Wavelengths

Many of Korg's core functions operate on uniformly spaced wavelength sequences (note: all wavelengths have units of Å). For every such such function expects the wavelength sequence(s) can be specified in any of the following ways:

• a pair of values (λstart, λend) specifies the bounds of wavelength sequence and a triple (λstart, λend, λstep) allows customization over spacing. (λstart, λend) is equivalent to (λstart, λend, 0.01)
• a vector of N pairs and/or triples, (λstart1, λstop1[, λstep1]), ..., (λstartN, λstopN[, λstepN]) specifies N disjoint wavelength sequences.

Examples

It is instructive to consider a few illustrative examples. Although these examples use Korg.synth, but apply equally to all functions that take wavelength parameters, e.g. Korg.synthesize, Korg.Fit.fit_spectrum, Korg.RV_prec_from_noise, and Korg.apply_LSF.

First let's consider a simple continuous wavelength range:

figure() # hide #md
wls, flux, _ = synth(Teff=5777, logg=4.44,
                    wavelengths=(5050, 5100)) # wavelength range (in Å)
plot(wls, flux)
ylabel("flux")
xlabel(L"$\lambda$ [\AA]")
gcf() # hide #md

In many cases, you want to compute spectra in noncontinuous chunks. We can specify these chunks by passing a vector of pairs.

figure() # hide #md
wls, flux, _ = synth(Teff=5777, logg=4.44,
                     wavelengths=[(5050, 5080), (5090, 5100)])
plot(wls, flux)
ylabel("flux")
xlabel(L"$\lambda$ [\AA]")
gcf() # hide #md

The spacing for any wavelength sequence specified by a pair is 0.01 Å. When you specify a sequence with a triple, the third element is used to overwrite the spacing with an arbitrary value. Let's sample every 5 Å in the first window and every 3 Å in the second (while this sampling is too coarse for accurate spectra, these choices make it easy to visualize differences from the previous example).

figure() # hide #md
wls, flux, _ = synth(Teff=5777, logg=4.44,
                     wavelengths=[(5050, 5080, 5), (5110, 5120, 2)])
scatter(wls, flux)
ylabel("flux")
xlabel(L"$\lambda$ [\AA]")
gcf() # hide #md

Gotchas and edge cases

• For a given tuple, (λstart, λstop, λstep), λend may not always be included. For example, if λstop - λstart is not an integer multiple of λstep, the last last wavelength produced will be less than λstop.
• Wavelength sub-ranges must not overlap. For example, [(5000, 5010, 0.01), (5010, 5020, 0.2)], will fail, because 5010 is duplicated. Likewise, [(5000, 5010, 0.02), (5009.9, 5020, 0.2)], will also fail
• Wavelenth sub-range must be sorted. For example, [(6000, 7000), (3000, 4000)] will fail.
• All tuples must satisfy the invariants 1 < λstart < λstop and 0 < λstep <= λstop - λstart. Violations of these invariants will produce undefined behavior.
  • Historically, Korg inferred (in an inconsistent manner) that wavelengths have units of cm when λstart<1.0. You should not expect/rely upon this behavior.
  • The Korg developers reserve the right to freely modify the behavior when the invariants are violated, at any time (i.e. in a minor release).
• Because Julia's range objects are not guaranteed to produce bitwise identical ranges across Julia versions, Korg's wavelengths also lack this guarantee.

Internals: Korg.Wavelengths

Functions that take wavelength parameters as arguments all pass them immediately to the Korg.Wavelengths constructor, which turns them into the internal representation, so the documentation for that function is also a good place to look if you are digging into the Korg internals.