Research

Semitransparent perovskite cells for upconversion

A silicon or perovskite cell throws away every photon below its bandgap. Triplet–triplet annihilation upconversion recovers some of them by combining two low-energy photons into one the cell can absorb — but the upconverter has to sit behind the cell, which means the light has to get through the back contact first. A normal device has a hundred nanometres of evaporated silver there.

So the back contact has to become transparent, and the only practical way to do that is sputtering a conducting oxide. Sputtering is violent: energetic species from the plasma reach the organic layers underneath and damage them long before you have a usable film. That conflict is the whole problem.

Silver stays in both. On the semitransparent devices it is patterned at the edge as a contact rather than covering the active area. Sb₂O₃ and SnO₂ buffer routes between C₆₀ and the oxide are still being explored.

What I did

Fabricated 120+ substrates end to end — glovebox spin coating through thermal evaporation — then sputtered IZO and ITO top contacts on the Lesker, with thickness on the Dektak and transmission on the Cary 6000i.

What came out of it

• The group's first working semitransparent cells.
• A two-step soft-sputter recipe — 37 W seed, then 100 W bulk, deposited directly onto the BCP — that cut median series resistance 4× and raised shunt resistance 2.7×, taking the best device from 10.6% to 13.9% PCE at over 80% transmittance beyond 750 nm.
• A root-cause failure analysis of a batch that came out shorted. Bare-glass witness samples through four-point probe, UV-Vis and photoluminescence traced the yield limiter to evaporation-mask step coverage; a masking split then gave 8/8 working devices against 4/8 without.
• A full-factorial DOE crossing C₆₀ (5–25 nm) and Sb₂O₃ (5–15 nm) thickness. Best stack in the tested range was 25/5 nm — at a corner of the design space, so the true optimum may sit outside it — with remaining losses attributed to series resistance.

Sputtered IZO on bare glass, patterned through a shadow mask. Test pieces like these are how a recipe gets characterised before it goes anywhere near a device.

Poster

Fabrication and Optimization of Semitransparent Perovskite Solar Cells for Upconversion Integration — Tung D. Nguyen, Tyler K. Colenbrander, Daniel N. Congreve. Presented at the Stanford EE REU symposium, 2026.

Electro-thermal transport in scaled interconnects

The Wiedemann–Franz law ties a metal's thermal conductivity to its electrical conductivity, and almost every compact thermal model of on-chip wiring leans on it. In heavily scaled lines there is good reason to expect it to fail. What we want to know is whether that failure actually changes anything once the wire is sitting in a real back-end stack, rather than modelled on its own.

What I'm doing

Computing Lorenz numbers from thermal conductivity and resistivity across films from 27 nm to 10 µm, and building a coupled electro-thermal finite-element study in COMSOL — electric currents, heat transfer and Joule heating together — to see how much the deviation moves predicted self-heating. I taught myself COMSOL for this project.

This is a work in progress. A technical report is in preparation.

Where it goes next

Extending the model to Al/WSe₂ contacts, and from there toward AlN, other two-dimensional materials, and topological semimetals — the candidate materials for interconnects and contacts once copper stops scaling well.