SOLAR FLUX ATLAS FROM 296 TO 1300 NM I. INTRODUCTION This atlas presents the solar flux spectrum observed with high reso- lution and high signal-to-noise using the Fourier transform spectrometer (FTS) on the McMath Solar Telescope at Kitt Peak National Observatory. The FTS allowed us to greatly improve the signal-to-noise and wavelength coverage compared to the earlier high resolution atlas produced by Beckers, Bridges, and Gilliam (1976) using a grating spectrograph at Sacramento Peak. Information about the observations is summarized in Table 1. Eight scans, odd-numbered 1 to 15, were made of the integrated solar disk with overlapping bandpasses that cover the spectrum from 296 to 1300 nm. High points from these scans are plotted in Figures 1 and 2 for frequency and wavelength. At the same time scans, even-numbered 2 to 16, were made for a standard lamp. The ultraviolet scans have resolving power of approximately 348000 and the infrared scans approximately 522000. Peak signal-to-noise for each scan ranges from 2600 to 9000 as listed in Table 1. At a particular wavelength the noise can be best estimated from the spectrum itself. II. WAVELENGTH REDUCTION The frequency scale for each scan must be normalized by a multipli- cative factor near unity that depends on the setup and alignment of the FTS. Ideally this factor would be determined by observing through a gas absorption cell for a gas with well-known lines in the bandpass. We determined this factor from the spectrum itself by using the OU2D line at 688.38335 nm and defining its air wavelength to be that given by Pierce and Breckenridge (1973). The spectra were converted from frequency to air wavelengths using Edlen's (1966) expression. As this line appears only in scans 11 and 13, it could not be used for the other scans. Scan 15 was set by aligning terrestrial lines common to scan 13. Since the other scans did not have suitable terrestrial lines, they were corrected in the solar rest frame by aligning relatively clean solar lines in overlapping scans. The solar rest frame was determined from radial velocities com- puted with a program that we obtained from Mark Reid of the Center for Astrophysics. It has errors on the order of 4 m/s because it does not include planetary perturbations. The radial velocities for the beginning, end, and middle of each scan are listed in Table 1. The change in radial velocity while a scan was being made resulted in about 0.2 km/s broadening of the solar lines. We used the middle value in shifting the spectra. Operationally, the frequency correction factor was also treated as a Doppler shift and specified in km/s. These values are listed in Table 1. Considering all the fitting and shifting, our final wavelengths in scan 1 may have errors as large as 0.1 km/s. If the wavelength of any one line in each scan can be determined more accurately, the whole scan can be rescaled to that accuracy. III. REDUCTION TO RESIDUAL FLUX To produce a residual flux spectrum we made a zero-point correction, divided out the standard lamp, fitted a pseudo-continuum to each scan, and then we combined the scans together into a continuous whole. Non-linearities in the detector introduce systematic errors in the strong, low-frequency components of the Fourier transform that produce a varying zero-point error in the spectrum. This error can be avoided by using a narrow bandpass to reduce the signal. We corrected the spectrum by fitting linear segments to opaque terrestrial line centers that were wide enough that ringing was not a problem. Ringing is clearly visible in the expanded scale plots in Figure 7 for terrestrial lines that are not resolved. Only scans 11, 13, and 15 had terrestrial lines that allowed this correction, and for those scans the maximum corrections were 0.45%, 0.25%, and 0.90%, respectively. We estimate the residual error to be about 0.10% . Scans 1 to 9 were not corrected and we cannot rule out the possibility of 0.25% zero-point errors. In particular, the cores of the Ca II H and K lines are uncertain by this amount. In conjunction with each solar scan, a standard lamp was scanned with the spectrometer in the same configuration. We divided the solar scans by the lamp scans to remove the effects of any small features in the bandpass limiting filters that were used with each scan. Figure 3 illustrates our method for subjectively fitting a pseudo- continuum to each scan. For each scan the highest point out of every 100 points was plotted and a curving pseudo-continuum was subjectively fitted to the highest high points. Each scan was normalized to its pseudo-con- tinuum so that it became residual flux with the continuum level at 1.0. Note in Figure 3 that the pseudo-continuum is obviously far from being the real continuum for scans 1 to 7. Figure 4 illustrates the subjective fitting together of the scans into a continuous whole. We made plots of the overlapping scans and chose a high region where the scans coincided as the crossover point. The wave- lengths where the cuts were made are 329.897, 378.2914, 401.965, 473.8, 576.5, 753.9, 999.7 nm. These plots also give an indication of repro- ducibility and noise level but these are regions near half-maxima of the bandpasses where the bandpasses have steep slopes of opposite sign. The quality of the data at the centers of the bandpasses is much higher. Ideally measurements like these should be taken with rectangular band- passes that overlap in order to avoid these problems. There are systematic differences between scans. First, zero-point corrections could be made only on scans that had completely opaque ter- restrial lines that are supposed to produce zero signal. No zero-point correction could be made for the blue and ultraviolet scans and we cannot rule out errors as large as 0.25% . If neighboring scans have different, non-negligible errors the residual fluxes are inconsistent across the join. Second, these data are of such high quality that very weak terres- trial features can be significant. Neighboring scans were taken through different atmospheric conditions with different amounts of water vapor and ozone. For example, at 402.07, 402.14, and 999.9 nm the spectra are actually different for this reason. Finally, the merged spectrum has solar wavelengths so terrestrial lines are Doppler shifted from each other in neighboring scans by the amounts listed in Table 1. The final residual flux spectrum is plotted twice, in Figure 5 at 40 nm per page, and in Figure 6 at 4 nm per page in the visible and at 8 nm per page in the infrared. In Figure 6 the top 10% is also plotted at an expanded scale that shows weak features. IV. IRRADIANCE CALIBRATION The calibration is based on the irradiance derived by Neckel and Labs (1984). Their data are listed in Table 2 and shown in Figure 7. Their bandpasses provide continuous coverage from 329 to 666 nm and sporadic points out to 1247 nm. From the atlas we computed the mean value of residual flux R for each bandpass assuming that the bandpasses were exactly rectangular. The width and R for each bandpass are listed in Table 2. The actual bandpasses were slightly rounded and have some wavelength error, but errors from these effects are inconsequential. We divided the observed irradiance by the mean R to find the irradi- ance pseudo-continuum level listed in Table 2. Our value of R includes the terrestrial lines. Neckel and Labs' corrections for terrestrial line absorption in their observations are also listed in Table 2. As they chose their bandpasses in the red and infrared to have few or no terrestrial absorption lines, the correc- tions vary from 0 to 2% and are typically on the order of 0.1% . We plan to determine this for our scans by synthesizing both the solar and the terrestrial spectrum, but we have not yet done so. We have assumed that our terrestrial absorption is approximately the same as that of Neckel and Labs' and that ours is negligible when theirs is negligible. Figure 7 includes a plot of the pseudo-continuum values corrected for this absorption. Ideally this should be a smooth curve, but, in fact, it has structure at the several per cent level. The dip at 600 nm is probably caused by our poor treatment of the ozone absorption there, which, by accident, falls between scans 9 and 11. The ozone feature is the Chappuis bands which run from 450 to 750 nm. The cross-section is tabulated by Vigroux (1953) and is roughly plotted in a figure in Goody (1964). The maximum absorption coefficient is 5.E-21 cmD2U which should produce about 5% absorption at 600 nm in our atlas. Our signal-to-noise is high enough that we should be able to see the structure. Vigroux also gives data between 407 and 447 nm. Apparently there are no measurements of ozone cross-sections between 350 and 400 nm and longword of 750 nm. At our signal-to-noise, ozone may be significant at all wavelengths. We would very much like to see a thorough, high-resolution laboratory study. In the infrared we subjectively fitted an expression for the pseudo-continuum to the measured points for the range 746.5 to 1300 nm, IRRADIANCE=0.80933E18/W**5/ (EXP(1.4384E7/W/(6892.094-2.691159*W+.001669325*W**2))-1.) for wavelength W in nm and irradiance in ;W/cmD2U/nm. Below 330 nm we pieced together a calibration using Broadfoot (1972) whose long wavelength limit is 319.9 nm and Arvesen et al. (1969) who start at 300 nm and go up. We plotted Broadfoot, Arvesen, and Neckel including overlaps and concluded that the Arvesen wavelength scale was shifted by +0.4 nm and that their irradiance should be reduced by about 10 per cent to be in agreement with Neckel and Labs and with Broadfoot. We made Neckel-like 2.1 nm bands for Broadfoot and 2 nm bands for Arvesen et al. These are listed in Table 2 and plotted in Figure 7. We used Broadfoot for 296 to 310 nm and Arvesen et al. multiplied by 0.9 from 312 to 328 nm. We do not trust this ultraviolet calibration to better than 25 per cent, both from the observations and from the very poor pseudo-continuum fit in this region as shown in Figure 3. To convert the atlas from pseudo-residual flux to irradiance, we computed a pseudo-continuum value, either from the fit in the infrared longward of 746.5 nm, or from linearly interpolated bandpass values, and then multiplied by the pseudo-residual flux. Table 3 lists the calibration of the pseudo-continuum level of the atlas in ;W/cmD2U/nm for every 0.05 nm in the visible and every 0.10 nm in the infrared. One ;W is 10 ergs/s. For flux in ergs/cmD2U/s/nm at the surface of the sun multiply the irradiance by 10.*(1.495985E13/6.9598E10)**2 = 462020. For astrophysical flux F divide the flux by ". For flux moment H divide the flux by 4". V. MAGNETIC TAPE LISTING A magnetic tape listing of the atlas may be obtained by sending a 2400 foot tape to Dr. Robert Kurucz, Center for Astrophysics, 60 Garden Street, Cambridge, Massachusetts 02138. The tape will be written 9 track, 1600 bpi blocked ASCII with 80 characters per record and 50 records per block. The first file will contain wavelength, residual flux pairs. The second file will be a program that can read the tape and convert it to irradiance if desired. The third file will be the residual flux inter- polated to uniform point spacing in wavelength. VI. REFERENCES Arvesen, J.C., Griffin, R.N., and Pearson, B.D. Appl. Opt., 8,2215-2232,1969. Beckers, J.M., Bridges, C.A., and Gilliam, L.B. 'A high resolution spectral atlas of the solar irradiance from 380 to 700 nm', AFGL-TR-76-0126, 340pp., 1976. Broadfoot, A.L., Ap.J., 173, 681-689, 1972. Edlen, B. Metrologia 2, 71-80, 1966. Goody, R.M. Atmospheric Radiation, Oxford Univ. Press, London, 436pp, 1964. Neckel, H. and Labs, D. Solar Physics, 90, 205-258, 1984. Pierce, A.K. and Breckenridge, J.B., 'The Kitt Peak Table of Solar Spectrum Wavelengths', Contribution No. 559, 1973. Vigroux, E. Annales de Physique, 8, 709-762, 1953.