SUMER/SOHO OBSERVATIONS OF HYDROGEN ... - IOPscience

The Astrophysical Journal, 503:L95–L100, 1998 August 10 q 1998. The American Astronomical Society. All rights reserved. Printed in U.S.A.

DYNAMIC BEHAVIOR OF THE UPPER SOLAR ATMOSPHERE: SUMER/SOHO OBSERVATIONS OF HYDROGEN LYMAN LINES Werner Curdt Max-Planck-Institut fu¨r Aeronomie, D-37191 Katlenburg-Lindau, Germany

and Petr Heinzel Astronomical Institute, Academy of Sciences of the Czech Republic, CZ-25165 Ondrˇejov, Czech Republic Received 1998 March 17; accepted 1998 May 21; published 1998 July 17

ABSTRACT We present first observations of the temporal evolution of hydrogen Lyman lines, made by the Solar Ultraviolet Measurements of Emitted Radiation (SUMER) spectrometer on the Solar and Heliospheric Observatory. A time series of about 33 minutes was obtained on 1997 June 5. The entrance slit has crossed a quiet-Sun region of 1150. 3 with two internetwork structures (cells) and the bright network regions. A data set of 59 spectra with 30 s exposure time was obtained, using the SUMER tracking system to compensate for the solar rotation. For ˚ and the three Lyman lines Ly5, Ly9, our analysis, we have selected a Lyman continuum window around 907 A and Ly15, which are formed at different depths in the upper chromosphere. In the cell interiors, we have detected significant periodic intensity variations with a Fourier transform power peak at 3.3–3.5 minutes, which is consistent with 3 minute internetwork oscillations. They seem to be associated with spatially unresolved “clusters” of grains. In the bright network regions, we detect slower oscillations of 6.9–7.6 minutes. These waves seem to propagate upward as we deduce from a phase shift between the three Lyman lines studied. The phase velocity was estimated to be roughly 3 km s21 in the network. Finally, we discuss the potential usefulness of the hydrogen Lyman lines for diagnostics of the temperature structure of the upper solar atmosphere. Our observations, in particular the fact that we see all Lyman lines in emission all of the time, put certain constraints on the temperature gradients above the region in which numerical simulations do predict a decrease of the mean kinetic temperature. Subject headings: Sun: chromosphere — waves eral advantages. First, the line cores are supposed to be formed in the upper chromosphere or at the base of the transition region, above the heights so far considered in simulations. The series spans the range of heights at which strong temperature gradients can be expected (at least this is the case for static VAL models) and therefore can be used to diagnose the time evolution of the temperature structure. The line cores of higher Lyman lines are sensitive to the kinetic temperature as demonstrated by Heinzel, Schmieder, & Vial (1997), and this is because the upper levels of the corresponding transitions are strongly coupled to the kinetic temperature. These lines are also relatively strong and thus easily observable. Since hydrogen is the most abundant species and contributes—at these heights—almost entirely to the electron density, it represents the basic plasma component. Lyman lines also play an important role in the total energy balance of the upper atmosphere. From simulations, the time-dependent Lyman line intensities can be synthesized and compared with observations. In this Letter, we present first observations of the temporal evolution of the hydrogen Lyman lines and its continuum, made by the SUMER spectrometer on SOHO. We obtained a data set from the slit crossing a quiet-Sun region with typical cell and network structures, and we discuss the periodicities seen there.

1. INTRODUCTION

New theoretical studies that have appeared during the last few years put into question hitherto widely accepted semiempirical models of the solar atmosphere. Radiation-hydrodynamical simulations of the nonmagnetic internetwork structure (cell interior) reveal remarkably good agreement between the observed and computed temporal variation of the Ca ii H line (for a review see Carlsson & Stein 1997). In such models, the equations of radiation hydrodynamics are solved for the lower atmosphere (up to 1800 km), which is driven by vertically propagating acoustic waves. Ab initio simulations predict significant temporal variations of the temperature in a highly dynamical atmosphere and show that the mean temperature may in fact be decreasing with height, in contrast to what one gets from static semiempirical models of Vernazza, Avrett, & Loeser (1981) (VAL models). Higher in the atmosphere, selected UV lines of different species (neutral or singly ionized) exhibit a similar 3 minute oscillatory behavior as Ca ii lines but are observed to be in emission all of the time (Carlsson, Judge, & Wilhelm 1997; see also Judge, Carlsson, & Wilhelm 1997 for the network behavior). These observations, made by the Solar Ultraviolet Measurements of Emitted Radiation (SUMER) spectrometer on the Solar and Heliospheric Observatory (SOHO), are thus not consistent with the above-mentioned simulations. This lead Carlsson et al. (1997) to ask, “Does the time-averaged gas temperature continue to decrease with height, as in the case of an ab initio, one-component nonmagnetic model?” From all that was said above, we see that new observations and simulations are highly needed, namely those that will tell us something about the upper atmosphere. In this respect we have considered the Lyman series of hydrogen, which has sev-

2. OBSERVATIONS

SUMER is a stigmatic, high-resolution normal-incidence spectrometer that covers the spectral range between 400 and ˚ . With a spatial resolution close to 10 (about 715 km 1610 A ˚ (with subpixel reson the Sun) and spectral pixels of 40 mA olution), it provides a unique opportunity to observe the whole hydrogen Lyman series plus the Lyman continuum. The L95

L96

CURDT & HEINZEL

˚ ), raster image of the studied region taken at Fig. 1.—CDS He i (584 A 11:48:40 UT. A cross represents the position of the SUMER slit and its length (courtesy of B. Schmieder and the CDS consortium).

Vol. 503

SUMER capabilities are described in detail by Wilhelm et al. (1995), and the in-flight performance is presented in Wilhelm et al. (1997) and Lemaire et al. (1997). In the framework of our “Lyman Program,” we have so far obtained four time series from different targets on 1997 June 5 and 7. In this Letter, we analyze the data of the first series obtained on 1997 June 5. Unfortunately, we could not observe at the disk center due to a constraint on the pointing of the instrument, limiting the slit position to a corridor in the west at that time. We used the 100 # 120 00 slit, centered at position X 5 6960. 2 and Y 5 1580. 3 (SOHO coordinates system, cos v 5 0.664). Starting at 11:56:49 UT, we have obtained a set of 59 spectra with 30 s exposure time and an image repetition rate of about 33.5 s. A few extra seconds were required for the solar rotation tracking. The whole sequence thus lasted for 1984 s (≈33 minutes). The exposure time and the length of the series were dictated by the requirements on photon statistics, the on-board storage capability, and constraints of the telemetry rate. Four times during the whole time sequence, the on-board SUMER solar rotation tracking system has offset the X-pointing by 00. 75 to compensate for the solar rotation. Therefore, the final position was X 5 6990. 2 with unchanged Y-pointing. We have observed a quiet-Sun region close to a narrow faint filament. The northern part of the slit has crossed the cell interior region, while the central part of the slit was placed over a bright network structure. The southern section of the

Fig. 2.—Time-averaged spectrum (300 s exposure) of the studied region. The upper panel shows the line profiles summed over the whole slit length (intensities are in arbitrary units; note that Ly5 is placed on the bare part of the detector). The lower panel indicates the intensity variations along the slit—one can clearly distinguish two bright network structures and dark internetwork regions.

No. 1, 1998

DYNAMIC BEHAVIOR OF UPPER SOLAR ATMOSPHERE

L97

Fig. 3.—Ly5 central pixel (line-center) intensity map. We see the same morphology as in Fig. 2, but resolved in time with 30 s exposures.

slit cuts another dark internetwork region and finally ends with a network brightening. This morphology was checked first on Meudon Ha and Ca ii K3 spectroheliograms from 1997 June 5 and also using a co-aligned CDS raster image in He i ˚ ) obtained just prior to our observations (see Fig. 1). (584 A We can not definitely exclude that rather weak and narrow parts of the close-by filament could possibly contaminate the southward cell emission. However, as will be shown, the dynamical behavior of both cells is almost the same, which indicates that we see two similar quiet-Sun cells. The spectra were acquired by the detector B on its central band of 120 spatial # 1024 spectral pixels. This covered the ˚ , including all Lyman lines spectral band from 903 to 943 A from Ly5 (here 5 means the line serial number) up to the series limit, plus part of the Lyman continuum. This is depicted in Figure 2, which shows an average spectrum taken with 300 s exposure time after the high cadence sequence. The darkest section in the spectrum corresponds to the cell interior (internetwork region), the bright part represents the network enhanced emission. A few pixels remain unexposed due to a small offset between the image of the slit and the read-out window. In order to avoid the co-pointing complication, we did not

convert slit pixel positions into solar coordinates. We keep the SUMER coordinate system of 00. 961 pixels with pixel numbers increasing from north to south. 3. DATA REDUCTION

After decompression, the data set has been flat-field corrected. The standard procedure for flat-field correction produced residual errors, employing the flat-field exposures available. Therefore, the total data volume was taken to define an average spectrum, which was then compared to a median-10 filtered version to compensate for detector nonuniformities. This method was suggested by C. deBoer (1997, private communication) and yields good results in similar cases. The geometrical distortion of the detector has been corrected using the destretching procedure of T. Moran, which also contains a compensation for the small inclination of the slit image relative to the detector pixel direction. The wavelengths have been calibrated by a second-order least-squares fit using several unblended O i lines in our spectral window as wavelength standards. The averaged profile of our spectral window is also displayed in Figure 2. The most prominent lines are marked

L98

CURDT & HEINZEL

Fig. 4.—Same as Fig. 3, but averaged by a smooth-4 noise filter and shown as three-dimensional plot. Here we can clearly see the intensity variations with about 3 minute periods in darker cell regions and about 7 minute periods in the bright network. For corresponding pixel positions, refer to the text.

and their identification and wavelengths are given. The Ly5 line and a part of the Lyman continuum are placed on the bare parts of the detector, which has a reduced sensitivity in this spectral range. Therefore, the Ly5 line intensity and the con˚ used in our analysis had to be tinuum intensity around 907 A corrected for the KBr/bare sensitivity ratio of 2.326. After this correction, the relative intensities of the studied Lyman lines can be compared without any radiometric calibration. Absolute intensities will only be required for subsequent papers devoted to a comparison with numerical simulations. For disk observations, the levels of scattered light or other parasitic signals are negligible. Thus, the noise in our data is only given by photon statistics. On average, we collected 230, 210, and 170 counts per exposure and pixel integrating the lines Ly5, Ly9, and Ly15, respectively.

Vol. 503

Fig. 5.—A comparison of the line-center intensity variations in Ly5, Ly9, and Ly15 lines using all 59 spectra of the whole data set. For comparison, we also include the Lyman continuum variations.

where these Lyman lines are formed, the chromospheric structures may appear more diffuse. Just by counting the number of intensity peaks over the observed time period of 33 minutes, one can arrive at a periodicity close to 3 minutes inside dark cell regions. For the bright network seen around pixel 60, the periods seem to be longer, about a factor of 2. It is important to note that the emission profiles of the optically thick Lyman lines span relatively large atmospheric depths (see, e.g., Fig. 1 in Vernazza et al. 1981) and, therefore, in oscillation studies one should better use the line-center intensities rather than the integrated emission, which would allow intensity variations to mix waves from different depths.

4.1. Cell Interior 4. RESULTS

The data from 1997 June 5 show significant temporal variations of the intensities and shapes of all Lyman lines observed. For this exploratory analysis, we have selected Ly5 ˚ ), Ly9 (920.963 A ˚ ), Ly15 (915.329 A ˚ ), and the (937.803 A ˚ , formed at difLyman continuum between 905.2 and 909.6 A ferent chromospheric heights. Ly5 is partly blended by the ˚ ) line, but its temporal variations are quite O i (937.8405 A ˚ ), which is less blended (for similar to the Ly6 line (930.748 A detailed line identifications, see Curdt et al. 1997). We have calculated the center of gravity of the average profiles of these lines using all 59 spectra and define these as line center. Intensity values for this noninteger position can be derived by linear interpolation between two pixels. In Figure 3, we show the temporal evolution of the Ly5 line center intensity applying a smooth-3 noise filter. Between spatial pixels 0 and about 30, we identify an internetwork region that is relatively dark. In the central part of the slit, we see the bright network, then further to south, the morphology seems to repeat. Averaged by a smooth-4 noise filter, we show the same Ly5 central intensity as a three-dimensional plot in Figure 4. Inside a dark internetwork region, we note a weak brightening around pixel 20 which exhibits clearly an oscillatory behavior. As a working hypothesis, we ascribe this feature to a cluster of spatially unresolved oscillatory grains. The typical size of Ca ii K grains is 0.5–1.5 Mm (Rutten 1995); our spatial resolution along the slit was about 20. Moreover, higher in the atmosphere

We have analyzed time variations of the three Lyman lines Ly5, Ly9, and Ly15 for two dark internetwork regions. In Figure 5, we display the central intensity variations for these lines (in arbitrary units, i.e., with different offset and scaling) for the spatial position at pixels 93 to 95. The Lyman continuum ˚ ) is also shown. The intensity (integrated from 905.2 to 909.6 A intensity values have been averaged over 3 pixels along the slit, and finally, a smooth-3 noise filter has been applied. As a side effect this measure against the Poisson noise smears all information on short-term temporal variations. Note that the line-center intensity is decreasing with increasing serial number all of the time. In the long term, the intensity seems to be constant. We have performed a Fourier analysis on these data and obtained a power peak in the frequency range of 4.8–5.1 mHz. This corresponds to periods between 3.5 and 3.3 minutes. Note that in the internetwork region higher Lyman lines are rather weak, and thus their variations are not so well resolved—see Ly15 (for a better photon statistics, we would need a longer exposure which, however, lowers the sampling over expected periods). In this particular plot, it is difficult to identify any expected phase shift between the four curves. The other cell was studied at pixel position 18 to 20 (possible cluster of grains), and we have arrived at quite similar periodicities, although the amplitudes of individual oscillations were varying in time, and thus a Fourier analysis was difficult to perform (for such cases we plan to use wavelet analysis).

No. 1, 1998

DYNAMIC BEHAVIOR OF UPPER SOLAR ATMOSPHERE

L99

4.2. Bright Network Figure 6 shows the same behavior as the previous figure but now for the bright network region located around the slit center (pixels 52 to 54) (same noise filtering). At first sight one can recognize much higher mean intensities and larger oscillation amplitudes. As expected from Figure 4, the Fourier analysis yields lower frequencies between 2.2 and 2.4 mHz. The corresponding periods of 7.6–6.9 minutes are about twice as long compared to periods found inside the cell interior. In the case of network, one can possibly identify a phase shift between oscillations in the three lines, indicating an upward propagating wave (higher serial members are formed deeper), while the continuum seems to be strongly correlated to the Ly5 intensity—see Figure 6. We have also processed the data from a close-by position at pixels 45 to 47. While the power-peak frequency is in the same range (i.e., same oscillations), there is only little correlation between peaks in these two spatial positions. Therefore, the oscillation pattern seems to vary on a spatial scale of several arcseconds, but the periods remain the same. We have also considered the risk of parasitic signals from the solar rotation tracking, but we found our power peaks at positions different enough to exclude a potential instrumental origin. 5. DISCUSSION AND CONCLUSIONS

We have presented first observations of a dynamical behavior of higher Lyman lines in the quiet solar atmosphere. Inside the cells we have detected intensity oscillations in selected Lyman lines with periods of 3.3–3.5 minutes, although this is based only on a short interval, so that other similar periods can be expected. These oscillations seem to correspond to 3 minute oscillations in the chromospheric internetwork regions, namely the bright grains (Carlsson & Stein 1997; Kalkofen 1998). If the 3 minute oscillations that we have detected are indeed related to the grains, we conclude that they arise from a “cluster” of spatially unresolved grains and that these clusters have a diameter of several arcseconds. This seems to be consistent with 30–80 diameter blobs reported by Carlsson et al. (1997). In the bright network regions, on the other hand, we found periods of 6.9–7.6 minutes, which are approximately twice as long. This is consistent with periodicities of about 7 minutes or longer reported for the magnetic network by Lites, Rutten, & Kalkofen (1993) (see also the discussion by Kalkofen 1998). However, the wave patterns detected in network positions as close as a few arcseconds are not very well correlated. This may indicate that we see individual (but unresolved) magnetic flux tubes with the same kind of waves but random phase shifts. These waves seem to propagate upward, as we can deduce from a phase shift between three observed Lyman lines (Fig. 6). A mean time delay between Ly5 and Ly15 intensity maxima is around 36 s. Assuming that the height difference between the regions in which the centers of these lines are formed is of the order of 102 km, we can estimate the phase velocity to amount very roughly to 3 km s21. Note that in the network study of Judge et al. (1997), periodicities of the order of 7 minutes or longer are not reported due to the missing solar rotation tracking in their observations. Since we have detected also a part of the Lyman continuum (see Fig. 2), we show its temporal variations in Figures 5 and 6. They correspond to variations of the intensity integrated from ˚ . While in Figure 5 the Lyman continuum 905.2 to 909.6 A variations roughly follow those of the Ly15 line (formation

Fig. 6.—Same as Fig. 5, but for the network pixels. A phase shift between the three lines seems to be present, in particular around 700 and 1600 s.

region is about the same), in the network (Fig. 6) we see rather good correlation with the Ly5 line. The network is geometrically complicated and one can possibly see the Lyman continuum formation region in phase with the Ly5 one. However, these conclusions are very preliminary and must be confirmed by analyzing longer series of data taken at different places. Numerical simulations of Carlsson & Stein (1997) predict temporal variations of the temperature inside the cell interiors (grains). Temperature increases and decreases with height in the chromosphere exhibit an oscillatory character. Note that these simulations are restricted to chromospheric heights below 1800 km. The problem is how these simulations can be extrapolated to the upper chromosphere and lower transition region in which our observed Lyman lines are supposed to be formed (i.e., their line cores). Since we observe all these Lyman lines in emission all of the time, a continuation of the temperature decrease to these atmospheric layers is rather problematic. As Carlsson et al. (1997) state, there is “something important missing from the calculations,” and we believe that our Lyman lines observations will help in identifying possible causes of such discrepancies. From semiempirical models of Vernazza et al. (1981), we have estimated the range of formation heights for the Lyman lines and continuum studied, using the model VAL-C (for which the ground-level populations are given). Using the corresponding temperature structure of the VAL-C model, one gets the line-center formation heights and the corresponding temperatures for the studied lines: 2200 km and 24,000 K for Ly5, 2109 km and 12,300 K for Ly9, 2097 km and 8970 K for Ly15, ˚ continuum. It is clear and 2050 km and 7660 K for the 907 A that the true dynamical chromosphere will differ from such models, but this example is given here just to demonstrate the diagnostic power of the hydrogen Lyman series. Analysis of time series data on these lines is thus promising and can tell us, together with new improved simulations, more about the dynamical nature of the upper chromosphere and its temperature structure. Here it is interesting to note that our observation that the line-center intensities are decreasing with the increasing line serial number all of the time, together with an assumption that the factor b1 5 n1 /n1∗ is always increasing with height, lead automatically to a temperature increase at all times. This is due to the strong coupling of upper-level populations to the kinetic temperature. To confirm these first results on Lyman lines, we plan to

L100

CURDT & HEINZEL

process the other time series from 1997 June 5 and to obtain new data with high-rate telemetry in order to cover a longer time interval. A detailed study of the temporal evolution of the line profiles that exhibit reversal features and peak asymmetries as well as Doppler shifts will also be a subject of our future work. The SUMER project is financially supported by DLR, CNES, NASA, and the ESA PRODEX Programme (Swiss contribution). SUMER is part of SOHO, the Solar and Heliospheric

Vol. 503

Observatory, of ESA and NASA. This study was partially completed in frame of the SOHO GI Programme. We highly appreciate the help of the MEDOC team, and in particular of J.-C. Vial, during the campaign. We would like to thank Klaus Wilhelm for his comments during the preparation of this paper and Mats Carlsson for the fruitful discussion on data reduction procedures. Part of this work was done when P. H. was visiting MPAE in Lindau. He was supported by GACˇR (grant 205/96/1199) and by the project K1-003-601 of the Academy of Sciences of the Czech Republic.

REFERENCES Carlsson, M., & Stein, R. F. 1997, in Solar and Heliospheric Plasma Physics, Lecture Notes in Physics Vol. 489, ed. G. M. Simnett, C. E. Alissandrakis, & L. Vlahos (Berlin: Springer), 159 Carlsson, M., Judge, P. G., & Wilhelm, K. 1997, ApJ, 486, L63 Curdt, W., et al. 1997, A&AS, 126, 281 Heinzel, P., Schmieder, B., & Vial, J.-C. 1997, Proc. Fifth SOHO Workshop (ESA SP-404; Noordwijk: ESA), 427 Judge, P. G., Carlsson, M., & Wilhelm, K. 1997, ApJ, 490, L195 Kalkofen, W. 1998, in ASP Conf. Ser, Cool Stars, Stellar Systems, and the Sun, 10th Cambridge Workshop, ed. J. A Bookbinder & R. A. Donahue (San Francisco: ASP), in press

Lemaire, P., et al. 1997, Sol. Phys., 170, 105 Lites, B. W., Rutten, R. J., & Kalkofen, W. 1993, ApJ, 414, 345 Rutten, R. J. 1995, Proc. Fourth SOHO Workshop (ESA SP-376; Noordwijk: ESA), 151 Vernazza, J. E., Avrett, E. H., & Loeser, R. 1981, ApJS, 45, 635 Wilhelm, K., et al. 1995, Sol. Phys., 162, 189 ———. 1997, Sol. Phys., 170, 75