Two shelves in one room. Lamps of equal output, 255 µmol/m²/s on both. Same seed, same substrate, same ten days. One difference: the second shelf ran almost twice as much blue in its spectrum.
Anthocyanins jumped 78 percent. Yield came in at 1990 g/m² against 1980. That is a 0.5 percent gap, and the statistics cannot see it.
Worth keeping that pair of numbers in mind before your next shopping trip for a grow light.
What was actually compared
The study came out of Andrea Ertani’s group at the University of Turin and was published on April 27, 2026 in Frontiers in Plant Science. They grew garden cress on hemp fiber mats in 50 x 30 cm trays, 35 g of seed per tray. Darkness until germination, then a 14 hour light and 10 hour dark photoperiod at 72-77 °F (22-25 °C) and 50-70 percent humidity. Nutrient solution went in only for the last three days before harvest; before that, plain water. Harvest on day ten, cut about 5 mm above the substrate.
There were two treatments, and they differed in the ratio of channels: blue, green, red, far red.
| Treatment | Blue | Green | Red | Far red |
|---|---|---|---|---|
| Shelf 1 | 13 | 15 | 61 | 11 |
| Shelf 2 | 24 | 12 | 56 | 8 |
An honest correction belongs here. Blue rose from 13 to 24, but far red fell from 11 to 8 at the same time, green from 15 to 12, and red from 61 to 56. The shares always add to one hundred, so moving one channel in isolation is impossible. The accurate phrasing is this: two spectra were compared, one of them bluer. Pinning the whole result on blue is more than the design supports, and the authors do not do it.
The scale noticed nothing
| Measure | Shelf 1 | Shelf 2 | Significant? |
|---|---|---|---|
| Height, cm | 4.12 | 4.24 | no |
| Fresh weight, g per tray | 94.88 | 97.77 | no |
| Yield, g/m² | 1990 | 1980 | no |
| Dry weight, g | 5.14 | 6.18 | no |
| Dry matter, % | 5.42 | 6.34 | no |
Not one significant shift. Dry weight in that table differs by 20 percent, 5.14 against 6.18, and still falls short of the threshold: there were three replicates per treatment, and the spread between trays swallows the difference.
Height deserves a separate note. Far red on the second shelf dropped from 11 to 8 percent, a third of its share. That channel is the one usually responsible for stretching, so less of it should mean a shorter plant. Instead the numbers read 4.12 against 4.24 cm, no difference. Over ten days on cress the effect never had time to show.
The authors add a caution of their own about dry matter. The shift from 5.42 to 6.34 percent may reflect a change in tissue composition, meaning hydration rather than extra biomass. They say so plainly, and with three replicates that is the right level of care.
Color and chemistry moved a lot
| Measure | Shelf 1 | Shelf 2 | Change |
|---|---|---|---|
| Anthocyanins, mg per 100 g fresh weight | 4.20 | 7.46 | +78% |
| Phenolic index | 13.78 | 20.99 | +52% |
| Total chlorophyll, µg/mg | 3.97 | 3.64 | no difference |
| Carotenoids | below the detection limit in both treatments | ||
Anthocyanins and phenolics rose significantly. The absolute numbers matter more here than the percentages: 4.20 against 7.46 mg per 100 g. This is the pigment that gives the purple cast, and a gap that size is visible to the eye without any instrument.
Nitrate stayed put at 10.17 and 10.56 mg/kg fresh weight. Both figures are tiny, and the spectrum did nothing to them. If that topic interests you, we cover it separately in nitrates in vegetables.
Inside the plant, 787 proteins were rearranged
The most interesting part of the paper sits in the proteomics. The team identified 3,161 proteins and checked how many changed in abundance between the shelves. The count came to 787: 151 went up, 636 went down.
The steepest drops landed on the photosynthetic machinery itself. The large subunit of Rubisco, the photosystem I apoprotein and the photosystem II D2 protein all fell by a factor of one hundred. Enzymes for anthocyanin synthesis and stress response proteins rose in parallel.
Now set that against the table above. Chlorophyll unchanged, carbon unchanged, nitrogen unchanged, C/N unchanged, weight unchanged. The plant ran an internal overhaul across hundreds of proteins, shifted resources out of photosynthesis and into defense, and walked out of it weighing the same.
The bluer spectrum made the plant work differently. It did not change how much of it there was.
Why a bluer spectrum raises anthocyanins
Plants read light through separate receptors tuned to separate parts of the spectrum. Cryptochromes handle blue; phytochromes handle red and far red. A change in the proportion between channels registers as a change in circumstances, and the plant retunes its metabolism to match.
The proteomic data show where the resources went. Under the bluer shelf, dihydroflavonol reductase rose, the key enzyme in anthocyanin synthesis. Enzymes on the neighboring branch of the same phenylpropanoid pathway, the one leading to lignin, fell at the same time. Both branches draw on shared precursors, and the plant redirected them toward pigment.
The second trace is stranger: the cryptochrome receptor itself declined. This is a known effect, documented in Arabidopsis back in the late 1990s. Under blue light at 20 to 30 µmol, CRY2 protein falls more than tenfold within an hour. The receptor works and gets consumed.
The general blue light response is described the same way throughout the literature: the plant raises stress tolerance and cuts its investment in photosynthesis and growth. The first half of that bargain produced the anthocyanins. The second half should have cost weight, and at 255 µmol it did not. There was enough light for both shelves.
So maybe intensity is the lever
A reasonable objection: Ertani held output constant, so of course nothing moved. Weight comes from the amount of light, not its color. Let us check that too.
A Chilean group led by Mónica Flores ran the mirror experiment in 2024. One spectrum throughout, three intensities: 120, 160 and 210 µmol/m²/s. Eight subjects, made up of cabbage, kale, mizuna and mustard, each in a green and a red cultivar.
| Crop and cultivar | 120 µmol | 160 µmol | 210 µmol | Verdict |
|---|---|---|---|---|
| Cabbage, green | 384 | 349 | 386 | no difference |
| Cabbage, red | 510 | 433 | 483 | less under stronger light |
| Kale, green | 621 | 609 | 584 | no difference |
| Kale, red | 632 | 637 | 580 | less under stronger light |
| Mizuna, green | 417 | 417 | 426 | no difference |
| Mizuna, red | 346 | 343 | 339 | no difference |
| Mustard, green | 497 | 542 | 428 | less under stronger light |
| Mustard, red | 556 | 391 | 574 | more under stronger light |
The tally: one cultivar out of eight gained weight from the extra light. Three lost. Four did not react. Output went up 75 percent, and the scale registered it in exactly one case.
Does this mean light does not matter
No, and the line needs drawing carefully. Both experiments worked in a range where light was already sufficient, from 120 to 255 µmol/m²/s. Inside that corridor, adding output or changing color barely touches weight.
Below the corridor the picture changes. On a windowsill in November or under a weak household lamp, microgreens stretch, pale and fall over, and every added micromole pays. The only question is where the paying stretch ends. Judging by these two papers it ends earlier than most people assume, and certainly before 255 µmol. Telling this kind of light shortage apart from an overlong blackout or excess heat once the stems are already falling over is the subject of the wiki on why microgreens get leggy.
So the advice to go brighter and bluer works for anyone currently sitting in the dark. For everyone else it is an expense. If a fixture does need replacing, compare photons per joule rather than watts; the wiki on PPE and grow light efficiency shows how to run that number.
What more light does do
Two things, both of them practical.
Tissue gets denser. Dry matter percentage climbs reliably under high intensity. Green mizuna at 210 µmol came in 50.6 percent higher than at 120. Same direction as the non-significant shift in Ertani’s data: more light makes the tissue denser rather than heavier.
The plant gets shorter. Weak light stretches the hypocotyl in every species tested except red kale. Green mustard measured 6.6 cm at 120 µmol and 3.1 cm at 210, a gap of 113 percent.
Which leads somewhere counterintuitive for a grower. Under insufficient light, microgreens look better. They are taller, softer, and the tray looks fuller. The weight is the same or higher, the tissue is simply thinner. The eye misleads in both directions here, and the scale is what to trust.
Where the weight actually lives
Neither experiment varied seeding density. But the seeding rate in each can be worked out from their own methods sections, and the spread is telling.
| Ertani, 2026 | Flores, 2024 | |
|---|---|---|
| Tray | 50 x 30 cm | 64 x 35 cm |
| Seed per tray | 35 g | 8 g |
| Seeding rate | 233 g/m² | 36 g/m² |
| Yield | about 1985 g/m² | 340-640 g/m² |
Different species, different substrate, different timing, so this is no controlled comparison and no proof that density delivers that particular gain. The order of magnitude is still visible: seeding rate differs by a factor of 6.5, yield by roughly four. For scale, spectrum moved yield by 0.5 percent and intensity by 27 percent at most, in both directions.
How much light you actually need and how to calculate the daily dose is covered in our piece on DLI and PPFD. What the channel percentages on a lamp’s spec sheet mean is explained in the spectrum reference. For the hours and PPFD your crop needs, use the DLI and PPFD calculator.
What this means when buying a light
First, translate that 0.5 percent into something you can feel. The gap between shelves came to 10 g per square meter. A rack of twenty 50 x 30 cm trays is three square meters, so 30 g per cycle. On a ten day cycle that works out to roughly one kilogram a year across the whole rack. And the statistics called that difference indistinguishable from zero.
- You buy a tunable spectrum for color and secondary metabolites. Anthocyanins, phenolics, antioxidant activity, a darker leaf. For a restaurant account or premium positioning, that is real value.
- You do not buy it for weight. In this experiment the bluer spectrum produced a 0.5 percent difference, and that one failed significance.
- If you sell by weight, look at PPFD, coverage area and uniformity across the tray. Those three parameters decide how much you cut.
- A seller promising a yield gain from spectrum is either citing something else or overstating. Ask for the number and the conditions it came from.
- Channel percentages on a spec sheet should add to one hundred. If a lamp is described as “full spectrum” with no numbers attached, comparing it to anything is impossible.
The limits of this data
Ertani worked with one species, one intensity, one photoperiod and three replicates per treatment, over ten days of growth. Carrying the conclusion straight across to lettuce, basil or sunflower is not supported.
Flores had eight subjects but a narrow intensity range of 120 to 210 µmol. What happens at 400 or 600 went unmeasured. The paper’s own conclusion is cautious: the response depends on the particular species and cultivar.
Neither of them measured flavor, shelf life after cutting, or what any of this costs. Anthocyanins and phenolics describe color and antioxidant activity, and substituting flavor for them would be wrong.
Summary
Two controlled experiments, eight species and two spectra add up to one picture. Light governs what the plant turns out to be: color, pigments, tissue density, length. How much of it grows is something light barely decides within the working range.
The practical takeaway is simple. Bring your lighting up to an adequate level and then stop fiddling with it. From there the lever sits elsewhere: the seed, the rate you sow it at, the substrate, the water, the temperature, and the day you choose to cut. One more side of the fixture decision sits outside the spectrum: greenhouse lighting maker Heliospectra went bankrupt with 1.69 months of cash left.
