Why the spectrum advice contradicts itself
Search this and you'll find two confident answers. Fixture makers say you need full spectrum — white light with far-red and UV, because red-and-blue alone gives you weak, stretched plants. Independent horticulture writers say plants mostly use red and blue and the rest is decoration. Both are stated as settled fact, and the split runs suspiciously close to who is selling a light.
The argument is settleable, because a photon's usefulness is two measurable things multiplied together: what it costs to emit, and how well the plant uses it. Nobody publishes that multiplication, so here it is.
A red photon is cheaper than a blue one
A photon's energy is inversely proportional to its wavelength, so a long, red photon takes less energy to produce than a short, blue one. That sets a hard ceiling on how many photons a watt can ever buy — physics, not engineering, and no fixture will ever beat it:
| Band | Peak | Ceiling, µmol/J | What it does |
|---|---|---|---|
| UV-A | 385 nm | 3.22 | Stress signaling — thicker leaves, more pigments and aromatics. Not photosynthesis. |
| Blue | 450 nm | 3.76 | Compact growth: opens stomata, thickens leaves, keeps internodes short. |
| Green | 550 nm | 4.60 | Penetrates the canopy — reaches leaves the red and blue never get to — and drives photosynthesis nearly as well as red. |
| Red | 660 nm | 5.52 | The workhorse: highest quantum yield and among the cheapest photons to make. |
| Far-red | 730 nm | 6.10 | Drives photosynthesis alongside PAR (the Emerson effect), triggers stem and leaf expansion, and signals flowering in daylength-sensitive crops. |
From E = hc/λ. Check it against something you already know: at 550 nm this gives 4.60 µmol/J, which is the familiar "1 watt of PAR is about 4.6 µmol per second" conversion.
Deep red at 660 nm tops out at 5.52 µmol/J and blue at 450 nm at 3.76 — red is 1.47× cheaper per photon before anyone builds anything.
So when two fixtures advertise different µmol/J figures, some of that gap is diode quality and some is simply how much red is in the mix. A manufacturer can raise its headline number by shifting the spectrum redder — the same trick as selling a smaller portion and quoting a lower price. It is not fraud, and it isn't quality either.
It also tells you how close the industry is to the wall. The best efficacy on our light calculator's reference list is 2.8 µmol/J, which is about 51% of the deep-red ceiling. There is real headroom left, but not the tenfold kind — anyone promising it is selling something.
Green is not the wasted color
The other half is biology. McCree measured photosynthetic response per absorbed photon across the spectrum, averaged over 22 species, and the resulting curve is the reference everyone cites and few reproduce:
| Wavelength | Relative quantum yield | Band |
|---|---|---|
| 400 nm | 0.60 | Blue |
| 450 nm | 0.72 | Blue |
| 500 nm | 0.71 | Green |
| 550 nm | 0.83 | Green |
| 600 nm | 1.00 | Red |
| 650 nm | 0.96 | Red |
| 700 nm | 0.60 | Red |
Normalized to 1.00 at the peak, which falls at 600 nm — in the orange-red, not the blue. Digitized from the published curve, so treat these as good to a couple of percent.
Notice where blue sits. At 450 nm the relative yield is 0.72; at 550 nm — the middle of the "wasted" green band — it is 0.83. Green light is better used than blue light, per photon absorbed. The reason leaves look green is that they reflect and transmit some green, but "some" is not "all", and the green that does get absorbed works nearly as hard as red.
Now multiply the two halves together — photons per joule × how well they're used — and the ranking that falls out is not the one you've been told:
| Band | Photons per joule | Quantum yield | Photosynthesis per joule |
|---|---|---|---|
| Red (660 nm) | 5.52 | 0.95 | 5.23 |
| Green (550 nm) | 4.60 | 0.83 | 3.82 |
| Blue (450 nm) | 3.76 | 0.72 | 2.71 |
A relative index — the physical ceiling times the quantum yield — for ranking colors, not for predicting what a fixture emits. UV and far-red are absent because they fall outside the range McCree measured, and guessing there would defeat the point.
Red leads, which nobody disputes. But green comes second, 1.41× better per joule than blue, and red is 1.93× better than blue — nearly double. Blue is the most expensive photon in the box and among the least efficiently used. You still need it, because it controls plant shape and without it plants stretch and go soft. You need it as a structural additive, not as half the spectrum.
Everything above is per photon delivered. Making green photons is a different problem: there's a well-known gap in LED physics between efficient blue diodes and efficient red ones, and green LEDs sit right in it at a fraction of the wall-plug efficiency of either. Efficient green LEDs essentially do not exist.
That's why white LEDs are built as blue diodes with a phosphor coating that converts some blue into everything else. You pay a conversion loss for that green — which is the real reason red-and-blue fixtures post better efficacy numbers than white ones, and it has nothing to do with plants preferring the colors.
The useful conclusion is narrower and more practical: the green in your white fixture is not being wasted, so do not pay extra to eliminate it, and do not believe a chart that draws green as a hole.
Two fixtures, two definitions of PAR
One more trap before you compare spec sheets. PAR has traditionally meant 400–700 nm. The extended definition, ePAR, runs 400–750 nm, because far-red photons that the old boundary excluded do drive photosynthesis when they land alongside ordinary PAR light — the Emerson enhancement effect, where adding far-red to a PAR background raises photosynthesis by about as much as adding more PAR would.
Far-red at 730 nm sits outside the traditional range and inside the extended one — it is, essentially, the entire disagreement between the two standards. Which means a fixture with far-red diodes posts a higher efficacy figure under ePAR than under PAR, for exactly the same hardware.
| Band | Counted by PAR (400–700) | Counted by ePAR (400–750) |
|---|---|---|
| UV-A (385 nm) | no | no |
| Blue (450 nm) | yes | yes |
| Green (550 nm) | yes | yes |
| Red (660 nm) | yes | yes |
| Far-red (730 nm) | no | yes |
UV is counted by neither, which is worth remembering when a fixture's marketing leans on it.
So: two efficacy numbers are only comparable if they were measured over the same range, and plenty of spec sheets don't say which they used. If one fixture advertises far-red diodes and a suspiciously good µmol/J, that is the first thing to check. The boundary is also fuzzier than it sounds — far-red LEDs aren't monochromatic, so a 711 nm diode puts a substantial share of its output back inside traditional PAR, and a 746 nm one puts a chunk beyond even ePAR.
Does the spectrum need to change by stage?
Less than you'll be told. The stage-by-stage spectrum charts — blue for veg, red for flower — descend from the era when that meant swapping a metal halide lamp for a high-pressure sodium one, because those were the two lamps available and they happened to differ that way. It was a description of the hardware, not a prescription for the plant, and it got copied forward into an age when one fixture does everything.
What survives the move to LED is narrower, and it's about shape rather than yield:
| Stage | What actually helps | What to ignore |
|---|---|---|
| Seedlings and clones | Enough blue to stop them reaching. Stretched seedlings are nearly always too little light, not the wrong color. | Any spectrum change. Fix the intensity first. |
| Vegetative | Blue keeps internodes short and leaves thick — the difference between a squat plant and a floppy one. | "Veg spectrum" premiums. A white fixture already has the blue. |
| Flower | Red carries the photosynthesis, and it's the cheapest photon, so a redder mix is genuinely more efficient here. | Switching fixtures. The gain is a few percent against the cost of a second light. |
| Daylength-sensitive crops | Far-red at the end of the day can pull flowering forward, and far-red at 730 nm is the signal the plant reads. | Far-red as a yield claim. It's a signal first and a photon second. |
The honest summary: blue controls shape, red does the work, far-red sends messages. None of it outranks having enough light in the first place.
This is also the answer to the purple panels. "Blurple" fixtures were red and blue diodes only, and they existed for exactly the reason this page has been circling — those are the two colors LEDs make efficiently, so dropping everything else flattered the efficacy number. They mostly died out because white diodes caught up, because you cannot diagnose a sick plant under purple light, and because the green they omitted turned out to be doing useful work. One practical leftover: a lux meter or phone app cannot read a blurple panel, since those are calibrated for something like daylight — a caveat the DLI calculator carries on its lux input for the same reason.
What to actually buy
- Buy photons first, spectrum second. The gap between too little light and enough light dwarfs every spectrum effect on this page. Size the fixture with the light calculator and confirm the daily total on the DLI calculator before you spend a minute comparing spectra.
- A white full-spectrum fixture in the 3.0 µmol/J class is the right default. Not because plants need white — they don't — but because it's efficient, it lets you see problems in real color, and the green you're "wasting" is working harder than the blue.
- Don't pay a premium for UV. It's outside both PAR definitions, the yield evidence is thin, and it damages tissue and eyes. Outside PAR entirely, damages tissue and eyes, and the evidence for yield is thin. Skip it until everything else is right.
- Treat far-red as a feature with a footnote. Genuinely useful, genuinely photosynthetic — and the single easiest way for a spec sheet to flatter itself. Check which range the efficacy was measured over.
- Ignore any chart showing green as a gap. It's drawn from chlorophyll absorption in a test tube, not from photosynthesis in a leaf, and the two are not the same measurement.
And if a plant looks wrong under a light you've already sized correctly, spectrum is near the bottom of the list of likely causes. Temperature and humidity move growth far more than a spectral tweak will — start with the VPD calculator and the humidity guide.
Sources: photon energy per wavelength is E = hc/λ, which yields λ ÷ 119.625 µmol/J and reproduces the standard 4.6 µmol/J at mid-PAR. The relative quantum yield curve is McCree's action spectrum (1972), averaged across 22 crop species and normalized to its peak, digitized here to the nearest few percent. PAR and ePAR ranges, the Emerson enhancement result and the note that far-red diodes emit across the boundary follow Apogee Instruments' published explanation of the 400–750 nm standard. Band roles are the consensus of horticultural lighting sources; the per-joule ranking is ours, and is a relative index rather than a fixture specification, for the reason given under the table. Reference efficacies are the same pinned list the light calculator uses.