Double Axis Grating Solar Eclipse

SpectroBurstTM spectra are generated using stacked, mutually rotated gratings. Typically, the gratings are double-axis gratings, featuring diffracting spots spaced equally in the x and y directions. During the Great American Solar Eclipse of August 21, 2017, we at SpectroClick tried to get pictures of the eclipse through our gratings. The attempt through a Nikon D50 digital camera was a disaster; the images were blurry and useless. Pictures through a cell phone camera with a handheld SpectroBurst Viewer looked better, but still didn’t have the pizzazz we wanted. Fortunately, the second Great American Solar Eclipse occurred on April 8, 2024. The weather cooperated; in Olney, IL there was some high cirrus cloud but otherwise perfect clarity and temperature. We mounted a Samsung S20 5G phone on a tripod that was tall enough that we didn’t need to do contortions to see the display screen and make on-the-fly adjustments. A software package, SolarSnap, available for both Android and iPhone users, was well designed to capture images during both partial and full phases of the eclipse. Here we show a few of the images. Unlike most discussions of the eclipse, we use the solar images to elucidate how double axis gratings and spectrometers using such gratings behave. Interested in the science of the solar corona? You’ll need to look elsewhere!

The solar spectrum is complicated. Stellarnet, another spectrometer company, has posted a video showing the changes in the sun’s spectrum going through a partial eclipse (their home base in Florida did not see totality). Have a look! They used an array detector and a smaller entrance aperture than we did. Small objects can be imaged with higher wavelength resolution than large objects. The sun is a fairly large object, subtending 0.5°. The corona is even larger; as its gossamer strands extend millions of miles into space, its apparent size depends on camera exposure. Students at the University of Montana Research Experiences for Undergraduates program got a beautiful image showing the corona’s spectrum during the 2017 eclipse.

Note that longer wavelengths are to the left in their image. Wavelengths of some of the prominent lines are: Hα, 656.3 nm (red). He I, 584.3 nm (yellow). Hβ, 486.1 nm (aqua). Hγ, 434.0 nm (blue).

Let’s look at an image of the partially-eclipsed sun through a double-axis grating, about four minutes before totality started:

At first glance, all 8 of the spectra dispersed from the crescent sun look very similar. However, the width of each spectrum is a projection of the crescent along the dispersion direction. The horizontal and diagonal spectra are narrower than the vertical spectra, and the wavelength resolution is poorer. Why? Because the dispersion direction where the light source appears smallest is in the vertical direction. It’s even cuter than that. Let’s do a cut and paste to put the upper and lower vertically-dispersed orders adjacent to each other and also paste the picture of the crescent sun adjacent to each spectrum.

Look carefully at the yellow section of the spectrum. In both cases, the curvature follows the curvature of the solar crescent. In the left-hand spectrum (from the upper part of the original image), yellow curves away from green towards red. In the right-hand spectrum (from the lower part of the original image), yellow curves from red towards green. If we averaged the spectrum across the width of the dispersed image, in the left-hand case, yellow (~585 nm) and redder wavelengths would blur. In the right-hand case, yellow blurs with green. This illustrates an important point about spectrometer design: image distortion blurs spectra, and whether the blur is due to the finite size of the light source, coma in the spectrograph, or distortion in the optics, resolution is NOT simply set by the dispersion of the instrument.

The images look grainy, don’t they? whether that is due to the cirrus clouds, weak signals that software scales to fill the 8 bit intensity depth of JPG images, or some other cause is not obvious. What is obvious is that noise is evident in the images. What isn’t so obvious is that ALL images contain noise; they look clean to humans only when the noise is low compared to the signal.

Now for the main event: spectra from the corona during totality!

The moon creates the dark hole in the middle of the corona. That hole can be seen in spectra that are not saturated (the two vertical spectra are saturated, and the hold is filled in). If we estimate that the corona extends across 1° of arc, then the first order spectra (horizontal or vertical) cover about 3° of arc and wavelength resolution is perhaps 75 nm. We can’t see individual spectral lines; we see only continuum or individual lines blurred across a significant fraction of each spectrum. The 21/2 orders (the diagonals) have slightly better resolution. But resolution as good as that Stellarnet spectrum? Not even close.

Why do SpectroClick spectrometers claim 5 nm resolution when these pictures obviously show nowhere near that resolution? The entrance aperture for our instruments is 50 μm across. That subtends an angle of about 0.1°. That’s about an order of magnitude smaller than the corona, so the resolution is an order of magnitude finer.

There’s a lot more detail we can pull from these images. See this PDF discussion that goes into those details.

Comments are closed.