Solar energy research has one problem that most people outside the field do not know about. The best performing solar cells today use lead. Lead is toxic, and once these panels are damaged or thrown away, that lead can leak into soil and water. As someone working in photovoltaic research, this bothered me enough to spend a long stretch of time trying to find a way around it.
That work turned into a published paper in Solar Energy Materials and Solar Cells, and I want to walk through how it actually came together.
The Starting Point
Silicon solar cells power most of the world right now, but they are expensive to manufacture at the efficiency levels we need. Perovskite solar cells became exciting because they are cheap to produce and can reach very high efficiency. The catch is that almost all high performing perovskite cells rely on lead as the main light absorbing material.
I worked with a team to explore a lead free alternative called methylammonium tin iodide, often written as CH3NH3SnI3. Tin can replace lead in this structure, but tin based cells have historically underperformed. Researchers before us had pushed efficiency into the mid twenties using different material combinations, but nobody had cracked 30% with this specific compound in a realistic, well optimized design.
Building the Simulation
We did not build a physical solar cell in a lab for this study. Instead we used a simulation tool called SCAPS-1D, which lets researchers model how a solar cell will behave based on the materials and layers you choose. This is common practice before committing to expensive physical fabrication, because it lets you test dozens of material combinations quickly and cheaply.
Our proposed structure stacked several layers together. On one side we used zinc sulfide as the electron transport layer instead of the more commonly used cadmium sulfide. Cadmium sulfide is toxic too, so replacing it fit our overall goal of keeping the design as clean as possible. On the other side we tested three different hole transport layers, and molybdenum trioxide came out ahead of the other two.
The reason zinc sulfide worked so well comes down to how the energy bands line up between layers. A bad alignment creates a barrier that traps electrons and causes them to recombine before they can generate usable current. Zinc sulfide created a smoother pathway for electrons compared to cadmium sulfide, which directly reduced that recombination loss.
Testing Every Variable We Could
Once we had a promising structure, we did not stop there. We systematically varied thickness of the perovskite layer, carrier density, defect density at the material interfaces, operating temperature, and the work function of the back metal contact. Each of these variables can quietly wreck a solar cell's performance if left unoptimized, so we tracked how each one shifted the output.
We also looked at charge recombination more closely using Nyquist plots and lattice mismatch calculations between layers, since a mismatch at the atomic level between two connected materials creates defects that hurt performance over time.
Bringing Machine Learning Into the Process
This is the part of the project I found most interesting. After running all these simulations, we had a large dataset showing how different physical parameters affected the final efficiency. Instead of just reporting the best result, we trained a linear regression machine learning model on this data to figure out which physical parameters actually mattered the most.
The model identified defect density as the single most influential factor, responsible for close to 29% of the variation in performance. That kind of finding is genuinely useful for other researchers, because it tells you where to focus your optimization effort instead of tweaking everything at once.
What We Achieved
After optimizing the structure, our best design reached an efficiency of 32.57%, with a fill factor of 87.58%, an open circuit voltage of 1.08 volts, and a short circuit current density of 34.36 mA per square centimeter. This put our lead free design ahead of most previously reported numerical results for this same tin based perovskite material.
The paper has since been cited by other groups working on similar lead free perovskite structures, which is a good sign that the approach is useful beyond just our own study.
Why This Work Matters
Solar technology only becomes truly sustainable if we stop trading one environmental problem for another. A highly efficient solar cell that still relies on toxic heavy metals is not a complete solution. This project showed that a lead free and cadmium free design can compete with, and in this case exceed, many previously reported efficiency numbers, without giving up on being clean and low cost.
Combining physics based simulation with machine learning also felt like a preview of how materials research is heading. Simulation tells you what works. Machine learning tells you why it works and where to look next. Using both together made our research process faster and more precise than relying on either one alone.
I'm Mirza Md Shakil, working in photovoltaic device research with a focus on lead free perovskite solar cells, numerical simulation, and machine learning applications in materials science.
Top comments (0)