1: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%

2: %%

3: %% This is a sample manuscript marked up using the

4: %% AASTeX v5.x LaTeX 2e macros.

5: 

6: %% The command below calls the preprint style

7: %% which will produce a one-column, single-spaced document.

8: %% Examples of commands for other substyles follow. Use

9: %% whichever is most appropriate for your purposes.

10: 

11: \documentclass[12pt,preprint]{aastex}

12: 

13: %% manuscript produces a one-column, double-spaced document:

14: 

15: %\documentclass[manuscript]{aastex}

16: 

17: %% preprint2 produces a double-column, single-spaced document:

18: 

19: %\documentclass[10pt,preprint2]{aastex}

20: 

21: %\usepackage{graphicx}

22: %\usepackage{natbib}

23: 

24: %

25: %	hector's

26: 

27: \def \etal {et al.}

28: \def \apj {ApJ}

29: \def \apjs {ApJS}

30: \def \apjl {ApJ}

31: \def \solphys {Solar Phys.}

32: \def \pasj {Pub. Astron. Soc. Japan}

33: \def \aap {A\&A}

34: \def \arcsec{\hbox{$^{\prime\prime}$}}

35: 

36: % for astrobib

37: %\newcommand{\citeauthor}[1]{\def\citeauthoryear##1##2##3{\rm ##1}\cite{#1}}

38: %\newcommand{\citeyear}[1]{\def\citeauthoryear##1##2##3{\rm ##3}\cite{#1}}

39: 

40: \newcommand{\citeN}[1]{\citeauthor{#1} (\citeyear{#1})}

41: \newcommand{\citeNP}[1]{\citeauthor{#1} \citeyear{#1}}

42: \newcommand{\citeyearNP}[1]{\citeyear{#1}}

43: \newcommand{\citeANP}[1]{\citeauthor{#1}}

44: 

45: %% --- Some useful spectral definitions

46: 

47: \newcommand{\CaII}{\ion{Ca}{2}}

48: \newcommand{\NaI}{\ion{Na}{1}}

49: \newcommand{\FeI}{\ion{Fe}{1}}

50: \newcommand{\OI}{\ion{O}{1}}

51: \newcommand{\NiI}{\ion{Ni}{1}}

52: 

53: %% --- Setting to make eqnarrays look better. Unfortunately, AASTex

54: %%     does not link to AMS styles which have better mechanisms to

55: %%     deal with tabbed equations.

56: 

57: %\setlength{\arraycolsep}{1pt}

58: 

59: 

60: \shortauthors{Centeno \& Socas-Navarro}

61: 

62: 

63: %---------main text

64: %

65: \begin{document}

66: 

67: %

68: \title{A new approach to the solar oxygen abundance problem}

69: 

70: 

71: \author{R. Centeno}

72:    	\affil{High Altitude Observatory, NCAR, 3080 Center Green Dr, Boulder, CO 80301, USA}

73: 	\email{rce@ucar.edu}

74: 

75: \author{H. Socas-Navarro}

76:    	\affil{High Altitude Observatory, NCAR\altaffilmark{1},

77: 	3080 Center Green Dr, Boulder, CO 80301, USA}

78:         \affil{Instituto de Astrof\'\i sica de Canarias, Avda V\'\i

79:           a L\' actea S/N, La Laguna 38205, Tenerife, Spain}

80: 	\email{navarro@ucar.edu}

81: 

82: \altaffiltext{1}{The National Center

83: 	for Atmospheric Research (NCAR) is sponsored by the National Science

84: 	Foundation.}

85: 

86: \newcommand{\EO}{$\epsilon_{\rm O}$}

87: \newcommand{\lgEO}{${\rm log}\epsilon_{\rm O}$}

88: \newcommand{\EFe}{$\epsilon_{\rm Fe}$}

89: \newcommand{\lgEFe}{${\rm log}\epsilon_{\rm Fe}$}

90: \newcommand{\ENi}{$\epsilon_{\rm Ni}$}

91: \newcommand{\lgENi}{${\rm log}\epsilon_{\rm Ni}$}

92: 

93: %\date{}%

94: 

95: \begin{abstract}

96: In this work we present new data that sets strong constraints on the

97: solar oxygen abundance. Our approach, based on the analysis of

98: spectro-polarimetric observations, is almost model-independent and

99: therefore extremely robust. The asymmetry of the Stokes~$V$ profile of

100: the 6300 \AA\, [\OI] and \NiI\, blend is used as an indicator of the

101: relative abundances of these two elements.  The peculiar shape of the

102: profile requires a value of \EO=730$\pm$100 ppm~(parts per million),

103: or \lgEO=8.86$\pm$0.07 in the logarithmic scale commonly used in

104: Astrophysics. The uncertainty range includes the model dependence as

105: well as uncertainties in the oscillator strenghts of the lines.  We

106: emphasize that the very low degree of model dependence in our analysis

107: makes it very reliable compared to traditional determinations.

108: \end{abstract}

109:    

110: \keywords{ Sun: abundances --

111:            Sun: magnetic fields --

112:            Sun: atmosphere --	

113: 	   techniques: polarimetric --

114:            line: profiles}	

115: 

116: \section{Introduction}

117: 

118: Oxygen is the third most abundant chemical element in the Universe,

119: after Hydrogen and Helium, and the one that is almost exclusively 

120: produced by nuclear fusion in stellar interiors. Its abundance in the

121: Sun was thought to be well established since the 1980s (\EO=850$\pm$80

122: parts per million, relative to hydrogen, ppm; \citeNP{anders-grevesse89}; more

123: recently \EO=680$\pm$100 \citeNP{grevesse-sauval98})\footnote{ In this

124:   paper we use a linear abundance scale, given in ppm, because it is a

125:   more convenient unit in our approach. Some times, however, the

126:   traditional Astrophysical logarithmic scale is given to allow for

127:   easy comparison with previous work (${\rm log}\epsilon=12+\log_{10}

128:   [\epsilon \times 10^{-6}]$).}.  

129: However, a recent work (\citeNP{asplundetal04}) using a

130: new 3D hydrodynamical model of the solar atmosphere (as well as

131: updated atomic and molecular data) recommends a revision of the O

132: abundance to a lower value of \EO=457$\pm$50 ppm (\lgEO=8.66 $\pm$

133: 0.05). The revised solar composition is said to fit better within its galactic

134: environment but it also creates a serious problem, namely it ruins the

135: exceptionally good agreement between various predictions of solar interior

136: models and properties inferred from helioseismology (see, e.g., \citeNP{basu-antia2008}).

137: 

138: The chemical composition of celestial bodies is not a directly

139: measurable quantity. It is deduced by fitting observations with

140: synthetic spectral profiles using a particular model atmosphere

141: (temperature, density ans so forth are prescribed), which makes it a

142: strongly model-dependent process.  Thus, the observations that lead to

143: the O abundance are not conclusive and arguments exist both in favor

144: and against the revision. The controversy on whether the proposed

145: revision should be adopted and the doubts that it would cast on

146: stellar structure and evolution models is serious enough that it is

147: often referred to as the solar oxygen crisis (\citeNP{ayresetal06}).

148: 

149: %%%librarian

150: %%%label: basu-antia2008

151: %%%authors: Basu, Antia

152: %%%year: 2008

153: %%%refereed: yes

154: 

155: 

156: 

157: %%%librarian

158: %%%label: anders-grevesse89

159: %%%authors: Anders, Grevesse

160: %%%year: 1989

161: %%%refereed: yes

162: %%%endlibrarian

163: %%%librarian

164: %%%label: grevesse-sauval98

165: %%%authors: Grevesse, Sauval

166: %%%year: 1998

167: %%%refereed: yes

168: 

169: %%%librarian

170: %%%label: asplundetal04

171: %%%year: 2004

172: %%%authors: Asplund, Allende Prieto

173: %%%refereed: yes

174: %%%title: granulation lines O    

175: 

176: 

177: 

178: Both the traditional and the proposed revision of the O abundance have

179: been obtained with similar sets of observations

180: (an irradiance spectrum and an average disk center spectrum) and

181: very similar techniques (fitting the equivalent width of atomic

182: and molecular abundance indicators with synthetic spectra). The main

183: difference prompting the revision is the model atmosphere employed for

184: the synthesis. The traditional abundance is obtained when a

185: semi-empirical one-dimensional (1D) model is used (e.g., that of

186: \citeNP{holweger-muller74}), whereas the new low abundance is obtained

187: %%%librarian

188: %%%label: holweger-muller74

189: %%%author: Holweger, Muller

190: %%%year: 1974

191: %%%refereed: yes

192: when using a three-dimensional (3D) hydrodynamical simulation of photospheric

193: convection (\citeNP{asplundmodel}).

194: % Recently, however, another

195: %group derived a high O abundance using a different 3D model (\citeNP{}).

196: 

197: %%%librarian

198: %%%label: ayresetal06

199: %%%author: Ayres

200: %%%title: profiling, abundances

201: %%%year: 2006

202: %%%refereed: yes

203: 

204: %%%librarian

205: %%%label: asplundmodel

206: %%%author: Asplund, nordlund, trampedach, stein

207: %%%title: hydrodynamical model atmospheres evidence

208: %%%year: 1999

209: %%%refereed: yes

210: 

211: One criticism that has been formulated (\citeNP{ayresetal06}) concerning

212: the \citeN{asplundetal04} work is that a 3D theoretical model is not

213: necessarily superior to a 1D semi-empirical one when used to fit

214: observations. This is a legitimate concern that deserves some

215: consideration. An attempt to resolve the issue was made by

216: \citeN{SNN07},

217: %%%librarian

218: %%%label: SNN07

219: %%%author: Socas-Navarro, Norton

220: %%%title: oxygen

221: %%%year: 2007

222: %%%refereed: yes

223: who employed spatially-resolved observations of \FeI \, lines to derive a

224: 3D semi-empirical model (thus combining the advantages of both

225: strategies) and used it to analyze simultaneous spectra of the \OI \,

226: infrared triplet at 7770 \AA. In this manner they derived \EO \, at each spatial

227: position, as opposed to previous single-valued determinations, obtaining

228: an average \EO=426 ppm. While this result supports the new low

229: abundace, the authors also pointed out that the spatial distribution

230: of \EO \, exhibits an unsettling degree of structure, suggesting that

231: the model is still less than perfect. This is not entirely surprising

232: since the \OI \, infrared triplet is strongly affected by non-LTE

233: effects and it is possible that the line formation physics is not yet

234: completely well understood.

235: 

236: In this work we take a novel approach and analyze Stokes~$V$ 

237: spectral data of the forbidden [\OI] 6300 \AA \,

238: line. Its formation physics is much simpler than that of the infrared

239: triplet, although it is a very weak line and is blended with a \NiI

240: \, line of similar strength. This blend, rather than being a problem,

241: turned out to be beneficial for our study because, even though it

242: complicates the determination of \EO, it provides us with a nearly

243: model-independent value of the ratio of \EO/\ENi \, as we show below.

244: 

245: \section{Data analysis and results}

246: 

247: The observations employed here were acquired with the

248: Spectro-Polarimeter for Infrared and Optical Regions (SPINOR,

249: \citeNP{spinor}) on July 28, 2007. We obtained simultaneous

250: observations of the \FeI \, lines at 6301.5 and 6302.5~\AA \, and the

251: 6300.3~\AA \, blend of the [\OI] and \NiI \, lines.

252: %%%librarian

253: %%%label: spinor

254: %%%author: socas-navarro, elmore, pietarila

255: %%%title: spinor

256: %%%refereed: yes

257: The goal of our campaign was to observe a sunspot umbra because the

258: different Zeeman splitting of the lines would allow us to

259: differentiate them in the polarized spectrum. Since the Stokes~$V$

260: profile is normally antisymmetric, the spectral blend gives rise to a

261: complex structure that provides us with important clues as we explain

262: below. These lines are very weak and therefore it is important to

263: observe a strongly magnetized region (ideally a sunspot umbra) to

264: maximize the polarization signal. Besides, umbral profiles are very

265: symmetric as velocity gradients are almost nonexistent in the umbral

266: photosphere (below 100 ms$^{-1}$). Therefore, we can be certain that

267: any asymmetries we may encounter are due to the presence of the blend.

268: 

269: We observed a small

270: sunspot (solar activity was very low during our observing run) with

271: moderate seeing conditions that we estimated to be approximately

272: 1.5\arcsec. Data reduction and polarization calibration were carried

273: out using the standard procedures for this instrument, which include

274: correction for the instrumental polarization introduced by the

275: telescope. We selected the best datasets of July 28: two consecutive

276: scans 20 minutes appart early in the morning, which we averaged; and

277: a third scan taken later in the day. We shall refer to these three 

278: datasets as Maps 1, 2 and 3,

279: respectively. 

280: 

281: Figure~\ref{fig:profs} shows the average Stokes~$V$ observed in the 

282: umbra of Map 1. All three

283: datasets have a very similar average umbral profile. Also depicted in

284: the figure are synthetic line shapes of the O (blue) and Ni (green)

285: components of the blend computed separately as well as combined

286: (red). The model atmosphere used for the synthesis was obtained from

287: the inversion of the \FeI \, lines.  All syntheses and inversions

288: presented in this paper were carried out with the LTE code LILIA

289: (\citeNP{lilia}). Tests were conducted with another standard code (SIR, \citeNP{sir})

290: using several of the models, to ensure that the same results were obtained.

291: 

292: %%%librarian

293: %%%label: lilia

294: %%%authors: socas-navarro

295: %%%year: 2001-2001

296: %%%title: Stokes inversion techniques

297: %%%refereed: no

298: 

299: %%%librarian

300: %%%label: sir

301: %%%authors: ruiz cobo, del toro iniesta

302: %%%year: 1991-1992

303: %%%title: Stokes

304: %%%refereed: yes

305: 

306: Notice in Figure~\ref{fig:profs} how the [\OI] and \NiI \, lines have

307: similar amplitudes. However, the differences in their rest wavelengths

308: and effective Land\'e factors give rise to a peculiar and strongly

309: asymmetric shape of the blend when they are combined. 

310: Particularly interesting is the feature marked by the arrow

311: in the figure, exactly at the switchover point between the [\OI] and

312: the \NiI \, blue lobes. This feature is present in all three of our

313: datasets. 

314: %The vertical position of this feature depends on details of

315: %the model employed for the synthesis and the zero-point chosen in the

316: %calibration of the Stokes~$V$ spectrum. However, the presence and

317: %shape of this feature is almost insensitive to the model and depends

318: %critically on the ratio of O and Ni abundances (\EO/\ENi), which in this

319: %example is set to 200. This makes this feature an extremely interesting

320: %abundance indicator. 

321: %It is easy to see that, as one increases the

322: %amount of O, the amplitude of the blue curve grows accordingly. 

323: 

324: Consider the average of the slope of the red curve in the shaded area

325: (hereafter referred to as $m_{6300}$). If we had much more O than Ni,

326: the red curve would follow the blue curve and this slope would be

327: positive. Conversely, if we had much more Ni than O it would be the

328: green curve that dominates and the slope would be negative. In an

329: intermediate situation, $m_{6300}$ takes values between these two

330: limits. Figure~\ref{fig:samples} shows this behavior in detail for a

331: range of the ratio \EO/\ENi \, from low to high values. The

332: amplitude of the profile and the vertical position of the central

333: feature depend on the details of the model employed in the synthesis

334: as well as the polarization calibration. However, the shape of the

335: feature in the shaded area of Fig~\ref{fig:samples} (and therefore the

336: mean slope $m_{6300}$) is nearly model-independent, as we show below,

337: which makes this feature valuable from a diagnostic point of view.

338: 

339: 

340: 

341: Table~\ref{table:lines} lists the atomic parameters that we

342: employed. Accurate determinations of the oscillator strength exist for

343: all lines (\citeNP{Ogf}, \citeNP{Nigf}, \citeNP{Fegf}; see also the

344: database of the National Institute of Standards and Technology,

345: NIST\footnote{http://physics.nist.gov/PhysRefData/ASD/index.html}),

346: except for \FeI \, 6302.5 \AA . We determined the $\log(gf)$ of

347: this line using the following procedure. We first inverted a

348: spatially-averaged quiet-Sun profile of \FeI \, 6301.5 \AA . With the

349: resulting model we synthesized 6302.5 \AA \, varying its $\log(gf)$

350: until a satisfactory agreement with the observed profile was

351: attained. We note that the accuracy of the \FeI \, oscillator

352: strengths is not very important in our work. We use the \FeI \, lines

353: to determine a suitable model atmosphere. However, the arguments given

354: below are nearly model-independent. In fact, one could use standard

355: sunspot models published in the literature and still obtain the same

356: results as we demonstrate using the model of \citeN{maltbyetal}.

357: %%%librarian

358: %%%label: Ogf

359: %%%author: Fischer, Saha

360: %%%refereed: yes

361: %%%year: 1983-1983

362: %%%endlibrarian

363: %%%librarian

364: %%%label: Nigf

365: %%%year: 2002-2008

366: %%%title: Ni 6300

367: %%%endlibrarian

368: %%%librarian

369: %%%label: Fegf

370: %%%year: 1991

371: %%%author: Bard, Kock, Kock

372: %%%title: astrophysical interest

373: 

374: 

375: 

376: The electric dipole and the magnetic quadrupole components of the [\OI \,] feature 

377: are listed separately in the table, although the electric dipole contribution is insignificant.

378: The two \NiI \, components correspond to the two major isotopes 

379: ($^{58}$Ni and $^{60}$Ni): the relative abundances have been folded into the cited 

380: gf-values. The \OI \, line is well described in LS coupling but

381: the \NiI \, line exhibits some small departures (see the atomic level

382: information compilation by NIST). We took this small departure into

383: consideration in computing the polarized line profile. 

384: 

385: We computed many different models by inverting the \FeI \, lines

386: observed in the umbra of Maps 1, 2 and 3 (the profiles for Maps 1 and

387: 2 were averaged together to improve the signal-to-noise ratio since

388: they were taken very close in time) and using different values for the

389: Fe abundance between 20 and 40 ppm. For each one of these models we

390: synthesized the Stokes~$V$ spectrum (an example is shown in

391: Fig~\ref{fig:profs}) and computed $m_{6300}$ as the average slope over

392: a bandpass of 50~m\AA \, around the [\OI] line center (the vertical

393: shaded area in Fig~\ref{fig:profs}). When performing all of these

394: syntheses we added random perturbations to the oscillator strengths of

395: the lines. The amplitude of these perturbations was set to the

396: uncertainties in the values published in the literature (14\% and 16\%

397: for the \NiI \, and \OI \, lines, respectively). In this manner we could

398: test the impact of these uncertainties on our results. The last column

399: in Table~\ref{table:lines} lists the range of abundances for the

400: various elements involved in the synthesis that was used to produce

401: the figure.

402: 

403: Figure~\ref{fig:asymvsratio} (blue shaded areas) shows how $m_{6300}$

404: varies as a function of the O/Ni abundance ratio

405: (\EO/\ENi).  Note that all of these models produce a similar curve of

406: $m_{6300}$ vs \EO/\ENi, regardless of which dataset we used to derive

407: the model, what Fe abundance we considered, etc. The spread, delimited

408: by the shaded region, is due a small degree of model dependence as

409: well as the random variations in the O and Ni oscillator strengths. As

410: a further test, we repeated the computation using a standard model

411: umbra (model M of \citeNP{maltbyetal}). The model temperature was

412: first normalized to produce the observed continuum intensity

413: (otherwise the ratio of \OI \, to \NiI \, would be incorrect).

414: %%%librarian

415: %%%label: maltbyetal

416: %%%authors: maltby, kjeldseth-moe, avrett

417: %%%refereed: yes

418: %%%year: 1985-1995

419: %%%endlibrarian

420: The values of $m_{6300}$ produced with the \citeN{maltbyetal} model

421: are still inside the band that we obtained with the models from our

422: inversions. We may then conclude that the

423: calibration curve of Fig~\ref{fig:asymvsratio} is extremely robust, in

424: the sense that it exhibits very little sensitivity to the model

425: atmosphere employed to obtain it.

426: 

427: 

428: 

429: The horizontal dashed line in Fig~\ref{fig:asymvsratio} marks the

430: value $m_{6300}$ of the observed profiles (along with the

431: uncertainty due to observational noise), compatible

432: %(175,195 Map 3

433: %212,260) Map 12

434: with a ratio \EO/\ENi=210$\pm$24. Interestingly, the Ni abundance

435: currently is better determined than that of O because there are more

436: well-suited photospheric lines of Ni in the solar

437: spectrum. \citeN{solarcomposition} quote a current value of

438: \ENi=1.7$\pm$0.15 ppm, in good agreement with the

439: meteoritic abundance of \ENi=1.5$\pm$0.1. 

440: %%%librarian

441: %%%label: solarcomposition

442: %%%author: grevesse, asplund, sauval

443: %%%year: 2007-2007

444: %%%title: solar chemical composition

445: %%%endlibrarian

446: 

447: The abundance of {\em atomic} O that we obtain is

448: \EO=360$\pm$50~ppm.

449: 

450: Since we are observing a sunspot, one must also be aware that some of

451: the O may be in the form of molecules. Such molecular O does not

452: contribute to the formation of the 6300~\AA\ blend and has

453: been neglected thus far. According to \citeN{leemolec}, the most

454: %%%librarian

455: %%%label: leemolec

456: %%%author: Lee

457: %%%year: 1981

458: %%%title: molecular formation

459: %%%endlibrarian

460: abundant O molecule (and the only one with a non-negligible

461: concentration) in sunspot umbrae is CO, which contains nearly 50\% of

462: all the O. Instead of relying on this value, we made a full

463: calculation for the temperature of our sunspot. We felt that this was

464: important because, as noted above, the sunspot that we observed was a

465: small one and it might be too warm compared to standard models. Our

466: code solves the molecular chemical equilibrium for an arbitrary number

467: of molecules.

468: %%%librarian

469: %%%label: ST84

470: %%%authors: Sauval, Tatum

471: %%%year: 1984

472: %%%endlibrarian

473: The interested reader can find a detailed account of the molecular

474: calculation in a forthcoming research note. Here we shall simply state

475: that 35 diatomic molecules which potentially can affect the atomic O an Ni

476: partial pressures were included (\citeN{ST84}). The 6300~\AA\ blend forms nearly at the

477: same height as the continuum and therefore it is straightforward to

478: obtain the temperature there. One simply needs to take the continuum

479: intensity and equate it to the Planck function at that wavelength (as

480: a sanity check, our inversions of the \FeI\ lines yield approximately

481: the same temperature at $\tau_{5000}=1$). With this calculation we

482: obtained that all the molecules containing O and Ni can be neglected

483: except for CO, which carries about 51\% of the O nuclei at that

484: temperature. 

485: 

486: It is important to note that the molecular calculation that we have

487: solved is simply that of chemical equilibrium in LTE (no radiative transfer

488: or spectral synthesis). The LTE/CE approximation is valid owing to the high 

489: densities of the umbral deep photosphere. This calculation does not require a model.

490: Only the gas temperature is needed, which can be extracted directly

491: from the observations as explained above. Therefore, we are not

492: introducing any significant complication to our analysis. Putting all

493: these numbers together we conclude that the total amount of O in the

494: umbra is \EO=730$\pm$100~ppm, or \lgEO=8.86$\pm$0.07.

495: 

496: \section{Conclusions}

497: 

498: Oxygen is a very important element because of its high abundance and

499: also because other relevant elements that do not produce suitable

500: photospheric lines can only be measured relative to it. Unfortunately,

501: there are few O abundance indicators in the solar spectrum and they

502: all present complications of one kind or another and require very

503: detailed modeling to fit the observations. This means that virtually

504: every abundance determination is susceptible of criticism. Moreover,

505: it is important to keep in mind that abundance determinations

506: inherently are model-dependent.

507: 

508: In this work we present a new approach that circumvents these problems

509: because: (a) we use a spectral feature whose formation physics is well

510: known (LTE line formation, well-known oscillator strengths); and (b) the

511: relation between the parameter $m_{6300}$ and the ratio \EO/\ENi \, is

512: almost model-independent. By analyzing a polarization profile we are

513: insensitive to common systematic errors, such as flatfield

514: uncertainties (polarization is a differential measurement). The

515: polarization profile of the 6300~\AA \, feature is very

516: peculiar. Instead of analyzing properties such as the amplitude of the

517: profile, its width or depth, etc, which are sensitive to details of

518: the modeling, we can focus on its shape. Some features such as

519: $m_{6300}$ depend very critically on the O and Ni abundances. The fact

520: that the simplest O abundance indicator (only LTE atomic line in the

521: visible spectrum) is blended with another line of similar strength has

522: been viewed historically by many researchers as a rather unfortunate

523: coincidence. In the light of this new approach, however, one could now

524: argue the opposite. The abundances of O and Ni in the solar

525: photosphere are ``just right'' to produce the peculiar polarization

526: feature marked by the arrow in Fig~\ref{fig:profs}. If either the O or

527: the Ni line had dominated over the other then we would have had no

528: other option than fitting the profile amplitude or width, which

529: again would have been a model-dependent determination and therefore

530: subject to question.

531: 

532: The ratio \EO/\ENi=210$\pm$24 that we obtain is a very robust result

533: and this ratio, combined with the well-established Ni abundance and

534: taking into account that 51\% of the O is in the form of CO, turns out

535: to be incompatible with the low O abundance proposed by

536: \citeN{asplundetal04}. 

537: 

538: 

539: 

540: \acknowledgments 

541: The authors are very grateful to Dr Thomas Ayres for

542: pointing out the issue of the CO formation that we had missed in our

543: initial manuscript. Without his suggestion this paper would have

544: presented very different results.  Thanks are also due to the staff at

545: the Sacramento Peak observatory (Sunspot, NM, USA) of the National

546: Solar Observatory for their enthusiastic support of our observing

547: campaign.  Financial support by the Spanish Ministry of Education and

548: Science through project AYA2007-63881 is gratefully acknowledged.

549: 

550: 

551: \bibliographystyle{../../bib/apj}

552: \bibliography{ms.bib}

553: 

554: \clearpage

555: 

556: 

557: \begin{deluxetable}{cccccccc}

558: \tablewidth{0pt}

559: \tablecaption{Atomic parameters

560: \label{table:lines}}

561: \tablehead { Ion &Wavelength & Excitation & $\log(gf)$ & Configuration &

562:   Configuration & Land\'e & Abundance \\

563:              &   (\AA)     & potential (eV) &        & (lower)  &

564:   (upper) & factor & range (ppm) }

565: \startdata

566:    \OI & 6300.304 &  0   &    -9.776  &  $^3$P$_2$ & $^1$D$_2$ & 1.25

567:    & 200-630 \\

568:    \OI & 6300.304 &  0   &    -12.202  &  $^3$P$_2$ & $^1$D$_2$ & 1.25

569:    & 200-630 \\

570:    \NiI & 6300.335 & 4.266 & -2.250 & $^3$D$_1$ & $^3$P$_0$ & 0.51 & 1-3 \\

571:    \NiI & 6300.355 & 4.266 & -2.670 & $^3$D$_1$ & $^3$P$_0$ & 0.51 &

572:    1-3 \\

573:    \FeI & 6301.501 & 3.654 & -0.718 & $^5$P$_2$ & $^5$D$_2$ & 1.67 & 20-40\\

574:    \FeI & 6302.492 & 3.686 & -1.235 & $^5$P$_1$ & $^5$D$_0$ & 2.50 & 20-40\\

575: \enddata

576: \end{deluxetable}

577: 

578: \clearpage

579: 

580: \begin{figure*}

581: %\epsscale{2}

582: \plotone{f1.eps}

583: \caption{Stokes $V$ spectrum of a sunspot umbra. Black,

584:   solid: observed. Red, dash-dotted: Synthesis of the [\OI] , \NiI

585:   \, and \FeI \, lines combined. Blue, dashed: Synthesis of [\OI] \, and

586:   \FeI \, lines 

587:   only. Green, dashed: Synthesis of \NiI \, and \FeI \, lines only. All

588:   profiles are normalized to the average disk-center quiet-Sun

589:   continuum.

590: \label{fig:profs}

591: }

592: \end{figure*}

593: 

594: 

595: 

596: \begin{figure*}

597: %\epsscale{2}

598: \plotone{f2.eps}

599: \caption{Synthetic Stokes~$V$ profile (solid) of the [\OI]/\NiI \,

600:   blend for different values of the abundance ratio \EO/\ENi . The

601:   shaded area represents the wavelength range over which we average

602:   the slope of the profile to obtain $m_{6300}$. For very low ratios

603:   (panel a), the slope is negative. As the ratio increases (panels

604:   b, c) $m_{6300}$ becomes less negative until it reaches zero. For

605:   higher ratios $m_{6300}$ continues to increase (panels d, e, f) until

606:   the [\OI] line dominates the blend and the profile takes on the normal

607:   antisymmetric Stokes~$V$ shape of an isolated line.  The

608:   profile amplitude and zero point depend on details of the model

609:   employed for the synthesis and the data calibration. For this

610:   reason, the profiles in the figure have been renormalized to have

611:   the same amplitude as the observed one (dotted line). However, the

612:   shape of the central feature in the shaded area and the mean slope

613:   $m_{6300}$ are virtually model-independent.  Therefore we base our

614:   analysis on this feature. After correcting for molecular formation,

615:   and assuming \ENi=1.7, the ratios corresponding to the new and old

616:   O abundances would be 130 and 200, respectively. The corresponding

617:   profiles would be similar to those shown in panels b (new abundance)

618:   and d (old abundance).

619: \label{fig:samples}

620: }

621: \end{figure*}

622: 

623: \begin{figure*}

624: %\epsscale{2}

625: \plotone{f3.eps}

626: \caption{Mean slope $m_{6300}$ of the central spectral feature (vertical

627:   shaded area in Fig~\ref{fig:profs}) as a function of the

628:   abundance ratio between O and Ni (\EO/\ENi). The solid lines show

629:   the average curve obtained from many different inversions of the

630:   \FeI \, lines using different values for the Fe abundance, with

631:   \EFe \, in the range 20, 40 ppm. The blue shaded areas represent

632:   the 1-$\sigma$ spread for all of those models. The diamonds

633:   represent the same curve obtained when a standard sunspot model

634:   (model M of \citeNP{maltbyetal}) is employed. The horizontal dashed line

635:   shows the value for $m_{6300}$ determined from the observed

636:   profile. The orange band indicates the uncertainty of the

637:   measurement due to noise. All three datasets yield a $m_{6300}$ that

638:   is within the orange band.

639: \label{fig:asymvsratio}

640: }

641: \end{figure*}

642: 

643: 

644: \end{document}

645: 

646: 

647: 

648: 

649: 

650: 

651: 

652: 

653: 

654: 

655: 

656: 

657: