A deep dive into solar power

Sol Where sunlight comes from, what the heck a photon is, and how a sheet of silicon turns one into electricity.

The light hitting a solar panel left the Sun eight minutes ago. But the energy it carries was made deep in the Sun's core, probably tens of thousands of years ago, when four hydrogen nuclei became one helium nucleus and a sliver of their mass disappeared. This page follows that energy the whole way, from the core to the socket.

4.26 million tonnes / s
of the Sun's mass turned into light every second
8 min 19 s
for light to cross 150 million km to Earth
1,361 W/m²
of sunlight arriving at the top of Earth's atmosphere
4.3 × 1021
photons landing on each square metre of ground per second, in full sun
01 · The map

A 700,000-kilometre trip, in five stops

The Sun is a ball of mostly hydrogen, about 109 times Earth's width, held together by its own gravity. That gravity squeezes the centre hard enough to make it hot, and the heat is what makes it shine. Energy is made in one small region at the middle and then has to get out. How it gets out changes on the way.

Click a layer, or step through them in order.

02 · Where the energy comes from

Four hydrogens in, one helium out, and some mass goes missing

The core of the Sun is about 15.7 million °C and roughly 150 times denser than water. At that heat, atoms are stripped bare. What's left is a soup of loose electrons and bare nuclei, and for hydrogen the bare nucleus is a single proton.

Protons are positively charged, so they repel each other. Every so often, two of them hit hard enough to fuse anyway. That starts a three-step chain, the proton–proton chain, which ends with four protons becoming one helium-4 nucleus. Step through it:

Step 1 of 3
proton neutron positron (e⁺) neutrino (ν) gamma-ray photon (γ)

The missing mass

Now weigh everything. Put four hydrogen atoms on one side of a scale and the helium atom they turn into on the other, and the helium side comes up light:

4 × hydrogen-1 (1.007825 u each)4.031300 u 1 × helium-44.002603 u missing0.028697 u · 0.71%

That 0.71% didn't go anywhere. It became energy, 26.7 million electron-volts per helium, by Einstein's E = mc². The c² is the whole story. The speed of light squared is about 9 × 1016, so a tiny amount of mass buys an enormous amount of energy. Pick an amount of hydrogen:

Fuse

Why the Sun doesn't just explode

At 15.7 million degrees, a typical proton has about 1.4 keV of kinetic energy. Pushing two of them into contact against their electric repulsion takes something like a thousand times more. By classical physics, fusion in the Sun should essentially never happen. It happens because protons are quantum objects. Every so often one tunnels through a barrier it has no business crossing.

Even then, step one also needs the weak nuclear force to turn a proton into a neutron at the moment of collision, which almost never works out. A given proton waits, on average, roughly ten billion years for its turn. That slowness is why the Sun lasts. Averaged over the core, it puts out only about 280 watts per cubic metre, less than the ~1,500 W/m³ your resting body gives off. The Sun is bright because it's enormous, not because it's intense.

03 · Getting out

Light takes the long way out

Fusion's energy starts as gamma rays and ends up as X-ray light in the core. The core is so dense that light gets only a tiny distance, often quoted as around a millimetre, before an electron or ion absorbs it or bounces it in a new, random direction. Then the same thing happens again, and again.

That's a random walk, and random walks are painfully slow. After N steps you are typically only √N steps from where you began. Travelling a distance of R steps takes about R² of them. Watch one:

One walker, 40 step-lengths from the edge

Scale that up to the Sun. From the centre to the bottom of the convective zone is about 490,000 km. With steps a millimetre long, that's roughly (4.9 × 1011)² ≈ 2 × 1023 steps. The trip takes tens of thousands of years, even though each step happens at the speed of light. The answer depends heavily on the step length, and the step length varies a lot with depth. Drag it:

1 mm

Estimate: t ≈ R² / (ℓ·c) with R = 0.7 solar radii. Detailed solar models, which track how density and opacity change with depth, give answers from roughly 10,000 to 170,000 years. Nobody can measure it directly, but every estimate lands in "a very long time".

It isn't the same photon

"A photon takes 100,000 years to escape" is a handy shorthand, but no single photon survives the trip. Each is absorbed and a new one emitted, and as the gas cools outward the energy gets split into more, weaker photons. One X-ray photon's worth of energy from the core leaves the surface as a few thousand photons of visible and infrared light. What really makes the journey is the energy.

One thing does get out immediately: neutrinos, made in step one of the chain. They barely interact with anything, so they cross the Sun in about two seconds and reach Earth eight minutes later. Around 60 billion of them pass through every square centimetre of you each second, day and night, straight through the planet.

04 · The main question

So what the heck is a photon?

Start with light as a wave. Light is a ripple in the electromagnetic field, the same field that makes magnets stick and lets radios work. Electric and magnetic fields keep regenerating each other and move through empty space at 299,792 km/s. The distance between crests is the wavelength. For visible light it's 380–750 nanometres, about a hundredth the width of a hair. Our eyes read the wavelength as colour.

Then comes the strange part. That wave can't carry just any amount of energy. It gains and loses energy only in fixed-size packets, and one packet is a photon. Packet size is set by frequency: E = h·f, where h is Planck's constant. Shorter wavelength means higher frequency, and that means bigger packets.

One photon's worth of light
550 nm

A photon has no mass and no size you could draw, and it can't slow down or sit still. It still carries energy and momentum, so sunlight pushes very gently on whatever it hits. Solar sails ride that push. It's also the basic unit of the story here. A solar cell doesn't respond to "brightness" in the abstract. It responds to photons, one at a time.

Wave or particle? Watch it happen

The best way to see what a photon is: send light through two narrow slits, one photon at a time, at a screen that records where each one lands. Each photon arrives as a single dot in a single place, like a particle. But let thousands pile up and they form stripes. That interference pattern only happens with waves passing through both slits and overlapping.

The double-slit experiment, one photon at a time
0 photons
Rate Which slit?

The honest answer

A photon is not a tiny ball, and it's not a wave in the ordinary sense either. It's the smallest possible change in the electromagnetic field at a given frequency. It spreads and interferes like a wave while it travels, but it deposits all its energy in one spot, all at once, when it's absorbed. The theory that describes this, quantum electrodynamics, is one of the most precisely tested ideas in science. It still doesn't offer a picture you can hold in your head. Physicists mostly stopped expecting one.

For solar power, one fact matters most: energy arrives in packets, and the packet size depends on colour. The rest of the story follows from that. For where visible light sits among radio, X-rays and the rest, see the electromagnetic spectrum.

05 · After eight minutes

What actually reaches the ground

The photosphere, the Sun's visible surface, is about 5,500 °C. Anything that hot glows with a broad, predictable spread of colours called a black-body spectrum. The Sun's peaks in green-blue and runs well into the infrared. The atmosphere then takes a cut. Ozone blocks almost all ultraviolet below 300 nm, and water vapour, oxygen and carbon dioxide each absorb particular bands in the infrared.

Sunlight, by wavelength
W/m² per nanometre · ASTM G173 reference spectra
Top of the atmosphere 1,348 W/m² At the ground, sun ~48° from overhead 1,000 W/m²
Show as a table

Two numbers are worth remembering. About 1,361 W/m² arrives above the atmosphere. That's the measured "solar constant"; the reference spectrum above sums to slightly less. About 1,000 W/m² reaches a sunny surface. That second number is the one solar panels are rated against. A "400-watt panel" makes 400 W at 1,000 W/m² and 25 °C.

Look at the ground curve again, because it's the menu a solar cell orders from. Only about 43% of the energy, and just 28% of the photons, are in light you can see. More than half the power arrives as infrared you can feel but not see.

06 · Photon → electron

Inside a solar cell: one photon, one electron

A solar cell is a thin wafer of semiconductor, usually silicon. In silicon, electrons can hold low energies (the valence band, where they're stuck holding atoms together) or higher energies (the conduction band, where they're free to move). Between the two is a forbidden range with no allowed states, the band gap. For silicon it's 1.12 electron-volts.

A photon with at least that much energy can lift one electron across the gap. That leaves a "hole" behind, which acts like a positive charge. The cell is built as two layers, p-type and n-type, doped with trace impurities. Where they meet, a permanent electric field forms that pushes the freed electron one way and the hole the other. Connect a wire and the electron goes round the circuit. That's the current.

Fire a photon into the junction
Photon

Band diagram: height is electron energy. Electrons roll downhill here; holes float uphill. The band gap below follows the material chosen in the next panel.

That animation shows both of the big losses. A photon below the gap passes straight through and is wasted. A photon above the gap frees exactly one electron no matter how much extra energy it brought, and the extra is lost as heat within about a trillionth of a second. A blue photon with 2.7 eV and an infrared one with 1.2 eV do the same job.

So the band gap is a trade-off. A small gap catches almost every photon but keeps only a little energy from each. A large gap keeps more from each photon but misses most of them. Slide it and see:

Material
1.12 eV
Where the ground-level sunlight's energy goes
…split by wavelength
Best possible efficiency vs band gap (Shockley–Queisser limit)

Computed live from the ground spectrum above, for an ideal cell at 25 °C. "Voltage & re-emission" is the part the simple one-photon-one-electron picture misses: a working cell's voltage always sits somewhat below its band gap, and a lit cell glows a little, giving some energy back as light. Click the right-hand chart to set the gap.

That ceiling, about 33.7% at a gap near 1.34 eV, is the Shockley–Queisser limit, worked out in 1961. Silicon's 1.12 eV gap is close to the best, with a theoretical ceiling around 33% (a bit lower, ~29%, once silicon's own quirks are counted). Real devices come close:

Typical rooftop silicon panel (module)20–23%
Best lab silicon cell≈ 27–28%
Perovskite-on-silicon tandem, lab (two gaps stacked)≈ 34–35%
Four-junction cell under concentrated sunlight, lab≈ 47%

The way past the ceiling is to stack cells with different gaps. A wide-gap top layer takes the high-energy photons, keeping more of their energy, and lets the rest through to a narrower-gap layer underneath. That's what tandem and multi-junction cells are, and why "perovskite" keeps turning up in solar news. Perovskites are crystals whose band gap can be tuned, which makes them a good top layer for silicon.

07 · Zooming back out

There is an absurd amount of it

Earth is a disc about 12,700 km wide catching sunlight at 1,361 W/m². That's about 173,000 terawatts, continuously. Humanity's entire energy use, including oil, gas, coal, nuclear, hydro, wood and everything else, averages about 20 terawatts. Sunlight delivers roughly nine thousand times that. In about an hour, sunlight brings Earth as much energy as the world uses in a year.

So how much ground would panels need? Choose a target:

Power
21%
220 W/m²

Panel area only. Real solar farms space rows apart and need roughly 2–3× more land, and storage for night-time is a separate problem. "All world energy" also overstates the need: much of that figure is waste heat from burning fuel, which electricity doesn't have to replace.

Almost everything is sunlight

Solar panels are the most direct way to use sunlight, but not the only one. Most of the energy we use came from the Sun by some route:

Food and woodPhotosynthesis. Plants catch photons with chlorophyll, using a similar trick of one photon lifting one electron, and store the energy in sugar.
Oil, gas and coalAncient photosynthesis, buried and cooked for millions of years. Burning it releases sunlight from the distant past.
Wind and wavesThe Sun heats the equator more than the poles and land more than sea. Air moves to even things out.
HydropowerSunlight evaporates seawater, lifts it into clouds and drops it on mountains. A dam spends that height.
Not sunlight: nuclear, geothermal, tidesUranium was forged in exploding and colliding stars before the Sun existed. Geothermal heat comes largely from radioactive decay inside Earth. Tides are driven by the Moon's gravity.

And the Sun is only halfway through. It has burned about half the hydrogen in its core in 4.6 billion years and has roughly five billion to go. That's slow enough that, for every practical purpose, the only real limit on solar power is how well we catch it.

Notes

Where the numbers come from

  1. Solar luminosity (3.828 × 1026 W), radius (695,700 km) and total solar irradiance (1,361 W/m²): IAU 2015 Resolution B3, nominal solar values. Mass loss is L/c² ≈ 4.26 × 109 kg/s. Light travel time is 1 au / c = 499 s.
  2. Core temperature (~15.7 million K), density (~150 g/cm³), photospheric temperature (5,772 K): NASA Sun Fact Sheet and IAU nominal values. Core power density ~276 W/m³ from the Standard Solar Model; resting human ≈ 100 W over ≈ 0.066 m³.
  3. Atomic masses of ¹H (1.00782503 u) and ⁴He (4.00260325 u): AME2020 Atomic Mass Evaluation. 0.028697 u × 931.494 MeV/u = 26.73 MeV. Step energies of the ppI chain and the ~2% neutrino share: standard nuclear astrophysics references (e.g. Adelberger et al., Rev. Mod. Phys. 83, 195, 2011).
  4. Photon escape time: Mitalas & Sills, "On the photon diffusion time scale for the Sun", ApJ 401, 759 (1992), ~170,000 yr. Lower estimates of ~104 yr come from cruder uniform-opacity models. The slider uses the simple estimate t = R²/(ℓc).
  5. Solar neutrino flux at Earth (~6.5 × 1010 cm⁻² s⁻¹, dominated by pp neutrinos): Standard Solar Model predictions confirmed by Borexino.
  6. Spectra: ASTM G173-03 reference spectra (extraterrestrial and AM1.5 global tilt, 37° tilted surface), as distributed with pvlib-python. Resampled to 5 nm bins (280–1200 nm), 10 nm (1200–2500 nm) and 25 nm (2500–4000 nm), preserving integrated power. Photon counts and band shares are computed from this data in your browser.
  7. Detailed-balance efficiency: Shockley & Queisser, J. Appl. Phys. 32, 510 (1961). Computed here for a 298 K cell under AM1.5G with the Boltzmann approximation for the cell's own emission, matching the tabulation in Rühle, Solar Energy 130, 139 (2016): 33.7% at 1.34 eV. The ~29% silicon-specific limit includes Auger recombination (Niewelt et al., Sol. Energy Mater. Sol. Cells 235, 111467, 2022).
  8. Record efficiencies: NREL Best Research-Cell Efficiency Chart and the Green et al. "Solar cell efficiency tables" series in Progress in Photovoltaics, as of 2025. Figures rounded because records keep moving.
  9. World primary energy ≈ 620 EJ/yr (≈ 19.7 TW): Energy Institute Statistical Review of World Energy 2024. World electricity ≈ 31,000 TWh/yr (≈ 3.5 TW): Ember Global Electricity Review. US electricity ≈ 4,300 TWh/yr: EIA. Earth's cross-section uses a mean radius of 6,371 km; Sahara ≈ 9.2 million km².