For decades, the guiding principle of solar cell engineering has been simple: cleaner is better. Remove every flaw, smooth every interface, and performance will follow.
At the High-Performance Computing Center Stuttgart, a team running simulations on Hawk — one of Europe’s most powerful supercomputers — set out to do exactly that. What they found in the data pointed somewhere else entirely.
Why solar cells still leave most sunlight on the table
Modern silicon solar panels convert roughly 22% of incoming sunlight into electricity. That number has climbed steadily over the past two decades, but it still leaves most of the sun’s energy unused. Understanding why requires a look at how silicon handles light.
When a photon strikes a silicon cell, it excites an electron, generating a small current. High-energy photons are the problem — violet light carries around 3 electron volts, far more than silicon can use. Silicon converts only about 1.1 eV into electricity. Everything else escapes as heat.
That heat loss is a double burden. It represents energy that was never captured, and it gradually degrades the panels themselves, shortening their working life.
Germany’s experience puts the scale of both progress and limitation in perspective. In 2000, solar supplied less than 1% of the country’s electricity. By 2022, that share had grown to roughly 11%, driven by subsidies and falling hardware costs. The growth is real — but so is the gap between what panels currently do and what physics suggests they could.
Tetracene: a molecule-thin layer that splits light differently
Researchers at the University of Paderborn, led by Prof. Wolf Gero Schmidt, have been exploring one promising path toward closing that gap. Their approach involves coating a silicon solar cell with a layer of tetracene — an organic semiconductor just one molecule thick.
Tetracene behaves differently from silicon when it absorbs a high-energy photon. Instead of wasting the surplus energy as heat, it undergoes a process called singlet fission. One high-energy exciton — a paired excited electron and the hole it leaves behind — splits into two lower-energy excitons. Those two excitons are then small enough for silicon to handle.
If they can be transferred across the tetracene–silicon interface, the underlying silicon layer can convert most of their energy into electricity. That is the key engineering challenge. In principle, Schmidt says, consistently applying singlet fission could boost solar cell efficiency by a factor of 1.4 — not a marginal gain, but a significant leap.
What the Hawk supercomputer revealed
To study how exciton transfer actually works at the interface, Schmidt’s team turned to Hawk, the supercomputer operated by the High-Performance Computing Center Stuttgart (HLRS). They ran ab initio molecular dynamics simulations — a computationally intensive method that tracks hundreds of atoms and their electrons simultaneously, advancing time in femtosecond steps.
The goal, initially, was straightforward: identify imperfections at the tetracene–silicon interface and find ways to eliminate them. A cleaner interface, the assumption went, would allow more efficient exciton transfer.
The simulations told a different story.
At the interface, some silicon atoms are not fully bonded to their neighbors. These incomplete bonds — called dangling bonds — are typically classified as defects, treated as sources of interference that degrade electronic performance. The AIMD simulations showed something else: silicon dangling bonds were actively helping excitons cross the interface. The team published the finding in Physical Review Letters.
Rethinking what a ‘defect’ means in semiconductor design
The result directly contradicted the prevailing direction of solar cell research, which has moved steadily toward cleaner interfaces, treating any deviation from atomic perfection as a problem to be solved.
Prof. Uwe Gerstmann, a collaborator on the project, argues the framing needs to change. He draws a parallel to donors and acceptors — deliberately introduced impurities already foundational to semiconductor electronics, the ones that make diodes and transistors possible. The idea that defects are inherently harmful, he suggests, is simply too narrow.
Dr. Marvin Krenz, the paper’s lead author, was direct about what the findings mean for the broader research community. The field has been working hard to remove defects at all costs. This paper suggests that approach may be leaving performance gains on the table.
The team’s new goal is to design interfaces that are, in their words, “perfectly imperfect.” Rather than eliminating dangling bonds, they want to place them deliberately — positioning them where they will do the most to encourage exciton transfer across the tetracene–silicon boundary.
What comes next for solar cell engineering
The immediate next step is more computing time. Future runs on high-performance systems will focus on mapping where dangling bonds should be placed to maximize and reliably reproduce the exciton transfer effect — not to produce a single breakthrough panel, but to build a systematic design framework.
Schmidt is measured in his expectations. The field has averaged roughly 1% annual efficiency improvement across solar cell architectures for decades, and he expects that pace to continue, with the dangling-bond insight contributing incrementally to it.
The collaboration between the University of Paderborn and Helmholtz Zentrum Berlin also signals something broader. Fundamental physics questions and applied engineering goals are increasingly being worked on together rather than in sequence — a structural shift in how the field organizes itself.
Perhaps the larger implication is about method. Supercomputer-driven simulation is proving capable of surfacing counterintuitive results that laboratory experiments, constrained by what researchers think to look for, might never find. As HPC resources grow more accessible, the capacity to be surprised by the data — and to follow where it leads — may become one of the most valuable tools in materials science.
Carlos is an engineer with strong expertise in technical and industrial topics. He previously worked at international companies such as Siemens and speaks Spanish, German, English, and Italian.
