( Mostly) Atomic and (Some ) Molecular Data for Analytic Stellar Spectroscopy

(Mostly) Atomic and (Some ) Molecular Data for AnalyticStellar Spectroscopy

Charles R. Cowley: U. Michigan

Saul J. Adelman: The Citadel

Donald J. Bord: UM Dearborn

Lots of help from many others!!

Outline of Talk

1. Ionization Energies for neutrals and firstfour ions.

2. Atomic Data Bases

4. Opacity: photoionization cross sections; TopBase

3. Partition functions

Visit our website:

http://astro.lsa.umich.edu/users/cowley/ionen.htm

http://www.hbcpnetbase.com/hbcp/

Or, a number of others for more general info:

http://physics.nist.gov/PhysRefData/

http://plasma-gate.weizmann.ac.il/DBfAPP.html

http://urania.astro.spbu.ru/staff/afk/-

AtDatCentre/DtBases/db.html

Part 1: Ionization Energies

For Atomic Data, conveniently arranged for spectral synthesis, two primary sources:

http://cfaku5.harvard.edu

http://www.astro.univie.ac.at/~vald/

There may be other instances in the history ofscience where so many have owed so much to so few, but I don’t know of them.

Both of these sites are under continuous revision.

Part 2: Atomic Data

Generally speaking: Either of these sources willprovide the basis for an encouraging synthesisof atomic features in most stars.

Details are another matter. Even within small wavelength intervals,of 5 to 10 A, there are generally a few features for which one cannot find an appropriate atomic line, even by modifying the oscillator strength by an order of magnitude or more.

This is typical in a region with molecular features, when using VALD,which has not (as of May 2002) yet included molecules.

But it is also common in other regions, and very probably due toincompleteness of atomic data: Either the lines are not in thedata bases, or the oscillator strengths are egregiously in error.

Here is a rather shocking plot comparing VALDand Kurucz-site log(gf)’s for Fe I

Standard deviation: 0.86Standard error: 0.05

The big differences are all for “second-generationlines,” lines not in the Multiplet Tables. So they arenot lines typically studied.

However, they are all from classifies levels, and would appear in any spectrum synthesis.

Generally speaking, one might prefer VALD forelements heavier than the iron peak.

Changes in the Kurucz data beyond what is onthe CDRoms (and therefore in VALD) are mostlyfor the important spectra of Fe I and Fe II.

This was the situation when I browsed the Kurucz site at the end of March 2002, and itit subject to change.

10.06.02

We have choices other than just RLKor VALD.

Check the NIST site, and hunt around among thelinks. An E-book is nearing completion, which will contain much useful information on the firsttwo spectra of elements from H to Es. There willbe wavelengths for some 11,000 lines, but gf’sfor only about 2500 lines.

A NIST project that is just beginning will have much more extensive data, including gf’s and hfs.

Of course, any improvements by NIST should beincorporated in VALD and Kurucz data bases!

Snooping on the internet can bring up a wealthof relevant information.

Here are just a few examples of places from whichone can get useful data:

Be advised that sometimes, one must get areference, and then pull up the journal electronically.Here is an incomplete example of what one can find.

The Main Problem is PERSON Power

Part 3: Partition functions

Focus on two aspects of partition functions:1. Completeness of low levels.

2. Treatment of levels above those known experimentally

--------------------------------------------------

Various algorithms are in use to give thepartition function u(T). Undoubtedly someare better than others. All must be basedon data that is sufficiently complete.Partition functions are important, algorithmsless so.

Recent examples:Added levels fromCowan code, orlaboratory work.REE III (1994)Ho II, W, Os(2002)

Mostly the changes aresmall—a few per cent,except in cases where ranks were D, E, or missing altogether.

For Ho II, nearly 50%: worth making!

Following a little-known study by Van’t Veer-Menneret(CR, 253, 380, 1961), we assert that the contributionto the partition function from levels above those typically known, ARE NOT IMPORTANT for the cal-culation of line strengths.

It is easy to see this in a simple case. We considerthe situation where the dominant ion is the first—virtually all atoms are first ionized. But, we areto compute a line arising from the first (neutral)spectrum.

We assert that in general, the partition functionfor the atom or ion (i) is unimportant for theline strength if the element is predominantly(i+1)-th fold ionized.

It is easy to see from the previous slide that thismust be approximately the case, but the assertioncovers a lot of parameter space, and should be checked for special cases.

Part 4: Photoionization cross sections“Opacity of opacity, saith the Preacher, opacityof opacity; all is opacity.” Ecclesiastes –

From a recently translated Agnostic gospel.

First ionization limit

Continuum

These linesare in Kurucztables.

Red level interactswith continuum

Solid line: TopBaseDotted: Hydrogenic

Cooperminimum

PECPEC:4s-4p(H and K)

It is straightforward to smooth these TopBasecross sections without changing the net absorption.This has been done by a number of people, andsystematically by Prieto (2002).

The next step, already being carried out, is toparameterize the smoothed fits for use in opacityroutines, such as si1op, al1op, fe1op, etc. usedin the Atlas codes. These routines give what Icall the “K-factor”:

Ca I 4s4d 1D

Dotted: Hydrogenic

Doted: Hydrogenic

Epilog: Molecular Spectra, et al.Best overall site: cfaku5.harvard.edu. New TiO, H2O

RadEn Database: www.chem.msu.su/eng/raden/

HITRAN: cfs-www.harvard.edu/HITRAN/

NIST: webbook.nist.gov/chemistry/form-ser.html

Atomic and Molecular Data and software: ccp7.dur.ac.uk

DREAM (REE): www.umh.ac.be/~astro/dream.shtml