Research

One question, three instruments.

The question never changes: exactly the bond you want, made cleanly, and understood. Synthesis asks it with photons and electrons; theory, from first principles; AI & automation, with machines that choose the next experiment.

The programme

Built to feed each other.

Most laboratories keep methodology, computation, and automation in separate rooms. Ours are wired in series: every reaction is a data point, every mechanism a model, every model the next experiment.

The circulation is concrete. Mechanisms computed in Direction II explain why Direction I’s mild couplings work at all; the descriptors they produce train the models of Direction III; and every automated run returns structured evidence that sharpens both. What follows is each direction in depth — and, at every joint, what it hands the next.

  1. Sustainable photo- & electrochemical synthesis Photons and electrons as traceless reagents; alternating current as a programmable reaction variable.
  2. DFT computation & mechanism Catalytic cycles, selectivity origins, and descriptors a model can read.
  3. AI & automated chemistry A dry – wet closed loop that chooses its own next experiment.

Method portfolio

Sustainable photo- & electrochemical synthesis.

Bond construction powered by the cleanest reagents chemistry has — photons and electrons. Five fields, one workshop: from dual-catalytic couplings to current studies of alternating-current reaction control.

Traditional coupling chemistry pays for its bonds with high temperatures and elaborate ligands. This programme pays with light and current instead: nickel-catalysed two- and three-component cross-couplings that run at room temperature under photochemical, electrochemical, and photoelectrochemical drive, assembling C – C and C – heteroatom bonds gently enough for the late-stage functionalization of complex molecules. Selectivity is engineered along complementary dimensions — photocatalyst triplet energy, ligand, energy-delivery platform, even the choice of electrode pair.

Photocatalysis · light/nickel dual catalysis

A visible-light photocatalyst converts photons into two currencies: single-electron transfer and triplet energy transfer. A nickel cycle spends them on bonds — aryl, alkyl, and heteroatom couplings that classical thermal catalysis reserves for harsher conditions.

In multicomponent settings the photocatalyst is a control dimension in its own right. Matching its triplet energy to the substrate switches trisubstituted olefins between their E and Z isomers — a stereochemical dial thermal chemistry does not offer, demonstrated in the group’s Nature Catalysis work and applied across cascade couplings since.

Figure slot Nat. Catal. 2019 — scheme: photocatalyst triplet energy switches E / Z

Electrocatalysis · nickelaelectrocatalysis

Electrochemistry makes the electrode the reagent: applied potential replaces stoichiometric oxidants and reductants, and paired electrolysis keeps whole couplings redox-neutral. Under nickelaelectrocatalysis, even weakly nucleophilic amines — anilines, sulfonamides, sulfoximines, carbamates, imines — undergo Buchwald–Hartwig-type arylation at room temperature.

The cell itself is programmable. Switching the electrode pair among graphite, nickel foam, zinc, and reticulated vitreous carbon steers 1,3-enynes into three distinct difunctionalizations; in collaborative work, the choice of electrolyte alone flips arenes between dearomative hydroalkylation and C(sp²)–H alkylation.

Figure slot Nat. Synth. 2023 — TOC: three products from three electrode pairs

Photoelectrocatalysis · where the two inputs meet

Photoelectrochemistry lets light and current share one cell: photons generate the radicals, electrodes turn over the metal cycle, and the two energy inputs are tuned independently. The group’s photo/electro/nickel triple catalysis delivered Z-selective olefins with no added photocatalyst — the platform’s starting point.

Its mature form treats the power supply itself as the instrument.

Figure slot Nat. Synth. 2025 — AC-PEC asymmetric coupling, up to 99% ee

Nickel-catalysed cross-coupling · multicomponent & stereodivergent

Beneath every platform runs the same workhorse: nickel. Sequential multicomponent reductive couplings assemble three partners in one operation, and ligand choice alone decides whether a 1,3-diene comes out trans or cis.

The platforms themselves turn out to be interchangeable evidence of one mechanistic core — one and the same coupling runs under thermal, photochemical, electrochemical, or mechanochemical drive.

Figure slot Angew. Chem. 2022 — TOC: one coupling, four energy platforms

Asymmetric catalysis · the hardest selectivity

The programme’s newest layer asks not which bond but which mirror image. Chiral nickel catalysis under alternating current delivered asymmetric sp³–sp² couplings of alcohols with up to 99% ee, applied to bioactive targets including a PDE2A inhibitor.

Asymmetric waveforms extend that control to reductive couplings: cathodic over-reduction and anodic metal deposition are suppressed while chemo-, regio-, E/Z and enantioselectivity are dialled independently. The same chemistry now delivers axially chiral scaffolds and phosphine ligands, scaled in a flow electrochemical cell built in-house.

Figure slot Nat. Commun. 2026 — TOC: waveform-controlled axial chirality
Reactions that work, waiting for their why. Next · Direction II — DFT computation & mechanism

The mechanism engine

DFT computation & mechanism.

Computation is the bridge between the group’s dry and wet halves — trained on the same chemistry it explains, and held to the bench’s standard: predictions must survive experiment.

A mechanism here is a working drawing, not a decoration for a paper’s last figure. DFT and wavefunction analysis map full catalytic cycles, rationalize what ligands and additives actually do, and trace chemo-, regio-, and stereoselectivity to their electronic origins — especially where the key intermediates are too fleeting to observe. The chemists running the calculations are the ones running the flasks.

Barriers, located — then removed

The founding computational result underwrites the whole synthesis programme: C–heteroatom reductive elimination from Ni(II) costs about 30 kcal/mol — impassable at room temperature. Oxidize the metal one step and the same elimination from Ni(III) costs about 5. Photocatalysis and electrochemistry are, mechanistically, machines for making Ni(III).

The same toolkit caught a barrier vanishing altogether: oxidative addition of alkyl halides to photoexcited Pd(0) proceeds without one, mapped jointly by experiment and computation in a cover-featured study.

reaction coordinate free energy 30 → 5 KCAL/MOL via Ni(II) · ~30 kcal/mol via Ni(III) · ~5 kcal/mol OXIDATION Ar–Ni–X C–X bond formed

Photo- or electrochemical oxidation to Ni(III) collapses the reductive-elimination barrier — the principle behind room-temperature coupling.

Prediction, then verification

Mechanism here is not post-hoc rationalization; it runs ahead of the bench. Full catalytic cycles are mapped until selectivity has an electronic address — how a photoredox–nickel–HAT triple cycle arylates the C–H bond beside nitrogen, and why the decatungstate–nickel pairing arylates but refuses to alkylate.

Where intermediates cannot be observed, computation proposes them: the olefin slow-release pool mechanism behind programmable ring-opening was first posited, and then validated, in silico. Calculations run both directions — explaining what the bench observed, and nominating what it should try next.

Figure slot computed transition state — the photoredox–nickel–HAT cycle, J. Am. Chem. Soc. 2020

From mechanisms to descriptors

Every mapped cycle leaves behind numbers: electronic, geometric, and steric descriptors that compress a molecule’s chemistry into something a model can read. They already earn their keep — a descriptor-driven campaign designed red-light photoredox catalysts predictively rather than by screening. Those descriptors are the currency the third direction trades in.

Mechanisms, compressed into numbers a model can read. Next · Direction III — AI & automated chemistry

Current focus

AI & automated chemistry.

Now building dry – wet closed loop Bayesian optimisation

The bottleneck of modern synthesis is no longer imagination: molecules we can design still wait months to be made. Our answer is a laboratory built to run, learn, and decide on its own.

We are building an AI- and automation-driven closed-loop laboratory — three systems behaving as one organism.

The brain

AI decision system

Bayesian optimisation and active learning weigh every result, and every descriptor computed next door, to propose the one experiment that carries the most information.

The nervous system

Laboratory operating system

One operating system schedules the stations, tracks every sample, and streams each measurement back as structured, machine-readable data. Nothing is transcribed; nothing is lost.

The hands & feet

Automated stations

Under inert atmosphere, robotic stations linked by AGV shuttles dispatch reagents, dose solids and liquids, run the reactions, and hand the crude to online analysis.

dispatchdosereactwork upanalyselearndecide

reaction modules photo · electro · thermal · high-pressure | online analysis GC · LC–MS

The playbook is proven. At KAUST the PI led the build-out of an automated synthesis-and-analysis platform with Chemspeed — dispensing, reaction, work-up, and analysis running as one workflow. At EIT the loop is being rebuilt at research-group scale, aimed at a condition space no one-variable-at-a-time habit could ever cover.

The point is not robots for their own sake. Machines make experiments reproducible, safe, and parallel; models make them informative. Chemists are left with the part machines cannot do — deciding which questions are worth a loop.

Published groundwork

Render slot the dry–wet laboratory, panoramic — stations, AGV line, photoreactor aglow
Render slot reaction module close-up — robotic hand-off at an electrochemical cell
The last step of every loop: back to the bench, with a better question. Around again · Direction I — synthesis

The loop keeps turning.

Explore the papers or current openings.

The publications page collects the papers cited above; the Join page lists current opportunities.

Join the group All publications