UVC reported in sunlight at Earth's surface is a measurement artifact

Maybe that last word should be 'anti-fact'

Before I start, I must apologise for the garbled text in my last post. If you didn’t manage to work out what I meant, please see my corrected version of record. That’s the version you see on the Substack App (a good reason to install it).

Herndon and co have been at it again. This time they claim to have ‘unequivocally’ measured large amounts of UVC reaching Earth’s surface. More even, than at some wavelengths in the UVA where absorptions by gases in the atmosphere are small. Where is it coming from then? To demonstrate how silly their assertion was, I asked my colleague, Ben Liley, if he could take the figure from their paper that they use to make this claim, and overlay the spectrum of sunlight arriving at the top of Earth’s atmosphere.

His update plot of irradiances as a function of wavelength is shown below. Notice that the y-axis is not linear. It has a logarithmic scale that spans 4 orders of magnitude. Ben’s new line is the orange one. Measurements similar to that orange line have been produced from multiple instruments viewing the Sun from outside Earth’s atmosphere since the satellite era began.

Herndon’s new results – shown by the red and black lines - all look plausible in the UVA region (i.e., at wavelengths between 315 and 400 nm). There, the effects of absorption due to gases in the Sun’s atmosphere – the ‘Fraunhofer lines’ - match the orange curve nicely. Their lines are a bit deeper because of their higher spectral resolution. And the steep drop-off towards shorter wavelengths in the UVB region is consistent with the drop-off expected due to absorption by ozone. That drop-off is similar to what you see in my mast-head plot above (which is plotted on a linear y-axis, rather than the log scale in Herndon’s plot).

But things clearly start to go wrong at wavelengths less than 300 nm. Instead of a continued decrease towards shorter wavelengths, the ‘measurements’ all show an increase. By wavelength 200 nm, their results exceed the solar output by more than a factor of twenty. Something must be wrong. At Earth’s surface, you just can’t get twenty times more radiation than what’s available at the top of the atmosphere.

In the real world, as I’ve shown previously, the measured values should be many orders of magnitude smaller than the orange curve throughout the UVC region because of absorption by gases like ozone and oxygen in Earth’s atmosphere. And, as Ben astutely noted, throughout that entire UVC region there are no corresponding Fraunhofer absorption lines visible in the red and black spectra. So it’s definitely not sunlight they’re measuring there.

To understand what’s going wrong, you need to know a bit about how these instruments work. Spectra like these are measured with optical instruments that include a dispersion element, like a quartz prism or a diffraction grating, to disperse the incoming white light into the colours of the spectrum (the rainbow of colours). The different wavelengths (colours) can be sampled in two ways:

1.     by rotating the prism or grating to allow different wavelengths of the light to sequentially pass through a fixed exit slit with a detector behind, or

2.     by imaging the spectrum directly onto an array detector that measures all wavelengths of interest at once.

In sunlight, array detectors give spurious results at wavelengths less than about 300 nm because the amount of radiation there is too small compared with the amount present at longer wavelengths. Due to imperfections in manufacturing gratings or prisms, there’s always some ‘stray light’ present inside the instrument. And when the signal gets too small, as it does at wavelengths less than 300 nm, the signal that arrives at the detector is dominated by that stray light.

The spectra measured at Earth’s surface by Herndon (red and black curves), and by D’Antoni and colleagues back in 2007 (green, blue and purple curves) all use array-detectors, rather than the photomultiplier detectors needed for accurate measurements of the Sun’s radiation arriving at the Earth’s surface in the UVB (and UVC) region.

Those increased values in the UVC region are due to a mixture of wavelengths of stray light from longer wavelengths. If those UVC values were truly representative of the wavelength scale, they would show the same Fraunhofer absorptions as in the orange curve. But they don’t. In fact, at all wavelengths less than about 300 nm, any fine-scale structure looks more like noise, which is what it is.

The apparent increase in deduced irradiance towards shorter wavelengths occurs because of the reduced sensitivity of the instrument at those wavelengths. To convert from the electric output of the diode current (say, in amps) to irradiances (say in Wm-2), you need to divide by the sensitivity (amps per Wm-2). You therefore have to divide by smaller and smaller numbers as the wavelength decreases.

That stray light is less of a problem with these most recent results than for D’Antoni’s earlier work, because they now use a double dispersion instrument. As we’ve previously commented, D’Antoni’s earlier published UVC values were too high by several orders of magnitude (factors of ten). At least the new results are a step in the right direction. For example, at 200 nm the most recent values are lower by about a factor of a hundred than the earlier results from D’Antoni, and are therefore closer to reality. But they’ve got a way to go. They’re still too high by a few orders of magnitude.

The problem is that even with their double dispersion instrument, there’s still a lot more stray light than with a true double monochromator where there’s a narrow intermediate slit between the two dispersing elements (i.e., the diffraction gratings or prisms). That slit transmits only the wavelength of interest at any point in the scan, which is then detected behind the exit slit with a photomultiplier. The spectrum is built up over time by rotating the dispersion elements to sequentially transmit each wavelength in turn through the exit slit to the detector. Instead, when using an array instrument that captures all of the spectrum at once, that intermediate slit has to be much wider – so allowing much more stray light to get through.

Here’s a suggestion for Herndon’s team.

While your instrument is set up to measure and display the spectrum for sunlight, insert into the light path, a glass filter that transmits no UVC, and see if the fraction of UVC ‘measured’ disappears as it should. I’ll bet a pound to a pinch of sh*t that it doesn’t.

To convince yourself, choose a filter from a reputable company where the spectral transmission of the filter glass has been measured in the laboratory. Bandpass filters like the ‘BG’ series manufactured by the Schott Glass Company in Germany would be ideal. They also produce many other long pass filters and neutral density filters that would do the same job. My colleague Germar Bernhard at Biospherical Instruments just up the road from you in San Diego probably has one you can borrow. If that’s too hard, you could try using a piece of ordinary glass as the filter. That should do the trick, but you might have to trust me in saying that it transmits no UVC (or you could search the web for verification).

I don’t expect an apology, but it would be good if Herndon and his team could at least set the record straight in the literature. Bad data are worse than no data.

But maybe not entirely useless in this case. At least their data show that D’Antoni’s bad results at 200 nm were too high by a factor of at least a hundred. Perhaps another retraction will be in order?

Note added 17 May 2023: As a reminder, here’s what the solar spectra should really look like in the UVC region, as we published way back in 2008. The plot shows that at 200 nm the amount reaching Earth’s surface at 200 nm is at least 4 orders of magnitude less than the extraterrestrial value. It’s actually many more orders of magnitude smaller. By my calculations, the irradiance at 200 nm will be at least another 15 orders of magnitude smaller (!!) than minimum of the y-axis on the plot below.