Implications of open circuit voltage of light-emitting diodes installed for plant cultivation

This study investigated a light sensor function of light-emitting diodes (LEDs), the use of which is expanding in plant cultivations. Unimodal spectral LEDs of four types with respective emission peak wavelengths of 464, 501, 634, and 849 nm, and a white LED with a bimodal spectrum with 455-nm primary and 574-nm secondary peaks were used for this study. Open circuit voltage (VOC) of up to 1-2 V was induced in the LEDs when they were irradiated with sunlight. The VOC value of the 634 nm LED saturated with a low photon flux density can be used for binary discrimination between daylight and night. Each LED had a VOC inducible threshold wavelength of incident light. By virtue of the wavelength threshold feature, existence of a plant leaf between a light source and the LEDs is detectable by comparing the VOC values of blue-green range LEDs (464 nm, 501 nm) and near-infrared (NIR) LEDs (849 nm). Under leaf shadow, the NIR LED VOC exceeded those of the blue-green range LEDs. Under natural incident sunlight, the VOCs of the blue-green range LEDs were greater than that of the NIR LED. Another function of LEDs in agricultural use has been demonstrated.


Introduction
Plant cultivation studies conducted in experimental chambers using light-emitting diodes LEDs as the sole lighting source have been continued Bula et al., 1991;Massa et al., 2008 since lettuce cultivation using red LEDs was planned for feeding astronauts fresh vegetables in a spacecraft Kliss and MacElroy, 1990 . The realization of blue LEDs Akasaki et al., 1993;Nakamura et al., 1993 , which enable emissions at the shorter wavelength peak of bimodal absorption spectra of chlorophylls, supported the role of LEDs as a light source for plant cultivation Hernándes and Kubota, 2016;Matsuda et al., 2004 . Subsequent declines of LED prices have led to commercial crop cultivation in closed, environmentally controlled facilities known as vertical farms Benke and Tomkins, 2017;O Sullivan et al., 2019 and plant factories with artificial lighting Kozai, 2013;Kozai et al., 2020 , which use numerous LEDs as light sources to drive photosynthesis, resulting in overwhelmingly high planting density, high growth rates, and high crop quality Orsini et al., 2020 . Field and greenhouse cultivations also use LEDs for purposes such as supplemental lighting for photosynthesis support Deram et al., 2014;Hikosaka et al., 2013 , photomorphogenesis modification Cope andBugbee, 2013;Jeong et al., 2014 , metabolite synthesis activation Kitazaki et al., 2018;Kopsell et al., 2014;Piovene et al., 2015 , andpest management Cochard et al., 2019;Stukenberg et al., 2015 . In addition to visible light of the 400 -700 nm wavelength range required for maintenance of the basic physiology of plants, far-red Shibuya et al., 2019;Zhang and Runkle, 2019 for photomorphogenesis and ultraviolet light Huché-Thélier et al., 2016 for metabolite synthesis are wavelength bands that improve crop quality.
In contrast to the original role of light emission, a photovoltaic effect is induced in an LED when light is illuminated onto the semiconductor die which usually converts the changes in electron energy levels into light Filippo et al., 2017;RayChaudhuri and Sen, 2009;Vannacci et al., 2018;Vannacci et al., 2019 . Taking advantage of the photovoltaic feature, atmospheric turbidity and precipitable water measurements Mims, 1992 , atmospheric optical depth measurements Acharya et al., 1995 , solar photometry and surface reflectance measurements RayChaudhuri and Sen, 2009 , power delivery and data transmission in a wireless and battery-less microsystem Haydaroglu and Mutlu, 2015 , distance and image sensors Vannacci et al., 2019 , and a vegetation monitoring sensor Kim et al., 2019;Ryu et al., 2010 have been realized using LEDs. The reported benefits of using LEDs as light sensors over photodiodes are affordability Acharya et al., 1995;Mims, 1992;RayChaudhuri and Sen, 2009;Ryu et al., 2010, small size Acharya et al., 1995Mims, 1992;RayChaudhuri and Sen, 2009;Ryu et al., 2010, sturdiness Acharya et al., 1995Mims, 1992;RayChaudhuri and Sen, 2009 , wavelength band selectivity Filippo et al., 2017;RayChaudhuri and Sen, 2009;Ryu et al., 2010;Vannacci et al., 2018;Vannacci et al., 2019 , and duality as an emitter and as a receiver Vannacci et al., 2018;Vannacci et al., 2019 . This study examined whether the photovoltaic effects might be used for monitoring light information in plant-cultivation environments. If some useful information could be acquired, then turned-off LEDs could also be used effectively as sensors for plant-cultivation management. LEDs can be installed proximately to plants Morrow, 2008 because of their characteristics of low heat radiation and small package size Schubert and Kim, 2005 . For this reason, LEDs can be installed not only on the ceilings and walls of horticultural facilities but also inside crop canopies Joshi et al., 2019 . If LEDs could be used as light sensors, then acquisition of light distribution information Ibaraki and Shigemoto, 2013 over a large plantation might become possible. Moreover, an LED might be used as a sensor for acquiring plant information related to optical parameters such as vegetation density.
For this study, open circuit voltages of unimodal spectral LEDs of four types with respective emission peak wavelengths of 464, 501, 634, and 849 nm were investigated when they were irradiated by other external lighting sources. Those LED lights cover the radiation wavelength ranges necessary for plant growth and development. In addition, the open circuit voltage of a white LED with a bimodal spectrum with a primary sharp peak of 455 nm and broad secondary peak of 574 nm was tested. First, we investigated the wavelength range of incident light that induces an open circuit voltage in the five types of LED. Next, the open circuit voltages of the LEDs irradiated with sunlight were measured. Finally, the open circuit voltages induced in the LEDs by incident light transmitted through plant leaves were investigated. Based on these data, we propose to use LEDs, which are usually used as a light source for plant cultivation practices, as light sensors that provide light environment information useful for plant cultivation management.

LED photovoltaic effects induced by impinging unimodal
spectral light Bullet-shaped, 5-mm-diameter LEDs of five types were assumed to be used as a light source for plant cultivation. Although LED package types have some variations, including surface-mount type which has been distributed commonly in plant cultivation practices, bullet-shaped LEDs were chosen for this study because their electrodes have ease of connectivity with test cables and test bread boards. Their performance as light sensors was tested to elucidate a new role of LEDs when they are turned off. Emission spectra of the five LED types were measured using a spectroradiometer MS720, Eko Instruments Co., Ltd., Tokyo, Japan Fig. 1a . The wavelengths λ PR s, which represent the peaks of unimodal spectral photon flux density SPFD of four LED types were, respectively, 464 nm NSPB510AS; Nichia Corp., Tokushima, Japan , 501 nm NSPE510DS; Nichia , 634 nm NSPR510CS; Nichia , and 849 nm TSHG8400; Vishay Intertechnology Asia Pte. Ltd., Tampines, Singapore . The bimodal spectral peaks of a white LED NSPW510DS-D1; Nichia were a 455 nm primary peak and a 574 nm secondary broad peak.
Using the values of these emission peak wavelengths, the five LED types are represented as LED 464 , LED 501 , LED 634 , LED 849 , and LED 455/574 . Figure 1b depicts the forward voltage V F -current I F characteristics of the five LED types. The LED 634 and LED 849 , which emit longer wavelength radiation, have I F increment at V F s of less than 2 V, whereas LED 455/574 , LED 464 , and LED 501 , which emit shorter wavelength radiation, have I F increment at V F s of greater than 2 V. The relation between V F and I F of an ideal p-n junction diode is given as Shockley equation Shockley, 1949 , where I O is the magnitude of saturated dark current when a reverse voltage is applied to the diode Fig. 2 and where q, k, and T respectively represent the unit charge amount, Boltzmann constant, and thermodynamic temperature. When the diode is irradiated with light to induce photovoltaic power, current I R flowing in the opposite direction to I F becomes qV F I R = I L -I F = I L -I O { exp -1 } 2 kT by taking the I R direction as positive Cuevas et al., 2000 . In that equation, I L is the generated current flowing through the p-n junction semiconductor by means of light irradiation. I L is proportional to both the number of photons N having V OC inducible wavelengths and the surface area A of the p-n junction semiconductor die receiving the light Cuevas et al., 2000 . Therefore, accurate light-sensing applications of p-n junction semiconductors usually monitor I L values RayChaudhuri and Sen, 2009 . However, the I L values generated in an LED are too diminutive to be measured using conventional and reasonable instruments. The application purpose of this study is to use photovoltaic effects of neumerous turned-off LEDs already installed ubiquitously around crops to obtain light information across the plantation. Therefore, measurement of LED voltage is more practical than that of photo-generated current. According to equation 2, the open circuit voltage V OC between the LED electrodes at I R = 0 with incident light becomes et al., 2000;Shockley, 1949 . Using coefficients a and b, equation 4 can be rewritten according to the following equation as the relation between the photon flux density and V OC that can be measured by the growers using conventional instruments.
In that equation, E eff is the photon flux density in the V OC inducible wavelength range. The photon to voltage conversion characteristics of LED 464 , LED 501 , LED 634 , LED 849 , and LED 455/574 can be expressed using equation 5 with experimentally obtained coefficients a and b.
To find the coefficients a and b, we conducted an experiment in which the five type LEDs were irradiated with unimodal spectral artificial light from outside using various other type LEDs. In a dark room, 3-mm-diameter bullet-shaped LEDs of 17 types Epitex Inc., Kyoto, Japan , respectively having specific wavelength λ PE at each radiation peak were placed respectively close to the spectroradiometer. The emitted SPFD was measured when each maximum rated forward current was applied to the respective 17 LEDs. One of the 3 mm LEDs of 17 types, as an emitter, was faced to one of the 5 mm LEDs as a light receiver. The V OC induced in the 5 mm LED was measured 34401A; Agilent Technologies Inc., California, USA when the 3 mm LED emitted light Fig. 3 .

LED photovoltaic effects induced by impinging sunlight
Assuming that LEDs are used for outdoor and greenhouse cultivations, V OC s of the 5 mm type LEDs irradiated with sunlight were measured. During 9 : 00 -12 : 00 on 20 October 2020, LED 464 , LED 501 , LED 634 , LED 849 , and LED 455/574 were fixed toward the zenith on the rooftop of a building of Shimane University Matsue Campus. The V OC values of each type LED were measured at 30 s intervals using a data acquisition unit 34972A; Keysight Technologies Inc., California, USA . Sunlight SPFD of wavelengths of 350 -1050 nm and photosynthetic photon flux density PPFD in the 400 -700 nm wavelength range were also measured at 30 s intervals using the spectroradiometer. The sky condition was fine. Assuming that LEDs are installed near crops, the LEDs are expected to receive light that has been transmitted through plant leaves or which was reflected by leaves. To test V OC responses under such conditions, one leaf blade of lettuce Lactuca sativa was placed over the LEDs and the spectroradiometer at 9 : 54. The V OC induced by the light transmitted through the leaf and SPFD of the leaf transmitted light were measured.
To confirm the reproducibility of phenomena observed on October 20, V OC s of the LEDs irradiated with sunlight were measured again on the same rooftop on 26 October 2020. The V OC s of the LEDs fixed toward the zenith at 9 : 00 -17 : 30 were measured at 3 s intervals 34972A . The SPFD and PPFD of sunlight were measured at 60 s intervals using the spectroradiometer positioned proximate to the LEDs. The weather was fine. At 10 : 22, a lettuce leaf covered the LEDs and the spectroradiometer. Table 1 presents V OC s induced in the 5 mm LEDs when they were irradiated with light emitted by one of the 3 mm LEDs of 17 types operating at their maximum rated forward current. The V OC s were induced when excitation light had a shorter wavelength than the inherent threshold wavelength of each 5 mm LED. Figure 4 depicts SPFDs of the 3 mm LEDs and SPFDs that can be emitted by the receiver-side 5 mm LEDs. Figure 4a depicts the SPFDs of light emitted from the 3 mm LEDs respectively having λ PE = 387, 408, 445, 494, 525 nm that induced V OC s in LED 464 and the SPFD emitted from λ PE = 594 nm LED that did not induce V OC in LED 464 . The SPFDs emitted by the 3 mm LEDs with λ PE = 387, 408, 445 nm distributed completely on the shorter wavelength side of the LED 464 spectrum. They induced V OC of around 2.3 V in LED 464 Table 1 . The spectral distributions of λ PE = 494, 525 nm ranged in the longer wavelength side than the LED 464 spectral peak. The shorter wavelength side of those spectra partially overlapped  In this way, when a photon having wavelength shorter than or overlapping with the unimodal 5 mm LED emission spectrum impinges the 5 mm LED, the photon energy exceeding the semiconductor bandgap is transferred to a holeelectron pair Cuevas et al., 2000 . Consequently, V OC is induced in the 5 mm LED. V OC is not induced if the impinged photon energy is less than the bandgap. This phenomenon is consistent with data reported by Vannacci et al. 2019 who presented data of overlapping emission and absorption spectra of an LED. Their spectral data clearly demonstrated the same cut-off at the spectral longest wavelengths, whereas a broader distribution of the absorption spectrum extended to the shorter wavelength side by the Stokes shift Martin et al., 1999;Yang et al., 1993 . Accordingly, the longest wavelength limit of each emission spectrum of the four unimodal LEDs is defined as the voltage-inducible threshold wavelength λ th Table 2 . For LED 455/574 , which has a bimodal emission spectrum, the spectral shape forming the 455 nm primary-peak is assumed to be approximately symmetrical. Then, the wavelength interval from λ PR = 455 nm to the shortest-wavelength emission limit is assumed to be as long as the interval from the λ PR to the longest-wavelength emission limit of the primary-peak. The λ th of LED 455/574 was defined as the longest-wavelength limit of the assumed primary-emission spectrum.

Open circuit voltage of LEDs irradiated by unimodal spectral light
Photons emitted from the 3 mm LEDs having λ PE = 408, 445, 614, 790, and 408 nm respectively induced the maximum V OC s in LED 464 , LED 501 , LED 634 , LED 849 , and LED 455/574 Table 1 . Each pair of the 3 mm LEDs and the 5 mm LEDs [ i.e. 408 nm λ PE , LED 464 , 445 nm λ PE , LED 501 , 614 nm λ PE , LED 634 , 790 nm λ PE , LED 849 , and 408 nm λ PE , LED 455/574 ] was faced to the other so that the 3 mm LED acts as the light emitter and the 5 mm LED as the receiver Fig. 3 . The radiated photon flux density was changed by regulating the power applied to the 3 mm LEDs. The relation between the effective photon flux density E eff having shorter wavelengths than λ th and the induced V OC was found experimentally Fig. 5 . Values of photon flux density emitted from each 3 mm LED were measured using the spectroradiometer positioned near each LED. When the E eff axis was plotted logarithmically, three distinct regions, designated as Zones I, II, and III, were found respectively as representing the V OC conditions of being zero, of sharp inclination, and of saturation. Because V OC s  Figure 6 depicts sunlight PPFD and V OC measured on 20 October 2020. Scud cumulus clouds occasionally covered the sun at around 11 o'clock. At 9 : 54, the lettuce leaf covered the LEDs and the spectroradiometer, causing the PPFD and V OC values to drop sharply and temporarily. The progress of V OC s was classified into three types: saturated, follow-up, and invariant in response to the PPFD increase over time Fig. 6a . The V OC s of LED 464 , LED 501 , and LED 849 were saturated, respectively, at 1.9 V, 1.9 V, and 1.0 V when the PPFD values were greater than 1100 μmol m -2 s -1 . These V OC values were slightly less than the respective saturation voltage values Fig. 5 obtained by the unimodal light irradiation. The V OC of LED 455/574 followed the PPFD fluctuation to some degree without saturation. The V OC of LED 634 remained at a constant value of 1.5 V without being affected by the leaf and cloud cover. Figure 7 depicts sunlight PPFD, and V OC measured on 26 October 2020. The maximum value of the PPFD was 1345 μmol m -2 s -1 at 11 : 43 : 00. Although it was a sunny day with no visible clouds, V OC and PPFD dropped sharply when the LEDs were covered with the lettuce leaf in the morning and when the sun positioned westerly in the late afternoon, during which time direct sunlight was hidden temporarily behind some university buildings and structures. The V OC s of LED 464 and LED 849 were saturated respectively at 1.9 V and 1.0 V, when the PPFD values were greater than 1100 μmol m -2 s -1 . These saturation voltages resemble those obtained during the October 20 experiment. The V OC s of LED 501 and LED 455/574 followed the progress of PPFD to some extent without saturation. The LED 634 V OC saturated at 1.5 V until PPFD decreased to 52 μmol m -2 s -1 at 16 : 50 Fig. 7a . The V OC dropped sharply at sunset. Results show that the LED 634 V OC can be used for binary discrimination between daytime and nighttime for outdoor cultivation. Fig. 8 depicts the relation between incident E eff and induced V OC of the LEDs of five types when incident sunlight irradiated the LEDs without any disturbance on 26 October 2020. The relation between E eff and V OC varied depending on the LED type. The V OC -E eff relational functions obtained using equation 5 were drawn in Fig. 8. Coefficients a and b are presented in Table 2. Because the LED 634 V OC was saturated at E eff = 40 μmol m -2 s -1 or greater and the function property changed Fig. 8c , a and b were found separately for Zones II and III, as in the case of the unimodal light experiments. The deviation between equation 5 and the measured values increased when sunlight became the light source. One possible cause of the deviation is the wavelength dependence of the photovoltaic effects Filippo et al., 2017;Haydaroglu and Mutlu, 2015;Vannacci et al., 2018;Vannacci et al., 2019 .  . Therefore, the sensitivity of V OC depends on the incident light spectrum. The V OC s were induced by photons having wavelengths shorter than λ th . Even for lights having the same E eff value, the wavelength ranges of included photons differ greatly between unimodal light and sunlight. In the sunlight spectrum with wavelengths shorter than λ th , photons having V OC sensitive wavelengths and photons having wavelengths that are less V OC sensitive coexist. When the proportion of the less-sensitive photons becomes dominant in sunlight E eff , the sunlight induces a lower V OC value than the V OC sensitive unimodal light having the same E eff value as that of the sunlight.

Open circuit voltage of LEDs irradiated with light transmitted through plant leaves
Covering the LEDs and the spectroradiometer with a single lettuce leaf decreased PPFD from 1196 μmol m -2 s -1 at 9 : 53 : 30 to 302 μmol m -2 s -1 at 9 : 54 : 00 on October 20 Fig. 6b . Figure   9a depicts the incident light spectra on the LEDs at 9 : 53 : 30 and at 9 : 54 : 00. By means of photon absorption by chlorophylls, SPFD decreased sharply at less than 500 nm and around 680 nm. As a result, V OC s of LED 464 , LED 501 and LED 455/574 having short λ th s were decreased substantially Figs. 6b and 10a . In contrast, 73 of the sunlight SPFD was available in the lettuce leaf shadow in the wavelength band of 700 -900 nm Fig. 9a . The transmitted light spectrum was concentrated at 700 -900 nm by the lettuce leaf covering. For this reason, the relation between V OC and E eff of LED 849 was closer to that obtained by the unimodal light irradiation with λ PE = 790 nm Fig. 5i than that obtained by sunlight irradiation Fig. 8d . Behind the lettuce leaf, 927 μmol m -2 s -1 of E eff Table 3 irradiated LED 849 . The E eff value was sufficient to saturate V OC Fig. 5i . Therefore, the LED 849 V OC was unaffected by the lettuce leaf shadow Fig. 6b, Fig. 10a . Even in the lettuce leaf shadow, LED 634 was irradiated with E eff = 295 μmol m -2 s -1 Table 3 , which was sufficient to saturate V OC Fig. 8c . Accordingly, the V OC value of LED 634 was unaffected by the lettuce leaf shadow Fig. 6b, Fig. 10a . At 11 : 31 : 00, PPFD decreased to 446 μmol m -2 s -1 because the sunlight was shaded by clouds Fig. 6c . The spectral change at this event is presented in Fig. 9c. In the shadow of clouds, the proportion of longer wavelength components decreased to a relative degree. Consequently, LED 849 V OC was decreased Fig.  6c, Fig. 10d   Because E eff = 243 μmol m -2 s -1 was obtained Table 3 , the   The instantaneous shading data (b) were unstable because the leaf swaying in the wind 567 was held by hands.  Table 2. a LED 464 , b LED 501 , c LED 634 , d LED 849 , and e LED 455/574 Fig. 9. Spectral photon flux density SPFD of sunlight incident on the LEDs. a SPFD of incident light on the LEDs at 9:53:30 just before covering the LEDs with the lettuce leaf on 20 October 2020, SPFD at 9:54:00 during the covering, and their ratio; b SPFD of incident light on the LEDs at 10:20:00 just before covering the LEDs with the lettuce leaf on 26 October 2020, SPFD at 10:22:00 during the covering, and their ratio; c SPFD of incident light on the LEDs at 11:29:00 just before the sun was obscured by scud clouds on 20 October 2020, SPFD at 11:31:00 when the sun was obscured by the scud cumulus clouds, and their ratio; d SPFD at 16:00:00 on 26 October 2020 LED 634 V OC retained saturation Fig. 8c . The V OC saturation value Fig. 10c was almost the same as the V OC obtained at 10 : 22 in the lettuce leaf shadow Fig. 10b . Therefore, it is not possible to distinguish from the V OC values of LED 634 whether the PPFD attenuation is attributable to the solar irradiance decrease or shading by leaves. In contrast, a clear attenuation in the LED 849 V OC value was obtained when PPFD decreased to 235 μmol m -2 s -1 at 16 : 00 by the sunlight natural decrement Fig. 10c , compared to the other LED 849 V OC value obtained when PPFD was reduced to 270 μmol m -2 s -1 as the result of lettuce leaf shadow Fig.   10b . Under natural sunlight conditions, the magnitude relation of V OC s was LED 464 , LED 501 > LED 849 irrespective of the sunlight intensity white bars in Fig. 10 . However, the V OC magnitude relation was LED 849 > LED 464 , LED 501 in the shadow of lettuce leaves. Accordingly, the magnitude comparison between the LED 849 V OC and the V OC s of LED 464 and LED 501 can discriminate the difference between light attenuation causes, one of which is attributable to natural sunlight attenuation by changes in solar altitude and cloud covering, and the other one of which is attributable to shading by plant leaves.

Conclusions
We investigated the light environment information obtained from the open circuit voltages of the turned-off LEDs. Few studies have reported the potential use of this information in plant cultivation practices. This study clarified that the open circuit voltage reflects the photon flux density around plants. Furthermore, results suggest that LED open circuit voltages can be used to extract characteristics and variations of qualitative light distribution in a cultivation space by taking advantage of numerous LED installations near the plants.
An LED with open circuit voltage that saturates at a low photon flux density can be used for binary discrimination between daylight and night. A threshold wavelength of light exists for inducement of the open circuit voltage to an LED. By virtue of this feature, existence of plant leaves between a light source and the LEDs can be detected by comparing the open circuit voltage relations between blue-green range LEDs and NIR LEDs.