DETERMINATION OF THE FORMATION TEMPERATURE OF Si IV IN THE SOLAR TRANSITION REGION

THE ASTROPHYSICAL JOURNAL, 477 : L119 L122, 1997 March 10 1997. The American Astronomical Society. All rights reserved. Printed in U.S.A. DETERMINATION OF THE FORMATION TEMPERATURE OF Si IV IN THE SOLAR TRANSITION REGION G. A. DOSCHEK, 1 J. T. MARISKA, 1 H. P. WARREN, 1 K. WILHELM, 2 P. LEMAIRE, 3 T. KUCERA, 4 AND U. SCHÜHLE 2 Received 1996 October 8; accepted 1996 December 26 ABSTRACT Using spectra obtained with the Solar Ultraviolet Measurements of Emitted Radiation (SUMER) spectrometer flown on the Solar and Heliospheric Observatory spacecraft, we deduce the temperature of formation of the Si IV ion in the solar transition region from the Si IV ultraviolet spectral line intensity ratio, 3p 2 P 3 2 3d 2 D 3 2,5 2 3s 2 S 1 2 3p 2 P 1 2, and compare the result to the temperature predicted under the assumption of ionization equilibrium. The wavelengths are as follows: 2 D 3 2,5 2, 1128.325, 1128.340 Å; 2 P 1 2, 1402.770 Å. Ratios are derived for typical features of the quiet Sun, such as cell center and network, and are systematically higher than those predicted at the 6.3 10 4 K ionization equilibrium temperature of formation of Si IV. For most solar features the ratios imply a temperature of formation of about 8.5 10 4 K. The ratios for the faintest features imply a temperature of formation of up to 1.6 10 5 K. It is not clear, however, that all the discrepancies between the measured and theoretical ratios are due to a temperature effect. Accurate temperature measurements are important since a large discrepancy from ionization equilibrium has significant implications for the physics of the transition region, such as the possible presence of nonthermal electrons. Subject headings: Sun: transition region ultraviolet: stars 1. INTRODUCTION Most spectroscopic analyses of the solar transition region and corona invoke the assumption of ionization equilibrium to derive physical parameters, such as the emission measure distribution, as a function of electron temperature. Determining the validity of this assumption is crucial for a proper understanding of the physics of the solar atmosphere. For example, an anomalously high electron temperature compared with the ionization equilibrium temperature may imply the presence of high-energy nonthermal electrons, or the occurrence of dynamical processes on timescales that are short compared with the ionization and recombination timescales. Dufton, Kingston, & Keenan (1984) suggest that an enhanced tail on the Maxwellian velocity distribution may exist at temperatures near the temperature of formation of Si IV. Their argument is based on a temperature-sensitive Si III line ratio. Alternatively, a discrepancy between measured and predicted electron temperatures may simply reflect errors in the atomic parameters upon which the theoretical equilibrium temperatures are based, and these errors can propagate through to other derived parameters used to test physical models of the solar atmosphere, such as the emission measure distribution. The assumption of ionization equilibrium can be checked by measuring the temperature at which an ion is formed, using a temperature-sensitive spectral line intensity ratio. Unfortunately, there are few such ratios for which reliable measurements can be made within the ultraviolet (UV) and extreme-uv (EUV) spectral windows of most solar spectrometers flown on space missions up to the present time. The paucity of useful temperature diagnostics is partly a result of 1 E. O. Hulburt Center for Space Research, Code 7670, Naval Research Laboratory, 4555 Overlook Avenue SW, Washington, DC 20375. 2 Max-Planck-Institut für Aeronomie, Postfach 20, D-37189 Katlenburg- Lindau, Germany. 3 Institut d Astrophysique Spatiale, Unité Mixte CNRS Université Paris 11, Bâtiment 121, F-91405 Orsay, France. 4 Applied Research Corporation, Code 682.3, NASA Goddard Space Flight Center, Greenbelt, MD 20771. L119 the diagnostic technique. A line ratio for which both lines are produced by electron impact excitation (the primary excitation mechanism in the solar case for almost all UV and EUV emission lines) is temperature sensitive only if the difference in excitation energy of the two lines is comparable to the thermal energy of the exciting electrons. Spectroscopically, this condition usually results in a large wavelength separation between the two diagnostic lines; hence, we encounter instrumental problems. However, there are some exceptions to this general rule, and this paper is concerned with one of these exceptions. Temperature-sensitive line ratios in the sodium isoelectronic sequence have been discussed for Mg II by Feldman & Doschek (1977a) and for Al III by Doschek & Feldman (1987). This sequence was first suggested for temperature diagnostics by Flower & Nussbaumer (1975). The diagnostic ratios are the ratios of the 3p 3d transitions relative to the 3s 3p transitions. Because at solar atmospheric densities the 3d levels are excited from the 3s levels and not the 3p levels, the temperature sensitivity depends on the energy difference between the 3d and 3s levels relative to the energy difference between the 3s and 3p levels. However, because the 3d levels must decay to the 3p levels and not the 3s levels, the wavelength difference between diagnostic line pairs is not as great as implied by the energy differences involved. For Mg II, all the lines fall at about the same wavelength, and the wavelength separation increases with increasing atomic number. For Al III, Doschek & Feldman (1987) found a temperature close to, but slightly below, the ionization equilibrium temperature predicted by Arnaud & Rothenflug (1985). The line pairs for Al III fall near 1610 Å (3p 3d) and 1860 Å (3s 3p). The next ion is Si IV, for which the line pairs occur near 1130 Å(3p 3d) and 1400 Å (3s 3p). This is a desirable ion for a temperature measurement, since the predicted temperature of Si IV formation is close to the temperature of minimum emission measure in the solar atmosphere. Unfortunately, the shorter wavelength lines fall below H I Ly, where the reflectance of aluminum-coated mirrors begins to decrease rapidly. As a result, the calibration of the Naval Research Laboratory (NRL) S082-B slit spectrograph on the Skylab space station is

L120 DOSCHEK ET AL. Vol. 477 too poorly known for accurate intensity measurements of the Si IV 3p 3d lines. The same applies to the NRL high-resolution telescope spectrograph. In the case of the Harvard College Observatory spectroheliometer on Skylab, the spectral resolution was too low to separate the Si IV 3p 3d lines from neighboring lines of Fe III. It is now possible to measure both the 1130 Å and the 1400 Å Si IV line intensities accurately and with the required spectral resolution by using spectra obtained from the Solar Ultraviolet Measurements of Emitted Radiation (SUMER) spectrometer flown on the ESA NASA Solar and Heliospheric Observatory (SOHO) spacecraft. This instrument is described in detail by Wilhelm et al. (1995, 1996a). In this Letter, we discuss SUMER spectra of a quiet-sun region and determine Si IV temperatures for various solar features. 2. SPECTROSCOPY AND DATA ANALYSIS The precise wavelengths and transitions of the Si IV line pair used in this work are as follows: 3s 2 S1 2 3p 2 P 1 2, 1402.770 Å; 3p 2 P 3 2 3d 2 D 3 2,5 2, 1128.325, 1128.340 Å. The 2 S 1 2 2 P 3 2 line at 1393.76 Å was considered too strong to be placed on the most sensitive part of the SUMER detector, and the 2 P 1 2 2 D 3 2 line at 1122.486 Å is hopelessly blended (Feldman & Doschek 1977b). The theoretical intensity ratio I(1128.325 Å 1128.340 Å) I(1402.770 Å) as a function of electron temperature was obtained from Keenan, Dufton, & Kingston (1986). For the Si IV observations, a quiet-sun region was observed near solar center, using a 1 300 slit. The spectral resolution for the Si IV lines is about 43 må per detector pixel. The spatial resolution along the slit is about 1. The observing procedure was to obtain a sequence of spatially displaced spectra by stepping the slit in the east-west direction in increments of 0"76. Each spectrum is 50 detector pixels wide (2.15 Å) in the dispersion direction. Sixty-one such spectra were obtained, each with an observing time of 32 s. This procedure was followed first for the 1130 Å lines and then for the 1400 Å lines. Combining the 61 spectra provides spatial resolution in the east-west direction over a distance of 0"76 61 46"4, although the spatial resolution is convolved with Doppler effects and any temporal evolution that occurred while the sequence of observations was made. Several additional observations in each wavelength range were taken within 60 s of each other so that the temporal variations in the line ratios could be studied. Automatic solar rotation compensation software ensured that the same solar regions were being observed for both the 1130 and 1400 Å lines. This was subsequently checked by visual inspection of the overall images of the quiet-sun region in the 1128.3 and 1402.77 Å lines. These images are shown in Figure 1 (Plate L3), which shows that there is a good qualitative correspondence between features in the 1128.3 Å lines and features in the 1402.77 Å line. To examine the dependence of the line ratio on intensity, we have constructed composite spectra for six quiet solar features: faint cell center, average cell center, average quiet Sun, average network, bright network, and very bright network. Intensity histograms are used to classify individual spectra in a manner similar to that used by Reeves (1976) and Vernazza, Evrett, & Loeser (1981). Representative network and cellcenter spectra of the 1128 Å lines are shown in Figure 2. There is an obvious enhancement of the Si IV lines relative to the Fe III lines in the network spectrum, compared with the cellcenter spectrum. This contrast enhancement is expected from previous studies that used Skylab spectra (e.g., Feldman, Doschek, & Patterson 1976; Reeves 1976). The brightest features in the SUMER spectra are quite small in spatial extent, =1 in size. For example, spectra obtained for adjacent pixels on either side of one of the brightest pixels along the length of the slit show a considerable reduction in the enhancement of the Si IV lines relative to the Fe III lines compared with the spectrum of the brightest pixel. To obtain accurate intensity ratios, the Si IV lines at 1128 Å were deconvolved from the Fe III lines using a least-squares fitting technique, assuming Gaussian distributions for the line profiles after subtracting the nearby continuum. It should be noted that for the very brightest regions the presence of explosive events or jets may lead to additional uncertainty in the derived intensity of the 1128 Å lines, since any enhancement in the wings of the line profile will be blended with the adjacent Fe III lines. The derived intensity ratios depend on the instrumental sensitivity at 1128 and 1403 Å. We assume the most recent instrumental calibration curves for detector A (Wilhelm et al. 1996b), which indicate intensities of 78 and 356 ergs cm 2 s 1 sr 1 Å 1 for each 1 count s 1 pixel 1 at 1128 and 1403 Å, respectively. Calibration curves derived before launch are given in Wilhelm et al. (1995). No significant changes in sensitivity were detected during the first 4 months of operation (Wilhelm et al. 1996b). 3. RESULTS AND DISCUSSION The ratios of the Si IV lines are given in Table 1 for the features mentioned in 2. In the network, the ratio is roughly constant with an average value of 0.095 (from intensities in energy units). However, at average quiet-sun and cell-center intensities, the ratio increases rapidly and reaches a maximum value of 0.168 in the cell interiors. Note that the larger ratios imply a higher temperature of formation than smaller ratios. The measurement uncertainties in these ratios are due to three primary effects: the uncertainty in the effective instrument area between the two Si IV wavelengths, the uncertainty due to the temporal separation of the observations, and the uncertainty due to counting statistics. The relative calibration uncertainty is estimated to be 35% at 1403 Å and 15% at 1128 Å. The counting uncertainty is 4% or less. The uncertainty due to the temporal separation of the observations is difficult to estimate. The line ratios derived from the relatively small number of spectra taken within 60 s of each other are very close to those derived from the composite spectra. Agreement is particularly good for cell-center and quiet-sun regions. At network intensities the line ratios agree to within 10% 15%. The temperature of maximum ion concentration for Si IV given by Arnaud & Rothenflug (1985) is 6.3 10 4 K. From Keenan et al. (1986), the theoretical Si IV ratio is 0.059 at this temperature. Thus the predicted ratio for an isothermal plasma is about a factor of 2.8 smaller than that observed in cell centers and 1.6 smaller than that observed in the network. The observed ratio for network features implies a temperature of about 10 5 K. An isothermal plasma has been assumed in the above discussion. However, because the two lines have different

No. 2, 1997 FORMATION TEMPERATURE OF Si IV L121 FIG. 2. Average quiet-sun network and cell-center spectra near the 1128 Å Si IV lines. The other three lines are transitions in Fe III (see Feldman & Doschek 1977b). The wavelength scale is not precise. excitation energies, their maximum emitting efficiencies occur at somewhat different temperatures, and therefore the observed ratio depends to some extent on the emission measure distribution of the solar atmosphere near a temperature of 10 5 K. To check the effect of the emission measure distribution on the observed line ratio, we adopt the typical emission measure distribution for the solar atmosphere derived by Raymond & Doyle (1981). We fold the theoretical temperature-dependent factors for the line intensities through this emission measure distribution for each line and then take the ratio of the resultant integrals. This ratio should be the observed line intensity ratio from a plasma with a typical quiet-sun emission measure distribution that is in ionization equilibrium. The predicted ratio we derive, taking account of Feature TABLE 1 Si IV Line Ratios I(1128.33 Å) a (ergs cm 2 s 1 sr 1 ) I(1402.77 Å) a (ergs cm 2 s 1 sr 1 ) Ratio Faint cell center... 3.96 23.53 0.168 Average cell center... 6.40 49.01 0.131 Average quiet Sun... 10.51 97.74 0.107 Average network... 17.73 179.62 0.099 Bright network... 26.97 282.80 0.095 Very bright network... 36.98 407.41 0.092 a See 2. the differential emission measure, is 0.042, which is about a factor of 2.2 below the measured ratio for network features and a factor of 3.0 below the cell-center ratio. It is difficult to find a satisfactory explanation for the increase in the line ratio for features that correspond to quiet-sun and cell-center regions. The brevity of a Letter precludes a detailed discussion, but we have ruled out as a possible cause direct and cascade radiative recombination into the 3p and 3d levels or absorbing (continuum absorption) clouds that for some reason occupy only cell-center regions. The blending of the 1128.3 Å lines as a result of the emergence of very cold lines in cell-center region spectra cannot be ruled out but seems unlikely. The Si IV profile is well represented by a single Gaussian, as are the neighboring Fe III lines; therefore, the blend would have to be very nearly perfect in wavelength. Furthermore, to bring the observed ratios into agreement with theory, the blend would have to be as intense as the Si IV 1128.3 Å lines. Thus, the theoretically predicted Si IV line ratio is not in satisfactory agreement with observation. Our results are qualitatively consistent with the Si III results reported by Dufton et al. (1984), but are not consistent with the Mg II and Al III results reported by Feldman & Doschek (1977a) and Doschek & Feldman (1987), respectively, and more recently for C III by Doschek (1997). It is possible that some of the discrepancy lies in the computed ionization equilibrium temperature for Si IV.

L122 DOSCHEK ET AL. There are several other temperature diagnostics available with SUMER that were not previously available. Some of these diagnostics are also density dependent, and effects such as Ly continuum absorption must be considered. However, analysis of these additional data should help in deciding whether a real problem exists in the physics of the transition region, or whether the problem is instead in the atomic physics. The SUMER project is financially supported by DARA, CNES, NASA, and the ESA PRODEX program (Swiss contribution). The NRL authors thank the SUMER SOHO team for enthusiastic encouragement and support in obtaining the observations. The NRL group also thanks the SOHO project for financial support, and G. A. D. acknowledges support from NASA Supporting Research and Technology grant W-18218. Arnaud, M., & Rothenflug, R. 1985, A&A, 60, 425 Doschek, G. A. 1997, ApJ, 476, 903 Doschek, G. A., & Feldman, U. 1987, ApJ, 315, L67 Dufton, P. L., Kingston, A. E., & Keenan, F. P. 1984, ApJ, 280, L35 Feldman, U., & Doschek, G. A. 1977a, ApJ, 212, L147. 1977b, A&A, 61, 295 Feldman, U., Doschek, G. A., & Patterson, N. P. 1976, ApJ, 209, 270 Flower, D. R., & Nussbaumer, H. 1975, A&A, 42, 265 REFERENCES Keenan, F. P., Dufton, P. L., & Kingston, A. E. 1986, A&A, 169, 319 Raymond, J. C., & Doyle, J. G. 1981, ApJ, 247, 686 Reeves, E. M. 1976, Sol. Phys., 46, 53 Vernazza, J. E., Evrett, E. H., & Loeser, R. 1981, ApJS, 45, 635 Wilhelm, K., et al. 1995, Sol. Phys., 162, 189. 1996a, Sol. Phys., submitted. 1996b, Appl. Opt., submitted

FIG. 1. Images of a quiet-sun region in the 1128.3 Å (left) and 1402.77 Å (right) Si IV lines. DOSCHEK et al. (see 477, L120) PLATE L3