Long-range π–π stacking Interfacial passivation
Molecular-dynamics snapshots of the mPy and pPy interfacial layers on a perovskite surface, with a bar chart of the number of π-stacking couples each molecule forms
Symmetry decides the packing. On the perovskite surface mPy lies flat and disordered, while pPy stands upright and stacks through its aromatic cores — five π-stacking couples to mPy's one (Nature Synthesis, 2025).

Direction 02

Perovskite Photovoltaic Technologies

Perovskite photovoltaic technologies

Taking the inverted (p–i–n) architecture as the main line, we engineer both the buried interface and the top interface: through molecular hybridization, surface transformation and control of crystallization kinetics, we improve open-circuit voltage, fill factor and long-term stability at the same time.

On stability, our concern is not only whether a device lasts long, but where the degradation mechanisms come from. A 2023 Nature paper proposed homogenizing the vertical cation composition to suppress phase segregation; a 2024 Nature paper used buried-interface molecular hybridization to improve interfacial contact; and a 2025 Nature Materials paper revealed how PCBM dimer formation drives device degradation and offered a strategy to suppress it.

Inverted p–i–n devices Buried interface engineering Crystallization kinetics Stability mechanisms
Molecular model of a 3:1 Me-4PACz : BA/NA/TA mixed layer bridging a NiO surface and a FAPbI3 perovskite, drawn across a 10 nm by 10 nm cell with a 2 nm anchoring layer
The buried interface, molecule by molecule: a 3:1 mixture of Me-4PACz with BA/NA/TA bridges the NiO substrate and the FAPbI3 perovskite, shown across a 10 nm × 10 nm cell with a ~2 nm anchoring layer (Nature, 2024).

Direction 03

High-Purity Precursors & Perovskite Microcrystals

High-purity precursors & perovskite microcrystals

Device reproducibility begins with the raw materials. The group has developed a perovskite precursor synthesis route that uses water as the solvent, replacing conventional organic-solvent processes that are costly and environmentally burdensome, while markedly improving precursor purity and batch-to-batch consistency — published in Science in 2024.

Building on this, we continue to advance the precise synthesis and compositional control of perovskite microcrystals, studying crystal growth kinetics, defect chemistry and impurity control to provide a stable and reliable materials foundation for high-efficiency devices.

Aqueous synthesis High-purity precursors Precise microcrystal synthesis Batch consistency
The pH windows of the aqueous precursor route, and a kilogram-scale batch of the purified precursor it produces
Aqueous synthesis of perovskite precursors: the pH window that sets the purification step — below pH 1.9 the route gives pure δ-FAPbI3 microcrystals — and the kilogram-scale batch it delivers (Science, 2024).

Direction 04

Lead Halide Perovskite Recycling

Lead halide perovskite recycling

For perovskite photovoltaics to become a genuinely "green" energy technology, two questions must be answered: where does the lead go, and can the materials return to the production line? The group is systematically studying closed-loop recycling pathways for lead halide perovskites — separating, purifying and regenerating perovskite precursors and functional-layer materials from end-of-life devices, while exploring the valorization of solvents and additives.

Alongside this comes greener fabrication: reducing the use of toxic solvents, lowering energy consumption and improving material utilization, so that scalable manufacturing also holds up on environmental grounds.

Closed-loop lead recovery Solvent valorization Green fabrication Life-cycle assessment
A four-step closed recycling loop around a PbI2, glass and copper hub: collection and delamination, dissolution and separation, selective precipitation, and re-synthesis into a new perovskite absorber
The closed loop we are working towards: end-of-life modules are collected and delaminated, the perovskite layer is dissolved away from the glass and metal electrodes, PbI₂ is selectively precipitated and purified, and the recovered PbI₂ goes back into a new absorber alongside fresh salts.

Direction 05

AI for Perovskite Photovoltaics

AI-assisted photovoltaic research & development

The compositional space and process window of perovskites are enormous — manual trial-and-error cannot exhaust them. The group explores a data-driven approach to R&D: structuring formulations, process parameters and device performance into datasets, and combining them with machine-learning models for high-throughput screening and process optimization, concentrating experimental resources on the most promising directions.

This direction is deeply coupled with experiment — the models produce actionable experimental suggestions, not conclusions on paper.

Data-driven formulation High-throughput screening Process optimization Experiment–model loop
The screening funnel — search space, structured dataset, machine learning, prioritised experiments — above the device stack the model predicts on, a predicted-versus-measured plot and the measurement-to-data closed loop
The experiment–model loop, end to end: millions of candidate compositions are condensed into a structured dataset, a structure–property model ranks which experiments to run next, and the measurements feed straight back into the data. The panels below show the device stack the prediction targets, how closely the model tracks measurement, and the loop itself.

Platform

Experimental platform & characterization

Drawing on the first-class shared research platform of the Function Hub at HKUST (Guangzhou) and the Wilson Tang Brilliant Energy Science and Technology Lab (BEST Lab), the group has a complete experimental loop from materials synthesis to device validation.

Materials synthesis & purification

Gloveboxes, Schlenk lines and solvent purification systems support the controlled synthesis of precursors and functional molecules.

Device fabrication

Spin-coating, blade-coating and slot-die coating, covering small-area devices through to module-scale validation.

Optoelectronic characterization

J–V measurements, EQE, steady-state and time-resolved PL, transient absorption spectroscopy and other carrier-dynamics characterization.

Stability assessment

MPPT tracking, damp-heat ageing at 85 °C / 85% RH, and light and thermal cycling reliability tests.

Structure & composition analysis

XRD, SEM/TEM, XPS, NMR and liquid chromatography, supporting analysis from crystal structure to chemical reaction mechanisms.

Computing & data

First-principles calculations, molecular dynamics and machine-learning modelling, forming a cross-validation loop with experiment.

About the images: the research images on this page come from the group's published papers and their official press releases. Copyright belongs to HKUST (Guangzhou) and the respective journals; they are used for academic showcase purposes only.

Interested in one of these directions?

Whether your strength is synthesis, devices, characterization or computation, the group welcomes you to join and collaborate.

View open positions Browse publications