Research

Boosting Perovskite Solar Cell Efficiency Through Triplet-Triplet Annihilation Upconversion

A material can only absorb light with energy larger than its bandgap, and that material will be essentially transparent to most of the light below that energy. In the context of solar cells, this means that a very significant amount of light being emitted by the Sun is not being absorbed and turned into usable power.

This is where upconversion comes in. The idea is to put a layer behind the cell that takes two low-energy, sub-bandgap photons and combines their energy into one higher-energy photon that the perovskite can absorb. So instead of losing that infrared, you can convert and absorb part of it.

The problem is the silver at the bottom of our cell, and, as we all know, silver is reflective and opaque. This applies to the sub-bandgap photons as well. These low-energy photons would pass straight through our perovskite, but instead of going to the upconversion layer at the back, they're being blocked by the silver.

This motivates us to make solar cells with a transparent contact, and hence, the semitransparent perovskite solar cell. But another problem exists here: to deposit this transparent conductive oxide (TCO) to serve as our contact, we have to use a physical vapor deposition method called sputtering. This is a very energetic process that can damage the underlying solar cell layers, which prompts us to add in a buffer to protect our cells from sputtering damage. However, these layers can add electrical resistance to the cell, significantly impeding the performance of our cell.

My task is to:

• Find the optimal material and thickness for this buffer layer, such that it is thin enough to not add significant series resistance, but at the same time is able to protect the underlying layers from sputter damage.
• Optimize the sputtering recipe to ensure both material transparency and electrical conductivity.

Both opaque and semitransparent devices have silver as the final electrical contact. The difference is that silver is only deposited on the edges of the substrate, instead of covering the active area (and serving as the back reflector) in the opaque devices.

What I did

Fabricated 135 substrates, 109 of them full devices, and characterized their performance. This process involves glovebox spin coating (2PACz, perovskite), thermal evaporation (C₆₀, buffer layers, silver) and sputtering (IZO and ITO). I then used a solar simulator to characterize electrical performance, profilometry to check film thicknesses, photoluminescence to look for PL quenching, and UV-Vis to see the transparency of individual films and the whole stack.

What came out of it

• Working semitransparent perovskite solar cells!
• A two-step soft-sputter recipe, with a 37 W seed layer, then 100 W for the bulk layer, deposited directly onto the BCP. Replacing single-step IZO on an Sb₂O₃ buffer with this recipe on BCP cut median series resistance 72% and raised shunt resistance 3.4×, taking the best device from 10.9% to 13.9% PCE.
• A root-cause failure analysis of a batch that came out shunted, with 53 of 64 pixels shorted. A controlled rerun with bare-glass witness samples traced the recovery to tooling-factor and sputter-source differences; in that batch, a masking split then gave 8/8 passing pixels against 2/8 without.
• A full-factorial DOE crossing C₆₀ (5–25 nm) and Sb₂O₃ (5–15 nm) thickness. The best stack in the tested range was 25/5 nm. Furthermore, I tested various thicknesses for SnO₂ and BCP as other candidates for the buffer layer.
• Designed a new deposition mask in Onshape. After doing an experiment where we altered the factory-made deposition mask using Kapton tape and saw yield going from 25% to 100%, alongside higher shunt resistance, I designed a new mask for a more permanent (and better looking) solution than Kapton tape.

Latest run

In the previous soft-sputter run, thinner BCP did better: the best pixels were 13.9% at 6 nm, 12.8% at 10 nm and 10.7% at 15 nm. The latest run compared soft sputter on four buffers, with best pixels of 10.8% on 3 nm BCP, 10.4% on 5 nm SnO₂, 9.1% on 6 nm BCP and 6.6% on 5 nm Sb₂O₃, so SnO₂ clearly outperformed Sb₂O₃ at equal thickness and thin BCP stayed competitive. These results are provisional, as the run likely had three confounds: the BCP evaporation tooling factor was off by about 50%, the IZO target showed damage, and the opaque Ag controls also fell short of their usual ~18% (best 15.9%). Next, I'll rerun on a new IZO target.

Sputtered IZO on bare glass, patterned through a shadow mask. This is the only way we are able to confidently know how much IZO has been sputtered onto our devices.

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.

Effects of Interlayer Dielectric Thermal Conductivity and Scaling on the Wiedemann-Franz Law Approximation in Metal Interconnects by Electro-thermal FEM Simulation

What I'm doing

As CMOS technology evolves to scale down the size of individual devices, so has the thickness of the metal interconnects connecting them together. As their dimensions approach the electron mean free path (MFP), boundary scattering introduces an additional, thickness-dependent contribution to the total scattering rate, causing resistivity (ρ) to rise faster than thermal conductivity (κ) decreases and the Wiedemann-Franz (WF) law to break down.¹

To analyze the effect of scaling on electro-thermal properties of metals, we calculate the modified Lorenz number (L* = κρ/T) for Cu and Al thin films with thickness from 27 to 10,000 nm using data extracted from literature.^1–7 We found L* to be close to the Sommerfeld value at bulk thickness but several times larger in the nanometric regime, indicating that as interconnect dimensions approach their electron MFP, the proportionality between electrical and thermal conductivity provided by the WF law becomes less reliable.^1, 3

This result is derived from measurements on thin films, but thin films alone do not make up devices. Therefore, I taught myself COMSOL and built a coupled electro-thermal FEM model.

What started as thin-film measurements from literature turned into a simulation study of a multi-level back-end-of-line interconnect stack embedded in a range of dielectrics, swept from conventional oxide to higher thermal conductivity candidates. Each dielectric material case is solved twice, once with WF-derived κ and once with measured κ, to see how much that choice changes predicted self-heating.

This is a work in progress.

References

1. [1] Islam, M.R. et al. Evaluating size effects on the thermal conductivity and electron-phonon scattering rates of copper thin films for experimental validation of Matthiessen's rule. Nat. Commun. 15, 9167 (2024).
2. [2] Feng, B., Li, Z. & Zhang, X. Prediction of size effect on thermal conductivity of nanoscale metallic films. Thin Solid Films 517, 2803–2807 (2009).
3. [3] Cui, X., Hutt, D.A. & Conway, P.P. Evolution of microstructure and electrical conductivity of electroless copper deposits on a glass substrate. Thin Solid Films 520, 6095–6099 (2012).
4. [4] Karoui, A. Aluminum ultra thin film grown by physical vapor deposition for solar cell electric nanocontacts. ECS Trans. 41, 21–28 (2011).
5. [5] Boiko, B.T., Pugachev, A.T. & Bratsychin, V.M. Method for the determination of the thermophysical properties of evaporated thin films. Thin Solid Films 17, 157–161 (1973).
6. [6] Brandt, R. & Neuer, G. Electrical resistivity and thermal conductivity of pure aluminum and aluminum alloys up to and above the melting temperature. Int. J. Thermophys. 28, 1429–1446 (2007).
7. [7] Touloukian, Y.S., Powell, R., Ho, C. & Klemens, P. Thermophysical properties of matter, Vol. 1: Thermal conductivity — metallic elements and alloys (IFI/Plenum, 1970).