B
Chemical Composition of the Solar System

By Section:

1. Photospheric Abundances
2. Solar Flare Ion Abundances
3. Coronal Emission Lines
4. Carbonaceous Chondrite Abundances

The principal sources of solar system abundances are:

1. photospheric absorption lines
2. solar flare ions
3. coronal emission lines
4. analyses of carbonaceous chondritic meteorites.

1. Photospheric abundances
The solar photosphere is generally believed to have the composition of average solar system material. This belief is justified because matter is not convectively exchanged between the surface and the regions of nuclear energy generation in the deep interior. This stratification is theoretically well­established, and, if not present, measurable quantities of nuclei with low nuclear binding energy (Li,Be,B) would not be expected in the photosphere. The actual quantitative abundances of these nuclei have probably been modified by thermonuclear reactions, either during an early, totally­convective, stage of the Sun or slowly over the age of the solar system at the base of the present surface convection zone. The destruction of solar D, well documented by Apollo data (Epstein and Taylor, 1973), has also occurred by these processes. Nevertheless, independent of details, the fact that any LiBeB survive makes it very unlikely that any other elements in the photosphere have suffered thermonuclear modification of their elemental or isotopic abundances.

The identification of photospheric composition with that of the average solar system assumes that the accretion of the Sun occurred without any fractionation of solids and gases. This is an issue of major importance, but testable with Genesis data ( Document C).

Despite being comparatively free from interpretational ambiguities, photospheric abundances have in the past been relatively uncertain because: (a) detailed knowledge of the physical state of the solar atmosphere is required to relate the energy removed in an absorption line to the photospheric abundance (e.g. Unsold and Baschek, 1983), (b) the atomic constants (transition probabilities) required for an abundance analysis have not been known to sufficient accuracy and (c) lines of minor and trace elements are insufficiently resolved in the forest of major element lines. Steady progress has been made over the past 20 years on all of these traditional problems, and the quoted errors in the most recent compilations (e.g. Anders and Grevesse, 1989; Grevesse, 1992) are relatively small. It has been clear, at least since Anders (1971), that average solar system abundances are known to an accuracy much better than a factor of 2. At present a realistic goal should be an accuracy of ±10% (standard deviation), relative to either Si or H (proposal Section 2.A.2).

Meteoritic abundance data are being interpreted at this level (see e.g. review articles in Kerridge and Matthews, 1988), and it would be important to have average solar system abundances, independent of meteorite data, to the same level of accuracy. A complete set of average solar system elemental abundances consists of 82 entries, and, even taking the quoted errors from Anders and Grevesse at face value, 18/82 photospheric abundances are known to the ±10% standard. It is important to know the other 64/82 accurately; moreover, it may be premature to place full confidence in the photospheric error estimates from Anders and Grevesse. For example, Anders and Grevesse reported a 40% difference between the meteoritic and photospheric Fe abundance which was far outside quoted error limits. This difference stimulated extensive re-evaluations of the photospheric Fe abundance with the latest studies (e.g. Holweger et al., 1991 and Biemont et al., 1991) revising the photospheric abundance downward by an amount much larger than the quoted errors by Anders and Grevesse and back in agreement with the meteoritic value. It is clear that alternative sources of solar abundances than the photosphere are still required.

2. Solar Flare Ion Abundances
Precise data are available for many elements for solar flare ions (e.g. Breneman and Stone, 1985; Garrard and Stone, 1993). Solar flare ions are much higher energy (MeV relative to keV for the solar wind) and are presumably emitted by different mechanisms. A serious complication is that relative abundances vary with ion energy for many flares. Cook et al. (1984) show that, even when flares are selected which have energy­independent compositions, there are systematic flare­to­flare variations. Furthermore, the compositions of energy­independent flares all show systematic deviations from photospheric abundances, and these deviations appear to correlate with first ionization potential (FIP) or more likely, first ionization time (Geiss, 1985). In addition there are fractionations which correlate with atomic charge/mass and are responsible for the flare­to­flare variations. To obtain a set of "coronal" abundances, empirical corrections for the charge/mass fractionation were applied by Breneman and Stone based on normalizing the average solar flare ion abundances for elements with first ionization potential < 9 eV to the corresponding photospheric values. To obtain estimates of average solar system abundances primarily from solar flare ion data Breneman and Stone applied a second, larger, correction (constant factor) to the solar flare ion data for elements with high first ionization potential (FIP) based on the photospheric O abundance relative to the low first ionization potential elements. Breneman and Stone proposed that the previously accepted photospheric C abundance was wrong, and this revision has subsequently been accepted (e.g. Anders and Grevesse). These corrections are plausible; nevertheless, there it no good way to test the assumptions made in the Breneman­Stone analysis, which, moreover, are not independent of photospheric data. It would be of considerable interest to have an independent solar wind data set for comparison. Moreover, in the data (for large flares) used by Breneman and Stone or Garrard and Stone (1993) the charge/mass corrections were relatively small, but in a wider sampling of flares having a wide range in intensity, Reames et al. (1990) find very large compositional variations (e.g. factor of 50 in Fe/C), which appear to be primarily dominated by the charge/mass type of fractionations. Given required corrections of this magnitude, it appears difficult to invert solar energetic particle data to obtain accurate solar abundances.

It should be emphasized that the complications of solar flare abundances do not automatically apply to the solar wind as well. A solar wind analog of the solar flare charge/mass fractionations has recently been proposed based on Ulysses data [Geiss et al., 1994], and there is a well-established FIP fractionation analogous to that seen by Breneman and Stone (Document G.2). Further, there is a well­documented and intriguing difference between some ²²Ne/²⁰Ne ratios in solar flares and the solar wind ( Document A) Although this difference is not understood, it clearly demonstrates that the same fractionation mechanisms do not apply to solar flares and to the solar wind.

3. Coronal Emission Lines
Beautiful spectra of coronal emission lines of elements between O and Ca have been obtained by high energy photon (EUV) detectors flown on spacecraft (e.g. Veck and Parkinson, 1981). Due to the high temperature of the corona, many of these lines are from one­ or two­electron atoms, and the required transition probabilities can be calculated with some confidence. Nevertheless, different measurements give different abundances with variations typically by a factor of 2­3. The source of the variability is most likely the inability to model accurately the physical heterogeneities (temperature, electron density) in the coronal regions from which the observed EUV photons originate.

4. Carbonaceous Chondrite Abundances
The most precise solar system abundances come from carbonaceous chondritic meteorites, although these abundances are obviously restricted to "non­volatile" elements. Empirically, "volatile" elements appear to be H, CNO, and the noble gases. A priori, there is no justification for the composition of any meteorite to be identified with average solar system composition. The use of abundances based on the CI carbonaceous chondrite subtype is based on two simple (but strong) arguments (e.g. Anders, 1971, Anders and Ebihara, 1982, Anders and Grevesse, 1989, Burnett et al., 1989): (a) The CI relative abundances (normalized to Si) agree with photospheric absorption line abundances to within the quoted errors in the latter in almost all cases, and (b) For the heavier elements (above iron) the CI elemental abundances (when broken down into individual isotopic abundances) are smooth functions of mass number for odd mass nuclei.

(a) It has been very striking that, as the accuracy of both data sets has improved in the past 30 years, the agreement has steadily converged. From Table 2 in Anders and Grevesse, for 73 elements for which CI­photosphere comparisons can be made, only 11 cases disagree by more than twice the quoted photospheric error. Precise (±10% or better) photospheric abundances are quoted for 15/73 elements and 9 of these 15 agree with the CI abundances to better than 10%. Although suspicions about photospheric error estimates seem justified (see Section 1, above), the agreement must be respected.

(b) Because of nuclear pairing forces, nuclei with even mass numbers are more stable than nuclei with odd mass numbers and are observed to be systematically more abundant; consequently abundance plots must be made for even or odd mass number nuclei separately. The odd mass curves are smoother and are usually used as a basis for discussion. This is because there are smooth trends in nuclear properties for odd­mass nuclei. The argument here, made originally by Suess and emphasized by Anders for many years, is that, if the true solar system abundances were not smooth functions of mass number, then it is very unreasonable that a process which fractionated CI chondrites from average solar system composition during their formation would produce a smooth abundance curve. Elemental abundances are less smooth below mass 60, presumably because nuclear binding energies are less regular as a function of mass. Consequently, the smoothness argument holds best for heavier elements. The most recent compilation of CI heavy element abundances by Anders and Grevesse is relatively smooth. Burnett et al., (1989) conclude that the smoothness argument definitely permits variations in CI and average solar system abundances of 10­30%, although it appears very unlikely that any CI heavy element abundance differs from the average solar system value by more than a factor of 2. For some astrophysical applications this is sufficiently accurate; however, for cosmochemistry and planetary science applications even the 10­30% variations are not acceptable, as discussed above. For example the factor of 1.3­1.5 difference in Fe/Si between CI and photospheric abundances quoted by Anders and Grevesse will cause a tremendous difference in two planets (e.g. Anderson, 1988).

Figure B1. Comparison of photospheric and CI meteorite Ge and Zn abundances. PIXE data: Burnett et al. (1989); Ebihara et al. (1982), Kallemeyn & Wasson (1981), Anders & Grevesse (1989).

In cosmochemistry even small, e.g. 10%, variations in composition of a given meteorite from an accurate set of solar system abundances could result in significant conclusions about conditions in the early solar system (compare, e.g. Larimer and Wasson, 1988, Burnett et al., 1989). The interpretation of such small differences is somewhat compromised at present. For example Figure B1 shows that there are variations among various samples of CI chondrites for Ge and Zn. Which sample gives the average solar system composition? This question cannot be answered; some data selection is required to produce a CI abundance for compilation, as is apparent from the location of the Anders and Grevesse point. Even though absolute concentrations are plotted and individual CI specimens can vary in volatile (e.g. water) content, the spread in the data cannot be explained by volatile variations. The location of photospheric abundances would be an important starting point in the interpretation of such CI intersample variations, but the uncertainties are too large. For example, interesting Se-Zn variations also exist (Burnett et al., 1989), but Se cannot be measured in the photosphere. The photospheric Zn abundance is in good accord with most of the analyses but the photospheric Ge abundance is significantly lower than all CI analyses.

Some elements show even larger intersample variations in CI chondrites. For example Rocholl and Jochum (1993) report almost a factor of 5 variation in U contents among ~0.1 g CI samples. Thus, for the CI concentration of U to give the solar system abundance one must depend on a very strong convergence of the mean of the various analyses. There is no a priori argument to support this convergence.

Finally, meteoriticists worry about the fact that CI chondrites have experienced a high degree of aqueous alteration (e.g. McSween, 1993). One might have expected that the material which provides the average chemical composition of the solar system would be more pristine.

In summary, for clean interpretations of meteorite chemistry it is important to have an independent set of solar system abundances based entirely on solar data (Burnett et al., 1989). These considerations set the desired standards of accuracy for a solar wind sample return mission to be about 10%. for elemental abundances. Overall this discussion has established that there is a need for improved solar abundances. In Document H we show that solar wind abundances can play a major role in satisfying this need.