The cell that cannot keep a third of the sunlight
Assumes: Light arrives in lumps, and brightness only changes how many · The brightness no lens can increase
The best silicon solar cells made in laboratories convert about 27 per cent of sunlight into electricity. Commercial panels on roofs manage around 22. It is natural to assume that the gap between those numbers and 100 is a matter of engineering — of purer silicon, better contacts, less reflection — and that a perfect cell would capture almost everything. It would not. A perfect silicon cell, with no defects, no reflection and no resistance, in sunlight above the atmosphere, would reach about thirty per cent and stop.
That limit was worked out by William Shockley and Hans-Joachim Queisser in 1961, and the remarkable thing about it is how little it assumes. It needs the Sun’s spectrum, the cell’s temperature, a band gap, and two facts: that light is absorbed in whole photons, and that anything able to absorb light must also emit it. Everything else follows, and the limit it gives is a property of light, not of silicon.
One photon, one electron
Light arrives in lumps showed that light gives up its energy to electrons in whole photons, the energy of each set by its frequency alone. A semiconductor has a band gap, which what happens when the wells get close traced to the levels of its atoms spreading into bands as the atoms crowd together: a range of energies its electrons cannot have, between the filled states that bind them and the empty states in which they can carry current. A photon with less energy than the gap cannot lift an electron across it, and passes through. A photon with more lifts one electron, exactly one, into the conduction band, and the electron arrives there with whatever excess energy the photon had.
That excess does not last. An electron high in the conduction band collides with the vibrating lattice and sheds its extra energy as heat in a fraction of a picosecond, falling to the bottom of the band, where it waits, typically for microseconds, to be collected. Whatever the photon brought, the electron keeps one gap’s worth.
Sunlight’s spectrum is broad, spread from the infrared to the ultraviolet, and a single gap cuts it in two. Below silicon’s gap of 1.12 electronvolts lies 23 per cent of the Sun’s power, in the infrared, which silicon cannot absorb at all. Above it, each photon contributes one gap’s energy; the blue and ultraviolet photons, carrying two or three times the gap, waste most of what they bring. The shaded areas are the three shares. What is left for the cell to use, at best, is 44 per cent of the sunlight.
That is called the ultimate efficiency, and it is a pure consequence of photons. A material that absorbed light continuously, rather than in quanta, could in principle extract the whole of the energy; a photon cannot hand over part of itself and keep the rest for another electron. The flash the eye counts found the same indivisibility at the other end of the scale, in a retina that registers single photons: what a photon does to a molecule is all or nothing.
The gap that balances two losses
The choice of gap is a compromise between the two losses. A narrow gap absorbs nearly all of sunlight’s photons but keeps little of each one’s energy; a wide gap keeps most of each photon’s energy but lets most of the photons through.
The numbers along the curve show the trade directly. Germanium, with a gap of 0.67 electronvolts, absorbs 76 per cent of sunlight’s photons and could carry 77 milliamperes per square centimetre, but each electron is worth so little that its limit is 21 per cent. A material with a 2.5 electronvolt gap would keep a great deal from each photon it absorbs and absorb only a tenth of them, for 15 per cent. Between them the curve is broad and flat: anything from 1.1 to 1.45 electronvolts comes within half a percentage point of the peak, which is why silicon, gallium arsenide and cadmium telluride, chemically unlike one another, all sit at the top. The flatness also means that nothing about the exact value of a material’s gap can rescue a cell from the limit; the loss is built into using one gap at all.
The dashed curve, the ultimate efficiency, peaks near 1.1 electronvolts at 44 per cent. Silicon sits almost exactly on the peak, which is luck rather than design: silicon was used for solar cells because it was the semiconductor the electronics industry already knew how to make. But the dashed curve assumes that every electron delivers its full gap as voltage, and no cell does. The solid curve is the real limit, and it is lower by a third across the whole range.
Every absorber is an emitter
The second fact is the one that turns 44 into 30. The glow that says nothing about the surface followed Kirchhoff’s law: a body that absorbs light of some colour well must, at the same temperature, emit light of that colour just as well, or it could be used to drive heat from cold to hot. A solar cell is a strong absorber above its gap, so it is also an emitter above its gap. In the dark at room temperature its emission is minute, because a 300 K body emits almost nothing at the energy of near-infrared photons. Under illumination it emits more.
The reason is that the electrons pumped into the conduction band by sunlight can fall back across the gap and give out a photon. The rate at which they do so rises exponentially with the voltage across the cell, since voltage measures how far the population of excited electrons exceeds its dark value. At zero voltage, all the pumped electrons flow out as current. As the voltage rises, more of them recombine and radiate instead, until at the open-circuit voltage the recombination exactly balances the generation and no current flows. The cell’s current is therefore
with the flux of solar photons above the gap and the flux of photons the cell would emit above the gap as a blackbody at its own temperature . Power is current times voltage, and the best operating point lies where the product peaks, somewhat below the open-circuit voltage. Doing this calculation at every gap gives the solid curve, with its peak of 30.5 per cent at 1.26 electronvolts.
The surprising part is that the limit is set by emission, a process nobody wants and no engineer designs in. A cell made of perfect material with no defects at all loses its electrons by glowing, and its voltage is the voltage at which that glow carries away as many electrons as sunlight supplies. A better material cannot remove that loss; it can only remove the others, by making radiation the only way electrons recombine.
That has a consequence that sounds like a paradox. A cell that reaches the limit is one in which every recombination produces a photon and every photon produced escapes, so at open circuit the cell glows with the whole of the light it has absorbed in excess of what it would emit in the dark. The best possible solar cell is also the best possible light-emitting diode. The record-holding gallium arsenide cells are, measured as LEDs, extraordinarily efficient emitters, while a silicon cell, whose indirect gap makes it a feeble emitter, loses its electrons to defects and to Auger collisions instead and cannot approach the limit as closely. Holding a solar cell’s voltage up means getting its light out, which is the opposite of what intuition about a device meant to capture light suggests.
A count and a free energy
The power a cell delivers is current times voltage, and the two factors come from different places. The current is a count. Sunlight above the atmosphere brings photons per square metre per second with energies above silicon’s gap, and each makes one electron, so the most current a silicon cell can carry is that number times the electron’s charge, 53 milliamperes per square centimetre. No property of the material raises it: a photon below the gap is not counted and a photon far above it counts once. The current depends on the gap and the spectrum and nothing else, and the electrons that pay in fixed amounts is the same counting applied to electrons striking atoms rather than photons striking a crystal.
The voltage is a free energy. It measures how much each collected electron is worth, and a cell holding a voltage delivers per electron, which is always less than the gap. For an ideal silicon cell in sunlight above the atmosphere the open-circuit voltage is 0.882 volts, and the best operating point is at 0.792 volts, where the cell’s own emission is still small enough not to eat much of the current. The ratio of the power at that point to the product of the largest current and the largest voltage, called the fill factor, is 87 per cent. Multiply 53 milliamperes per square centimetre by 0.882 volts by 0.87, and divide by the 136 milliwatts per square centimetre that sunlight brings, and the result is the 30 per cent of the figure. Each of the three factors has a reason; none of them is a matter of engineering.
Why the voltage falls short of the gap
The open-circuit voltage is the clearest place to see what the second fact costs. A photon above the gap delivers its electron to the band edge, so it might seem the cell could hold a voltage equal to the gap. It cannot, and the shortfall has two parts that thermodynamics would recognise.
The first is a Carnot factor. Sunlight is radiation from a body at 5,772 kelvin, and a cell at 300 kelvin converting it is a heat engine running between those temperatures. In the limit where the cell absorbs a narrow band of photons at the gap, the voltage it can hold is the gap times , for silicon a loss of 58 millivolts. That is the factor the temperature an engine really takes its heat at found limiting any engine fed from a hot source, in a converter with no moving parts.
The second is larger and stranger. Sunlight arrives from a disc that fills one 46,000th of the cell’s sky. The cell’s own emission goes out into the whole hemisphere above it. So for every photon it receives from the Sun’s narrow cone it can emit into a cone 46,000 times larger, and a voltage at which the emission rate per unit solid angle matched the Sun’s would let the cell lose 46,000 times too many photons. The voltage must fall until emission into the whole hemisphere balances absorption from the small cone, and the cost is , 278 millivolts at room temperature. This is the dilution of sunlight paid for in volts. A smaller term, 76 millivolts, is returned because a hotter source supplies more photons above the gap in proportion to their exponential factor, and what remains is the detailed-balance voltage to within a few hundredths of a volt.
Buying back the solid angle
The brightness no lens can increase showed that no arrangement of lenses and mirrors can make the image of the Sun brighter than the Sun’s surface, but that it can make the Sun appear larger, up to the point where it fills the whole sky the receiver sees. The concentration factor that achieves this is of the Sun’s angular radius, 46,200, and it is the largest any optical system can reach in air.
Concentrating sunlight shrinks the second loss. Every factor of ten in concentration multiplies the photon flux tenfold without changing the cell’s emission into its hemisphere, and the voltage rises by , about sixty millivolts. At full concentration the étendue term disappears entirely, and the efficiency limit rises to 40.7 per cent. The best gap also moves lower, because with less voltage lost there is more value in collecting the infrared photons a wide gap discards. Concentrator cells under hundreds of suns are, for this reason, the most efficient solar cells ever made, though almost all of their gain comes from stacking several gaps rather than from concentration alone.
There is an equivalent way to buy back the solid angle without concentrating: restrict the angles into which the cell can emit. A cell behind a filter that reflects back everything outside the Sun’s cone would emit only into that cone and hold the same voltage as one under full concentration. The cone light has to find to get out showed how high-index materials already restrict their emission by total internal reflection, trapping light inside, and the most efficient single-junction cells, of gallium arsenide, exploit exactly that: by reflecting their own emitted photons back into the material to be reabsorbed, they recycle the emission and approach the radiative limit closely.
A hot cell gives up voltage
The cell’s temperature appears in two places: in the Carnot factor and, more importantly, in the dark emission rate. A warmer cell emits far more light above its gap, so its recombination current rises and the voltage at which it balances the solar generation falls.
The limit falls by about 0.045 percentage points per kelvin. A panel in full sun on a still summer day runs at fifty to seventy degrees Celsius, and its efficiency is measurably lower than on a cold bright day; real silicon panels lose about 0.3 to 0.4 per cent of their output per kelvin, a larger relative loss than the ideal figure because real cells’ recombination is dominated by defects whose rates also rise with temperature. The coldest bright places, high mountains in spring, are where panels give the most for the light they receive.
What the limit assumes
The calculation treats the Sun as a 5,772 kelvin blackbody seen from above the atmosphere. Real sunlight at the ground has passed through air and lost part of its infrared to water vapour and carbon dioxide, and the standard terrestrial spectrum gives a limit of 33.7 per cent at 1.34 electronvolts, slightly higher than here because the atmosphere removes more of the photons a cell would waste than of those it would use. The cell is treated as a perfect step absorber: every photon above the gap absorbed, none below. Real absorption edges have tails, which below the gap, where there is nothing to absorb traced to an exponential that extends over several decades, and those tails add slightly to the emission and lower the voltage. And the only recombination allowed is radiative; real cells also lose electrons to defects, at their surfaces, and by Auger processes in which one electron’s energy is handed to another, which in silicon, an indirect-gap semiconductor that radiates weakly, sets the practical limit nearer 29 per cent.
The deeper assumptions are the two the limit is built on: one gap, and one electron per photon. Relax either, and the limit changes. A stack of cells with different gaps, each taking the slice of the spectrum it suits, divides the thermalisation loss among them; with infinitely many gaps under full concentration the limit becomes 86.8 per cent, the efficiency of an ideal engine driven by sunlight. And a material in which a high-energy photon can make two electrons, by a process called singlet fission in some organic molecules and multiple-exciton generation in some nanocrystals, recovers part of the excess the single-gap cell throws away as heat.
Still open: whether one photon can be made to pay twice
Stacked cells are established; multi-junction cells under concentration have passed 47 per cent. They are expensive, and the cheaper route to beating the single-gap limit would be a single material that pays twice for a high-energy photon. Singlet fission has been demonstrated, and attached to silicon it has been shown to inject extra electrons; whether it can be made efficient and stable enough to raise the efficiency of a practical cell is still being worked out, as is whether hot electrons can be collected before they cool, which would bypass the thermalisation loss entirely. Both would leave the photon picture of the limit intact. They change what a photon can be made to do.
The habit worth carrying away is to find the loss that perfection does not remove. A single-gap cell keeps one gap’s energy from each absorbed photon and none from those below the gap, which caps it at 44 per cent of sunlight; and since every absorber also emits, its voltage falls short of the gap by a Carnot factor and by , the price of the Sun filling a 46,000th of its sky, bringing the limit to about 30 per cent at a 1.26 eV gap, recovered to 40.7 per cent only by concentrating the Sun until it fills the sky. A third of sunlight is lost before any engineering begins. The rest of the shortfall is the engineering.
Part 8 of 8
This essay is one argument about Photon. The others:
The objects named here
The third axis, after the field and the reading path: the things themselves, and every essay that touches each one.
Band gapBlackbody radiationCarnot efficiencyDetailed balanceEtenduePhoton
- The glow that carries a voltage band gap, detailed balance, photon
- The work a diluted beam will not do carnot efficiency, detailed balance, etendue
- The flicker that is half wave and half particle blackbody radiation, photon
- The product doping cannot move band gap, detailed balance