Look at any textbook on botany and you will find this maxim: plants respond to optical radiation in the spectral range of 280 nm to 800 nm. Period, end of discussion. The question is, how was this spectral range (sometimes referred to as Photobiologically Active Radiation, or PBAR) determined?
This question addresses issues beyond mere academic curiosity. Recent studies, both in the laboratory and in the field, have shown that ultraviolet-C radiation – optical radiation with wavelengths shorter than 280 nm – offers significant economic benefits for horticultural applications. The effects and potential economic benefits of near-infrared radiation – optical radiation with wavelengths longer than 800 nm – have yet to be explored.
The reason for choosing 280 nm as the lower limit of PBAR is simple: plants in the wild are simply not exposed to ultraviolet-C radiation. As shown by the terrestrial solar spectrum in Figure 1, the atmospheric ozone layer effectively blocks any significant amount of ultraviolet-C radiation (100 nm to 280 nm) from reaching the Earth’s surface.
The logic of excluding UV-C radiation from the definition of PBAR may be sound, but it has had unintentional consequences. We have had the ability to produce UV-C radiation for ultraviolet germicidal irradiation (UVGI) applications for over a century (e.g., Kowalski 2009), and low-pressure mercury vapor lamps generating 254 nm UV-C radiation have been used in hospitals and food processing facilities for disinfection purposes since the 1930s. However, it has only been in the past decade or so that UV-C irradiation (mostly using 254 nm UV-C lamps) has been studied and commercialized for pre- and post-harvesting applications in horticulture (e.g., Aarrout et al. 2020, Urban et al. 2016).
The use of UV-C radiation in commercial horticultural applications, including in open fields, greenhouses, and enclosed vertical farms, is proving to have important economic benefits in terms of plant health and reducing spoilage post-harvest. This therefore begs the question: by excluding near-infrared radiation (NIR) from the definition of PBAR, what (if anything) are we missing?
Far-red and the Phytochromes
To understand why 800 nm was chosen, we first need to look at the phytochromes, a class of photoreceptors that control numerous functions in higher plants, including seed germination, shade avoidance, photomorphogenesis, stem elongation, branching, circadian rhythms, root growth, and flowering times (e.g., Smith 2000 and Wang et al. 2015).
A phytochrome molecule has two isoforms, or states. Its ground state, designated Pr, preferentially absorbs red light with a peak spectral absorptance at approximately 660 nm. Upon absorbing a red photon, the molecule undergoes a conformational change to become the Pfr isoform. Left in the dark, this isoform will eventually revert to the Pr ground state. However, the molecule will also revert to the ground state if it absorbs a far-red photon with a peak spectral absorptance at approximately 725 nm.
The Pfr isoform regulates physiological changes in plants, and so it represents the biologically active form of phytochrome. It is, in other words, a biological switch. The relative concentration of Pr to Pfr will depend on the ratio of red to far-red light (expressed as R:FR) incident upon the plant leaves, and the plant will respond accordingly (although often in a species-specific manner).
The role of red and far-red light (which is nowadays defined by ASABE 2017 as the spectral region of 700 nm to 800 nm) was discovered by Borthwick et al. (1952). They determined that red light in the region of 525 nm to 700 nm promoted the germination of lettuce seeds (Lactuca sativa L.) with a peak spectral response at 660 nm, while far-red light in the region of 700 nm to 820 nm inhibited germination, with a peak spectral response of roughly 720 nm.
The spectral absorptances of Pr and Pfr were measured in vitro by Butler et al. (1964), Gardner and Graceffo (1982), and Sager et al. (1988), with moderately similar results. Today, the high-resolution (2 nm) dataset of Sager et al. is most commonly referenced.
Of note however is the spectral limit for these datasets: 800 nm. Visible light spectroradiometers typically have a spectral range of 350 nm to 800 nm. Wider spectral ranges are possible, but at the cost of reduced spectral resolution. Thus, while near-infrared spectroradiometers are available, they typically have spectral ranges on the order of 650 nm to 1100 nm. The decision therefore to define 800 nm as the limit of PBAR may have been dependent in part on the limitations of laboratory equipment.
This would not appear to be a serious issue, however, as the spectral absorptances of both Pr and Pfr clearly do not extend significantly beyond 800 nm … so why bother looking?
A review of the academic literature identified only three papers that considered the effect of NIR on plants. Flint and McAlister (1936) studied the effect of different spectral bands on the inhibition of lettuce seed germination, but it is unclear from their paper what data points were used to generate the plot shown in Figure 3. (The spectral bandwidth was approximately 20 nm.) Even if the small circles represent five peak wavelengths, however, the inhibition of 80 percent at 800 nm is contrary to the Pfr spectral absorptance at this wavelength.
Schäfer et al. (1982) and Johnson et al. (1995) studied the effects of NIR on sprouting common oat (Avena sativa) seeds grown without visible light. Using NIR light-emitting diodes with peak wavelengths of 916 nm (nominal 880 nm) and 958 nm (nominal 935 nm), Johnson et al. noted morphological changes such as coleoptile elongation, advanced leaf emergence and increased gravitropic response (Figure 4).
From a horticultural perspective, these are curious but not particularly useful effects of NIR on cereal grasses grown in darkness. These studies leave the question of whether NIR has any effect on higher plants grown under electric lighting unanswered.
NIR and Daylight
What is surprising about this question is that we should already have answers for any plant species grown in greenhouses versus the same species grown under LED lighting in vertical farms. Most greenhouse glazing is either soda-lime glass with almost constant spectral transmittance from 350 nm to 2800 nm (Figure 5), or polycarbonate panels with constant spectral transmittance from 390 nm to 1100 nm. Referring to Figure 1, the relative radiant power of the solar spectrum from 800 nm to 900 nm is 19 percent of that from 400 nm to 800 nm. The greenhouse glazing has only minimal impact on the spectral power distribution of visible light and NIR (and hence the R:FR ratio) inside the greenhouse.
In practice, however, it would be difficult to conduct experiments comparing the effects of daylight versus electric lighting. It would, for example, be necessary to match both the photosynthetic photon flux density (PPFD) and the spectral power distribution of daylight at the plant canopy. In addition, the inherent variability of daylight and consequent changes in PPFD would further complicate the experiments.
NIR and LED Grow Lights
An obvious solution is to grow plants under LED grow lights with and without additional NIR radiation sources. A 260-watt VIPARSPECTRA TC600 LED grow light system was therefore chosen to investigate the issue. This luminaire features twelve spectral bands provided by 10 quasimonochromatic and two phosphor-coated white light LEDs:
3000K white light
7500K white light
with the relative spectral power distribution, measured with a calibrated Ocean Optics STS-VIS spectroradiometer, shown in Figure 6.
This luminaire was chosen specifically for its inclusion of 730 nm far-red LEDs. The ratio of red light (herein defined as the spectral band of 645 nm to 685 nm) to far-red light (herein defined as the spectral band of 710 nm to 750 nm), varies from approximately 1.1 in full sunlight to roughly 0.2 in the shade, where the red light is screened by the chlorophyll in the blocking leaves. The problem with most LED grow lights is that, without far-red LEDS, their R:FR ratios can be extremely high. Using a grow light with significant far-red output thus ensures that the plants are not responding to any residual PBAR emitted by the NIR radiation sources. (The VIPARSPECTRA luminaire had a R:FR ratio of 6.61.)
The first experiment consisted of growing petunias (Petunia x hybrida) from seed in peat moss at 20° C. The seeds sprouted in seven days under diffuse daylight before being transferred to the growth rack (Figure 7). One group of seedlings was additionally irradiated with 850 nm NIR radiation from a 4-watt Tendelux AI4 IR Illuminator with a 90-degree beam spread and an integral infrared bandpass filter. The grow light was operated with a photoperiod of 12 hours, while the NIR source was energized continuously.
A near-infrared spectroradiometer was not available to measure the spectral power distribution between 800 nm and 900 nm. However, measurements with the visible light spectroradiometer showed that any residual emission from the 850 nm LEDs below 800 nm was less than 0.3 percent of the VIPARSPECTRA peak emission.
The seedlings grown under NIR appeared to develop more quickly, but the growth rate equalized as the leaves matured after seven weeks, However, the mature leaves (8 weeks) grown under NIR were noticeably darker, suggesting a greater concentration of chlorophyll.
The most noticeable morphological difference was that the plants grown under NIR were considerably more compact (Figure 8).
The second experiment consisted of growing sweet basil (Ocimum basilicum v. ‘Sweet Genovese’) from seed in potting soil at 20° C. Two groups were grown under separate VIPARSPECTRA grow light systems, with one group irradiated by two 18-watt CMVision CM-IRP6-850 IR illuminators with a 90-degree beam spread (Figure 9). Measurements with the visible light spectroradiometer again showed that any residual emission from the 850 nm LEDs below 800 nm was less than 0.3 percent of the VIPARSPECTRA peak emission. (The VIPARSPECTRA grow light system for the seedlings without NIR had an R:FR ratio of 7.14.)
The VIPARSPECTA grow lights were operated with a photoperiod of 12 hours, while the CMVision NIR sources were energized continuously.
The seeds without NIR irradiation germinated within one week, with a success rate of approximately 50 percent. However, the seeds with NIR irradiation germinated within two weeks, and with a success rate of approximately 35 percent.
After six weeks, there was a marked difference in stem elongation, where the plants with NIR irradiation were approximately twice as tall (Figures 10A and 10B). In addition, the plants with NIR irradiation had larger and lighter-colored leaves (Figures 11A and 11B).
Due to the lack of quantitative spectral power distributions for the NIR irradiation of the plant canopy and the differences in NIR irradiance between the two experiments, the results observed above must be considered incomplete pending further study. However, a number of tentative conclusions can be drawn regarding constant 850 nm irradiation:
Germination of Ocimum basilicum seeds is inhibited.
The shade avoidance response of Petunia x hybrida appears to antagonized, resulting in a more compact plant architecture.
The shade avoidance response of Ocimum basilicum is clearly promoted, resulting in stem elongation and thus taller plants.
The leaf area of Ocimum basilicum is increased.
The chlorophyll production in the leaves of Petunia x hybrida appears to be slightly increased.
The chlorophyll production in the leaves of Ocimum basilicum appears to be moderately inhibited.
One possible explanation for these observations is that the biologically active isoform of phytochrome (Pfr) has an unreported spectral absorption band in the region of 800 nm to 900 nm. This would have the effect of reducing the Pfr content in the leaves, which in turn reduces the chlorophyll concentration (e.g., Kreslavski et al. 2018) and increases the leaf area (e.g., Boccalandro et al. 2009). This could explain the Ocimum basilicum results, but not the Petunia x hybrida results.
NIR absorption by phytochrome Pfr could further explain the inhibition of Ocimum basilicum seed germination, as demonstrated by Flint et al. (1936) and Borthwick et al. (1952) with far-red light.
The shade avoidance response of Ocimum basilicum can also be explained by NIR absorption, as this is equivalent to lowering the R:FR ratio. The plant stems will then elongate in order for the upper leaves to receive as much full sunlight as possible.
This does not, however, explain the apparent antagonism of the shade avoidance response of Petunia x hybrida. One possibility is that the lower NIR irradiance used in the experiment elicits a nonlinear response by phytochrome Pfr, although this would be more likely to have a null rather than a negative effect.
Another possibility is that there is an undiscovered photopigment with a spectral absorption band beyond 800 nm (Johnson et al. 1995). Alternatively, there may be an undiscovered spectral absorption band in one of the cryptochromes or other known photopigments involved in photomorphogenesis.
Whatever the case, it is evident that more research is required. The PBAR limit of 800 nm has been assumed for at least the past seven decades, but then the effects of UV-C radiation on plants remained unexplored for an equal length of time. The effect of near-infrared radiation of plants is a topic that deserves to be explored.
ASABE. 2017. ANSI/ASABE S640 JUL2017, Quantities and Units of Electromagnetic Radiation for Plants (Photosynthetic Organisms). St. Joseph, MI: American Society of Agricultural and Biological Engineers.
Aarrout, J., and L. Urban. 2020. “Flashes of UV-C Light: An Innovative Method for Stimulating Plant Defenses,” PLoS ONE 15(7):e0235918. DOI: 10.1371/journal.pone.0235918.
ASTM International. ASTM G173-03(2020), Standard Tables for Reference Solar Spectral Irradiances: Direct Normal and Hemispherical on 37° Tilted Surface. West Conshohocken, PA: ASTM International; 2020.
Boccalandro, H. E., et al. 2009. “Phytochrome B Enhances Photosynthesis at the Expense of Water-Use Efficiency in Arabidopsis,” Plant Physiology 150(2):1093-1092. DOI: 10.1104/pp.109.135509.
Borthwick, et al. 1952. “A Reversible Photoreaction Controlling Seed Germination,” Proceedings of the National Academy of Science 38:662–666. DOI: 10.1073/pnas.38.8.662.
Flint, L. H., and E. D. McAlister. 1936. “Wave Lengths of Radiation in the Visible Spectrum Inhibiting the Germination of Light-Sensitive Lettuce Seed,” Smithsonian Miscellaneous Collections 94(5):1-11.
Gardner, G., and M. Graceffo. 1982. “The Use of a Computerized Spectroradiometer to Predict Phytochrome Photoequilibria under Polychromatic Irradiation,” Photochemistry and Photobiology 36(3):349-354. DOI: 10.1111/j.1751-1097.1982.tb04385.x.
Kowalski, W. 2009. Ultraviolet Germicidal Irradiation Handbook: UVGI for Air and Surface Disinfection. Berlin, Germany: Springer-Verlag.
Kreslavski, V. D., et al. 2018. “The Impact of the Phytochromes on Photosynthetic Processes,” BBA – Bioenergetics 1859:400-408. DOI: 10.1016/j.bbabio.2018.03.003.
Kusuma, P., and B. Bugbee. 2021. “Far-red Fraction: An Improved Metric for Characterizing Phytochrome Effects on Morphology,” J. American Society of Horticultural Scientists 146(1):3-13. DOI: 10.21273/JASHS05002-20.
Sager, J. C., et al. 1988. “Photosynthetic Efficiency and Phytochrome Photoequilibria Determination Using Spectral Data,” Trans. ASAE 31(6):1882-1889.
Schäfer, E., et al. 1972. “In vivo Measurement of the Phytochrome Photostationary State in Far Red Light,” Photochemistry and Photobiology 15(5):457-464. DOI: 10.1111/j.1751-1097.1972.tb06257.x.
Smith, H. 2000. “Phytochromes and Light Signal Perception by Plants – An Emerging Synthesis,” Nature 407:585-591. DOI: 10.1038/35036500.
Urban, L., et al. 2016. “Understanding the Physiological Effects of UV-C Light and Exploiting its Agronomic Potential Before and After Harvest,” Plant Physiology and Biochemistry 105:1-11. DOI: 10.1016/j.plaphy.2016.04.004.
Wang, H., et al. 2015. “Phytochrome Signaling: Time to Tighten Up the Loose Ends,” Molecular Plant 8:540-551. DOI: 10.1016/j.molp.2014.11.021.
Whereas human vision relies on five opsins as photoreceptors, most plants have a wide variety of photopigments that are responsive to optical radiation from 280 nm to 800 nm. Beyond photosynthesis, plants rely on this radiation to control photomorphogenesis, phototropism, shade avoidance, and both circadian and circannual rhythm entrainment.
Quasimonochromatic LEDs have proven a boon for botanists in that the molecular genetics of these responses can be elucidated with precisely controlled spectral power distributions (SPDs). In terms of photopigments, cryptochromes, for example, respond to blue light, while phytochrome responds to the R:FR ratio of red (approximately 660 nm) to far-red (approx. 735 nm) light.
The problem is that botanists do not define what is meant by “blue,” “green,” “yellow,” “red,” or “far-red” visible light, while ultraviolet radiation is broadly defined as UV-A and UV-B. Consequently, it is difficult to replicate laboratory experiments without knowing the SPD of the horticultural light source.
This paper proposes an LED “color” specification that represents a given SPD using a small number of radial basis functions, to provide a metric for comparing biologically similar SPDs. It further introduces a trainable fuzzy logic SPD classifier that can compare biologically similar SPDs for specific horticultural applications.
When the first high-pressure sodium (HPS) lamps were introduced in the late 1960s, they were quickly adopted by commercial greenhouse operators as a means of providing supplemental electric lighting. This made it economically possible to grow vegetables and flowers throughout the year in controlled environments. They had luminous efficacies, ranging from 100 to 150 lumens per watt, they were available in sizes ranging from 400 to 1,000 watts, and they could be incorporated in luminaire housings designed to withstand the heat and humidity of greenhouses.
One disadvantage of HPS lamps is they produce mostly yellow light with fixed spectral power distributions (SPDs). This is not particularly important for plant photosynthesis, as most plants can take advantage of optical radiation within the spectral range of 400 to 700 nm. Horticulturalists often refer to the “McCree curve,” which plots average photosynthesis efficiency versus wavelength for a variety of field-grown crops (McCree 1972). As shown in Figure 1, the spectral output of HPS lamps is near the peak of the McCree curve.
The problem is that while the yellow light of HPS lamps may be good for photosynthesis, plants have a wide range of photopigments that respond to optical radiation from 280 nm to 800 nm (often referred to as “photobiologically active radiation,” or PBAR). These responses include:
Photomorphogenesis – any change in the morphology (i.e., shape) or composition of a plant or its components that is induced by optical-radiation exposure
Photoperiodism – response of a plant to daily (circadian) or seasonal (circannual) changes in optical-radiation exposure
Photosynthesis – conversion of “photosynthetically active radiation” (PAR) into chemical energy stored as carbohydrates to fuel plant activities
Phototropism – any self-actuated change in the orientation of a plant or its components toward or away from optical radiation
Secondary metabolite production – organic compounds not directly involved in plant growth, development, or reproduction, including compounds used as medicines, flavorings, pigments, and drugs
Shade avoidance – a set of responses to being shaded by other plants, including changes in morphology, flowering times, and allocation of resources
While many of these responses have been known or suspected for decades, it was difficult for botanists to study them in the laboratory without suitable light sources. This changed, however, with the introduction of horticultural luminaires with high-flux quasimonochromatic light-emitting diodes (LEDs). Somewhat serendipitously, the absorption spectra of chlorophyll A and B have peaks that correspond with those of approximately 450-nm InGaN and approx. 660-nm AlInGaP LEDs (Figure 2). Today, the photon efficacy (measured in micromoles of PAR photons per Joule, rather than in lumens) of LED modules is typically greater than equivalent 1,000-watt HPS lamps.
While commercially available horticultural luminaires with blue and red LEDs (producing so-called “blurple” light) are now successfully competing with traditional HPS luminaires, botanists’ attention has turned to the capabilities of multichannel LED luminaires with controllable SPDs. Over 500 academic studies over the past decade have investigated the effects of different wavelength ranges on plants and their absorption by photopigments (Table 1).
280 nm – 315 nm
Sec. metabolism Shade avoidance Phototropism
315 nm – 400 nm
Chlorophylls Cryptochromes Phototropin Phytochromes Zeitlupe family
Sec. metabolism Photomorphogenesis
400 nm – 500 nm
Carotenes Chlorophylls Cryptochromes Phytochromes Zeitlupe family
Photosynthesis Sec. metabolism Shade avoidance Phototropism Photoperiodism
500 nm – 575 nm
Photosynthesis Sec. metabolism Shade avoidance
575 nm – 610 nm
Yellow – orange
Photosynthesis Sec. metabolism
610 nm – 700 nm
Photosynthesis Photomorphogenesis Sec. metabolism Shade avoidance Photoperiodism
700 nm – 800 nm
Photomorphogenesis Shade avoidance Photoperiodism
Table 1 – Plant Responses to Optical Radiation
The problem is that while UV-A and UV-B are formally defined in the scientific literature (e.g., ISO 2007), the visible color names are colloquial and based on human visual responses. The title of one paper in particular illustrates this issue: “Green light drives leaf photosynthesis more efficiently than red light in strong white light: Revisiting the enigmatic question of why leaves are green” (Terashima 2009). For anyone interested in either replicating the experiments or extrapolating their results, what are “green,” “red,” and “white” light?
The color name “far red,” which refers to the spectral range of 700 nm to 800 nm, is formally defined in terms of horticulture (ASABE 2017). It is important in terms of shade avoidance and photoperiodism, where plants rely on two isoforms of phytochrome to detect the ratio of red to far red (R:FR) optical radiation (Sager et al. 1988), but there is no equivalent definition of “red” (Figure 3). Luminaire manufacturers are now offering products with 660-nm red and 735-nm far-red LEDs to induce or delay flowering in ornamental plants (e.g., Craig and Runkle 2013), but many previous horticultural studies have relied on daylight alone or daylight and incandescent lamps to explore the effects of varying R:FR. How should these studies be interpreted in terms of modern horticultural lighting practices with LED-based luminaires?
Horticultural researchers have recognized that the use of colloquial color names is a problem. Many papers describe their experimental methods in detail, including light source SPD plots, names of specific luminaire products, and occasionally tabulated SPDs. This still leaves open, however, the problem of interpreting the results in terms of other optical radiation sources with similar SPDs.
One proposed product label for horticultural light sources is shown in Figure 4 (Both et al. 2017). By avoiding the use of color names, this proposal eliminates any dependence on the human visual system. However, the arbitrary separation of the PBAR spectral range into 100-nm wide bands ignores the distinct responses of plants to UV-A and UV-B radiation, as well as the response of plants to narrower changes in wavelength. For example, Johkan et al. (2012) provide an example wherein the growth of lettuce under quasimonochromatic radiation from “green” LEDs with center wavelengths of 510 nm, 520 nm, and 530 nm varies markedly depending on the center wavelength for the same photosynthetic photon flux density (Figure 5).
The spectral absorptance characteristics of the primary plant photopigments chlorophyll A and B, b-carotene, and phytochrome (Figure 2) suggest that their absorptances vary very rapidly with changes in wavelength. However, these data represent the spectral absorptance of the pigment extracts dissolved in solvents (i.e., in vitro). As shown by Moss and Lewis (1952), a combination of the structural complexity of the leaves, screening by other photopigments, and the presence of accessory photopigments have the effect of broadening the spectral absorptance characteristics of the photopigments in vivo. Studies such as those of McCree (1972) have shown that in general, plants are reasonably tolerant of small changes in the center wavelengths of quasimonochromatic radiation. (Johkan et al.  was likely an exception in that photosynthesis probably occurred due to b-carotene rather than chlorophyll A or B, with longer wavelengths of green light being incapable of exciting this photopigment.)
In view of this and other studies, it is clear that any attempt to characterize the SPDs of horticultural luminaires needs to take into consideration the responses of plants to changes in center wavelength of quasimonochromatic light sources, and more generally to horticultural luminaires with both quasimonochromatic and broadband radiation sources.
In relation to this, Maloney (1986) discusses the physical basis of spectral reflectance distributions from natural objects, including organic materials. These distributions are band-limited by molecular interactions and superimposed vibrational and rotational patterns, with the result that the number of parameters needed to adequately represent spectral reflectance distributions in visible light (i.e., 400 nm to 700 nm) is five to seven. Westland et al. (2000) came to a similar conclusion based on statistical studies of reflectance spectra, noting that the spectral reflectance distributions of most natural surfaces form a set of band-limited functions with a frequency limit of approximately 0.02 cycles per nm. This implies that visible light reflectance spectra can be adequately represented using six to 12 basis functions (e.g., Westland and Ripamonti 2004).
A small number of radial basis functions (e.g., Buhmann and Jäger 2000) can therefore be used to approximate a real-valued function (such as an SPD) as a weighted sum of the basis functions. As an example, the set of Gaussian functions , where , and for , can be used to approximate any SPD from 350 nm to 800 nm where the functions are separated by 25 nm (Figure 6).
An advantage of this method is that the set of basis function weights is much smaller than the set of enumerated values for a measured SPD. Rather than referring to “red,” “green,” “blue,” or “white” light, horticulturalists can state the values of a set of basis function weights. Moreover, a useful approximation of the original SPD significant to the needs of horticultural lighting can be reconstructed from these weights.
As an example, Figure 7 shows 13 radial basis functions over the visible light range of 400 nm to 700 nm that are each multiplied on a per-wavelength basis by the SPD of a 4000-K white light LED to yield the basis function weights.
Figure 8 shows a reconstruction of the LED SPD using a cubic spline curve with the weights as knots. The reconstruction is clearly different from the original SPD, but in terms of predicting plant responses, it is likely adequate.
Horticultural Spectral Sensor
Referring to Figure 6, each basis function can be seen as the responsivity of a radiant flux meter in combination with a Gaussian bandpass filter with a center wavelength of xi. Combining the unweighted outputs of the 19 filtered meters results in a flat response from 375 nm to 775 nm. Presented with an arbitrary SPD, the filtered meter outputs represent the appropriate weighting for the basis functions to approximate the SPD. Moreover, the absolute values of the filtered meter outputs can be used to estimate the absolute spectral irradiance incident on the meters, from which can be calculated the absolute irradiance in watts per square meter and the photosynthetic photon flux density (PPFD) in micromoles per square meter per second.
Figure 9 shows the spectral responsivities of an AS7262 6-channel visible light spectral sensor with a spectral range of approximately 430 nm to 570 nm, as manufactured by ams AG of Premstaetten, Austria. While the spectral responsivities are not ideal Gaussian functions, it is clear that an instrument to directly measure radial basis function weights can be fabricated using existing technology.
An advantage of this system and method in terms of horticultural light sources is that the spectral power distributions can be unambiguously measured and expressed as a small set of numbers, regardless of the SPD complexity. If the representations of two SPDs are similar, the horticulturalist may be assured that they will likely have the same biological effect on a plant species. As an example, white light fluorescent lamps typically exhibit a combination of continuum and line spectra, whereas white light LEDs typically exhibit a narrow peak emission near 450 nm and a broad continuum from the blends of green- and red-emitting phosphors (e.g., Figure 7). Regardless, if their sets of radial basis function weights are similar, the two light sources may also be regarded as similar with respect to horticultural applications.
Fuzzy Logic SPD Classifier
The unanswered question is, what does “similar” mean in the context of comparing two or more SPD representations? With 13 to 19 radial basis function weights as parameters, it becomes impractical to formulate a table of rules for comparison purposes. The solution to this problem is a fuzzy logic classifier.
Fuzzy logic is often seen as a mathematical means of representing vagueness and imprecise information when making decisions, where input signals are “fuzzified” by mapping their precise values to a set of fuzzy membership functions. Referring to Figure 10 as an example, a radial basis function with weight 0.85 has membership 0.60 in “high” and membership 0.40 in “very high.” (Triangular membership functions are used in this example, but trapezoidal and sigmoid functions may also be used.)
Referring to Figure 7, the fuzzification of the set of 13 radial basis function weights results in Table 2.
Table 2 – Fuzzification of 4000-K White Light LED SPD
Table 2 – Fuzzification of 4000K white light LED SPD.
The set of n fuzzified weights for a given SPD is then submitted to a fuzzy if-then rule system. Given any two fuzzified weights x1 and x2 as inputs, a typical fuzzy rule will be:
where there are multiple output classes.
Each rule calculates a “vote” t that is determined by degree of membership m for each fuzzified weight:
and where the fuzzy AND operator is implemented as the minimum of the two membership values.
Once all of the rules have been processed, their votes are aggregated:
for all votes.
This is arguably the simplest possible implementation of a fuzzy logic classifier. There are other methods for calculating and aggregating votes that are likely better for the purpose, but it is the principle that is of interest. What a fuzzy logic classifier accomplishes is a framework for representing expert knowledge of the effect of similar but different SPDs on plant growth and health, taking into consideration the plant species, plant growth stage, plant environmental conditions, and other parameters. In a sense, the fuzzy if-then rules formalize what is known about plant responses to optical radiation (e.g., Table 1) and classify horticultural luminaire SPDs accordingly.
While light-emitting diodes have provided botanists with the ability to generate precisely controlled SPDs for their research, the use of colloquial color names in their published papers has made it difficult to interpret and summarize their research results for the horticultural industry. This paper therefore proposes the use of a small number of radial basis functions to represent SPDs for horticultural lighting purposes, based on the observation that the absorption characteristics of photopigments in vivo limits the need for more-detailed SPDs. A proposal for a horticultural spectral sensor that measures radial basis function weights directly is also introduced.
Finally, a fuzzy logic classifier is proposed as a means of representing expert knowledge gained from horticultural research using fuzzy if-then rules, thereby resolving the problem of determining the similarity of two or more SPDs for horticultural lighting purposes.
ASABE. 2017. ANSI/ASABE S640 JUL2017, Quantities and Units of Electromagnetic Radiation for Plants (Photosynthetic Organisms). St. Joseph, MI: American Society of Agricultural and Biological Engineers.
Both A-J et al. 2017. Proposed product label for electric lamps used in the plant sciences. Hort Technol. 27(4):544-549.
Buhmann, MD, Jäger J. 2000. On radial basis functions. Acta Numerica. 9:1-38.
Craig DS, Runkle ES. 2013. A moderate to high red to far-red light ratio from light-emitting diodes controls flowering of short-day plants. J Am Soc Horticult Sci. 138(3):167-172.
Kuehni RG. 2003. Color Space and Its Divisions. Hoboken, NJ: John Wiley & Sons.
ISO. 2007. ISO 21348:2007(E). Space Environment (Natural and Artificial – Process for Determining Solar Irradiances. Geneva, Switzerland: ISO.
Johkan M et al. 2012. Effect of green light wavelength and intensity on photomorphogenesis and photosynthesis in Lactuca sativa. Environ Exper Botany. 75:128-133.
Maloney L. 1986. Evaluation of linear models of surface spectral reflectance with small number of parameters. J Opt Soc Am. 3(10):1673-1683.
McCree KJ. 1972. The action spectrum, absorptance and quantum yield of photosynthesis in crop plants. Agric Meteorology. 9:191-216.
Moss RA, Lewis WE. 1952. Absorption spectra of leaves. I. The visible spectrum. Plant Physiol. 27(2):370-391.
Sager JC et al. 1988. Photosynthetic efficiency and phytochrome equilibria determination using spectral data. Trans ASABE. 31(5):1882-1889.
Terashima I et al. 2009. Green light drives leaf photosynthesis more efficiently than red light in strong white light: Revisiting the enigmatic question of why leaves are green. Plant Cell Physiol. 50(4):684-697.
Westland S et al. 2000. Colour statistics of natural and man-made surfaces. Sensor Rev. 20(1):50-55.
Westland S, Ripamonti C. 2004. Computational Color Science Using Matlab, Chapter 10. Chichester, UK: John Wiley & Sons.
These are the first half-dozen of over 200 online articles that were over a period of approximately two weeks following the publication of this Tel Aviv University press release, dated December 14, 2020:
with the subtitle, “Groundbreaking research finds UV-LED diodes efficiently and cheaply disinfect social spaces.”
“Groundbreaking” research? This has a touch of hyperbole, but let’s see …
On September 10, 2020, the respected Journal of Photochemistry & Photobiology, B: Biology published this paper:
Gerchman, Y., et al. 2020. “UV-LED disinfection of Coronavirus: Wavelength effect,” J. Photochemistry & Photobiology B: Biology 212 (2020) 112044 (DOI: 10.1016/j.photobiol.2020.112044).
The paper is open-access, for which the publisher deserves due credit for making its COVID-19-related research papers freely available.
The paper’s abstract is interesting:
“UV light-emitting diodes (UV LEDs) are an emerging technology and a UV source for pathogen inactivation, however low UV-LED wavelengths are costly and have low fluence rate. Our results suggest that the sensitivity of human Coronavirus (HCoV-OC43 used as SARS-CoV-2 surrogate) was wavelength dependent with 267 nm ~ 279 nm > 286 nm > 297 nm. Other viruses showed similar results, suggesting UV LED with peak emission at ~286 nm could serve as an effective tool in the fight against human Coronaviruses.”
but the introduction is more informative:
“Numerous studies have examined the sensitivity of different microorganisms (including viruses) to UV LED at different wavelengths as detailed in Table 1, for suspended viruses. However, no study to date has examined the efficiency [sic] of UV LEDs at different wavelengths on the inactivation of the human corona virus. Here, we have used the human coronavirus OC43 (HCoV-OC43) as a surrogate to the SARS-CoV-2, to develop a dose-response curve for UV-LEDs at various wavelengths.”
.. and here we need to pause in order to put these statements into context. The authors referenced twelve previous studies in their Table 1, but the key phrase here is “UV LED.” If we generalize this to “ultraviolet (UV) radiation,” there are many more studies of the relationship between wavelength and the efficacy (not “efficiency”) of UV radiation in inactivating viruses. In fact, the first study was published 144 years ago (Downes and Blunt 1877). The virucidal action spectrum for UV radiation was first established by Rivers and Gates (1928) and Sturm et al. (1932).
Figure 1 shows the action spectra (DIN and IES) for germicidal ultraviolet radiation applications, such as upper-room air and municipal water disinfection, that have been widely adopted by the CIE, IES, ACGIH, NIOSH, DIN, and other standards organizations:
What is rarely mentioned is that these action spectra are based on laboratory results with the Escherichia coli bacterium (e.g., Gates 1930). It is not a coincidence that the peak response near 265 nm corresponds with the peak spectral absorptance of deoxyribose nucleic acid (DNA) – UV radiation disrupts the genetic code of viruses, bacteria, and fungi, thereby preventing them from reproducing (e.g., Hollaender and Oliphant 1944).
The peak spectral response of different viruses and other pathogens may therefore vary by perhaps five nanometers or so (e.g., Linden 2001). However, the DIN and IES action spectra remain applicable for practical applications of germicidal UV radiation.
… which brings us back to the current paper of Gerchman et al. (2020). The results presented in the paper are summarized in Figure 2, where the dose refers to the UV irradiance multiplied by the exposure time, and the horizontal “limit of quantification” line represents the dose required to achieve log-three (99.9 percent) inactivation of the virus colony:
We may compare this to the IESNA germicidal response curve as enumerated in CIE 155:2003, Ultraviolet Air Disinfection, relative to the peak response at 265 nm:
Gerchman et al.
1.00 ± 0.09
0.82 ± 0.03
0.46 ± 0.05
0.19 ± 0.05
Table 1 – Spectral response comparison.
The error bars shown in Figure 2 represent standard deviations in the results, which the authors explained as “an artifact due to lack of precision in enumerating the low number of [virus] survivors.” It is refreshingly surprising to see such honesty in published results; this sort of information is usually confined to supplementary material so as to not “confuse” the reader with data that might weaken the paper’s conclusions.
Whether there are errors of fact or the authors used questionable experimental procedures is not a topic that I as a science journalist (among other things) am academically qualified to comment on. What I will say, however, is that the paper itself has been carefully constructed and well-written, and is a model of academic writing.
So far, so good. However, there are no surprises here – the OC43 coronavirus appears to be somewhat less susceptible to longer wavelengths than E. coli bacteria, but this does not invalidate the applicability of germicidal response functions shown in Figure 1. The research is if anything no more than a confirmation of accepted scientific fact – viruses are susceptible to ultraviolet radiation, with a peak response near the DNA peak spectral absorptance of 265 nm.
The question then is, why did the editors of the Journal of Photochemistry & Photobiology decide that the paper was worthy of publication? While it is impossible to speak on behalf of the editors, one answer is that there is often value in the publication of negative results.
If the research had shown that the OC43 coronavirus was highly susceptible to longer wavelengths of UV radiation, that would have been stunning news that would have more than justified publication of the paper. Instead, the study merely confirmed that the existing standard germicidal response curves are generally applicable to HCoV-OC43, and (presumably) to SARS-CoV-2.
The academic value of the paper is therefore in describing what appears to be a carefully designed and executed series of experiments that yielded negative results. It informs other researchers of what has been done, and so allows them to direct their research efforts elsewhere.
There have to date been over 54 thousand academic papers relating to the SARS-CoV-2 virus that have been published in biomedical and life science journals. The probability of any one article coming to the attention of the public is basically infinitesimal … if it were not for university media relations officers.
The role of the media relations officer is to present the often-arcane details of academic research to the public. For both public and private universities, being seen in positive terms by the public is key to obtaining financial support from both public and private institutions. In other words, the role of a media relations officer is that of a marketing professional.
It is often a difficult job – how do you take a paper with a random title such as, “Genetic diversity of the Plasmodium falciparum GTP-cyclohydrolase 1, dihydrofolate reductase and dihydropteroate synthetase genes reveals new insights into sulfadoxine-pyrimethamine antimalarial drug resistance” (DOI:10.1371/journal.pgen.1009268), and present it to a public more interested in YouTube celebrities and sports figures? You begin your press release with a catchy title such as, “New mutations in malaria parasite encourage resistance against key preventative drug,” but the only qualification for the job is usually a bachelor’s degree in journalism. The media relations officer, through misunderstanding their interview with the researcher or lack of knowledge and experience, may fail, sometimes spectacularly.
Tel Aviv University
I have no desire to name and shame the media relation officer at Tel Aviv University in Israel, but I must say something. Here are a few select quotes for the aforementioned press release of December 14th, 2020:
“Researchers from Tel Aviv University have proven that the coronavirus can be killed efficiently, quickly and cheaply using ultraviolet (UV) light-emitting diodes (UV-LEDs). This is the first study in the world conducted on the disinfection efficiency of a virus from the family of coronaviruses using UV-LED irradiation at different wavelengths or frequencies.”
This may be technically true, but the significance of this study is grossly overrated in calling it, “the first study in the world.”
“In the study, the researchers tested the optimal wavelength for killing the coronavirus, and found that a length of 285 nanometers was almost as efficient in disinfecting the virus as a wavelength of 265 nanometers, requiring less than half a minute to destroy more than 99.9% of the coronaviruses.”
It does not take a mathematician to understand that a relative efficacy of 46 ± 5 percent is not “almost as efficient.”
“The entire world is currently looking for effective solutions to disinfect the coronavirus … The disinfection systems based on LED bulbs, however, can be installed in the ventilation system and air conditioner, for example, and sterilize the air sucked in and then emitted into the room.”
This is true, but it has nothing to do with the research paper. The popular press has been awash with stories about germicidal disinfection using radiation, a technology that has been in commercial use since 1909 (von Recklinghausen 1914). Economics currently favour low-pressure mercury vapour lamps that emit ultraviolet-C radiation at 254 nm, but rapid progress is being made in the development of more efficient and inexpensive ultraviolet LEDs. Again, nothing to do with this paper.
Quoting one of the paper’s authors from the press release, “We discovered that it is quite simple to kill the coronavirus using LED bulbs that radiate ultraviolet light, but no less important, we killed the viruses using cheaper and more readily available LED bulbs, which consume little energy and do not contain mercury like regular bulbs.”
This is … painful. Even if the author meant specifically the OC43 coronavirus and ultraviolet radiation generated by light-emitting diodes rather than mercury-vapour lamps, stating that anything was “discovered” is simply ludicrous.
It should further be noted that the radiant efficacy of commercially-available 285 nm UV-LEDs is currently on the order of one percent. This may be compared to that of mercury-vapour lamps, with efficacies on the order of 40 percent. The advantage of UV-LEDs is that it is much easier to direct their emitted radiation into narrow beams, a requirement for upper-room air disinfection devices.
“Last year, a team of researchers led by Prof. Mamane and Prof. Gerchman patented a combination of different UV frequencies that cause dual-system damage to the genetic load and proteins of bacteria and viruses, from which they cannot recover – which is a key factor that is ignored.”
This is one of those, “Wait, what?” moments, where an entirely different and much more relevant press release could have been written about this technology. The patent application in question is US Patent Application 20200255305, “Method and Device for Water Disinfection,” that uses two sources to simultaneously emit UV-C and UV-B radiation. Patent applications are not peer-reviewed, however, and so the information needs independent verification and much more detail concerning the experiments behind the invention.
The primary problem is that the media relations officer was clearly struggling to understand the issues and write a coherent and informative press release. At best, there appears to have been “a failure to communicate.”
Whatever misinformation and confusion there may have been in the press release, it won the lottery in being selected by the mass media for amplification. Looking at only the first six of some 200 article titles:
“99.9% of Covid-19 virus dead in 30 seconds with UV LEDs, says Tel Aviv research.”
No, neither the paper nor the press release made any claims regarding the SARS-CoV-2 virus that causes the COVID-19 disease. It explicitly stated that the HCoV-OC43 virus (which is one of the many viruses responsible for the common cold) was chosen as a surrogate for the SARS-CoV-2 virus, and that “… our future work will confirm these results by testing the impact of LEDs and their combinations on SARS-CoV-2.”
“You can kill Covid with a flick of a switch, study shows.”
The paper never suggested such an outlandish idea. It instead focused on the relative dose of ultraviolet radiation at different wavelengths to inactivate (not “kill”) a particular virus. The irradiance levels employed in the experiments would be totally impractical for surface disinfection in the real world.
“Tel Aviv research: 99.9% of COVID-19 virus dead in 30 seconds with UV LEDs.”
The paper uses the noun “seconds” just once, referring to “… up to 60 s for 267 and 279 nm and up to 90 s for 286 and 297 nm.” The key metric is dose – ultraviolet irradiance multiplied by exposure time. Whoever wrote this headline simply invented the number as clickbait.
… and so it goes, like some sinister version of the children’s game Chinese whispers. Perfectly reasonable and valid scientific information is endlessly repeated and distorted from paper to press release to mass media articles. Like most such events, the story will have a half-life measured in weeks to a few months before it is forgotten. Unfortunately, the misinformation spreads like a virus, mutating at each step of transmission while driven by the need for favorable press coverage and website advertising revenue … and we are all the poorer for it.
Downes, A., and T. P. Blunt. 1877.” Research on the effect of light upon bacteria and other organisms,” Proc. Royal Society of London 26:488-500.
Gates, F. L. 1930. “A study of the bactericidal action of ultra violet light: III. The absorption of ultra violet light by bacteria,” J. General Physiology 14:31-42.
Hollaender, A., and J. W. Oliphant. 1944. “The inactivating effect of monochromatic ultraviolet radiation on influenza virus,” J. Bacteriology 48:447-54.
Linden, K. G. 2001. “Comparative effects of UV wavelengths for the inactivation of Cryptosporidium parvum oocysts in water,” Water Science & Technology 34(12):171–174.
Rivers, T., and F. Gates. 1928. “Ultra-violet light and vaccine virus. II. The effect of monochromatic ultraviolet light upon vaccine virus,” J. Experimental Medicine 47:45-49.
Sturm, E., et al. 1932. “Properties of the causative agent of a chicken tumor. II. The inactivation of the tumor-producing agent by monochromatic ultra-violet light,” J. Experimental Medicine 55:441-444.
von Recklinghausen M., 1914. “The Ultra-Violet rays and their application for the sterilization of water,” J. Franklin Institute 178(6):681–704.
A word of caution: I am going to be annoyingly pedantic here, but with good reason. The lighting industry has a century-long history of introducing unfamiliar technologies using familiar terminology. We later come to regret our choice of words when it becomes necessary to express precisely what we mean.
Consider, for example, the term luminance. We understand this today to mean “luminous flux per unit solid angle per unit projected source area,” which we express in candela per square meter (where a candela is one lumen per steradian), or sometimes nits. Seventy years or so ago, however, it was common to refer to “brightness,” which today is considered a subjective attribute of a point light sources (and not to be confused with “lightness,” a subjective attribute of an area light source).
The lighting industry at first distinguished between “brightness” and “photometric brightness,” but eventually accepted “luminance” as the companion of radiance. (We will not talk about “helios,” an alternative proposed and rejected in the 1940s.) In the process, it also deprecated alternate units of measurement for luminance, including stilbs, lamberts, apostilbs, skots, brils and foot-lamberts.
π x 107
π / (0.3048)2
π x 10-4
π x 103
Table 1 – Deprecated luminance units
Luminance and candela per square meter – life is so much easier when we can agree on the terminology!
Courtesy of the current pandemic, the lighting industry is becoming all too familiar with a technology that has been in use since the 1930s – germicidal lamps emitting ultraviolet radiation that inactivate bacteria and viruses. Sadly, we are once again using familiar terminology that we will later have to deprecate in order to express what we really mean.
It does not have to be like this, however. It may be as futile an exercise as the legend of King Canute the Great ordering the tide to stop, but we can at least examine what terminology we should be using when discussing ultraviolet radiation as a means of disinfection.
We begin with the term ultraviolet, which designates a region of the electromagnetic spectrum beyond that of visible light with wavelengths of approximately 400 nm to 700 nm. The ultraviolet region is divided by the International Lighting Commission (CIE) into three subregions:
315 nm – 400 nm
280 nm – 315 nm
100 nm – 280 nm
Table 2 – Ultraviolet spectral regions
Here, however, is the point: “ultraviolet” is an adjective. It makes as much sense to refer to “ultraviolet” as it does to refer to “blue” – blue what? Just as we invariably refer to “blue light” in lighting design, we should refer to ultraviolet radiation. (For the record, light is “the natural agent that stimulates sight and makes things visible”; all else – including light – is electromagnetic radiation.)
From here, we move on to the term germicidal. This refers to the effect of UV radiation on pathogens, including viruses, bacteria, and fungi. Ultraviolet photons have enough energy to disrupt the DNA and RNA of bacteria and fungal cells, preventing the organisms from reproducing and thus inactivating them; eventually, they die. Viruses are not technically alive, but in disrupting their genetic code, the UV radiation prevents them from invading living cells and replicating, and thus inactivates them as well.
Unfortunately, this has led to the increasingly popular term germicidal UV, abbreviated GUV. Respecting the previous argument, a better term is ultraviolet germicidal irradiation (UVGI). This has a long history of use in the ultraviolet disinfection community, and the lighting industry should respect it as being unambiguous.
For anyone in the lighting industry, a product that emits visible light is a luminaire. … except that ultraviolet radiation is not light.
There are several types of ultraviolet radiation sources:
Low-pressure mercury-vapor arc lamps
These are basically linear or compact fluorescent lamps without phosphor coatings to convert ultraviolet radiation into visible light, and fused quartz or “soft” glass tubes that are transparent to UV-C radiation. They emit essentially monochromatic UV-C radiation with a wavelength of 254 nm. (The tubes are usually designed to block 185 nm radiation, which can generate toxic ozone.) They are typically used for “upper-room” air disinfection (CIE 2003), HVAC air ducts, and mobile ultraviolet disinfection platforms and robots.
Medium-pressure mercury-vapor arc lamps
These are similar to high-intensity discharge (HID) lamps, but feature broadband radiation centered around 250 nm. They are mostly used for water disinfection purposes.
Pulsed xenon lamps
These are basically high-power electronic flash lamps which emit intense broadband radiation across the entire ultraviolet spectrum with a peak near 230 nm. They are commonly used with mobile disinfection platforms in hospitals.
Krypton-chlorine (KrCl*) excimer lamps emit mostly ultraviolet radiation with a peak wavelength of 222 nm, and are being explored as a safer alternative to low-pressure mercury-vapor arc lamps.
Microplasma emitters are based on the essentially the same technology formerly used to produce plasma television screens, but are designed to emit ultraviolet radiation with a peak wavelength of 222 nm.
Ultraviolet-emitting LEDs are becoming increasing available, although to date with exceedingly low radiant efficiencies of one to four percent and peak wavelengths no shorter than 265 nm for commercial devices. However, the roadmap for the development of UV-C LEDs promises much greater efficiencies in the foreseeable future (Krames 2020).
For germicidal purposes, the optimal wavelength for disrupting the DNA of pathogens is 265 nm (Figure 1). Once again, however, this is electromagnetic radiation and not light. It may therefore be preferable to refer to UVGI sources rather than “luminaires.”
This should be an easy one. For proper UVGI systems design, we will need to know the intensity distribution of the UVGI sources, just like we have luminous intensity distributions for luminaires and photometric data reports.
Measuring these characteristics in the laboratory is not easy. Integrating spheres measuring two meters or more in diameter are often used to measure the luminous flux generated by luminaires, but the typical barium sulphate (BaSO4) coating is not sufficiently reflective for ultraviolet radiation measurements. Microstructured polytetrafluoroethylene (PTFE), perhaps best known as Labsphere Inc.’s diffuse reflectance standard Spectralon, has acceptable reflectance (> 95 percent at 250 nm), but it must be machined rather than applied as a coating.
Regardless, what is being measured is the radiant intensity distribution of the UVGI sources, with corresponding radiometric data reports. This further implies that these reports should be based on the ANSI/IES TM-33-18 luminaire optical data document specification rather than ANSI/IES LM-63-19 (and earlier editions) and EULUMDAT file formats, which are specific to photometric data (IES 2018). (The Italian UNI 11733:2019 standard is a mirror document to ANSI/IES TM-33-18.) This document specification was developed specifically with ultraviolet radiometry in mind.
More generally, we should always use radiometric units when referring to the performance of UVGI systems:
Luminous intensity (cd)
Radiant intensity (μW/sr)
Radiant fluence (mJ/cm2)
Table 3 – Photometric versus radiometric terminology
The International Bureau of Weights and Measurements defines radiometric measures in terms of watts, meters, and steradians, but UVGI systems typically express measurements in microwatts (μW) and square centimeters (cm2) for convenience (BIPM 2019).
If, as a professional lighting designer, you did a double take on the term radiant fluence, it is understandable – what is this? ANSI/IES RP-16-17 (ANSI 2017) defines “radiant fluence” as “the omnidirectional radiant energy externally incident on an elementary sphere per unit cross-sectional area in time Δt,” or more succinctly:
where dω is the differential solid angle, da is the cross-sectional area of the sphere, and Δt is the exposure time. (I warned you that I was going to be annoyingly pedantic.)
Fortunately, this metric is much simpler conceptually than its formal definition implies, and it is crucial to an understanding of UVGI system performance.
The SARS-CoV-2 virus is transmitted primarily by respiratory droplets and aerosols that are generated by the simple acts of breathing, talking, sneezing, coughing and singing. The same is true, however, for many other viruses, bacteria, and fungi, including those responsible for the common cold, influenza, measles, chickenpox, and tuberculosis. Respiratory droplets, which are typically larger than 5 μm and consist mostly of water, fall to the ground rapidly after being produced. Aerosols, on the other hand, are relatively dry and may persist in the air for several hours.
Imagine then a droplet or aerosol as a transparent sphere holding the viruses or bacteria in solution. Seen from any direction, the cross-sectional area is defined by the equation for the area of a circle: π x r2, where r is the sphere radius.
Now think of a parallel beam of ultraviolet radiation in this direction. The beam has a measurable radiance L that is expressed in microwatts per steradian per square centimeter (μW/sr-cm2). Conceptually, it is the radiant power of the ultraviolet photons passing through the transparent sphere in the given direction at a specific moment in time.
If we sum the beam radiance for all possible directions, we have the irradiance of the sphere. We can make the diameter of the sphere infinitesimally small, in which case we have the spherical irradiance (aka the fluence rate) of the elementary sphere at a point in free space, measured in microwatts per square centimeter (μW/cm2).
It is not however the irradiance that is important for germicidal action, but rather the radiant dose or fluence. The aerosol particle may drift through the air and thus be exposed to varying levels of spherical irradiance. The longer the particle is exposed to radiation, however, the more likely it is that a UV photon will intersect the DNA or RNA of a virus (which typically has a diameter of less than 100 nanometers) and disrupt it.
Fluence then is the spherical irradiance (measured in μW/cm2) of the particle integrated (i.e., summed) over time (measured in seconds). One joule (energy) is one watt (power) times one second. For UVGI purposes is typically expressed in millijoules per square centimeter (mJ/cm2).
Fluence is defined specifically for aerosols (and more generally points in free space), but the concept of radiant dose also applies to the irradiation of surfaces. For a uniform distribution of radiance (i.e., equal in all directions), the irradiance of a surface is 1 / π times that of spherical irradiance without the surface at that point.
Choosing Radiant Dose
When the UV dose results in a 90 percent disinfection (10 percent survival), it is referred to as D90. For UVGI applications where higher disinfection rates are needed, D99 (i.e., 99 percent disinfection) is often used. To completely disinfect a surgical tool, it is common to expect 99.9999 percent disinfection. There may still be a few active viruses or bacteria that remain, but their density is too low to achieve an infection or replicate in a colony.
The decrease in survival with larger UV doses is exponential – if 10 mJ/cm2 results in 90 percent disinfection, 20 mJ/cm2 results in 99 percent, 30 mJ/cm2 in 99.9 percent, and so on. This is an example of exponential decay, and so it is more convenient to refer to the disinfection in logarithmic units: D90 is log-1 , D99 is log-2, and so on to log-6 for complete disinfection.
There are however complications in determining an appropriate UV dose for inactivating pathogens. Table 4 (adapted from Kowalski 2009) presents the average radiant doses required to achieve D90 disinfection for pathogens:
Air – Low RH
Air – High RH
Table 4 – Average UV dose in mJ/cm2 required to achieve D90 inactivation. (RH – relative humidity, expressed in percent.)
The first complication is that the pathogen species and its environment matters. Hundreds of studies have reported on the susceptibility of many different pathogen species to either monochromatic (254 nm) or broadband UV irradiation, where the pathogens are: a) suspended in water; b) cultured on a growth medium on a surface (e.g., a Petri dish); or c) suspended in airborne droplets or aerosol particles of various sizes. As shown in the table, the relative humidity of the air can have a very significant effect on vegetative bacteria (but not viruses).
The second complication is that Table 4 represents the average dose for D90 disinfection. Depending on the pathogen species and environment, the required dose can vary by a factor of 100 or more from these averages. Designing a UVGI system with an average radiant dose does not guarantee protection from specific pathogens. The choice of narrowband versus broadband radiation sources also matters, especially for bacteria.
The third complication is that once D90 disinfection has been achieved, it is often the case that the surviving 10 percent of the population is an order of magnitude more resistant to UV irradiation. For example, if it takes 10 mJ/cm2 to achieve D90 disinfection, it may take 40 mJ/cm2 to achieve D99 disinfection.. As shown in Figure 2, such pathogens are said to have a “two-step” rare constant for their susceptibility to UV-C radiation.
A fourth complication is that when vegetative bacteria and fungi have been exposed to monochromatic UV radiation as a means of disinfection, subsequent exposure to visible light may enable the “killed” cells to repair their DNA and recolonize, thereby increasing their survival rate by 10 to 100 times, especially if the pathogens are suspended in water or present on surfaces in the presence of high relative humidity. (Viruses do not seem to have the complexity needed to effect self-repair of their DNA.) This may be a concern if, for example, the exposed surfaces of a hospital room are decontaminated with a mobile UV disinfection robot, but the room is flooded with direct sunlight thereafter.
A fifth and final complication is that once a surface has been infected with respiratory droplets, any bacteria may find sufficient resources to begin colonizing the surface. The surface may be continuously irradiated by, for example, UV‑C radiation from a microplasma emitter or UV-LED array, but if the irradiance is too low, the surface may not achieve even D90 decontamination, regardless of the exposure time.
There is admittedly a considerable amount of information here that extends beyond the bounds of ultraviolet radiation terminology. It is needed, however, to put the terminology used for decades by the ultraviolet disinfection community into context. Ultraviolet radiation is not visible light, and so the lighting community needs to both understand and respect the terminology when adopting UVGI system design practices for building safety and human health. If we do not use the correct terminology, we risk (and deserve) a plague of apostilbs, brils, lamberts, skots and other annoyances later down the road.
Lighting designers will be familiar with the illuminance of a planar surface, which is measured in lumens per square meter (or foot). The irradiance of a planar surface by a germicidal radiation source is conceptually the same, except that it is measured in watts per square meter (W/m2). Most designers, however, will not be familiar with the concept of the spherical irradiance (or equivalently, fluence rate) of a point in space. This is an essential metric for air disinfection by germicidal lamps, and so it needs to be understood.
Figure 1 shows a parallel beam of UV radiation A with a cross-sectional power density of ΦA watts of power (flux) per square meter (W/m2) irradiating a surface oriented perpendicular to the beam. The beam can be made infinitesimally narrow, thereby irradiating a point on the surface, with irradiance EA = ΦA W/m2.
Another parallel beam B with ΦB watts per square meter intersects the surface with incidence angle θ, the surface irradiance due to this beam being EB= (ΦBcos θ) W/m2. If we sum the contribution of beams from all possible directions above the surface of the plane, we obtain the surface irradiance ES. For germicidal applications, this irradiance is typically measured in microwatts per square centimeter (µW/cm2).
Irradiance is an appropriate metric for pathogens cultured on a plate, but aerosols are miniscule droplets of water suspended in air. If these droplets have been expelled by someone with an infectious respiratory disease such as tuberculosis or influenza, they may contain bacteria or viruses. These pathogens can be inactivated by UV radiation, but there is no planar surface that is being irradiated, and so the concept of irradiance does not apply.
Figure 2 represents the droplet as a transparent sphere S that does not absorb or refract UV radiation. The sphere has radius r, and so has a cross-sectional area π r2. If this sphere is irradiated by a beam A with cross-sectional power density ΦA W/m2, its irradiance is the same as the beam power density (again typically measured in µW/cm2). The diameter of the sphere can be made infinitesimally small, yielding an elementary sphere – a point in space – while the irradiance remains constant.
If we sum the contribution of beams from all possible directions of an imaginary sphere surrounding the point, we obtain the spherical irradiance of the point in space. This is also referred to as the fluence rate, which, when multiplied by the exposure time (and assuming a constant rate), yields the fluence of the droplet (typically measured in millijoules per square centimeter, mJ/cm2). If the fluence rate varies (as when the droplet moves through a three-dimensional UV radiation field), the fluence is the sum of the fluence rates multiplied by the chosen time intervals, which may vary from milliseconds to hours.
Finally, the term germicidal dose applies to both the irradiance of surfaces and the spherical irradiance of aerosols, but the terms fluence and fluence rate apply only to aerosols.
About Those Standards …
The concept of spherical irradiance may be simple to understand, but only if you ignore the official definitions. The CIE International Lighting Vocabulary has taken the approach that more is better, offering a choice of:
Luminous spherical exposure (see 17-1028)
Spherical illuminance (see 17-1245)
Fluence (see 17-1028)
Fluence rate (see 17-1245)
Radiant fluence (see 17-1028)
Radiant fluence rate (see 17-1245)
Radiant spherical exposure
Photobiological fluence (see 17-1028)
Photobiological fluence rate (see 17-1245)
Photon fluence (see 17-1028)
Photon fluence rate (see 17-1245)
Photon spherical exposure (see 17-1028)
Photon spherical irradiance (see 17-1245)
In other words, spherical irradiance (or fluence rate) and fluence with different spectral weightings such as V(λ) for luminous quantities and photobiological action spectra, including photon quantities for horticultural applications.
Trying to understand the CIE definition of fluence rate can be a challenge:
Quantity defined by the formula:
where dΩ is the solid angle of each elementary beam passing through the given point and Le its radiance at that point.
This makes sense if you think of an “elementary beam” as being an infinitesimally narrow cone (which makes it equivalent to the infinitesimally narrow beam discussed above), but there is the unnecessary complication of defining the beam radiance in terms of point sources. It is conceptually much easier to begin with parallel beams that can be made infinitesimally narrow.
ANSI/IES RP-16-17, Nomenclature and Definitions, on the other hand, is decidedly more spartan:
Luminous fluence rate
Radiant fluence rate
Spectral radiant fluence
Spectral radiant fluence rate
In addition to being spartan, the IES definitions are both clear and concise:
Fluence rate: The omnidirectional radiant flux externally incident on an elementary sphere about the point, per cross-sectional area of the sphere.
Fluence: The omnidirectional radiant energy externally incident on an elementary sphere about the point, per cross-sectional area of the sphere.
The advantage of these definitions is that they make no mention of radiance or solid angles.
Measuring Spherical Irradiance
CIE-ILV 17-1245, Spherical Irradiance, includes this helpful note:
This is the appropriate radiometric quantity for describing a dose rate for a photobiological or photochemical effect in a scattering medium (e.g., light in skin). It is also the appropriate quantity for describing the irradiation of microorganisms. It is frequently used incorrectly as a substitute for irradiance in some publications.
This leads to the obvious question: how do you measure spherical irradiance?
One approach is to use a spherical actinometer, an instrument that consist of a hollow quartz sphere measuring a centimeter or so in diameter that is filled with a solution of ferrioxalate, persulfate, or iodide/iodate; their transmittance after exposure is linearly proportional to the UV-C fluence (e.g., Kowalski 2009).
Another option is to use a radiometer with a “scalar irradiance” (yet another synonym for fluence rate) collector and a narrowband ultraviolet filter, such as the AMOUR radiometer manufactured by Biospherical Instruments (San Diego, CA). However, such an instrument has a “blind spot” of approximately 80 degrees where the spherical Teflon collector is mounted on its connector shaft.
UPDATE 20/08/28 – The manufacturer has stated that their AMOUR radiometer is not capable of measuring UV-C radiation due to excess absorption by the Teflon collector.
It may be difficult to measure spherical irradiance, but it is possible to predict its three-dimensional distribution in space using modified lighting design and analysis software, including contributions of direct radiation from UV-C sources and interreflections from surfaces. However, as Pierre de Fermat said regarding his Last Theorem, “I have discovered a truly remarkable theorem … which this margin is too small to contain.” More details to follow …
Thanks to Dawn DeGrazio of the Illuminating Engineering Society for her review and comments.
Kowalski, W. 2009. Ultraviolet Germicidal Irradiation Handbook: UVGI for Air and Surface Disinfection. Heidelberg, Germany: Springer.