LineTAP

[Markus Demleitner]

Hi Franck,

On Thu, Apr 28, 2022 at 06:21:49PM +0200, Franck Le Petit wrote:
> Following Margarida nice presentation about LineTAP, here are some comments and questions. 
> I saw that Joule was proposed. Joule is a bit a strange unit for
> energy levels because 1 Joule is lot of energy whereas rotational
> levels or hyperfine structures contains very few energy. Using
> Joule means that we will have to manipulate quantities as 10E-XX
> where XX is a lot. 

Yeah, well... whatever unit you choose, in a multi-wavelength regime
you'll always have ugly numbers for *almost* everyone.  The advantage
of Joule is that they're ugly for *everyone*, so nobody can really
complain.

Hence, my vote would still be for Joule of energy.

> That is not an issue if users never see these quantities in Joule,
> if, for instance, they are converted in other units. 

That is certainly the intent; in general, I suspect LineTAP will be
almost exclusively used by clients of some sort hiding the actual
communication for users.  In those cases where people directly write
ADQL, there's ivo_specconv doing all the hard thinking.

> The most important point here, for energy levels and line
> wavelengths is the precision.  Spectroscopy is one of the domain of
> Physics where the precision on data is very large. And, in
> astrophysics we need this precision on spectroscopic data. 

On that point, note that LineTAP conceptually is something like a
basic index on XSAMS data; its purpose is to let people *find* lines,
and to aid that, clients ought to be able to do an order-of-magnitude
estimate of a line strength given basic physics (temperature, perhaps
pressure) and chemistry.

For anything beyond that, I'd say we'd point to XSAMS (where people
might opt to have pure-LineTAP services; but these then are really
only good for showing lines on a plot).

> 1 - The VO interfaces should not introduce a lost of precision for
> example manipulating very small numbers (energy of a hyperfine
> transitions in Joule) and very large numbers. 

As long as you don't add or substract, there's (essentially) no loss
of precision -- you maintain whatever mantissa you have.  The only
exception would be if the exponents were to exceed +/- 308 (for IEEE
doubles), and that, I'm sure, won't happen for any halfway reasonable
choice of base unit.

When you *do* add and substract, the units again don't really enter
into the numerics.  If the two values are more than, say, 10 orders of
magnitude apart, IEEE doubles will lose precision if the smaller
value has more than, say, 6 digits.  No choice of units will help
with that.

But again, I'd plea that we keep to the use cases I'd like to define
for LineTAP (see the docs), and with them I do not see the need for
this kind of computation or precision.

> We should avoid the situation where each data provider and each
> client has to implement unit conversion system. We can be sure
> there will have mistakes. 

True -- but given we want to serve data all across the electromagntic
spectrum (though I have to admit I doubt we're ready for lines in
gamma resulting from nuclear transitions), you just cannot do without
conversions in general.  One might argue it's nice if *some*
community (current optical, which I guess guided the choice of
Ångström in VAMDC and hence current LineTAP) doesn't have to convert,
but I'd consider that a minor advantage.

> 2 - Zero point energy 
>
> But, this quantity is useful in chemistry not in spectroscopy. So,
> why do we need it for LineTAP ? It is not an issue if we say that
> the H2 v=0 J=0 level is at 0.00000 K and the HD v=0 J=0 level is
> also at 0.00000 K even if these two zeros are not the same. Indeed,
> for spectroscopy, we will never do operations between energy levels
> of H2 and HD. So, each molecule as its 0.000 level energy that is
> not the same between molecules. But, for our problem of line
> indentification that is not an issue. 
> 
> So in my opinion, LineTAP should not deal with zero point level energy. 

Sure.  But we want to estimate the line strength, for which we have
to estimate the population, for which we need the energy above the
ground state (assuming Boltzmann statistics, that is, but that we
simply cannot help).

That's about as far as my (non-spectroscopist) physics take me.  How
this relates to the XSAMS zero points I can't really tell, but I'd be
grateful if experts helped me out there.

> 3 - Einstein coefficients, Oscillator strengths and others
> 
> In the presentation, it was said that one can go from one of these
> quantities to the other ones by simple conversions. That is true
> but for some molecules that can be quite tricky and create debates
> between specialists. 
> 
> Moreover, to go from one quantity to another one, one often need
> another quantity that is the degeneracy of the levels. For example,
> to convert an oscillator strength in Einstein coefficients : 
> Aki = 6.6702E15 / lambda**2 * g_i / g_k * f_ik
> 
> So, we need the degeneracies of the levels and those one depends on
> the physical effects that are considered (ex : hyperfine structure
> or not).
> To understand a numerical value for the degeneracies, one need to
> know the quantum numbers. 
> 
> So, how does LineTAP deal with that or, more precisely, how will it
> provide enough information to users so that they can use properly
> the data ? 

That is the central question we're poring over now, and again my
non-spectroscopist physics desert me.

Given the use case "estimate (emission/absorption) line strength
assuming LTE and sufficient physics": Is there ever the need to look
at unweighted oscillator strengths?

Thanks,

            Markus