Photoperiodism was described a century ago by Garner and Allard (1920), who showed with soybean and tobacco that some plants flower only when days are shorter than a critical length. It is the unbroken dark period that the plant measures, and a brief exposure to light in the middle of the night can cancel the inductive effect of a short day. For breeders the response is quantitative more often than absolute. In soybean and in several other annual crops the rate of progress towards flowering can be described as a function of mean temperature and photoperiod, with a base phase that responds to temperature alone and a photoperiod-sensitive phase after it (Hadley et al., 1984; Summerfield et al., 1991). Two practical points follow: short days must be accompanied by warm conditions to be effective, and genotypes differ, so that a schedule tuned on one variety may be slow for another.

The study closest to the present purpose is that of Jähne et al. (2020), who used LED-controlled light schemes for the short-day crops soybean, rice and amaranth. With a ten-hour photoperiod and a spectrum enriched in blue and deprived of far-red light they obtained short, sturdy soybean plants and estimated that up to five generations a year were achievable. Other groups have reached fast turnover by different means. Nagatoshi and Fujita (2019) combined carbon dioxide supplementation with germination of immature seed in soybean. In pigeonpea, germination of immature seed together with single seed descent was shown to shorten the cycle substantially (Saxena et al., 2017), and in peanut a commercial programme adopted continuous light under controlled temperature (O'Connor et al., 2013). For chickpea, a long-day species, Samineni et al. (2020) reported up to seven generations a year using extended photoperiod and early seed harvest without any tissue culture step. A recent review of the field stresses that protocols remain crop-specific and that the cost of facilities limits adoption in the regions where these legumes are bred (Wanga et al., 2021).

Light-emitting diodes convert electricity to light efficiently, emit little radiant heat towards the canopy and can be made in narrow or broad spectra (Morrow, 2008; Massa et al., 2008). They can therefore be hung close to the plants, which is what makes a shelf-based system possible. Spectrum affects plant form as well as photosynthesis. Blue light restrains stem elongation, while a low ratio of red to far-red light is read by phytochrome as shade and promotes stretching (Franklin and Whitelam, 2005). In temperate long-day pulses far-red enrichment has been used deliberately to hasten flowering (Croser et al., 2016), but for short-day crops grown at high density the compact habit produced by a far-red-poor spectrum is the more useful property. Bugbee (2016) cautions that total light captured by the canopy generally matters more to growth than fine adjustment of spectral ratios, a helpful principle when the lamps on offer are ordinary white ones.

Any room that can be made fully dark will serve: a store room, a shipping container or a screened shed lined with blackout fabric. Darkness is the non-negotiable requirement, because stray light from a corridor or a street lamp during the night may be enough to delay flowering. Shelving of the kind used in warehouses carries the plants in tiers, each tier lit by lamps fixed to the underside of the shelf above. Reflective sheeting on the walls and around the shelves returns light that would otherwise be lost. Lamps should be chosen for their output per unit of electricity and for a cool white colour; purpose-made horticultural fixtures are better but are not required. Light at canopy height should be measured with a quantum sensor if one can be borrowed, since the eye is a poor judge of intensity, and a simple lux meter calibrated once against such a sensor is adequate for routine checks afterwards.

A mechanical or digital timer switches the lamps to give a short day, with ten hours as the proposed starting value on the evidence from soybean. Short-day crops bring an unexpected economy here: the lamps run for less than half the hours needed in a long-day protocol, so the electricity bill for lighting is correspondingly lower. The night must not be interrupted. Watering, scoring and crossing are done during the light period, and a dim green torch should not be assumed to be safe without testing. Where mains supply is unreliable, an outage during the day merely costs growth, whereas a timer that resets and lights the room at night can undo induction, so timers with battery backup are worth their small extra cost.

Temperature should be kept warm and steady within the range the crop meets in its normal season. In hot climates the difficulty is cooling a closed room, and an extraction fan running during the light period, with a light-trap baffle so that it does not admit daylight, is usually the first measure. Plants are grown singly in small cells of a free-draining medium. Root restriction and close spacing reduce plant size and seed number, which is acceptable because single seed descent needs only one or two good seeds from each plant. Feeding is generous until flowering and then reduced so that the plant matures its first pods instead of continuing to branch.

In the field, several weeks pass between the point at which a legume seed can germinate and the point at which the pod is dry. The protocol recovers that time by harvesting pods once the seeds have filled and begun to lose their green colour, drying them in moving air at room temperature for a few days, and sowing immediately. Hard-seeded types may need the seed coat nicked. This step avoids embryo rescue, which demands sterile facilities and skilled staff, and follows the approach taken in pigeonpea and chickpea (Saxena et al., 2017; Samineni et al., 2020). How early pods can be picked without loss of germination has to be established for each crop.

Each cell carries a label that traces back to its cross and its plant in the previous generation. One seed from each plant is sown for the next generation and a second is kept as a reserve (Brim, 1966). No selection is practised during rapid advance, apart from the unavoidable loss of plants that fail to flower, and the proportion of each population lost in this way is recorded. That record matters, because a protocol that silently discards late genotypes will shift the population towards earliness without the breeder having chosen it.