Spectral Engineering: Which Blue is Best?

Abstract:

The growing body of evidence showcasing physiological effects of light mediated by ipRGCs (intrinsically photosensitive retinal ganglion cells) has led to the emergence of spectral engineering to balance the desired visual and physiological effects of light, generally based on time of day (i.e. daytime spectrum and nighttime spectrum). However, end users and designers are left to decide what of this spectral engineering is worthwhile and what is marketing fluff. In order to truly find out, we must investigate the interaction between the visual sensitivities and the physiological sensitivity.

Here I show that typical LED lighting which has a peak emission around 450nm, is aligned with the maximum sensitivity of the blue perception. Meaning that, much to the surprise of most people, LED light has the LEAST amount of blue light compared to any other technology (i.e. fluorescent, incandescent, etc) of the same color temperature, despite its “peaky” appearance. Conversely, for daytime illumination tailored spectrum can provide a significant benefit for any given color temperature, with 490nm peak illumination achieving the highest amount of daytime stimulation.

Color temperature provides some guidance for biological potential. Colder has higher capacity for biological stimulation, whereas warmer color temperatures have lower capacity. Combining these two concepts: when designing for nighttime, a standard 450nm peak will have the least biological stimulation, and you should strive for the warmest acceptable color temperature. For daytime stimulation, the coldest acceptable color temperature with a peak blue emission at 490nm would be optimal.

It should be noted that other parameters play a role in these physiological effects of light, such as intensity, spatial distribution, duration and timing. However, the goal of this article is simple to look at spectrum.

Background:

The discovery of a novel photoreceptor in the eye has changed our understanding of the role of light in our lives. In 2001, Dr. Bud Brainard showed evidence for a non-rod non-cone photoreceptor ̶ the ipRGC ̶ affected melatonin release [1]. However, it is now understood that these ipRGCs are responsible for several physiological effects which can be broken down into two categories – the direct and indirect effects of light [2].

The primary indirect effects are related to circadian synchronization and tracking seasonal changes, where light is understood to be the strongest signal for synchronizing the circadian clock [3]. This makes sense considering the solar cycle is by far the most prevalent and predictable signal we have. The problem we face today is that we spend most of our time indoors, removed from this signal, which is too dim to be considered daytime and too bright too be considered night [4, 5]. This lack of delineation between day and night has led to more night owl tendencies [6], which have been shown to lead to a whole host of concerns: decrease in learning and attention, increased risk of obesity, addiction and cardiovascular disease [5,7,8,9]. Moreover, this appears to be a widespread phenomenon touching many people. It has been shown that 87% of non-shift workers have some form of circadian dysfunction and the associated health risks previously mentioned [5].

The direct effects of light include increases alertness, working memory, and mood [2]. This is attributed to an increase in stimulation of our ipRGCs, with the majority of these affects shown to be mediated directly by melanopsin [10, 11] – a photopigment with a peak sensitivity around 490nm [12]. Unfortunately, we spend so much time indoors that we are not receiving enough of these beneficial light signals during the daytime.

From a lighting technology standpoint, these newly discovered responses have led to some confusion regarding the non-visual impacts of LEDs on our physiology. Some people will claim that LEDs’ contain a large blue peak making, sending too many daytime signals and thus deeming them unsuitable for nighttime use (see figure 1) [13]. Others showcase a trough in the melanopic (sky blue) region and cite this dip as LEDs’ fundamental shortcoming, not providing enough daytime signals, making them unsuitable for daytime use. In order to truly understand the answer, we must understand how color vision interacts with these ipRGCs.

Figure 1. Comparison of different light sources, in a peer-reviewed journal, can be misleading to over-emphasize the blue light in LED [13]

Lighting Technology to the Rescue:

Light technologies have tried to come up with a solution to this modern-day problem of insufficient circadian exposure in a handful of different ways. The most notable has been by offering light fixtures that change the color temperature of the light over the course of the day. More recently however, the understanding of this novel photopigment with peak sensitivity at 490nm has led to a focus on spectral engineering – Designing white light to either maximize melanopsin stimulation for daytime exposure or minimize it for nighttime exposure. Several lighting manufacturers offer claims that something significant has been done with their blue light content that will impact physiology, professing a certain “amount” of blue light or that the blue pump has been removed and replaced with a violet pump. Unfortunately, some of these claims come with their own unique definition of what blue light means, which only adds to the confusion about the role of ‘blue’ light in LEDs. A simpler definition of blue light is any radiation from 400-500nm [14].

As interest in healthy spaces and circadian lighting continues to grow, lighting manufacturers are looking to either maximize or minimize the amount of melanopic stimulation provided per unit of visual stimulus. This is calculated as an M/P ratio; – a weighting factor used to convert photopic lux (or fc) into melanopic lux (or fc). This is done such that an equal energy spectrum (flat line spectrum) will result in an M/P ratio of 1. It should be noted that CIE has a similar weighting function, Daylight Equivalency Ratio (DER), which weight such that D65 (6500K daylight) would achieve a ratio of 1.

In application, we want to maximize M/P or DER during the daytime and minimize it in the evening. Manufacturers “spectrally engineer” primarily by changing the location of the blue peak. The following data will demonstrate how manipulating the location of “blue” light to short wavelength blue light (i.e. 410nm) , or long wavelength blue, such as 490nm, may not actually have a critical impact on melanopsin stimulation. The reason for this is that there is an interaction with color vision that must be considered during this analysis.

Color Matching Functions:

Color matching functions (see figure 2A) are used to convert any spectral power distribution into a point on the color space diagram (i.e a color). Using a dot product of any spectra and we can determine the tristimulus values X, Y, and Z. Additionally, it should be noted that Y is also the luminous efficiency function. This is then converted to (x,y) on the CIE 1931 Color Space (figure 2B) via the following equations:

𝑥=𝑋 𝑋+𝑌+𝑍

𝑦=𝑌 𝑋+𝑌+𝑍

Increasing the value of any of the tristimulus variables has the following implications: – More Y is going to make the color appear more green, but also add more lumens. -More X is going to make the color appear more red.
-More Z is going to make the color appear more blue.

Figure 2. Top (A) is the color matching functions X (red), Y (green), and Z (purple) used to convert SPD into the (x,y) color point on the CIE 1931 color space diagram shown on the right with black body locus shown in black (B). Melanopsin is plotted in blue. Y (green) also serves as the luminous efficiency function.

In a traditional LED spectrum (figure 3) there appears to be “huge” peak at 450nm, similar to what is shown in figure 1, followed by a trough in the 480-500nm region, and a broad mound with a lot of energy around 550nm. This “huge” peak is actually a blue LED “pump” that is used to excite a broad phosphor or phosphors. The phosphor wants to have the majority of its weight around peak Y, as this will yield the most lumens. Interestingly at the blue peak of 450nm, the Y is very low which means there are very little lumens in 450nm light. Thus, in order to achieve the most efficiency at any given color point, we would want to use blue minimally and really just use it to shift the color point to the black body locus in order to create the desired white light. In fact, 450nm is near the peak sensitivity of Z, the “blue” direction. This blue peak is purposely located at 450nm in order to take advantage of our peak color perception sensitivity and consequently uses the LEAST amount of blue light. This is contrary to common perception, because those blue peaks are very tall, but do not contain a lot of energy in them. Note: cooler color temperature white light requires a larger shift and thus more blue light, while warmer color temperature light needs less blue light.

Figure 3. A 6500K LED overlaid on top of the color matching functions and melanopsin.

Blue Shifting Example:

An example of how a blue light is used to shift a color point is shown in figure 4. Here we have an amber phosphor and apply a 410nm peak, such that the peak emission of blue to the peak emission of amber is 1:1. We see a shift towards (even past) the black body locus that defines white light.

Figure 4. Amber phosphor (A) and its color point (via red arrow) on CIE 1931 color space (C). (B) The same amber phosphor is A, but with a 410nm blue LED use to shift the color point. The resulting color point is shown (via red arrow) on CIE 1931 color space (C).

Now let’s look at what happens if we shift this pump to longer wavelengths. Figure 5 shows that by shifting from 410nm to 420nm of same peak intensity, we achieve twice the color shift. We should notice two things. First, when the shift occurs from 410nm to 420nm, the Z has a much steeper slope than melanopsin. This means color point is being affected more than melanopsin.

Figure 5. Amber phosphor (A) and its color point (via red arrow) on CIE 1931 color space (C). (B) The same amber phosphor is A, but with a 410nm blue LED use to shift the color point. (C) The same amber phosphor is A, but with a 410nm blue LED use to shift the color point. The resulting color point is shown (via red arrow) on CIE 1931 color space (D).

Figure 6 showcases stepping that peak in 10nm increments. By going from 420nm to 430nm achieves another equally sized color shift as what was shown in figure 5. Now going to 440nm the shift becomes smaller as it begins to enter the “plateau region” of the Z color matching function. 450nm is not much different than 440nm. Then something interesting happens at 460nm – the shift starts to make a U-turn and heads back towards the black body

locus. We can see that 470nm has a smaller shift than 460nm (closer to the BBL), 480nm has a smaller shift than 470nm and 490nm has a smaller shift than 480. In fact, 490nm and 410nm have similar shift magnitudes.

Figure 6. Amber phosphor (A) and its color point (via red arrow) on CIE 1931 color space (K). (B) The same amber phosphor is A, but with a 410nm blue LED use to shift the color point. The resulting color point is shown (via red arrow) on CIE 1931 color space (K). (C) The same amber phosphor is A, but with a 420nm blue LED use to shift the color point. The resulting color point is shown (via red arrow) on CIE 1931 color space (K). (D) The same amber phosphor is A, but with a 430nm blue LED use to shift the color point. The resulting color point is shown (via red arrow) on CIE 1931 color space (K). (E) The same amber phosphor is A, but with a 440nm blue LED use to shift the color point. The resulting color point is shown (via red arrow) on CIE 1931 color space (K). (F) The same amber phosphor is A, but with a 450nm blue LED use to shift the color point. The resulting color point is shown (via red arrow) on CIE 1931 color space (K). (G) The same amber phosphor is A, but with a 460nm blue LED use to shift the color point. The resulting color point is shown (via red arrow) on CIE 1931 color space (K). (H) The same amber phosphor is A, but with a 470nm blue LED use to shift the color point. The resulting color point is shown (via red arrow) on CIE 1931 color space (K). (I) The same amber phosphor is A, but with a 480nm blue LED use to shift the color point. The resulting color point is shown (via red arrow) on CIE 1931 color space (K). (J) The same amber phosphor is A, but with a 490nm blue LED use to shift the color point. The resulting color point is shown (via red arrow) on CIE 1931 color space (K).

In order to quantify how this color shift interacts with melanopic content, we can compare color shift and M/P ratio (figure 7). We can then look at this interaction in terms of melanopic content per color shift (M/P ratio divided by x,y distance from phosphor point). Think of this as the most or least melanopic value we can get for that color shift. We find two interesting things from this exercise. First, there is no difference between melanopic contributions per color shift from the 410nm peak and 450nm peak (both have a unitless ratio of about 2.5). In addition, if we were to find the low point, it would actually be at 430nm, not 410nm, but the value of shifting those 20nm is probably minimal (1.7 versus 2.5). Moreover, we see that longer wavelengths really start to increase the value, where 490nm is about 3 times better than 450nm at providing melanopic content.

Figure 7. (A) shows the color shift distance in grey and the M/P ratio in yellow. (B) shows the M/P ratio per color shift. What does this mean?

1. The big scary 450nm blue peak actually contains the least amount of blue for any given color temperature (i.e. total color shift). This also means it contains close to the least amount of daytime signals (M/P) for any given color temperature. The only real way to reduce the blue any further is by shifting color temperature to be as warm as possible, which would be useful for nighttime settings.
2. Using cooler color temperatures of LED spectra with a 450nm peak is going to yield one of the worst M/P ratios for that color temperature. Using a peak emission closer to 490nm is going to get you the most benefit for any given color temperature. This means we can create melanopic rich white light that aligns with architectural lighting color temperature preferences for daytime spaces, instead of requiring the use of cooler color temperatures such as 5000K or 6500K in interior spaces that may be deemed undesirable.

Shown a different way in figure 8, if we normalize the color shift magnitude, we can compare what the spectra might look like to get the same shift magnitude as the 410nm shift shown in figure 4. We can see in figure 7A we need about a third of the peak of 410nm at 450nm to achieve the same magnitude color shift. When we look at what this means from an M/P ratio standpoint, in figure 7B, we see again that the M/P ratio stays pretty flat from 410nm to 450nm, and significantly increases from 460nm, with 490nm exhibiting the highest M/P ratio. Note: the magnitude for these shifts is the same, but the color temperature is slightly cooler with longer wavelength.

Figure 8. (A) shows the color shift distance in grey and the M/P ratio in yellow. (B) shows the M/P ratio per color shift. Note, the phosphor is the same in each SPD.

Concluding Statements:

What I have outlined above is that the blue “peak” we see in traditional LEDs is actually a method for energy efficiency to use the least amount of blue light for maximum visual effect. Moreover, it should be noted that “violet” pumps with blue phosphors can be used; these phosphors will be broader (i.e. less spiky), but ultimately will not achieve anything different than what has been shown here. If the energy in the violet diode is broader amongst the blues, it will exhibit a similar trend. If its peak is around 490nm, it will have more melanopic content than if it peaks at 450nm. Based on these findings we have the following recommendations for circadian lighting applications:

Nighttime Use: Despite a 450nm peak in traditional LEDs that may look as if its providing a significant amount of blue signal, closer analysis reveals that it in fact provides almost the least amount of melanopic signals of any blue peak spectrum that creates the same color white light. This means that going to shorter wavelength blues (or violets) does not achieve much melanopic benefit because we need significantly more of these shorter wavelengths to produce the same visual appearance. Moreover, producing significantly more of these short wavelengths require significantly more energy to do so. Thus, I believe that for nighttime use 450nm is the best blue. Additionally, I strongly recommend using the warmest acceptable color temperature to further limit the blue light content.

Daytime Use: Using a white light source with a 490nm peak will provide the most melanopic impact of any blue spectrum that creates the same white light. We recommend a high melanopic lux white light source while also using the coolest acceptable color temperature to maximize this blue content.

The graph below provides an illustrated summary for this paper (figure 9).

Figure 9. A review of figure 5 with daytime suitable and nighttime suitable spectral engineering. For nighttime use combine a 450nm peak with as warm of a color temperature as is acceptable. For daytime, you want to target a peak at 490nm and combine with as cool of a color temperature as is acceptable.

References:

1. Hughes, S., Jagannath, a, Brainard, G. C., Hanifin, J. P., Greeson, J. M., Byrne, B., Glickman, G., Gerner, E., & Rollag, M. D. (2001). Action spectrum for melatonin regulation in humans: evidence for a novel circadian photoreceptor. The Journal of Neuroscience : The Official Journal of the Society for Neuroscience, 21(16), 6405–12. http://doi.org/21/16/6405 [pii]
2. Rodgers, J., Hankins, M. W., Peirson, S. N., & Foster, R. G. (2016). Signalling by melanopsin (OPN4) expressing photosensitive retinal ganglion cells. Eye, 44(November 2015), 1–8. http://doi.org/10.1038/eye.2015.264
3. LeGates, T. A., Fernandez, D. C., & Hattar, S. (2014). Light as a central modulator of circadian rhythms, sleep and affect. Nature Reviews Neuroscience, 15, 443. Retrieved from https://doi.org/10.1038/nrn3743
4. Espiritu, R. C., Kripke, D. F., Ancoli-Israel, S., Mowen, M. A., Mason, W. J., Fell, R. L., … Kaplan, O. J. (1994). Low illumination experienced by San Diego adults: Association with atypical depressive symptoms. Biological Psychiatry, 35(6), 403–407. http://doi.org/10.1016/0006-3223(94)90007-8
5. Roenneberg, T., & Merrow, M. (2016). The Circadian Clock and Human Health. Current Biology, 26(10), R432–R443. http://doi.org/10.1016/j.cub.2016.04.011
6. Skeldon, A. C., Phillips, A. J. K., & Dijk, D.-J. (2017). The effects of self-selected light-dark cycles and social constraints on human sleep and circadian timing: a modeling approach. Scientific Reports, 7(September 2016), 45158. http://doi.org/10.1038/srep45158
7. Zarrinpar, A., Chaix, A., & Panda, S. (2015). Daily Eating Patterns and Their Impact on Health and Disease. Trends in Endocrinology & Metabolism, xx(2), 1–15. http://doi.org/10.1016/j.tem.2015.11.007
8. Roenneberg, T., Allebrandt, K. V., Merrow, M., & Vetter, C. (2012). Social jetlag and obesity. CurrentBiology, 22(10), 939–943. http://doi.org/10.1016/j.cub.2012.03.038
9. Sulli, G., Manoogian, E., Taub, P., Panda, S. (2018). Training the Circadian Clock, Clocking the Drugs,and Drugging the Clock to Prevent, Manage, and Treat Chronic Diseases. Trends Pharmacol Sci. 2018Sep;39(9):812-827. doi: 10.1016/j.tips.2018.07.003.Epub 2018 Jul 27.
10. Fernandez,D.Fogerson,P.Lazzerini,O.Thomsen,M.Layne,R.Severin,D.Zhan,J.Singer,J.Kirkwood,A. Zhao, H. Berson, D. Hattar, S. (2018). Light Affects Mood and Learning through Distinct Retina-BrainPathways. Cell. 2018 Sep 20;175(1):71-84.e18. doi: 10.1016/j.cell.2018.08.004. Epub 2018 Aug 30.
11. Gaggioni,G.,Maquet,P.,Schmidt,C.,Dijk,D.,&Vandewalle,G.(2014).Neuroimaging,cognition,lightand circadian rhythms. Frontiers in Systems Neuroscience, 8(July), 126.http://doi.org/10.3389/fnsys.2014.00126
12. Lucas,R.J.,Peirson,S.N.,Berson,D.M.,Brown,T.M.,Cooper,H.M.,Czeisler,C.A.,…Brainard,G. C. (2014). Measuring and using light in the melanopsin age. Trends in Neurosciences, 37(1), 1–9. http://doi.org/10.1016/j.tins.2013.10.004
13. Tosini,G.,Ferguson,I.,&Tsubota,K.(2016).Effectsofbluelightonthecircadiansystemand eyephysiology, (August 2015), 61–72.
14. Garcı,D.,Vicente-tejedor,J.,Marchena,M.,Ramı,L.,Sa,C.,Villa,P.De,&Germain,F.(2018). Removal of the blue component of light significantly decreases retinal damage after high intensity exposure, 1–18.