This page is about what happens inside one cell, at the level of photons, band structure and the V-I curve. If what you actually want is the path from the sun to the socket, panel to inverter to meter, that is the how solar panels work guide and it is the better place to start. Come back here when you want to know why silicon and not something else, and why a cell with a theoretical ceiling near 33% ships in modules at 22-24%.
The basic physics is the easy part. The hard part is how dozens of secondary effects (surface recombination, contact resistance, spectral mismatch, temperature coefficients) drag the theoretical 33% efficiency down to 22-24% in production, and the industry has spent decades chiseling them away one at a time.
The physics of solar is the easy part. The hard part is clawing back the 10 efficiency points that real-world losses steal from every silicon cell.
What Energy Does a Photon Need to Generate Electricity?
For silicon, a photon needs energy above 1.12 eV (the bandgap) to lift an electron from the valence band into the conduction band. Photons below that threshold pass straight through, useless for power. Photons above it are absorbed, but any excess energy beyond 1.12 eV converts to heat instead of electricity, a loss called thermalization. That 1.12 eV threshold corresponds to about 1,100 nanometres: UV and visible light absorb well but lose more to thermalization, near-infrared absorbs efficiently with minimal loss, and anything above 1,100 nm passes through unabsorbed.
Why silicon specifically? The solar spectrum at Earth's surface (AM 1.5G reference) peaks in the visible band around 500 nm. The Shockley-Queisser detailed balance analysis (1961) puts the optimal single-junction bandgap around 1.34 eV, where the limit peaks at 33.7% under AM 1.5G (Ruhle, 2016). The curve is flat near the top: silicon's 1.12 eV allows about 33%, and gallium arsenide (1.42 eV) allows about 33% too, which an integration of the ASTM G173 reference spectrum reproduces to within a few tenths of a point. Silicon won on economics rather than physics: abundant, well-understood, and endlessly scalable, while GaAs is rarer and harder to grow defect-free.
How Does the P-N Junction Separate Electron-Hole Pairs?
The P-N junction is the boundary between two silicon regions doped differently: phosphorus for N-type (extra electrons) and boron for P-type (holes). Electrons from the N-side diffuse into the P-side and holes diffuse the other way, leaving behind fixed ionized dopant atoms that create a permanent electric field, the "built-in field" of about 0.6-0.7V across roughly 0.5-2 micrometers. That region is the depletion zone. When a photon is absorbed inside or near it, the field splits the electron-hole pair: the electron accelerates toward the N-side, the hole toward the P-side. Further out, pairs can recombine before reaching the field.
In good mono-crystalline silicon, minority carrier diffusion length is comparable to or longer than the wafer, which is only about 120-140 micrometers thick in current production (ITRPV, reported by TaiyangNews), so most absorbed photons produce useful current. Multicrystalline silicon has grain boundaries that act as recombination sites and trim efficiency, which is a large part of why it has all but left the market.
One naming wrinkle worth knowing: most new cells are built on n-type wafers (phosphorus-doped base, boron-doped emitter), the reverse of the classic p-type PERC layout. ITRPV puts n-type at about 82% of wafers and TOPCon at roughly 75% of cell production in 2025. The junction physics is identical either way.
What Does the V-I Curve Tell You About a Cell?
The V-I curve plots voltage versus current under a set illumination. It's the fundamental characterization of a cell and the basis for everything an inverter does. A typical 400W panel under STC (1,000 W/m2, 25 deg C cell temperature) gives these key points:
| Parameter | Symbol | Typical value | What it means |
|---|---|---|---|
| Short-circuit current | Isc | 11-12A | Current when output is shorted, max possible current |
| Open-circuit voltage | Voc | 41-45V (60-cell panel) | Voltage with no load, max possible voltage |
| Maximum power voltage | Vmp | 34-38V | Voltage at maximum power point |
| Maximum power current | Imp | 10.5-11.5A | Current at maximum power point |
| Fill factor | FF | 0.78-0.84 | Imp x Vmp / (Isc x Voc), curve "squareness" |
The curve starts at Isc (zero voltage), stays roughly flat as voltage rises, then drops sharply near Voc. The maximum power point sits at the "knee" where the V and I product is largest, a "400W" nameplate means Vmp x Imp = Pmp at that point under STC. Fill factor for a good panel runs around 0.80, capturing 80% of the Isc x Voc rectangle. Temperature and irradiance both shift the curve: higher temperature mainly cuts Voc, why panels lose roughly 0.30%/deg C of power in TOPCon cells and about 0.35%/deg C in PERC, while lower irradiance cuts Isc proportionally without touching Voc much. See our TOPCon vs HJT vs PERC comparison for how cell technologies shift this curve.
How Does MPPT Find and Hold the Maximum Power Point?
Maximum Power Point Tracking is the inverter's algorithm for landing the string on the V-I curve's power peak. That peak drifts constantly with irradiance and temperature, so a fixed operating point bleeds energy. The most common algorithm is Perturb and Observe (P&O): measure V and I, compute power P = V x I, nudge the voltage 1-3%, then check whether power rose or fell, reversing if it fell. It oscillates around the peak, holding within 1-2% under stable light, though accuracy can drop to 95-97% under passing clouds. Fancier algorithms (Incremental Conductance, particle swarm) trade complexity for precision; SolarEdge HD-Wave and Enphase IQ8 claim 99%+ accuracy under fast-changing conditions.
Multiple MPPT channels matter for mixed orientations or shading, a residential string inverter usually has 2-3 channels, each tracking one string. Module-level electronics push this further: the SolarEdge P370 optimizer and Tigo TS4-A-O retrofit put a tiny MPPT controller on each panel, and the Enphase IQ8A microinverter converts DC to AC at each panel with its own MPPT. Is that worth it on every install? Not for an unshaded south-facing roof, where string-level MPPT captures nearly all available power. Module-level electronics pay back where shading or mismatch exist.
What Limits Real-World Cell Efficiency?
Work through the energy budget for a 1.12 eV cell under the AM 1.5G spectrum and the physics takes most of the sunlight before engineering gets a turn. Integrating the ASTM G173 reference spectrum reproduces the textbook numbers: below-bandgap photons carry about 19% of the incident energy and pass through unabsorbed, and thermalization (excess photon energy above 1.12 eV dumped as heat) costs another 32%. That leaves about 49% as energy at the bandgap, and the voltage a real junction can hold (a Voc well under the bandgap, set by radiative recombination) trims it to the Shockley-Queisser limit of roughly 33%. Silicon's own Auger recombination pulls the practical ceiling down to 29.4% (Richter et al., 2013).
The remaining gap to production is the engineering part: recombination at grain boundaries and unpassivated surfaces (TOPCon's tunnel oxide and HJT's passivation attack this), optical losses from reflection and busbar shading (anti-reflective coatings and texturing), resistive losses in fingers and ribbons, and temperature derating in operation, since summer roofs run 50-70 deg C, not the 25 deg C of STC.
The silicon cell record is 27.81% (LONGi's hybrid interdigitated back-contact cell, certified by ISFH in April 2025), already within about 1.6 points of the practical limit. Commercial modules run roughly 22-24% for volume TOPCon and HJT, with the best commercial silicon modules just under 25% (Fraunhofer ISE Photovoltaics Report, 2026 update). The remaining headroom for a single junction is small, so the next big step is stacking a second junction on top. Perovskite-silicon tandems passed 35% in a certified lab cell in 2026, a story covered in our piece on UV and the solar spectrum.
Summary
Solar panels make electricity by absorbing photons above silicon's 1.12 eV bandgap, separating the resulting electron-hole pairs across a P-N junction's built-in 0.6-0.7V field, and collecting the DC current through metallization on the front and back of the cell (NREL Photovoltaic Cell Physics). A typical 60-cell panel produces ~36V Vmp at 11A Imp, around 400W under STC. The inverter's MPPT continuously adjusts operating voltage to the V-I curve's power peak as conditions change, at roughly 97-99% efficiency. Real production efficiency runs 22-24% versus the roughly 33% Shockley-Queisser limit for silicon (29.4% practical), the gap split among below-bandgap photons, thermalization, recombination, optical losses, resistive losses, and temperature derating. For the materials and construction behind those cells, our piece on what solar panels are made of covers the build details.