1: %

2: % MS.TEX: Chaplin et al. paper on use of r_0/1 to constrain Z

3: %

4: % Revision History:

5: %

6: % 2006 November 3: started, WJC

7: %

8: % 2007 Feb, Jan 29: started SB

9: %

10: % 2007 Feb 3, very different version SB

11: %

12: % 2007 Feb 8: Started final tidy-up, new plots etc, WJC

13: %

14: % 2007 July 17: Dealing with referee's comments, WJC

15: 

16: 

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

18: 

19: %\documentclass{emulateapj}

20: 

21: \shorttitle{Low-$l$ solar $p$ modes and solar abundances}

22: 

23: \shortauthors{Chaplin et al.}

24: 

25: \begin{document}

26: 

27: \title{Solar heavy element abundance: constraints from frequency

28: separation ratios of low-degree $p$ modes}

29: 

30: \author{William J. Chaplin}

31: 

32: \affil{School of Physics and Astronomy, University of Birmingham,

33: Edgbaston, Birmingham B15 2TT, U.K.; w.j.chaplin@bham.ac.uk}

34: 

35: \author{Aldo~M.~Serenelli}

36: 

37: \affil{Institute for Advanced Study, Einstein Drive, Princeton, NJ

38: 08540; aldos@ias.edu}

39: 

40: 

41: \author{Sarbani Basu}

42: 

43: \affil{Department of Astronomy, Yale University, P.O. Box 208101, New

44: Haven, CT 06520-8101; sarbani.basu@yale.edu}

45: 

46: \author{Yvonne~Elsworth}

47: 

48: \affil{School of Physics and Astronomy, University of Birmingham,

49: Edgbaston, Birmingham B15 2TT, U.K.; y.p.elsworth@bham.ac.uk}

50: 

51: \author{Roger~New}

52: 

53: \affil{Faculty of Arts, Computing, Engineering and Sciences, Sheffield

54: Hallam University, Sheffield S1 1WB, U.K.; r.new@shu.ac.uk}

55: 

56: \and

57: 

58: \author{Graham~A.~Verner}

59: 

60: \affil{School of Physics and Astronomy, University of Birmingham,

61: Edgbaston, Birmingham B15 2TT, U.K.; gav@bison.ph.bham.ac.uk}

62: 

63: \begin{abstract}

64: 

65: We use very precise frequencies of low-degree solar-oscillation modes

66: measured from 4752 days of data collected by the Birmingham

67: Solar-Oscillations Network (BiSON) to derive seismic information on

68: the solar core.  We compare these observations to results from a large

69: Monte Carlo simulation of standard solar models, and use the results

70: to constrain the mean molecular weight of the solar core, and the

71: metallicity of the solar convection zone.  We find that only a high

72: value of solar metallicity is consistent with the seismic

73: observations.  We can determine the mean molecular weight of the solar

74: core to a very high precision, and, dependent on the sequence of Monte

75: Carlo models used, find that the average mean molecular weight in the

76: inner 20\% by radius of the Sun ranges from 0.7209 to 0.7231, with

77: uncertainties of less than 0.5\% on each value.  Our lowest seismic

78: estimate of solar metallicity is $Z=0.0187$ and our highest is

79: $Z=0.0239$, with uncertainties in the range of 12--19\%.  Our results

80: indicate that the discrepancies between solar models constructed with

81: low metallicity and the helioseismic observations extend to the solar

82: core and thus cannot be attributed to deficiencies in the modeling of

83: the solar convection zone.

84: 

85: 

86: \end{abstract}

87: 

88: \keywords{Sun: helioseismology - Sun: interior - Sun: abundances}

89: 

90: 

91: \section{Introduction}

92: \label{sec:intro}

93: 

94: In a series of papers, Asplund et al.  (2004, 2005b, 2005c) and

95: Allende-Prieto et al. (2001, 2002) have revised the spectroscopic

96: determinations of the solar photospheric composition. In particular,

97: their results have determined carbon, nitrogen and oxygen abundances

98: to be lower by about 25\% to 35\% than previous determinations

99: (Grevesse \& Sauval 1998; hereafter GS98). The revision of the oxygen

100: abundance leads to a comparable change in the abundances of neon and

101: argon. Additionally, Asplund (2000) has also determined a somewhat

102: lower value (10\%) for the photospheric abundance of silicon compared

103: to the GS98 value. As a result, all the elements for which abundances

104: are obtained from meteoritic measurements have seen their abundances

105: reduced by a similar amount. These measurements have been summarized

106: in Asplund et al. (2005a; hereafter AGS05), and the net result is that

107: the ratio of the mass fraction of heavy elements to hydrogen in the

108: Sun is $Z/X=0.0165$ (alternatively, $Z=0.0122$), about 28\% lower than

109: the previous value, $Z/X=0.0229$ ($Z=0.0169$) given by GS98.

110: 

111: The new low-abundance value for the heavy elements, albeit the result

112: of a much more sophisticated modeling of the solar atmosphere, has

113: given rise to discrepancies between helioseismic observations and

114: predictions from solar models constructed with the low value of

115: $Z/X$. Solar models constructed with the GS98 composition have shown a

116: remarkable agreement with the solar structure, as determined by

117: helioseismology techniques (Christensen-Dalsgaard et al. 1996; Bahcall

118: et al. 1997; Morel et al. 1999; Basu et al. 2000). However, when the

119: AGS05 composition is adopted in the solar models, the predicted

120: surface helium abundance is too low and the convective envelope too

121: shallow.  Additionally, the model sound-speed and density profiles

122: show a degraded agreement with their solar counterparts when compared

123: to predictions from models that use the older GS98 composition

124: (Bahcall \& Pinsonneault 2004; Basu \& Antia 2004; Bahcall et

125: al. 2005a, 2005b, 2005c, 2006; Delahaye \& Pinsonneault 2006).

126: 

127: The discrepancy between the low-$Z/X$ models and the helioseismic

128: observations has led to attempts to determine the solar metallicity

129: from helioseismic data alone, just as the solar helium abundance was

130: determined by helioseismology. Antia \& Basu (2006) used helioseismic

131: data to estimate a value for $Z$ in the solar convection zone of

132: $0.0172\pm 0.002$, i.e., closer to the GS98 value and much larger than

133: the AGS05 value. The uncertainty in their results arose from

134: uncertainties in the equation of state, and a lack of data on acoustic

135: modes of high angular degree ($ l > 200$).  In this paper we look to

136: the solar core, where uncertainties in the physics of the equation of

137: state and opacities are much lower, to try to constrain the solar

138: metallicity.

139: 

140: We make use of solar $p$-mode data derived from observations made by

141: the ground-based Birmingham Solar-Oscillations Network (BiSON; Chaplin

142: et al. 1996). The BiSON instruments make disc-averaged observations of

143: the Sun in Doppler velocity.  BiSON data can be used to determine very

144: precise frequencies of low-$l$ modes ($l \le 3$) that can be used to

145: probe the solar core. The possibility of using these data to shed

146: light on the solar abundance problem was explored by Basu et

147: al. (2007). Basu et al. made very specific combinations of the low-$l$

148: frequencies, the so-called `small frequency spacings' and `frequency

149: separation ratios', to compare models with the observations.

150: 

151: The \emph{small frequency spacings} of the low-$l$ $p$ modes are given

152: by the combination.

153:  \begin{equation}

154:  d_{l\,l+2}(n) = \nu_{n,l} - \nu_{n-1,l+2},

155:  \label{eq:fine}

156:  \end{equation}

157: where $\nu_{n,l}$ is the frequency of a mode of degree $l$ and radial

158: order $n$.  The fine spacings are determined predominantly by the

159: sound-speed gradient in the core. Using the asymptotic theory of $p$

160: modes it can be shown that (see e.g., Christensen-Dalsgaard \&

161: Berthomieu 1991)

162: \begin{equation}

163: d_{l\,l+2}(n)\simeq -(4l+6){\Delta_l(n)\over{4\pi^2\nu_{n,l}}}\int_0^R

164: {dc\over dr}{dr\over r},

165: \label{eq:smallsep}

166: \end{equation} 

167: where $R$ is the solar radius, and $\Delta_l(n)$ is the large

168: frequency spacing given by

169:  \begin{equation}

170:  \Delta_l(n) = \nu_{n,l} - \nu_{n-1,l}.

171: \label{eq:largesep}

172:  \end{equation}

173: The large frequency separation depends inversely on the sound-travel

174: time between the center and the surface of the Sun.  The frequencies

175: $\nu_{n,l}$ and $\nu_{n-1,l+2}$ are very similar and hence are

176: affected in a similar way by near-surface effects.  By taking this

177: difference in frequency a large part of the effects from the

178: near-surface uncertainties cancels out, making the spacings a useful

179: probe of the deep solar interior and core. Some residual effects do

180: nevertheless remain.  One way of reducing the effects of the

181: near-surface errors is to use the \emph{frequency separation ratios}.

182: The frequency separation ratios (Roxburgh \& Vorontsov 2003; Ot\'i

183: Floranes, Christensen-Dalsgaard \& Thompson 2005; Roxburgh 2005) are

184: formed from the small frequency spacings and large frequency spacings

185: of the modes.  The separation ratios are then constructed according

186: to:

187:  \begin{equation}

188:  r_{02}(n) = \frac{d_{02}(n)}{\Delta_1(n)},

189:  ~~~~~~~~r_{13}(n) = \frac{d_{13}(n)}{\Delta_0(n+1)}.

190:  \label{eq:rats}

191:  \end{equation}

192: Since both the small and large spacings are affected in a similar

193: manner by near-surface effects, these ratios are somewhat independent

194: of the structure of the surface.

195: 

196: Basu et al. (2007) showed that small spacings and separation ratios

197: for models constructed with the old GS98 composition match the

198: observed BiSON spacings and ratios much more closely than do the

199: spacings and ratios of models with the lower AGS05 composition. In

200: short, models constructed with higher metallicities compare better

201: with the BiSON data than do models constructed with lower

202: metallicities, although the level of agreement deteriorates when the

203: metallicity becomes very large. This indicates that we should be able

204: to determine solar metallicity using the spacing and ratio data.

205: 

206: In this paper, we therefore expand on the work of Basu et al. (2007),

207: and use the small spacings and separation ratios from BiSON data to

208: determine the metallicity of the Sun. We compare the observed spacings

209: and ratios with spacings and ratios of some 12,000 solar models. The

210: models, which were made for an extensive Monte Carlo simulation

211: (Bahcall et al. 2006), account for all the relevant uncertainties

212: entering standard solar model calculations.  From this comparison, we

213: show that it is possible to place extremely tight constraints on

214: $\mu_{\rm c}$, the mean molecular weight averaged over the inner

215: 20\,\% by radius (i.e., over most of the solar core) of the Sun.

216: Since the mean molecular weight in the core is related to the

217: metallicity at the surface (i.e., the convection zone), we can also

218: place reasonably precise constraints on the heavy element abundance,

219: $Z$.  Both $\mu_{\rm c}$ and $Z$ are measures of solar metallicity,

220: albeit for different regions of the Sun.

221: 

222: 

223: The rest of the paper is organized as follows. The observed data and

224: the models used are described in \S~\ref{sec:data}. In

225: \S~\ref{sec:lin}, we compare the BiSON separation ratios with the

226: separation ratios of two sequences of models made to test the response

227: of the spacings and ratios to $\mu_{\rm c}$ and $Z$.  In this section

228: we also show how we can obtain a seismic estimate of the solar

229: $\mu_{\rm c}$ and $Z$ by comparison of the BiSON and model spacings

230: and ratios.  In \S~\ref{sec:op} we determine the systematic errors

231: that arise due to uncertainties in the radiative opacities and the

232: equation of state --- two quantities are that not amenable to a Monte

233: Carlo type study.  In \S~\ref{sec:mc} we expand the BiSON-model

234: comparison by analyzing the results for a grand total of 12,000 Monte

235: Carlo solar models, computed by Bahcall et al. (2006). This analysis

236: allows us to test the impact of changes to several input solar model

237: parameters (in particular changes to the mixture of the heavy

238: elements) on the results.  Finally, we summarize our results in

239: \S~\ref{sec:summ}.

240: 

241: 

242: %\section{BiSON data}

243: \section{Data and models used}

244: \label{sec:data}

245: 

246: 

247: We have made use of Doppler velocity observations made by the BiSON

248: over the 4752-d period beginning 1992 December 31, and ending 2006

249: January 3. Frequencies were determined by fitting resonant peaks in

250: the power spectrum of the complete time series to yield estimates of

251: the low-$l$ frequencies (e.g., see Chaplin et al. 1999). Prior to

252: calculation of the small spacings and separation ratios, we removed

253: the solar-cycle shifts from the raw fitted low-$l$

254: frequencies. Details on the process used to remove the solar-cycle

255: shifts can be found in Basu et al. (2007) (see also Chaplin et

256: al. 2005). The BiSON spacings and ratios were then constructed from

257: these corrected frequency data, and the uncertainties on individual

258: fitted frequencies propagated in the usual manner to give the spacing

259: and ratio uncertainties. Table~\ref{tab:bison} contains those

260: solar-cycle-corrected BiSON mode frequencies which were used to

261: compute the BiSON frequency spacings and separation ratios analyzed in

262: the paper.

263: 

264: Basu et al. (2007) showed that the separation ratios depend on the

265: molecular weight, but they did not determine the exact dependence of

266: the separation ratios on the average mean molecular weight of the

267: core. To do so, we use two very different sets of models. The first

268: set of models, which we refer to as the \emph{test} models, consist of

269: two sequences of ten solar models each. One sequence of models was

270: constructed with the relative heavy element abundances of GS98, while

271: the second sequence was made with the relative heavy abundances of

272: AGS05.  To fix the $Z/X$ of a given model in either sequence, the

273: individual relative heavy element abundances of GS98 (or AGS05) were

274: multiplied by the same constant factor. This factor was then changed

275: from one model to another within the sequence.  All models in the two

276: sequences were constructed with the same nuclear reaction rates,

277: opacities, equation of state and diffusion rates.  These models were

278: constructed to test the dependence of the separation ratios on the

279: \emph{average} mean molecular weight, $\mu_{\rm c}$ and the

280: \emph{total} heavy element abundance, $Z$. For reference, we include

281: in this paper tables of low-$l$ mode frequencies for two standard

282: models. Table~\ref{tab:gs98} has frequencies for a model with the

283: exact GS98 abundance; while Table~\ref{tab:ags05} has frequencies for

284: a model with the exact AGS05 abundance.

285: 

286: 

287: The second set of models comprised a grand total of 12,000 models

288: created for a Monte Carlo study (Bahcall et al. 2006), and we refer to

289: these as the \emph{Monte Carlo} models.  The characteristics and

290: methods of computation of the models can be found in Bahcall et

291: al. (2006). Here, we summarize the salient points only.

292: 

293: For each solar model 19 input parameters were drawn randomly from

294: separate probability distributions for every parameter (see Bahcall et

295: al. 2006 for more details).  Seven of the input parameters were

296: nuclear reaction rates for important low-energy fusion reactions. The

297: solar age, luminosity, and diffusion coefficient (rate) were the next

298: three parameters. The final nine parameters were the abundances of

299: nine heavy elements: C, N, O, Ne, Mg, Si, S, Ar and Fe. For each solar

300: model, radiative opacity tables corresponding to the randomly

301: generated composition were computed and used. Radiative opacity tables

302: were generated using data and codes provided by the Opacity Project

303: group as presented in Seaton (2005). Low-temperature opacities (for

304: temperatures under $10^4\,\rm K$) were from Ferguson et al. (2005).

305: Variations of the abundances were dealt with in such a way as to give

306: four sequences of Monte Carlo models. Choices had to be made regarding

307: the underlying mixture, and the probability distribution for the

308: mixture. Two basic mixtures were used: The GS98 and the AGS05

309: mixtures. Probability distributions were then assigned on the basis of

310: two different estimates of the uncertainties in the abundances of the

311: nine individual elements: `conservative' (large) uncertainties, based

312: on differences between the abundances of the GS98 and AGS05 mixtures;

313: and `optimistic' (small) uncertainties, based on the uncertainties

314: quoted by Asplund et al.  (2005a). The content of the four sequences

315: of models may be summarized as follows:

316: 

317: GS-Cons --- These 5000 models were made with the `conservative'

318: (large) abundance uncertainties, centered on the GS98 mixture.

319: 

320: GS-Opt --- These 1000 models were made with the `optimistic'(small)

321: abundance uncertainties, centered on the GS98 mixture.

322: 

323: AGS-Cons --- These 1000 models were made with the `conservative'

324: (large) abundance uncertainties, centered on the AGS05 mixture.

325: 

326: AGS-Opt --- These 5000 models were made with the`optimistic' (small)

327: abundance uncertainties, centered on the AGS05 mixture.

328: 

329: We have calculated frequencies of low-degree modes for all the models

330: and constructed the small spacings and separation ratios in exactly

331: the same manner as for the observations. The model frequencies come

332: from adiabatic calculations. Since non-adiabatic effects are not

333: included, this leads to a well-known mismatch between the absolute

334: values of the adiabatic model frequencies and the observed

335: frequencies. However, as noted earlier, differences due to these

336: near-surface effects are reduced significantly by taking frequency

337: differences, and using the small spacings and separation ratios.

338: 

339: 

340: \section{Dependence of the separations on metallicity}

341:  \label{sec:lin}

342: 

343: In order to parametrize the relation between metallicity and the small

344: spacings and separation ratios, we begin by comparing the BiSON

345: spacings and ratios with the spacings and ratios of the \emph{test}

346: models. To show how the BiSON-model comparisons were made, consider

347: the analysis of the separation ratios. We calculated for each model

348: the differences between the observed BiSON ratios, $r_{02}(n)$ and

349: $r_{13}(n)$, and the model ratios, $r'_{02}(n)$ and $r'_{13}(n)$,

350: i.e.,

351:  \begin{equation}

352:  \Delta r_{l,l+2}(n) = r_{l,l+2}(n) - r'_{l,l+2}(n).

353:  \label{eq:diff1}

354:  \end{equation}

355: These differences were then averaged over $n$, for each of the $\Delta

356: r_{02}(n)$ and $\Delta r_{13}(n)$, to yield \emph{weighted mean

357: differences}, $\langle \Delta r_{02} \rangle$ and $\langle \Delta

358: r_{13} \rangle$:

359:  \begin{equation}

360:  \langle \Delta r_{l,l+2} \rangle =

361:  \frac{\displaystyle\sum_{n}^{} \Delta r_{l,l+2}(n) / \sigma^2_{r_{l,l+2}}(n)}{

362:        \displaystyle\sum_{n}^{} 1/\sigma^2_{r_{l,l+2}}(n)}.

363:  \label{eq:diff2}

364:  \end{equation}

365: The formal uncertainties of the BiSON spacings,

366: $\sigma_{r_{l,l+2}}(n)$, were used to weight the averages (with the

367: usual uncertainty-squared Gaussian weighting applied). We averaged

368: data over the ranges where good determinations of the separation

369: ratios were available, here $n=9$ to 25.

370: 

371: Fig.~\ref{fig:linmu} shows plots of the weighted mean differences of

372: the separation ratios (in \%) versus $\ln \mu_{\rm c}$, the natural

373: logarithm of the average mean molecular weight of the core.  The top

374: two panels show data for the GS98 set, and the bottom two panels show

375: data for the AGS05 set. The formal uncertainties on each point, which

376: come from the BiSON data, are not plotted. They are $0.046\,\%$ on

377: each $\langle \Delta r_{02} \rangle$; and $0.038\,\%$ on each $\langle

378: \Delta r_{13} \rangle$. Fig.~\ref{fig:lin} shows plots of the weighted

379: mean differences versus the natural logarithm of the surface heavy

380: element abundances, $\ln Z$. Again, the top two panels show data for

381: the GS98 set, and the bottom two panels show data for the AGS05 set.

382: 

383: When a straight line was fitted to the data in each plot (solid

384: lines), the fitting coefficients indicated that a linear dependence

385: was a good model for the data. The fit for the $Z$ data was in all

386: cases described by:

387:   \begin{equation}

388:   \langle \Delta r_{l,l+2} \rangle =

389:   \alpha_{l,l+2} + \beta_{l,l+2} \ln Z,

390:   \label{eq:fit}

391:   \end{equation}

392: where $\alpha_{l,l+2}$ and $\beta_{l,l+2}$ are, respectively, the

393: best-fitting estimates of the intercept and gradient of the straight

394: line.  A similar straight-line fitting model was used for the

395: $\mu_{\rm c}$ data, with $\ln \mu_{\rm c}$ used as the independent

396: variable.

397: 

398: The quality of the $Z$ fits degraded significantly when $Z$, rather

399: than $\ln Z$, was used as the independent variable. Visible departures

400: from a straight line were then observed.  The fits also deteriorated

401: when the fine spacings, rather than the separation ratios, were used.

402: We therefore devote the remainder of the paper to analysis of weighted

403: mean differences made from the separation ratios of the BiSON data and

404: the solar models.  These weighted mean differences have then been used

405: to determine the natural logarithm of $Z$, as opposed to $Z$

406: itself. We adopted a similar approach to our study of the mean

407: molecular weight in the core. Here, we used the $\ln \mu_{\rm c}$ of

408: the models as the independent variable for the plots. Again, this was

409: because we found that use of $\mu_{\rm c}$, rather than $\ln \mu_{\rm

410: c}$, degraded the quality of the fits somewhat, although not as much

411: as in the case of $Z$.

412: 

413: With reference to Figures~\ref{fig:linmu} and~\ref{fig:lin}, it is not

414: surprising that the two measures of metallicity affect the separation

415: ratios in similar ways.  The quantities $\mu_{\rm c}$ and $Z$ are

416: related, in that a higher $Z$ results in a higher $\mu_{\rm c}$. Two

417: models with the same $Z$, nuclear reaction rates, opacities and

418: equation of state can have different values of $\mu_{\rm c}$ only if

419: the diffusion rates are different in the two models.  It is, however,

420: not surprising that the dependence of the separation ratios on the two

421: parameters is somewhat different given that $\mu_{\rm c}$ also depends

422: heavily on the helium abundance in the core. All the models are

423: calibrated to have the same radius and luminosity at the solar age,

424: and hence differences in $Z$ generally give rise to differences in the

425: core helium abundance.

426: 

427: 

428:  \subsection{Seismic estimates of solar $Z$ and $\mu_{\rm c}$}

429:  \label{subsec:seis}

430: 

431: If the observational data are unbiased, the location on the abscissa

432: that marks where each best-fitting straight line passes through zero

433: on the ordinate will give us a `seismic' estimate of the average mean

434: molecular weight $\mu_{\rm c}$ averaged over the inner 20\,\% by

435: radius, and the surface abundance $Z$ of the Sun.  (Note that the $Z$

436: we refer to is always the present-day (solar age) surface $Z$.) These

437: locations are marked on the various panels of Figures~\ref{fig:linmu}

438: and~\ref{fig:lin} by the intersecting dotted lines. Conclusions drawn

439: from the test models will of course neglect any dependence of the

440: differences of the separation ratios on changes to other solar model

441: input parameters, including changes to the mixture of the heavy

442: elements. We go on to discuss the impact of such changes, and the

443: overall error budget, in later sections. Here we show explicitly how a

444: value for, and uncertainty on, $Z$ may be estimated from the

445: differences. The same procedures give results for $\mu_{\rm c}$.

446: 

447: From the best-fitting coefficients, we seek to find $\ln Z$ where

448: $\langle \Delta r_{l,l+2} \rangle=0$.  From Equation~\ref{eq:fit}, we

449: therefore have:

450:   \[

451:   \ln Z =  -\alpha_{l,l+2} / \beta_{l,l+2},

452:   \]

453: So that

454:   \begin{equation}

455:   Z = \exp[-\alpha_{l,l+2} / \beta_{l,l+2}].

456:   \label{eq:getz}

457:   \end{equation}

458: Application of Equation~\ref{eq:getz} to the $\langle \Delta r_{02}

459: \rangle$ and $\langle \Delta r_{13} \rangle$ data of each set (GS98 or

460: AGS05) will give us four seismic estimates of the solar heavy element

461: abundance. Using GS98 models we get $Z=0.01798$ and $Z=0.01774$ for

462: $\langle \Delta r_{02} \rangle$ and $\langle \Delta r_{13} \rangle$

463: respectively. The corresponding results using the AGS05 models are

464: $Z=0.01617$ and $Z=0.01611$. A similar analysis of $\mu_{\rm c}$

465: results in values of $0.7253$ and $0.7244$ when GS98 models are used,

466: and $0.7260$ and $0.7255$ when AGS05 test models are used.  These

467: estimates are listed in the fourth column of Tables~\ref{tab:linmu}

468: and \ref{tab:lin}, along with estimates of the goodness-of-fit and

469: uncertainties.

470: 

471: We make use of the observed scatter (variance) of the differences

472: $\langle \Delta r_{02} \rangle$ and $\langle \Delta r_{13} \rangle$

473: about their best-fitting straight line to estimate the uncertainty on

474: the seismic estimates of solar $\mu_{\rm c}$ and $Z$.  In short, we

475: translate the characteristic scatter on the ordinate into an implied

476: uncertainty on the abscissa. To determine the uncertainty in $Z$ (the

477: same procedure was used for $\mu_{\rm c}$) we first determine the set

478: of residuals about the best-fitting straight line, i.e., for each

479: point we compute

480:   \[

481:   \delta \langle \Delta r_{l,l+2} \rangle =

482:  \langle \Delta r_{l,l+2} \rangle - [\alpha_{l,l+2} + \beta_{l,l+2} \ln Z].

483:   \]

484: The variance of these residuals yields an estimate of their $1\sigma$

485: standard deviation, which we call $\sigma_{\langle \Delta r_{l,l+2}

486: \rangle}$. This characteristic uncertainty on the residuals may be

487: translated into an implied uncertainty on $\ln Z$ via the best-fitting

488: gradient, i.e.,

489:   \begin{equation}

490:   \sigma(\ln Z) = \sigma_{\langle \Delta r_{l,l+2} \rangle} /

491:   \beta_{l,l+2}.

492:   \label{eq:method2}

493:   \end{equation}

494: The equivalent $1\sigma$ limits on $Z$ are then bounded by $\exp{[\ln

495: Z+\sigma(\ln Z)]}$ and\\ $\exp{[\ln Z-\sigma(\ln Z)]}$. Our estimate

496: of the uncertainty is itself uncertain through the uncertainty on the

497: gradient, $\beta_{l,l+2}$. Thus we require that $N$ be large enough to

498: ensure that the best-fitting gradient (and our `look-up curve') is

499: well constrained. Here, gradients for the $Z$ fits were returned to a

500: fractional precision of better than 1\,\%.

501: 

502: The third column of Tables~\ref{tab:linmu} and~\ref{tab:lin} shows the

503: computed $\sigma_{\langle \Delta r_{l,l+2} \rangle}$ (in \%). The

504: fifth column gives the implied $1\sigma$ uncertainties on $\mu_{\rm

505: c}$ and $Z$, which we call $\sigma(\mu_{\rm c})$ and $\sigma(Z)$.

506: Here, the positive and negative uncertainties were the same, at the

507: level of precision of the data, because the fractional uncertainties

508: were so small. The sixth column shows the implied precision in the

509: determination of $\mu_{\rm c}$ and $Z$ (in \%). For each of the GS98

510: and AGS05 sets, we also combined the estimates from $\langle \Delta

511: r_{02} \rangle$ and $\langle \Delta r_{13} \rangle$ -- on the

512: assumption the estimates are independent -- to give the estimates

513: shown in the third and sixth rows of the tables.

514: 

515: Inspection of the $\mu_{\rm c}$ results in Table~\ref{tab:linmu} shows

516: that all results (both individual and combined) are consistent with

517: one another.  This is not surprising given the almost direct

518: dependence of the sound speed, and its derivative, on $\mu_{\rm c}$.  The

519: combined estimate obtained with the GS98 models differs from that

520: obtained with the AGS05 models by only $1\sigma$. As we shall see in

521: \S~\ref{sec:mc}, the uncertainties on the seismic estimates of

522: $\mu_{\rm c}$ do not increase much when other changes to the solar

523: model input parameters are considered.

524: 

525: Inspection of the $Z$ results in Table~\ref{tab:lin} shows that the

526: individual, and combined, estimates for solar $Z$ are all

527: significantly higher than the `low' Asplund et al. value of $Z \sim

528: 0.0122$. However, our combined GS estimate ($0.01785$) and its

529: combined AGS counterpart ($0.01611$) differ from each other by

530: $\approx 16\sigma$ (combined uncertainty). This difference might at

531: first glance be seen as a cause for concern and an indication that

532: systematic errors are much larger than the random errors caused by

533: uncertainties in the observed frequencies. However, we go on to show

534: in \S~5 that other systematic sources of error -- arising from the

535: sensitivity of the separation ratios to other parameters of the solar

536: models, including the relative mixture of the heavy elements -- mean

537: that realistic estimates of the uncertainties on $Z$ are actually

538: larger in size than the uncertainties given in

539: Table~\ref{tab:lin}. This is in stark contrast to what is found for

540: $\mu_{\rm c}$ (see previous paragraph).

541: 

542: Disagreement between the seismic results obtained from the GS98 and

543: AGS98 models can to a large extent be understood in terms of the

544: differences in the two mixtures.  The influence of the relative

545: mixture is clear from the fact that the separation ratios for the

546: AGS05 models and the GS98 models are different for the same value of

547: $Z$; the differences are less pronounced for $\mu_{\rm c}$. This

548: finding is not difficult to understand. For the calibrated solar

549: models used in this work, the dominant contribution to the $Z$ and

550: $\mu_{\rm c}$ of each model comes from different elements. For $Z$,

551: the dominant elements, in order of importance, are oxygen, carbon,

552: neon and nitrogen.  The value of $\mu_{\rm c}$ is determined by the mass

553: fractions of helium and hydrogen in the core. The abundances of

554: hydrogen and helium in the core depend strongly on the abundances of

555: heavy metals that contribute to the opacity in the core. These metals,

556: again in order of importance, are iron, silicon, sulfur and

557: oxygen. The difference between the GS98 and AGS05 mixture lies

558: predominantly in the relative abundances of oxygen, carbon, nitrogen

559: and neon, and much less so in the abundances of iron, silicon and

560: sulfur. This explains why for the same $Z$, $\mu_{\rm c}$ is different

561: for the GS98 and AGS05 models, as can be seen from Fig.~\ref{fig:zmu}.

562: 

563: Since the separation ratios depend basically on $\mu_{\rm c}$ and

564: temperature, the location at which $\langle \Delta r_{l,l+2} \rangle

565: =0$ occurs at slightly different values for the two sets of

566: models. The slopes of the $\langle \Delta r_{l,l+2} \rangle$-$\ln Z$

567: curves for the GS98 and AGS05 models (Figures~\ref{fig:lin}) are also

568: different for the same reason.  We investigate other sources of

569: systematic errors in \S~\ref{sec:mc}. The results in \S~\ref{sec:mc}

570: show that realistic estimates of the uncertainties on $Z$ are actually

571: larger in size than the uncertainties given in Table~\ref{tab:lin}.

572: 

573: 

574: \section{Uncertainties due to opacity and equation of state}

575: \label{sec:op}

576: 

577: The results obtained with the test models have two obvious

578: limitations. First, they do not test the impact of changes to the

579: relative mixture of the heavy elements, except to reveal that changes

580: in relative abundances matter. And second, the models test the

581: dependence of the separation ratios on $\mu_{\rm c}$ and $Z$ for a

582: \emph{fixed} set of solar model input parameters, with only the $Z/X$

583: varied. The study above does not deal with uncertainties in the

584: seismic estimates of solar $\mu_{\rm c}$ and $Z$ caused by other input

585: parameters, such as nuclear reaction rates, diffusion rates,

586: uncertainties in the relative mixtures, etc.  We investigate these

587: effects by conducting a Monte Carlo study, which is described below in

588: \S~\ref{sec:mc}. This study allowed us to test the impact of 19 solar

589: model input parameters on the results. There are, however, two

590: important inputs that are not amenable to a Monte Carlo study, and

591: these are radiative opacities and the equation of state (EOS). These

592: two quantities cannot be described by a single number and hence we are

593: forced to use a different approach to determine the uncertainties in

594: the separation ratios caused by uncertainties in opacities and EOS. We

595: use an approach similar to that used by Bahcall et al. (2006).

596: 

597: We determined the uncertainty introduced in $\left< \Delta

598: r_{l,l+2}\right>$ by the opacities as follows. We computed a pair of

599: solar models with the same input parameters and EOS, but one was made

600: with opacities from the OP project (Badnell et al. 2005) and the other

601: with opacities from OPAL (Iglesias \& Rogers 1996).  For this matched

602: pair of models we get $\left< \Delta r_{l,l+2}\right>_{{\rm OP},i}$

603: and $\left< \Delta r_{l,l+2}\right>_{{\rm OPAL},i}$ (here the

604: subscript {\it i} denotes the pair of matched models).  The unbiased

605: estimator for the variance of the difference is

606:  \begin{equation}

607:  s^2_i{(\left< \Delta r_{l,l+2}\right>({\rm opacity}))}= \left[\left<

608:  \Delta r_{l,l+2}\right>_{{\rm OP},i} - \left< \Delta r_{l,l+2}\right>_{{\rm

609:  OPAL},i} \right]^2/2.

610:  \label{eq:var}

611:  \end{equation}

612: and we adopt this quantity as the standard deviation squared,

613: $\sigma^2_i{(\left< \Delta r_{l,l+2}\right>{\rm (opacity))}}$.  In

614: order to obtain a more representative value for $\sigma_{\rm

615: opac}{(\left< \Delta r_{l,l+2}\right>)}$ we averaged this difference

616: over a set of $N=40$ matched pairs of models, where the 19 other input

617: parameters were varied for different matched pairs.  The final

618: expression we adopt for $\sigma_{\rm opac}{(\left< \Delta

619: r_{l,l+2}\right>)}$ is

620: \begin{equation}

621: \sigma_{\rm opac}{(\left<   \Delta  r_{l,l+2}\right>)}=   \sqrt{N^{-1}

622:   \sum_i{s^2_i{(\left<  \Delta r_{l,l+2}\right>({\rm opacity}))}} }.  

623: \label{eq:opac}

624: \end{equation}

625: An analogous procedure was used for the EOS, but in this case one

626: model in each pair was computed using the 2001 OPAL EOS (Rogers 2001,

627: Rogers \& Nayfonov 2002), while the other model was computed using the

628: 1996 OPAL EOS (Rogers et al.  1996).

629: 

630: In addition to the uncertainties on $\left< \Delta r_{02} \right>$ and

631: $\left< \Delta r_{13} \right>$ given by Equation~\ref{eq:opac}, we

632: applied the same procedure to compute the implied uncertainties for

633: the values of $\mu_{\rm c}$ and $Z$ predicted by the solar models.

634: Uncertainties in the opacity and EOS will affect the solar model

635: results for $Z$ and $\mu_{\rm c}$, i.e.  $Z$ and $\mu_{\rm c}$ will be somewhat

636: different for the two models in each of the matched pairs described

637: above.  The results are given in Table~\ref{tab:opac_eos}.

638: 

639: The impact on the separation ratios of uncertainty in the radiative

640: opacities is easy to understand --- changes in opacity cause changes

641: in temperature, which in turn change sound speed and its derivative,

642: thereby changing the separation ratios. However, since the opacity

643: uncertainties are small in the core, the overall effect is quite

644: modest. The impact of uncertainty in the EOS is more important. This

645: might seem surprising, until one realizes that the 1996 OPAL EOS did

646: not treat relativistic effects properly at temperatures and densities

647: relevant to the solar core. This results in a somewhat deficient core

648: structure (see Elliott \& Kosovichev, 1998). Since the deficiency is

649: mainly in the core, it will affect the low-$l$ separations used in

650: this work disproportionately and cause larger uncertainties in the

651: seismic estimates of solar $Z$ and $\mu_{\rm c}$.  The updated 2001

652: OPAL EOS has the correction put in. Thus the uncertainty in the

653: seismic estimates of solar $Z$ and $\mu_{\rm c}$ caused by EOS uncertainties

654: may be considered to be an upper limit to the EOS effects.

655: 

656: 

657: \section{The Monte Carlo study}

658: \label{sec:mc}

659: 

660: 

661: In this section we present results, using extended sequences of solar

662: models, which seek to address the influence of other input parameters

663: on the seismic solar $Z$ results reported in \S~\ref{subsec:seis}. The

664: bulk of the results come from tests on 12,000 solar models created for

665: a Monte Carlo study of the dependence of solar model characteristics

666: on different input parameters (see Bahcall et al. 2006).  As discussed

667: in \S~\ref{sec:data}, the models have 19 different input parameters

668: selected at random from a distribution of the inputs.

669: 

670: Scatter plots of the weighted mean differences of the separation

671: ratios (in \%) versus the natural logarithm of $\mu_{\rm c}$ are shown

672: in Figures~\ref{fig:GSmuMC} and \ref{fig:AGSmuMC} for all four sequences

673: of Monte Carlo models.  Figures~\ref{fig:GSMC} and~\ref{fig:AGSMC} show

674: the corresponding plots against $\ln Z$. Uncertainties caused by

675: uncertainties in the radiative opacities and EOS have been included by

676: adding to the data random components with normal distributions

677: characterized by the standard deviations given in

678: Table~\ref{tab:opac_eos}. The solid lines in each panel are the

679: best-fitting straight lines for the data. The dotted lines intersect

680: at the location along each best-fitting line where the weighted mean

681: difference is zero.

682: 

683: We again adopted a linear model for the analysis in the light of: (i)

684: the results on the test models (discussed in Section~\ref{sec:lin}

685: above); and (ii) the observed scatter on the plots, which precluded us

686: from imposing a more complicated fitting model. Detailed breakdowns of

687: fitting results for the Monte Carlo sequences are presented in

688: Table~\ref{tab:mcmu} (for $\mu_{\rm c})$, and Table~\ref{tab:mc} (for

689: $Z$).

690: 

691: The main results obtained from the Monte Carlo sequences are as

692: follows:

693: 

694: \begin{enumerate}

695: 

696: \item Before turning to discussion of the results on $\mu_{\rm c}$ and

697: $Z$, we consider first the $\left< \Delta r_{02}\right>$ and $\left<

698: \Delta r_{13}\right>$ of the model sequences. In the case of the

699: AGS-Opt sequence, the distribution of $\left< \Delta r_{02}\right>$ is

700: characterized by a mean value of $-2.05\,\%$ and a standard deviation

701: of 0.75\,\%.  For $\left< \Delta r_{13}\right>$ the corresponding

702: values are, respectively, $-1.80\,\%$ and 0.54\,\%.  These values

703: imply a difference with helioseismology measurements of $2.7\sigma$

704: and $3.3\sigma$ for $\left< \Delta r_{02}\right>$ and $\left< \Delta

705: r_{13}\right>$ respectively.  Note from Table~\ref{tab:opac_eos} that

706: the EOS has a large impact on the total standard deviations of the

707: $\left< \Delta r_{02}\right>$ and $\left< \Delta r_{13}\right>$

708: distributions.  As discussed previously, we are probably

709: overestimating the EOS uncertainties in the separation ratios.

710: Consequently, the 2.7 and 3.3-$\sigma$ differences should be

711: considered as robust upper limits to the real discrepancy.

712: 

713: On the other hand, for the GS-Opt sequence the mean and standard

714: deviation of the $\left< \Delta r_{02}\right>$ distribution are

715: $-0.43\,\%$ and 0.76\,\% respectively, while for $\left< \Delta

716: r_{13}\right>$ we get $-0.32\,\%$ and 0.53\,\% for the mean and the

717: standard deviation.  These numbers translate into differences of only

718: $\approx 0.6\sigma$ with the helioseismology measurements. It is worth

719: mentioning that we perform this comparison only for Monte Carlo

720: sequences having the optimistic choice of uncertainties, because the

721: aim is to compare helioseismology measurements with solar models that

722: adopt compositions (central values and uncertainties) given by the

723: solar abundance determinations, i.e. GS98 and AGS05.

724: 

725: \item As in the case of results obtained using the test models, the

726: average mean molecular weight in the inner 20\,\% by radius, $\mu_{\rm

727: c}$ is determined to much higher precision than $Z$ (results in

728: Table~\ref{tab:mcmu}). The precision in each of the combined measures

729: of $\mu_{\rm c}$ is better than 0.5\,\%. Even with the improved

730: precision, the estimates given by analysis of the four sequences of

731: models are in excellent agreement with one another and provide a very

732: robust determination of $\mu_{\rm c}$.  The largest difference between any

733: of the two combined measures is significant at only $\sim

734: 1\sigma$. The average of our four combined seismic estimates is

735: $\mu_{\rm c}=0.7226$.

736: 

737: \item From the AGS-Opt sequence of models, i.e.  that adopting the

738: AGS05 central values and uncertainties for the composition, we find

739: that solar models constructed with the AGS05 composition have an

740: average $\mu_{\rm c}$ of $0.7088 \pm 0.0029$ (1$\sigma$ uncertainty).

741: This value differs from our combined seismic estimates of solar

742: $\mu_{\rm c}$, which are given in Table~\ref{tab:mcmu}, by between 3.1

743: (for the combined AGS-Cons measure) and $3.7\sigma$ (for the combined

744: AGS-Opt measure) (where $\sigma$ is determined by adding in quadrature

745: the uncertainties from the observations and the uncertainty in the

746: theoretical distribution). On the other hand, in the case of models

747: adopting the GS98 composition, we derive an average value of $\mu_{\rm

748: c}=0.7203 \pm 0.0029$ for the optimistic (i.e., lower) uncertainties

749: on the abundance inputs.  All the combined seismic estimates agree

750: with this value to well within $1\sigma$.

751: 

752: \item All seismic estimates of solar $Z$ are high (see

753: Table~\ref{tab:mc}).  They lie noticeably above the low Asplund et

754: al. (2005a) (AGS05) value.  This is clear even from visual inspection

755: of the plots in Fig.~\ref{fig:AGSMC}, which shows results for models

756: with heavy-element mixtures based on the AGS05 values. The location on

757: the abscissa where the best-fitting lines pass through zero lie well

758: to the high-$Z$ side of each cloud of points.  None of the AGS-Opt

759: models has a mean weighted difference greater than zero. Estimation of

760: $Z$ from these data therefore amounts to an \emph{extrapolation},

761: rather than the \emph{interpolation} as is possible with the GS-Cons

762: and GS-Opt models.

763: 

764: From the AGS-Opt sequence of models, which has abundance uncertainties

765: consistent with Asplund et al. (2005a), we find that solar models

766: constructed with the AGS05 composition have an average $Z$ of $0.0125

767: \pm 0.0007$ (1$\sigma$ uncertainty).  This value differs from our

768: combined seismic estimates of solar $Z$, which are given in

769: Table~\ref{tab:mc}, by between $2.0\sigma$ (for the combined GS-Cons

770: measure) and $4.3\sigma$ (for the combined AGS-Opt measure) (where

771: $\sigma$ is determined by adding in quadrature the uncertainties from

772: the observations and the uncertainty in the theoretical distribution).

773: 

774: \item All seismic estimates of solar $Z$ are slightly higher than the

775: Grevesse \& Sauval (1998) (GS98) value, of $Z=0.0169$ (see

776: Table~\ref{tab:mc}).  This is clear from visual inspection of the

777: plots in Fig.~\ref{fig:GSMC}, which shows results for models with

778: heavy-element mixtures based on the GS values. The best-fitting lines

779: pass through zero in the high-$Z$ parts of the distributions of

780: points. Differences between the GS98 value and our combined seismic

781: estimates of solar $Z$ lie between $0.5\sigma$ (for the combined

782: GS-Cons measure) and $2.0\sigma$ (for the combined AGS-Opt measure).

783: 

784: \item All seismic estimates of solar $Z$ in Table~\ref{tab:mc} are in

785: good agreement with the seismic estimate of Antia \& Basu (2006),

786: which was $Z = 0.0172 \pm 0.0020$. The Antia \& Basu result was quoted

787: with a precision of just over 12\,\%. The precision in our four,

788: combined measures of $Z$ (rows~3, 6, 9 and 12 of Table~\ref{tab:mc})

789: ranges from $\sim 12\,\%$ to $\sim 19\,\%$. The largest difference

790: between any of the two combined measures is significant at only $\sim

791: 1\sigma$.

792: 

793: \end{enumerate}

794: 

795: The observed scatter in Figures~\ref{fig:GSmuMC} to \ref{fig:AGSMC}

796: deserves discussion, as do the differences between the results given

797: by the Monte Carlo sequence of models and the test models from

798: \S~\ref{sec:lin}.

799: 

800: The scatter in the computed $\langle \Delta r_{02} \rangle$ and

801: $\langle \Delta r_{13} \rangle$ about the best-fitting straight lines

802: -- as characterized by the $\sigma_{\langle \Delta r_{l,l+2} \rangle}$

803: in column~3 of Tables~\ref{tab:mcmu} and~\ref{tab:mc} -- far exceeds

804: the sizes of the formal uncertainties on the mean differences found in

805: \S~\ref{sec:lin}.  Recall that the formal uncertainty (a result of

806: uncertainties in the BiSON data) is only $0.046\,\%$ on each value of

807: $\langle \Delta r_{02} \rangle$; and $0.038\,\%$ on each value of

808: $\langle \Delta r_{13} \rangle$.  The uncertainties on the seismic

809: estimates of solar $Z$ and $\mu_{\rm c}$ computed from analysis of the

810: Monte Carlo data are therefore dominated by scatter introduced by the

811: input parameter choices for the solar models. As we discuss below, the

812: largest contribution to this scatter comes from the relative abundance

813: of heavy elements that contribute to $Z$.  It is worth pointing out

814: that while the uncertainties on the seismically estimated values of

815: solar $Z$, obtained with the Monte Carlo sequences, are much larger

816: (by over an order of magnitude) than the uncertainties obtained from

817: the test models, corresponding differences in the uncertainties on the

818: seismic values of $\mu_{\rm c}$, i.e., $\sigma(\mu_{\rm c})$, are just

819: a few times larger.

820: 

821: The $1\sigma$ standard deviation of the fitting residuals,

822: $\sigma_{\langle \Delta r_{l,l+2} \rangle}$, changes depending on

823: whether `conservative' (large) or `optimistic' (small) uncertainties

824: are used for the input abundances to the solar models.  The

825: $\sigma_{\langle \Delta r_{l,l+2} \rangle}$ values for the GS-Cons and

826: AGS-Cons sequences are between $\sim 20$ and $\sim 60\,\%$ higher than

827: the corresponding values of the GS-Opt and AGS-Opt sequences.  This is

828: not surprising --- the `conservative' GS-Cons and AGS-Cons sequences

829: cover, by their very nature, a large, `pessimistic' range of input

830: abundance values, thus the scatter due to the uncertainties in the

831: relative abundance is also larger than that for the `optimistic'

832: sets. The uncertainties that arise from using the GS-Cons and the

833: AGS-Cons models may therefore be treated as respectable \emph{upper}

834: bounds for $\sigma(Z)$ and $\sigma(\mu_{\rm c})$.

835: 

836: The gradients and intercepts of the best-fitting lines of the

837: relations between the separation ratios and the metallicity of the

838: test models (Figure~\ref{fig:lin}) are steeper and higher,

839: respectively, then those of their Monte-Carlo counterparts

840: (Figures~\ref{fig:GSMC}~and~\ref{fig:AGSMC}). The gradients are in

841: some cases steeper by as much as a factor of 3.  There are also

842: differences between one Monte Carlo sequence and another, with the

843: gradients differing by up to $\sim 35\,\%$; the gradients are

844: constrained typically to much higher precision. However, in spite of

845: these differences, the zero-crossing points, which serve to provide

846: the estimates of solar $Z$, return consistently robust values.

847: 

848: Again, it is not very difficult to explain why the gradients are

849: higher for the sets of test models compared to the sets of Monte Carlo

850: models, and why the gradients differ from one Monte Carlo set to

851: another.  The answer again lies in the impact of changes to the

852: relative mixture of heavy elements.  Each set of linear test models

853: has the same relative mixture of elements; these models represent the

854: case of perfect correlation between all the abundance uncertainties. A

855: given value of $Z$ gives a unique value of $\mu_{\rm c}$, and other

856: conditions in the core (e.g., temperature), and thus matches to a

857: unique value of $\langle \Delta r_{02} \rangle$ and $\langle \Delta

858: r_{13} \rangle$.  The Monte Carlo sequences simulate the opposite

859: situation; here, all the abundance uncertainties are assumed to be

860: independent of one another. The different dependences on relative

861: abundances of $Z$ and $\mu_{\rm c}$ imply that their variations in the

862: MC sequences become, to some degree, uncorrelated.  For example, large

863: changes in CNO elements give rise to large changes in $Z$ but have a

864: much more modest impact on $\mu_{\rm c}$ (and consequently on the

865: separation ratios). The result is that in the MC sequences the

866: one-to-one relation between $Z$, $\mu_{\rm c}$ and other conditions in

867: the core is lost, as is the one-to-one correlation between $Z$ and

868: $\langle \Delta r_{02} \rangle$ or $\langle \Delta r_{13}

869: \rangle$. The overall effect is to force the linear fitting gradients

870: to shallower values when a range of mixtures is admitted in the

871: models.  There are changes in the details of the mixtures between the

872: different Monte Carlo sequences, and hence some (albeit much more

873: modest) changes in gradient are seen from one sequence to another.

874: 

875: It is worth adding that changes to the diffusion rates do not alter

876: significantly the fitting gradients.  If one takes appropriate subsets

877: of the Monte Carlo sequences, one finds that the relation between the

878: rate of diffusion and $\mu_{\rm c}$ is always the same. And the

879: diffusion rate and $Z$ appear to be only very weakly correlated.

880: Changes to the diffusion do not affect the relation between the

881: fitting gradients of $\langle \Delta r_{02} \rangle$ or $\langle

882: \Delta r_{13} \rangle$ versus $\ln Z$ or $\ln\mu_{\rm c}$. The only

883: effects seen are changes to the intercept of the fits between the

884: separation ratios and the metallicities, which change the seismic

885: estimates of $Z$, and thus diffusion is a relevant source of

886: uncertainty. We find that physical inputs other than the relative

887: mixture, the diffusion coefficients, and the EOS have a much smaller

888: effect on the uncertainties of the estimated values of solar $\mu_{\rm

889: c}$ and $Z$.

890: 

891: 

892: 

893: \section{Results and Discussion}

894: \label{sec:summ}

895: 

896: We have used the frequencies of low-degree acoustic oscillations of

897: the Sun, determined by the BiSON network over a period of 4752 days,

898: to try and determine the metallicity $Z$ of the Sun. We did so by

899: comparing the frequency separarion ratios of a large set of solar

900: models, from four different Monte Carlo sequences, with the BiSON

901: observations.  Specifically, we used a weighted average of the

902: difference of the separation ratios. We find that in addition to giving

903: good constraints on the solar metallicity, which by definition is the

904: abundance of heavy elements in the solar convection zone, the

905: comparison provides an excellent means to determine the mean molecular

906: weight of the solar core.

907: 

908: The frequency separation ratios of the low-degree acoustic modes are

909: sensitive to the conditions in the solar core.  By using the weighted

910: mean differences of the separation ratios, we have obtained seismic

911: estimates for the mean molecular weight of the solar core ($\mu_{\rm

912: c}$).  The seismic estimates are robust and depend only very weakly on

913: the solar models used to construct the weighted mean differences. All

914: of the four sequences of Monte Carlo models used in this work allow us

915: to determine $\mu_{\rm c}$ to a precision of better than 0.5\%, and

916: estimates from different sequences are consistent with each other to

917: better than $1\sigma$.  This is true even for the seismic estimates

918: obtained using solar models with the AGS05 composition. These solar

919: models do not agree with helioseismic results on the solar sound-speed

920: and density profiles, surface helium abundance, depth of the

921: convective envelope and frequency separation ratios.  We have now

922: shown that solar models constructed with the AGS05 composition

923: (central values and uncertainties) fail to reproduce the seismically

924: determined average mean molecular weight of the solar core by more

925: than $3\sigma$.  On the other hand, solar models with the older GS98

926: composition have $\mu_{\rm c}$ well within $1\sigma$ of our seismic

927: estimate of $\mu_{\rm c}$ in the Sun.

928: 

929: We get estimates for the solar metallicity in the range between

930: $Z=0.0187$ and $Z=0.0229$ for the four Monte Carlo sequences with

931: uncertainties in the range of 12\% to 19\% on each measurement.  All

932: our seismic estimates for metallicity are consistent with each other

933: at about the $1\sigma$ level, and are higher than the solar

934: metallicity derived by Asplund et al. (2005) by between 2.1 and

935: $4.3\sigma$. Our estimates are also consistent with the seismically

936: derived value of $Z=0.0172\pm 0.0020$ obtained by Antia \& Basu

937: (2006). Finally, our estimates are slightly higher (by between 0.5 and

938: $2\sigma$) than the solar metallicity value recommended by Grevesse \&

939: Sauval (1998).

940: 

941: The Antia \& Basu (2006) results were obtained by looking at the

942: near-surface ionization zones. The signature of interest was the

943: change in the adiabatic index resulting from ionization of material,

944: which depends on the equation of state and the metallicity. The

945: biggest source of systematic error in the results of Antia \& Basu was

946: therefore the equation of state. Here, we have obtained very similar

947: results for solar $Z$ by looking at a region where very different

948: physical inputs matter. The main source of systematic error for this

949: study was the relative abundance of the heavy elements.  Thus errors

950: in the physical inputs used to construct the solar models are not the

951: reason why we obtain consistency high estimates of solar $Z$. As a

952: matter of fact, even when we use low-$Z$ models, we still get

953: estimates of solar $Z$ that are high.

954: 

955: When the AGS05 abundances were published, it was soon noticed that

956: solar models constructed with those abundances did not agree well with

957: the helioseismically determined properties of the Sun. In particular,

958: the most obvious discrepancy was in the position of the

959: convection-zone base.  It was speculated that this implied that the

960: physical inputs to the models, particularly the opacities, were

961: incorrect. The fact that the helium abundance in the convection zone

962: was also incorrect lead to the belief that the problems are localized

963: to the outer parts of the Sun.  Based on the results of this paper, we

964: conclude that this is not the case and that the problems extend to the

965: core.  It therefore seems unlikely that the origin of the discrepancy

966: lies in the simplified modeling of the regions close to the tachocline

967: or in the treatment of convection which is adopted in standard solar

968: model calculations.  While rotation and its associated mixing are

969: certainly not modeled in the standard solar models used in this work,

970: it should be noted that solar models that attempt to account for

971: rotation (and some degree of rotationally induced mixing) predict a

972: lower helium abundance in the solar core, which translates into a

973: lower average mean molecular weight (Palacios et al. 2006). Our

974: results seem to indicate that this would make the disagreement between

975: models with the AGS05 composition and helioseismology even worse.

976: 

977: 

978: \acknowledgements This paper utilizes data collected by the Birmingham

979: Solar-Oscillations Network (BiSON), which is funded by the UK Science

980: Technology and Facilities Council (STFC). We thank the members of the

981: BiSON team, colleagues at our host institutes, and all others, past

982: and present, who have been associated with BiSON. GAV acknowledges the

983: support of STFC.  SB acknowledges partial support from NSF grant

984: ATM-0348837.  AMS is partially supported through a John Bahcall

985: Membership and a Ralph E. and Doris M. Hansmann Membership at the IAS.

986: 

987: 

988: \begin{thebibliography}{}

989: 

990: \bibitem[Allende Prieto  et al.(2001)]{all01}Allende Prieto,  C., Lambert,

991:   D. L., \& Asplund, M.  2001, ApJ, 556, L63

992: 

993: \bibitem[Allende Prieto  et al.(2002)]{all02}Allende Prieto,  C., Lambert,

994:   D. L., \& Asplund, M. 2002, ApJ, 573, L137

995: 

996: \bibitem[Antia \& Basu(2006)]{hma06} Antia, H.M., Basu, S. 2006, ApJ,

997: 644, 1292

998: 

999: \bibitem[Asplund(2000)]{asp00a}Asplund, M. 2000, A\&A, 359, 755

1000: 

1001: \bibitem[Asplund  et al.(2004)]{asp04}Asplund,  M., Grevesse,  N., Sauval,

1002:   A. J., Allende Prieto, C., \& Kiselman, D. 2004, A\&A, 417, 751

1003: 

1004: \bibitem[Asplund   et   al.(2005)]{ags05}   Asplund,   M.,

1005:   Grevesse,  N.,  \&  Sauval, A.  J.  2005a,  in  ASP  Conf. Ser.  336,  Cosmic

1006:   Abundances   as   Records   of   Stellar  Evolution   and   Nucleosynthesis,

1007:   ed. T. G. Barnes III, \& F. N.  Bash (San Francisco: ASP), 25

1008: 

1009: \bibitem[Asplund et al.(2005)]{asp05c} Asplund, M., Grevesse, N.,

1010: Sauval, A. J., Allende Prieto, C., Blomme, R., 2005b,

1011: A\&A, 431, 693

1012: 

1013: \bibitem[Asplund  et al.(2005)]{asp05a}Asplund,  M., Grevesse,  N., Sauval,

1014:   A. J., Allende Prieto, C., \& Kiselman, D. 2005c, A\&A, 435, 339

1015: 

1016: \bibitem[Badnell et al.(2005)]{bad05} Badnell, N. R., Bautista, M. A.,

1017: Butler, K., Delahaye, F., Mendoza, C., Palmeri, P., Zeippen, C. J., \&

1018: Seaton, M. J. 2005, MNRAS, 360, 458

1019: 

1020: \bibitem[Bahcall et al.(2005a)]{bbps05} Bahcall, J. N., Basu, S.,

1021: Pinsonneault, M. H., \& Serenelli, A. M. 2005a, ApJ, 618, 1049

1022: 

1023: \bibitem[Bahcall et al(2005b)]{bbs05} Bahcall, J. N., Basu, S.,

1024: Serenelli, A.M. 2005b, 631, 1281

1025: 

1026: \bibitem[Bahcall \& Pinsonneault(2004)]{bap04} Bahcall, J.N.,

1027: Pinsonneault, M.H. 2004, Phys. Rev. Lett., 92, 121301

1028: 

1029: \bibitem[Bahcall et al.(1997)]{bah97} Bahcall, J.N., Pinsonneault,

1030: M.H., Basu, S., Christensen-Dalsgaard, J. 1997, Phys. Rev. Lett., 78,

1031: 171

1032: 

1033: \bibitem[Bahcall et al.(2005c)]{bsp05} Bahcall, J. N., Serenelli,

1034: A. M., \& Basu, S. 2005c, ApJ, 621, L85

1035: 

1036: \bibitem[Bahcall, Serenelli \% Basu (2006)]{bah06} Bahcall, J.N.,

1037: Serenelli, A.M., Basu, S., 2006, ApJSS, 165, 400

1038: 

1039: \bibitem[Basu \& Antia(2004)]{bA04} Basu, S., \& Antia, H. M. 2004,

1040: ApJ, 606, L85

1041: 

1042: \bibitem[Basu et al. (2007)]{bas07} Basu, S., Chaplin, W. J., Elsworth, Y.

1043: New, R., Serenelli, A. M., Verner, G. A., 2007, ApJ, 655, 660

1044: 

1045: \bibitem[Basu et al.(2000)]{basu02} Basu, S., Pinsonneault, M.H.,

1046: Bahcall, J.N. 2000, ApJ, 529, 1084

1047: 

1048: \bibitem[Chaplin et al.(1996)]{chaplin96} Chaplin~W.~J., Elsworth~Y.,

1049: Howe~R., Isaak~G.~R., McLeod~C.~P., Miller~B.~A., New~R.,

1050: van~der~Raay~H.~B.  \& Wheeler~S.~J., 1996, Sol. Phys., 168, 1

1051: 

1052: \bibitem[Chaplin et al(1999)]{chaplin99} Chaplin~W.~J., Elsworth~Y.,

1053: Isaak~G.~R., Miller~B.~A., New~R., 1999, \mnras, 308, 424

1054: 

1055: \bibitem[Chaplin et al.(2005)]{chaplin05} Chaplin~W.~J., Elsworth~Y.,

1056: Miller~B.~A., New~R., Verner~G.~A., 2005, \apj, 635, L105

1057: 

1058: \bibitem[Christensen-Dalsgaard \& Berthomieu(1991)]{jcd91}

1059: Christensen-Dalsgaard, J., Berthomieu, G. 1991, in Solar Interior and

1060: Atmosphere, eds. A.N. Cox, W.C.Livingston, M.S. Matthews, University

1061: of Arizna Press, Tuscon, p401

1062: 

1063: \bibitem[Christensen-Dalsgaard et al.(1996)]{jcd96}

1064: Christensen-Dalsgaared, J., D\"appen, W., Ajukov, S.V., et al.  1996,

1065: Science, 272, 1286

1066: 

1067: \bibitem[Delahaye \& Pinsonneault(2006)]{del06} Delahaye, F., \&

1068: Pinsonneault M. H. 2006, ApJ, 649, 529

1069: 

1070: \bibitem[Elliott \& Kosovichev(1998)]{el98} Elliott, J.R., Kosovichev,

1071: A.G. 1998, ApJ, 500, L199

1072: 

1073: \bibitem[Ferguson et al.(2005)]{fer05}Ferguson, J.~W., Alexander,

1074:   D.~R., Allard, F., Barman, T., Bodnarik, J.~G., Hauschildt, P.~H.,

1075:   Heffner-Wong, A., \& Tamanai, A. 2005, ApJ, 623, 585

1076: 

1077: \bibitem[Grevesse \& Sauval(1998)]{gre98} Grevesse, N., \& Sauval,

1078: A. J.  1998, in Solar composition and its evolution --- from core to

1079: corona, eds., C. Fr\"ohlich, M. C. E. Huber, S. K. Solanki, \& R. von

1080: Steiger, Kluwer, Dordrecht, p. 161

1081: 

1082: \bibitem[Iglesias \& Rogers(1996)]{opal96} Iglesias, C. A., \& Rogers,

1083: F. J. 1996, \apj, {464}, 943

1084: 

1085: \bibitem[Morel et al.(1999)]{mo99} Morel, P., Pinchon, B., Provost,

1086: J., Berthomiue, G.  1999, A\&A, 350, 275

1087: 

1088: \bibitem[Ot\'i Froranes et al.(2005)]{oti05} Ot\'i Floranes~H.,

1089: Christensen-Dalsgaard~J., Thompson~M.~J., 2005, \mnras, 356, 671

1090: 

1091: \bibitem[Palacios et al. (2006)]{pal06} Palacios, A., Talon, S.,

1092: Turck-Chieze, S., \& Charbonnel, C.  2006, to appear in Proceedings of

1093: the SOHO 18 / GONG 2006 / Workshop (ESA SP-624). ``Beyond the

1094: Spherical Sun: a new era for helio- and asteroseismology''.  7-11

1095: August, 2006.  Sheffield, UK.  Editor: K.Fletcher. astro-ph/0609381.

1096: 

1097: \bibitem[Rogers(2001)]{ro01} Rogers, F. J. 2001, Contrib. Plasma

1098: Phys., 41, 179

1099: 

1100: \bibitem[1996]{roge96} Rogers,~F.~J., Swenson,~F.~J., \&

1101: Iglesias,~C.~A. 1996, ApJ, 456, 902

1102: 

1103: \bibitem[Rogers \& Nayfonov(2002)]{opal02} Rogers, F. J., \& Nayfonov,

1104: A.  2002, \apj, 576, 1064

1105: 

1106: \bibitem[Roxburgh \& Vorontsov(2003)]{rox03} Roxburgh~I.~W.,

1107: Vorontsov~S.~V., 2003, A\&A, 411, 215

1108: 

1109: \bibitem[Roxburgh(2005)]{rox05} Roxburgh~I.~W., 2005, A\&A, 434, 665

1110: 

1111: \bibitem[Seaton (2006)]{seaton06} Seaton, M. J., 2005, MNRAS, 362,

1112: L1

1113: 

1114: \bibitem[Seaton \& Badnell (2004)]{seaton04} Seaton, M. J., Badnell, N. R.,

1115: 2004, MNRAS, 354, 457

1116: 

1117: \end{thebibliography}

1118: 

1119: \newpage

1120: 

1121: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

1122: 

1123: \begin{deluxetable}{lllll}

1124: \tablecolumns{5} \tablewidth{0pc} \tablecaption{Solar-cycle corrected

1125: BiSON frequencies (in $\rm \mu Hz$) used in the paper} \tablehead{

1126: \colhead{$n$}& \colhead{$l=0$}& \colhead{$l=1$}& \colhead{$l=2$}&

1127: \colhead{$l=3$}} \startdata

1128:  8&                     & $1329.629 \pm 0.004$& $1394.682 \pm 0.011$& $1450.986 \pm 0.038$\\

1129:  9& $1407.481 \pm 0.006$& $1472.841 \pm 0.005$& $1535.861 \pm 0.008$& $1591.575 \pm 0.014$\\

1130: 10& $1548.343 \pm 0.008$& $1612.717 \pm 0.006$& $1674.527 \pm 0.008$& $1729.109 \pm 0.017$\\

1131: 11& $1686.588 \pm 0.012$& $1749.296 \pm 0.008$& $1810.323 \pm 0.012$& $1865.281 \pm 0.019$\\

1132: 12& $1822.212 \pm 0.014$& $1885.089 \pm 0.010$& $1945.804 \pm 0.017$& $2001.218 \pm 0.022$\\

1133: 13& $1957.432 \pm 0.015$& $2020.798 \pm 0.014$& $2082.105 \pm 0.018$& $2137.781 \pm 0.025$\\

1134: 14& $2093.496 \pm 0.016$& $2156.781 \pm 0.017$& $2217.678 \pm 0.022$& $2273.521 \pm 0.031$\\

1135: 15& $2228.768 \pm 0.018$& $2291.980 \pm 0.014$& $2352.196 \pm 0.031$& $2407.660 \pm 0.034$\\

1136: 16& $2362.767 \pm 0.025$& $2425.587 \pm 0.019$& $2485.856 \pm 0.026$& $2541.677 \pm 0.032$\\

1137: 17& $2496.140 \pm 0.022$& $2559.196 \pm 0.022$& $2619.670 \pm 0.025$& $2676.191 \pm 0.031$\\

1138: 18& $2629.656 \pm 0.021$& $2693.332 \pm 0.021$& $2754.454 \pm 0.024$& $2811.352 \pm 0.029$\\

1139: 19& $2764.128 \pm 0.021$& $2828.097 \pm 0.021$& $2889.545 \pm 0.024$& $2946.981 \pm 0.029$\\

1140: 20& $2899.010 \pm 0.019$& $2963.306 \pm 0.020$& $3024.687 \pm 0.024$& $3082.319 \pm 0.035$\\

1141: 21& $3033.736 \pm 0.021$& $3098.127 \pm 0.022$& $3159.800 \pm 0.028$& $3217.712 \pm 0.040$\\

1142: 22& $3168.612 \pm 0.025$& $3233.147 \pm 0.026$& $3295.087 \pm 0.035$& $3353.387 \pm 0.054$\\

1143: 23& $3303.537 \pm 0.030$& $3368.495 \pm 0.031$& $3430.807 \pm 0.047$& $3489.430 \pm 0.071$\\

1144: 24& $3439.006 \pm 0.042$& $3504.195 \pm 0.040$& $3567.005 \pm 0.057$& $3626.022 \pm 0.101$\\

1145: 25& $3574.896 \pm 0.055$& $3640.347 \pm 0.052$&                     &                     \\

1146: 26& $3710.942 \pm 0.089$&                     &                     &                     \\

1147:   &                     &                     &                     &                     \\            

1148: \hline \enddata

1149: \label{tab:bison}

1150: \end{deluxetable}

1151: 

1152: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

1153: 

1154: \begin{deluxetable}{lllll}

1155: 

1156: \tablecolumns{5} \tablewidth{0pc} \tablecaption{Model frequencies (in

1157: $\rm \mu Hz$) for exact GS98 abundance} \tablehead{ \colhead{$n$}&

1158: \colhead{$l=0$}& \colhead{$l=1$}& \colhead{$l=2$}& \colhead{$l=3$}}

1159: \startdata

1160:  8&         & 1329.356& 1394.304& 1450.684\\

1161:  9& 1407.157& 1472.596& 1535.579& 1591.042\\

1162: 10& 1548.061& 1612.210& 1674.187& 1728.762\\

1163: 11& 1686.350& 1748.958& 1809.846& 1864.909\\

1164: 12& 1821.823& 1884.864& 1945.720& 2001.096\\

1165: 13& 1957.418& 2020.779& 2082.327& 2138.317\\

1166: 14& 2093.866& 2157.501& 2218.607& 2274.732\\

1167: 15& 2229.760& 2293.493& 2354.395& 2410.169\\

1168: 16& 2365.099& 2428.289& 2489.179& 2545.531\\

1169: 17& 2499.604& 2563.249& 2624.244& 2681.130\\

1170: 18& 2634.334& 2698.557& 2760.298& 2817.711\\

1171: 19& 2770.118& 2834.665& 2896.607& 2954.613\\

1172: 20& 2906.128& 2971.186& 3033.263& 3091.349\\

1173: 21& 3042.414& 3107.438& 3169.828& 3228.318\\

1174: 22& 3178.697& 3243.930& 3306.347& 3365.269\\

1175: 23& 3314.908& 3380.526& 3443.372& 3502.561\\

1176: 24& 3451.650& 3517.369& 3580.551& 3640.271\\

1177: 25& 3588.594& 3654.679&         &         \\

1178: 26& 3725.807&         &         &         \\

1179:   &         &         &         &         \\

1180: \hline 

1181: \enddata

1182: \label{tab:gs98}

1183: \end{deluxetable}

1184: 

1185: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

1186: 

1187: \begin{deluxetable}{lllll}

1188: \tablecolumns{5} \tablewidth{0pc} \tablecaption{Model frequencies (in

1189: $\rm \mu Hz$) for exact AGS05 abundance} \tablehead{ \colhead{$n$}&

1190: \colhead{$l=0$}& \colhead{$l=1$}& \colhead{$l=2$}& \colhead{$l=3$}}

1191: \startdata

1192:  8&         & 1328.009& 1392.406& 1448.604\\

1193:  9& 1405.414& 1470.904& 1533.686& 1588.631\\

1194: 10& 1546.296& 1609.957& 1671.676& 1726.210\\

1195: 11& 1684.076& 1746.706& 1807.136& 1861.847\\

1196: 12& 1819.239& 1882.140& 1943.031& 1998.140\\

1197: 13& 1954.946& 2018.007& 2079.174& 2135.089\\

1198: 14& 2090.926& 2154.618& 2215.457& 2271.133\\

1199: 15& 2226.755& 2290.126& 2350.930& 2406.628\\

1200: 16& 2361.873& 2425.004& 2485.506& 2541.695\\

1201: 17& 2496.103& 2559.770& 2620.763& 2677.364\\

1202: 18& 2631.043& 2695.017& 2756.563& 2813.952\\

1203: 19& 2766.599& 2831.216& 2892.902& 2950.568\\

1204: 20& 2902.576& 2967.404& 3029.432& 3087.386\\

1205: 21& 3038.792& 3103.708& 3165.737& 3224.098\\

1206: 22& 3174.772& 3240.032& 3302.407& 3360.995\\

1207: 23& 3311.132& 3376.466& 3439.208& 3498.371\\

1208: 24& 3447.692& 3513.433& 3576.329& 3635.823\\

1209: 25& 3584.526& 3650.497&         &         \\

1210: 26& 3721.747&         &         &         \\

1211:   &         &         &         &         \\

1212: \hline 

1213: \enddata

1214: \label{tab:ags05}

1215: \end{deluxetable}

1216: 

1217: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

1218: 

1219: \begin{deluxetable}{lcccccc}

1220: \tablecolumns{7} \tablewidth{0pc} \tablecaption{$\mu_{\rm c}$

1221: results for `linear test' models} \tablehead{ \colhead{Set}&

1222: \colhead{$N$}& \colhead{$\sigma_{\langle \Delta r_{l,l+2}

1223: \rangle}$}& \colhead{$\mu_{\rm c}$}& \colhead{$\sigma(\mu_{\rm

1224: c})$}& \colhead{$\sigma(\mu_{\rm c})/\mu_{\rm c}$}&

1225: \colhead{$\chi^2_{N-2}$}\\

1226: \colhead{}& \colhead{}& \colhead{(\%)}& \colhead{}& \colhead{}&

1227: \colhead{(\%)}& \colhead{}} \startdata

1228:   GS-Lin $r_{02}$& 10& 0.136& 0.7253& $\pm 0.0012$& 0.17& 9.9\\

1229:  & & & & & \\

1230:   GS-Lin $r_{13}$& 10& 0.099& 0.7244& $\pm 0.0010$& 0.14& 7.4\\

1231:  & & & & & \\

1232:   GS-Lin comb.& & &0.7248& $\pm 0.0008$& 0.11& \\

1233:  & & & & & \\

1234:  & & & & & \\

1235:  AGS-Lin $r_{02}$& 10& 0.108& 0.7260& $\pm 0.0009$& 0.12& 6.1\\

1236:  & & & & & \\

1237:  AGS-Lin $r_{13}$& 10& 0.089& 0.7255& $\pm 0.0008$& 0.11& 6.1\\

1238:  & & & & & \\

1239:  AGS-Lin comb.& & &0.7258& $\pm 0.0006$& 0.08& \\

1240:  & & & & & \\

1241: \hline

1242: \enddata

1243: \label{tab:linmu}

1244: \end{deluxetable}

1245: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

1246: 

1247: \begin{deluxetable}{lcccccc}

1248: \tablecolumns{7} \tablewidth{0pc} \tablecaption{$Z$ results for

1249: `linear test' models} \tablehead{ \colhead{Set}& \colhead{$N$}&

1250: \colhead{$\sigma_{\langle \Delta r_{l,l+2} \rangle}$}&

1251: \colhead{$Z$}& \colhead{$\sigma(Z)$}& \colhead{$\sigma(Z)/Z$}&

1252: \colhead{$\chi^2_{N-2}$}\\

1253: \colhead{}& \colhead{}& \colhead{(\%)}& \colhead{}& \colhead{}&

1254: \colhead{(\%)}& \colhead{}} \startdata

1255:   GS-Lin $r_{02}$& 10& 0.057& 0.01798& $\pm 0.00013$& 0.7& 1.7\\

1256:  & & & & & \\

1257:   GS-Lin $r_{13}$& 10& 0.034& 0.01774& $\pm 0.00008$& 0.5& 0.9\\

1258:  & & & & & \\

1259:   GS-Lin comb.& & &0.01785& $\pm 0.00007$& 0.4& \\

1260:  & & & & & \\

1261:  & & & & & \\

1262:  AGS-Lin $r_{02}$& 10& 0.056& 0.01617& $\pm 0.00011$& 0.7& 1.7\\

1263:  & & & & & \\

1264:  AGS-Lin $r_{13}$& 10& 0.050& 0.01604& $\pm 0.00011$& 0.7& 1.9\\

1265:  & & & & & \\

1266:  AGS-Lin comb.& & &0.01611& $\pm 0.00008$& 0.5& \\

1267:  & & & & & \\

1268: \hline

1269: \enddata

1270: \label{tab:lin}

1271: \end{deluxetable}

1272: 

1273: 

1274: 

1275: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

1276: 

1277: \begin{deluxetable}{lcccc}

1278: \tablecolumns{6} \tablewidth{0pc}  \tablecaption{Effective standard deviations

1279:   due to uncertainties in the radiative opacities and equation of state}

1280:    \tablehead{  \colhead{Quantity}  &  \colhead{$\Delta

1281:     r_{02}$} & \colhead{$\Delta r_{13}$} & \colhead{$Z$} & 

1282: \colhead{$\mu_{\rm c}$} \\ 

1283: \colhead{} & \colhead{(\%)} & \colhead{(\%)} & \colhead{(\%)} & \colhead{(\%)} } 

1284: \startdata

1285: $\sigma_{\rm opac}$ & 0.16 & 0.064 & 0.083& 0.077\\ 

1286: $\sigma_{\rm EOS}$ & 0.53 & 0.28 & 0.047& 0.031\\

1287: \enddata

1288: \label{tab:opac_eos}

1289: \end{deluxetable}

1290: 

1291: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

1292: 

1293: 

1294: \begin{deluxetable}{lccccc}

1295: \tablecolumns{6} \tablewidth{0pc} \tablecaption{$\mu_{\rm c}$

1296: results for `Monte Carlo' models} \tablehead{ \colhead{Set}&

1297: \colhead{$N$}& \colhead{$\sigma_{\langle \Delta r_{l,l+2}

1298: \rangle}$}& \colhead{$\mu_{\rm c}$}& \colhead{$\sigma(\mu_{\rm

1299: c})$}&

1300: \colhead{$\sigma(\mu_{\rm c})/\mu_{\rm c}$}\\

1301: \colhead{}& \colhead{}& \colhead{(\%)}& \colhead{}& \colhead{}&

1302: \colhead{(\%)}} \startdata

1303:   GS-Cons $r_{02}$& 5000& 0.82& 0.7234& $\pm 0.0051$& 0.7\\

1304:  & & & & & \\

1305:   GS-Cons $r_{13}$& 5000& 0.56& 0.7229& $\pm 0.0033$& 0.6\\

1306:  & & & & & \\

1307:   GS-Cons comb.& & & 0.7231& $\pm 0.0030$& 0.4\\

1308:  & & & & & \\

1309:    GS-Opt $r_{02}$& 1000& 0.65& 0.7233& $\pm 0.0049$& 0.7\\

1310:  & & & & & \\

1311:    GS-Opt $r_{13}$& 1000& 0.38& 0.7228& $\pm 0.0030$& 0.4\\

1312:  & & & & & \\

1313:    GS-Opt comb.& & & 0.7230& $\pm 0.0026$& 0.4\\

1314:  & & & & & \\

1315:  AGS-Cons $r_{02}$& 1000& 0.76& 0.7214& $\pm 0.0046$& 0.6\\

1316:  & & & & & \\

1317:  AGS-Cons $r_{13}$& 1000& 0.52& 0.7206& $\pm 0.0034$& 0.5\\

1318:  & & & & & \\

1319:  AGS-Cons comb.& & & 0.7209& $\pm 0.0027$& 0.37\\

1320:  & & & & & \\

1321:  AGS-Opt $r_{02}$& 5000& 0.63& 0.7239& $\pm 0.0047$& 0.6\\

1322:  & & & & & \\

1323:  AGS-Opt $r_{13}$& 5000& 0.37& 0.7225& $\pm 0.0029$& 0.4\\

1324:  & & & & & \\

1325:  AGS-Opt comb.& & & 0.7229& $\pm 0.0025$& 0.3\\

1326: \enddata

1327: \label{tab:mcmu}

1328: \end{deluxetable}

1329: 

1330: 

1331: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

1332: 

1333: 

1334: \begin{deluxetable}{lccccc}

1335: \tablecolumns{6} \tablewidth{0pc} \tablecaption{$Z$ results for

1336: `Monte Carlo' models} \tablehead{ \colhead{Set}& \colhead{$N$}&

1337: \colhead{$\sigma_{\langle \Delta r_{l,l+2} \rangle}$}&

1338: \colhead{$Z$}& \colhead{$\sigma(Z)$}&

1339: \colhead{$\sigma(Z)/Z$}\\

1340: \colhead{}& \colhead{}& \colhead{(\%)}& \colhead{}& \colhead{}&

1341: \colhead{(\%)}} \startdata

1342:   GS-Cons $r_{02}$& 5000& 1.03& 0.0189& $^{+0.0059}_{-0.0045}$& 28\\

1343:  & & & & & \\

1344:   GS-Cons $r_{13}$& 5000& 0.83& 0.0186& $^{+0.0052}_{-0.0040}$& 25\\

1345:  & & & & & \\

1346:   GS-Cons comb.& & & 0.0187& $^{+0.0039}_{-0.0030}$& 19 \\

1347:  & & & & & \\

1348:    GS-Opt $r_{02}$& 1000& 0.72& 0.0197& $^{+0.0057}_{-0.0045}$& 26\\

1349:  & & & & & \\

1350:    GS-Opt $r_{13}$& 1000& 0.52& 0.0192& $^{+0.0043}_{-0.0035}$& 21\\

1351:  & & & & & \\

1352:    GS-Opt comb.& & & 0.0194& $^{+0.0035}_{-0.0028}$& 16 \\

1353:  & & & & & \\

1354:  AGS-Cons $r_{02}$& 1000& 0.90& 0.0233& $^{+0.0072}_{-0.0055}$& 27\\

1355:  & & & & & \\

1356:  AGS-Cons $r_{13}$& 1000& 0.75& 0.0226& $^{+0.0062}_{-0.0049}$& 25\\

1357:  & & & & & \\

1358:  AGS-Cons comb.& & & 0.0229& $^{+0.0047}_{-0.0037}$& 18 \\

1359:  & & & & & \\

1360:  AGS-Opt $r_{02}$& 5000& 0.72 & 0.0220& $^{+0.0047}_{-0.0039}$& 20\\

1361:  & & & & & \\

1362:  AGS-Opt $r_{13}$& 5000& 0.50& 0.0219 & $^{+0.0038}_{-0.0032}$& 16\\

1363:  & & & & & \\

1364:  AGS-Opt comb.& & & 0.0219& $^{+0.0029}_{-0.0025}$& 12\\

1365: \enddata

1366: \label{tab:mc}

1367: \end{deluxetable}

1368: 

1369: 

1370: 

1371: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

1372: 

1373:  \begin{figure*}

1374:  \epsscale{1.0}

1375:  \plottwo{f1a.ps}{f1b.ps}\\

1376:  \plottwo{f1c.ps}{f1d.ps}

1377: 

1378: \caption{Upper two panels: weighted mean differences, $\langle \Delta

1379: r_{02} \rangle$ (upper left-hand panel) and $\langle \Delta r_{13}

1380: \rangle$ (upper right-hand panel), versus $\ln\mu_{\rm c}$, plotted

1381: for a sequence of 10 solar models. All models have the same relative

1382: mixture of heavy elements, corresponding the Grevesse \& Sauval (1998)

1383: mixture but differ in the total amount of metals.  The solid line in

1384: each panel is the best-fitting straight line. The dotted lines

1385: intersect at the location along each best-fitting line where the

1386: weighted mean difference is zero. Lower panels: as per upper panels,

1387: but for a sequence of 10 solar models having a relative mixture of

1388: heavy elements corresponding to the Asplund et al. (2005) mixture.}

1389: 

1390:  \label{fig:linmu}

1391:  \end{figure*}

1392: 

1393: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

1394: 

1395: 

1396: 

1397:  \begin{figure*}

1398:  \epsscale{1.0}

1399:  \plottwo{f2a.ps}{f2b.ps}\\

1400:  \plottwo{f2c.ps}{f2d.ps}

1401: 

1402:  \caption{Upper two panels: weighted mean differences, $\langle \Delta

1403:    r_{02} \rangle$ (upper left-hand panel) and $\langle \Delta r_{13}

1404:    \rangle$ (upper right-hand panel), versus $\ln Z$, plotted for a

1405:    sequence of 10 solar models. All models have the same relative

1406:    mixture of heavy elements, corresponding the Grevesse \& Sauval

1407:    (1998) mixture but differ in the total amount of metals. The solid

1408:    line in each panel is the best-fitting straight line. The dotted

1409:    lines intersect at the location along each best-fitting line where

1410:    the weighted mean difference is zero. Lower panels: as per upper

1411:    panels, but for a sequence of 10 solar models having a relative

1412:    mixture of heavy elements corresponding to the Asplund et

1413:    al. (2005) mixture.}

1414: 

1415:  \label{fig:lin}

1416:  \end{figure*}

1417: 

1418: 

1419: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

1420: \begin{figure}

1421: \plotone{f3.ps}

1422: \caption{The relation between $Z$ and $\mu_{\rm c}$ (the average mean

1423: molecular weight in the inner 20\% by radius) for models constucted

1424: with the GS98 relative heavy element abundances and the AGS05 relative

1425: heavy element abundances.}

1426: \label{fig:zmu}

1427: \end{figure}

1428: 

1429: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

1430: 

1431:  \begin{figure*}

1432:  \epsscale{1.0}

1433:  \plottwo{f4a_1.ps}{f4b_1.ps}\\

1434:  \plottwo{f4c.ps}{f4d.ps}

1435: 

1436:  \caption{Upper two panels: weighted mean differences, $\langle \Delta

1437: r_{02} \rangle$ (upper left-hand panel) and $\langle \Delta r_{13}

1438: \rangle$ (upper right-hand panel), versus $\ln\mu_{\rm c}$, plotted

1439: for the 5000 Monte Carlo `GS-Cons' solar models. These models were

1440: made with the `conservative' (large) abundance uncertainties, centered

1441: on the GS98 mixture. The solid line in each panel is the best-fitting

1442: straight line. The dotted lines intersect at the location along each

1443: best-fitting line where the weighted mean difference is zero. Lower

1444: panels: As per the upper panels, but for the 1000 Monte Carlo `GS-Opt'

1445: solar models. These models were made with the `optimistic' (small)

1446: abundance uncertainties, centered on the GS98 mixture.}

1447: 

1448:  \label{fig:GSmuMC}

1449:  \end{figure*}

1450: 

1451: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

1452: 

1453:  \begin{figure*}

1454:  \epsscale{1.0}

1455:  \plottwo{f5a.ps}{f5b.ps}\\

1456:  \plottwo{f5c_1.ps}{f5d_1.ps}

1457: 

1458:  \caption{Upper two panels: weighted mean differences, $\langle \Delta

1459: r_{02} \rangle$ (upper left-hand panel) and $\langle \Delta r_{13}

1460: \rangle$ (upper right-hand panel), versus $\ln\mu_{\rm c}$, plotted

1461: for the 1000 Monte Carlo `AGS-Cons' solar models. These models were

1462: made with the `conservative' (large) abundance uncertainties, centered

1463: on the AGS05 mixture. The solid line in each panel is the best-fitting

1464: straight line. The dotted lines intersect at the location along each

1465: best-fitting line where the weighted mean difference is zero. Lower

1466: panels: As per the upper panels, but for the 5000 Monte Carlo

1467: `AGS-Opt' solar models. These models were made with the `optimistic'

1468: (small) abundance uncertainties, centered on the AGS05

1469: mixture. Although visually the solid lines on the lower plots do not

1470: look like the best-fitting lines, they are the unbiased lines given by

1471: a least-squares fit.}

1472: 

1473:  \label{fig:AGSmuMC}

1474:  \end{figure*}

1475: 

1476: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

1477: 

1478:  \begin{figure*}

1479:  \epsscale{1.0}

1480:  \plottwo{f6a_1.ps}{f6b_1.ps}\\

1481:  \plottwo{f6c.ps}{f6d.ps}

1482: 

1483:  \caption{Upper two panels: weighted mean differences, $\langle \Delta

1484: r_{02} \rangle$ (upper left-hand panel) and $\langle \Delta r_{13}

1485: \rangle$ (upper right-hand panel), versus $\ln Z$, plotted for the

1486: 5000 Monte Carlo `GS-Cons' solar models. These models were made with

1487: the `conservative' (large) abundance uncertainties, centered on the

1488: GS98 mixture. The solid line in each panel is the best-fitting

1489: straight line. The dotted lines intersect at the location along each

1490: best-fitting line where the weighted mean difference is zero. Lower

1491: panels: As per the upper panels, but for the 1000 Monte Carlo `GS-Opt'

1492: solar models. These models were made with the `optimistic' (small)

1493: abundance uncertainties, centered on the GS98 mixture.}

1494: 

1495:  \label{fig:GSMC}

1496:  \end{figure*}

1497: 

1498: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

1499: 

1500:  \begin{figure*}

1501:  \epsscale{1.0}

1502:  \plottwo{f7a.ps}{f7b.ps}\\

1503:  \plottwo{f7c_1.ps}{f7d_1.ps}

1504: 

1505:  \caption{Upper two panels: weighted mean differences, $\langle \Delta

1506: r_{02} \rangle$ (upper left-hand panel) and $\langle \Delta r_{13}

1507: \rangle$ (lower right-hand panel), versus $\ln Z$, plotted for the

1508: 1000 Monte Carlo `AGS-Cons' solar models. These models were made with

1509: the `conservative' (large) abundance uncertainties, centered on the

1510: AGS05 mixture. The solid line in each panel is the best-fitting

1511: straight line. The dotted lines intersect at the location along each

1512: best-fitting line where the weighted mean difference is zero. Lower

1513: panels: As per the upper panels, but for the 5000 Monte Carlo

1514: `AGS-Opt' solar models. These models were made with the `optimistic'

1515: (small) abundance uncertainties, centered on the AGS05

1516: mixture. Although visually the solid lines on the lower plots do not

1517: look like the best-fitting lines, they are the unbiased lines given by

1518: a least-squares fit.}

1519: 

1520:  \label{fig:AGSMC}

1521:  \end{figure*}

1522: 

1523: 

1524: 

1525: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

1526: 

1527: \end{document}

1528: