On a farm in Sweden, during the 2024 growing season, two plots of broccoli grew under roofs made of magenta solar panels. A third plot grew in the open air. The panels were semi-transparent cadmium telluride (CdTe) modules, not the opaque dark-blue panels most farms and rooftops use. The trial, published on 2 October 2026 in Cell Reports Physical Science by Silvia Ma Lu and colleagues at Mälardalen University, is a field test of a familiar agrivoltaics promise: that the same land can produce food and electricity. The promise is real, but the details are where the work happens.
The light was dimmer than the label suggests
The two panel systems were nominally 50% and 70% transparent. Those labels are not the number a plant experiences. In the photosynthetically active radiation range — the part of sunlight that drives photosynthesis — the panels actually transmitted about 13.2% and 19.2% on average. So the broccoli was not growing in “half sun” or “70% sun”. It was growing in a much dimmer world.
That matters because shade is not a neutral backdrop. The panels also changed the microclimate: daytime heat and drought stress were lower underneath, and soil moisture followed a different pattern. The crop under the panels was not simply shaded. It lived under a roof that altered temperature, humidity, water loss and light at the same time. This is why the trial is interesting, and why it cannot yet separate the effects of the panels’ colour from the effects of the shelter they provide.
The 25 days are the headline for a grower
Open-field broccoli matured 79 days after transplanting. Broccoli under the panels was harvested on day 104. That is 25 extra days. The shaded plants produced less total aboveground biomass but more leaves, and they converted each unit of intercepted light into biomass more efficiently. In other words, they did not defy plant physiology. They grew more slowly and used the scarce light more carefully.
For a market grower, 25 days is not a rounding error. It can push harvest into a different price window, collide with pest pressure, and stretch cash flow. A similar head weight at the end does not erase that cost. The trial reports mean head fresh weight of about 188 g in the open field and about 169 g and 174 g under the two panel systems. The differences were not statistically significant, but with one plot per treatment, that is a weak kind of comfort. It is a signal worth following, not proof that the panels are yield-neutral.
Efficiency is not yield
The radiation-use efficiency under the panels was 2.8 and 4.5 times the open-field value. This is the number that travelled furthest in the press release, which described a 4.5-fold increase in how efficiently the plants used sunlight. The phrase is accurate in the narrow sense: the plants produced more biomass per unit of intercepted light. It is not a 4.5-fold yield increase. A smaller crop that uses light efficiently is still a smaller crop. If a farmer reads “4.5 times” as “four and a half times more broccoli”, they have read the wrong number.
Nutrients move with maturity
The nutrient results are a good lesson in reading a comparison by date. On day 79, the panel-grown broccoli was not yet mature, while the open-field crop was. At that point the panel crop had higher concentrations of calcium, magnesium and potassium than the mature open-field crop. By day 104, when the panel crop had caught up in maturity, the three macronutrients were broadly comparable. Iron, zinc, manganese and water-soluble carbohydrates were lower.
That pattern does not mean the panels “added” calcium and then “removed” iron. It means maturity, dilution and growing conditions were tangled together. A nutrient comparison between crops harvested on different days is not a clean test of panel colour. The day-104 comparison is the more relevant one for someone eating the broccoli, and it still comes from one site and one season.
What 1.18 does and does not mean
The team estimated a land equivalent ratio (LER) of 1.18. Under the model’s assumptions, one hectare producing both broccoli and electricity would match the output of 1.18 hectares farmed separately. That is a land-use metric, not an 18% revenue gain. It rests on assumptions about a one-hectare layout, panel density and land loss, and on a single site and single season with one plot per treatment. It is a route map, not a commercial feasibility guarantee.
The electricity side has its own gap. Ramesh Rayudu, a power-systems engineering expert at Victoria University of Wellington, noted that the experimental system was estimated to produce 224.7 MWh per hectare over the growing period, against 325.2 MWh per hectare for a conventional solar reference. The roughly 100 MWh per hectare difference would, depending on electricity and broccoli prices, require roughly 1.2 to 3.2 tonnes more broccoli per hectare to offset. He also raised the central uncertainty: the experiment cannot yet separate the effect of the panels’ colour or selected wavelengths from shade, temperature, humidity and other changes beneath the panels.
Where this leaves a grower
The authors call the systems research prototypes. Ma Lu said configurations similar to the prototype may currently suit smaller applications, such as community gardens or greenhouse roofs, rather than immediate deployment over large agricultural areas. She also said the results need validation across more growing seasons and larger scales. The team has begun lab work with magenta, red and blue panels to reduce interference from unfiltered light.
So the practical reading is not “put magenta panels over your fields”. It is narrower and more useful: a semi-transparent roof can slow a broccoli crop by 25 days while keeping head size roughly in the same range, and it can change nutrients in ways that depend on when you harvest. The next question is not whether the colour is beautiful. It is whether wavelength-selective panels earn their extra complexity compared with plain shade, and whether the electricity they give up is worth the broccoli they help produce. Those are field questions, and they need more than one season.