Astron. Astrophys. 351, 1139–1148 (1999) ASTRONOMY AND ASTROPHYSICS Temporal variability in the electron density at the solar transition region M.E. Pérez1 , J.G. Doyle1 , E. O’Shea2 , and F.P. Keenan2 1 2 Armagh Observatory, College Hill, Armagh BT61 9DG, Ireland (epp,jgd@star.arm.ac.uk) Department of Pure and Applied Physics, Queen’s University Belfast, Belfast BT7 1NN, Ireland (E.Oshea,F.Keenan@qub.ac.uk) Received 23 July 1999 / Accepted 6 October 1999 Abstract. The electron density as measured in the transition region of a coronal hole, a ‘quiet’ Sun region at disk center plus an active region shows variations of up to a factor of two at Te ∼ 1.5 105 K, lasting at most only a few minutes. There is remarkable agreement between the number of such variations, their temporal variability and duration in the coronal hole and ‘quiet’ Sun datasets, consistent with an earlier bright point study. There appears to be evidence of super-granular cells, with the increases in electron density occurring along the network boundaries. At some locations, periodicities of between 8 and 16 min are visible in the electron density variations. We associate these variations with the sites of explosive events. Key words: line: profiles – Sun: activity – Sun: corona – Sun: transition region – Sun: UV radiation The Solar Ultraviolet Measurements of Emitted Radiation (SUMER) instrument (Wilhelm et al. 1997) on SOHO provides the opportunity to observe the solar atmosphere in the spectral range from ∼500 to 1600 Å with high spectral and spatial resolution. In first order, the spectral resolution is ∼43mÅ, while ∼22mÅ is achieved in second order. The spatial resolution is approximately 1 arc sec in the E-W direction and 2 arc sec along the slit (N-S direction). It should be pointed out that only lines separated by less than 40 Å in first order, and 20 Å in second order, can be observed simultaneously with SUMER due to the size of the CCD. Surprisingly few lines can be used for density diagnostics, due to blending problems, the weakness of some lines, and the fact that possible useful lines cannot be observed simultaneously. The line pair most useful for diagnosing the transition region is probably O iv 1399.8/1401.2 (Wikstol et al. 1997). 2. Theoretical line ratios 1. Introduction There are abundant references to solar electron density (Ne ) diagnostics in the literature, with e.g. emission lines arising from transitions in O iv providing accurate determinations of Ne (Griffiths et al. 1999, Doschek et al. 1998, O’Shea et al. 1998, Spadaro et al. 1994, Dwivedi & Gupta 1991, Hayes & Shine 1987, Feldman & Doschek 1978). For instance, Hayes & Shine (1987) used the ratio of Si iv 1402.8 Å and O iv 1401.2 Å, and found that short-lived bursts typically showed electron density increases coupled with a small line shift to the red. They suggested this might be caused by ‘ micro-flares ’. Cheng (1980), analysing coronal loops in Fe xv & Fe xvi lines, found a density enhancement of ∼ 30% in a loop within 7 minutes, plus a slower variation over a longer time interval. He suggested that this increase in density could be due to mass ejection from lower regions, and the associated dissipation of the electric current associated with the resulting high-density twisted flux strands (Nakagawa & Stenflo 1979) contributing to the coronal heating. In this paper we use the O iv 2s2 2p2 P o → 2s2p4 P densitysensitive multiplet around 1400 Å to analyse time-series solar spectra. More precisely, we use the O iv 1399.8 Å and 1401.2 Å lines for our analysis. Send offprint requests to: M.E. Pérez The model ion for O iv consisted of the 8 energetically lowest LS states, namely 2s2 2p 2 P; 2s2p2 4 P, 2 D, 2 S, 2 P; 2p3 4 S, 2 D and 2 P, making a total of 15 fine-structure levels. Energies for all of these were obtained from Safronova et al. (1996). Electron impact excitation rates for transitions in O iv were taken from Zhang et al. (1994). For Einstein A-coefficients, the calculations of Nussbaumer & Storey (1982), Brage et al. (1996) and Dankworth & Trefftz (1978) were adopted for the 2s2 2p 2 P1/2 – 2s2 2p 2 P3/2 , 2s2 2p – 2s2p2 and 2s2p2 – 2p3 transitions, respectively. As noted by, for example Seaton (1964), excitation by protons may be important for fine-structure transitions. In the present analysis we have used the proton rates of Foster et al. (1996, 1997) for transitions within 2s2 2p 2 P and 2s2p2 4 P, respectively. Using the atomic data discussed above in conjunction with the statistical equilibrium code of Dufton (1977), relative O iv level populations and hence emission line strengths were calculated for a range of electron temperatures and densities. Details of the procedures involved and approximations made may be found in Dufton (1977) and Dufton et al. (1978). In Fig. 1 we plot the theoretical ratio R = I(1399.8 Å)/ I(1401.2 Å) as a function of electron density at the electron temperature of maximum O iv fractional abundance in ionisation equilibrium, log Te = 5.2 (Mazzotta et al. 1998). Given errors 1140 M.E. Pérez et al.: Temporal variability in the electron density at the solar transition region Table 1. Description of observational data Date Start UT End UT Pointing: X,Y Slit hni (arc sec2 ) Location X Width/Y Width Exposure time 10 July 1996 07:36:15 08:42:56 (630,-200) h6i 0.3 × 120 AR ∼7 × 82 20 s 17:09:42 18:16:42 (3,0) h4i 1.0 × 120 QS ∼10 × 112 20 s 14 July 1996 22:32:46 00:00:09 (3,0) h4i 1.0 × 120 QS ∼10 × 85 20 s Fig. 1. Theoretical O iv I(1399.8 Å)/I(1401.2 Å) line ratios, calculated at an electron temperature of log Te = 5.2. The present results are shown with a continuous line, and those from the CHIANTI database with a dashed line. of typically ±10% in the adopted atomic data (see references above), we would expect the theoretical values of R to be accurate to better than ±15%. We note that R is very insensitive to variations in the adopted electron temperature. For example, varying Te by 0.2 dex leads to a <1% change in the theoretical R ratio. Also shown for comparison in Fig. 1 are the values of R obtained from the CHIANTI database (Landi et al. 1999). An inspection of the figure shows that the current diagnostic calculations are quite similar to those from CHIANTI. We therefore adopt the former in the present analysis, but note that use of the CHIANTI ratios would not significantly affect our results nor discussion. 3. SUMER observations and data reduction 3.1. Data The data used here were obtained with SUMER on-board SOHO on 10 and 14 July 1996 (see Table 1). These datasets were taken in order to look for variations in electron density in the solar transition region, using the density sensitive line ratio of O iv 1399/1401. The pointing for our observations were centered on different regions in the Sun: one extended active region (AR), two ‘quiet’ Sun regions (henceforth QS1 and QS2) and one 01:07:23 02:14:04 (0,910) h4i 1.0 × 120 Northern CH ∼1 × 112 20 s Fig. 2. A SOHO EIT image obtained in Fe xv 284 Å on 14 July 1996 at 01:30 (courtesy of the EIT consortium). The SUMER temporal series for O iv were centered 910 arc sec from disk center, i.e., in the Northern CH shown in this zoom image. region in the Northern coronal hole (CH). We used slit number six for the AR dataset (0.3 × 120 arc sec2 ) and slit number four for the other datasets (1.0 × 120 arc sec2 ). All the datasets were taken with a 20 s exposure time, and each region was observed over a period of approximately one hour and seven minutes. These observations were taken in a sit-and-stare mode with the rotational compensation turned off. This meant that for the CH an area of approximately 1.5×120 arc sec2 was observed, since the rotational velocity in this region of the Sun is very low (∼1.5 arc sec in 67 minutes, see Fig. 2)1 . An area of 10 × 120 arc sec2 was covered over the observation period for the QS datasets at disk centre, and ∼7×120 arc sec2 for the AR dataset, (see Fig. 3 & Fig. 4). Detector A was used for all the datasets and the observations were taken in first order. Due to very low signal-to-noise or problems with detector sensitivity at the ends of the slit image, some positions at the top and/or the bottom of the slit where clipped out. For the AR dataset thirty positions at the Northern end and four positions at the Southern end were clipped out, so that the final dimensions are ∼7 × 82 arc sec2 . For the CH dataset the final dimensions are ∼1×112 arc sec2 after clipping low signal-to-noise pixels. For the QS the clipping depended on the dataset, and it was due to low signal-to-noise since the slit is centered in the detector. Four positions at the Southern end were clipped out for both datasets, so that for QS1 the dimensions 1 see http://star.arm.ac.uk/∼ambn/preprints.html for color plots M.E. Pérez et al.: Temporal variability in the electron density at the solar transition region 1141 value. Nevertheless, areas with measurable densities were found and they are discussed in Sect. 4. 3.2. Data reduction and calculation of errors Fig. 3. A SOHO EIT image obtained in Fe xii 195 Å on 10 July 1996 at 20:38 (courtesy of the EIT consortium). The SUMER datasets for O iv were centered at (3,0) arc sec in the disk center, i.e., in the QS shown in this zoom image. The SUMER rastered area(s) of ∼10 × 120 are over-plotted with a white rectangle. Fig. 4. A SOHO EIT image obtained in Fe xii 195 Å on 10 July 1996 at 20:38 (courtesy of the EIT consortium). The SUMER datasets for O iv were centered at (630,−200) arc sec, i.e., in the AR shown in this zoom image. The SUMER rastered area of ∼7 × 82 is over-plotted with a white rectangle. where reduced to 10 × 112 arc sec2 , but for QS2 the dimensions where reduced further to 10 × 85 arc sec2 after clipping out twenty-seven positions in the Northern end of the slit. The O iv 1401.16 Å line is blended with the S i 1401.51 Å transition (see Judge et al. 1998, for reference wavelengths), although in most areas in the Sun the S i feature is considerably weaker than O iv. The S i line was appreciable only in the ‘quiet’ Sun datasets. The O iv 1407/1401 ratio is also available from our data, but the O iv 1407.38 Å line is blended with the second order O iii doublet at 703.85 Å, and some preliminary analysis with this ratio showed that unblending the two features was difficult. Since the O iv lines we use here are not strong lines we used a binning in time of four minutes, plus a running mean along the slit of five pixels, to decrease the noise level of our data without losing a desirable spatial/time resolution. The low signal-to-noise of our data in the QS and CH regions made a reliable estimation of the electron density very difficult for some positions in our raster/temporal images. This, combined with the fact that the O iv 1399/1401 density-sensitive ratio is in the low density limit for a large fraction of the ‘quiet’ Sun and coronal hole spectra, were the main reasons why for these regions a large part of our density estimates were set to the minimum theoretical For the SUMER instrument, the process of data reduction involves three main steps: flat-fielding, de-stretching and radiometric calibration. Our dataset were automatically flat-field corrected on board. The de-stretching process is necessary in particular for the data located towards the edges of the detector due to various wavelength and spatial distortions (see Siegmund et al. 1994, Wilhelm et al. 1997). Other non-linearity effects that ought to be corrected in SUMER are dead-time effects and local gain depression. Dead-time effects of the detectors become significant for high total detector counts rates, for instance higher than 50 000 counts s−1 . The local gain depression is critical for intense lines with more than 10 counts s−1 pixel−1 . Detector noise is partly reduced by the flat field correction which corrects the readout noise and pixel-to-pixel variations. The line fitting has been carried out using the CFIT BLOCK subroutine (Haughan 1997). For all the datasets, only one Gaussian was used to fit either the O iv 1399 Å line or the O iv 1401 Å line. In the case of O iv 1401 Å, which has the weak line S i 1401.514 Å present in the QS and CH datasets, we checked using two Gaussians but found that the results were more reliable using only one. For the above corrections the basic IDL routines can be found from within the SUMER software tree.2 The other source of noise in our data is the photon-related statistical noise, which obeys a Poisson distribution. Poisson noise in the data is calculated as the square root of the number of counts per pixel. For the estimation of the errors that affect our final results we have to include errors in the line fitting parameters and the propagation of these errors into the line ratio. Finally, the 1σ uncertainty in the calculated values of the electron density are estimated from the theoretical curve (Fig. 1), by considering the corresponding 1σ variation in the observed ratio. The analysis of periodicities presented in Sect. 4 was carried out using the PERIODOGRAM.PRO routine given in the CDS software tree. This routine uses the method of Horne & Baliuna (1986) to calculate the periodogram. 4. Results 4.1. Coronal hole (CH) The Northern CH dataset, centered at (0, 910) arc sec, started at 01:07:23UT and ended at 02:14:04UT on 14 July 1996 and had an exposure time of 20s. Since our image is in fact a temporal series for the observational period (∼1h7min), the total area covered by this dataset was ∼1.5 × 112 arc sec2 . The variations of the electron density values for each position along the slit with time is shown in Fig. 5. The allowed range of values for the grey scale is between 3.6 109 and 2.5 1010 cm−3 , with the average electron density for the whole image being 6.8 109 cm−3 2 sohowww.nascom.nasa.gov/instruments.html 1142 M.E. Pérez et al.: Temporal variability in the electron density at the solar transition region Fig. 5. The O iv electron density variations in the CH dataset. Fig. 6. Electron density (left panel) and intensity (right panel) values for the CH dataset. These are the values corresponding to locations A and B in the slit image represented in Fig. 7. The typical 1σ error in log Ne is indicated on each plot, at the positions where the minimum and maximum values for the errors were found. (log Ne = 9.83+0.20 −0.17 ). This average was calculated only for the 37% density values greater than the low density limit. The data in Fig. 5 clearly show several individual density enhancements which are temporal in nature, lasting only a few minutes. In Fig. 6, we show the electron density and intensity variations for the O iv 1399 Å line as a function of time for two regions indicated in Fig. 7. As an indication, the typical 1σ error in log Ne is indicated on each plot, at the positions where the minimum and maximum values for the errors were found. In region A (from 928 to 924 arc sec North), on average we find variations in Ne between consecutive points in the E-W direction comparable to the mean errors in the derived electron density, and in many cases exceeding them. Something similar can be seen in region B (from 919 to 915 arc sec North), where we found variations of up to a factor of three in Ne , while the mean errors were approximately two times smaller. When checking for some kind of periodicity in the areas of our slit where the electron density could be estimated, we found evidence for periods of ∼8 and ∼16 min. For example, in the region between 914.6 to 918.5 arc sec South, we found a period at 8 min and as well as a longer period at 16 min. In this region the electron density ranged between 5 109 and 2 1010 cm−3 . Between 924.4 to 928.3 arc sec South, a period of 8 M.E. Pérez et al.: Temporal variability in the electron density at the solar transition region 1143 Fig. 8. Histograms of electron density values for all datasets. Fig. 7. A CH image in O iv 1399 Å resulting from the average over the whole observational period. The overploted regions A and B represent areas with large electron density variations (see Fig. 6 for further details). min was again visible in Ne (see Fig. 6). The overall distribution of electron density values in this dataset can be seen in Fig. 8. 4.2. Active region (AR) This dataset centered in the active region NOAA 7978, at (630,−200) arc sec on the solar disk started at 07:36:15UT and ended at 08:42:56UT on 10 July 1996. It had an exposure time of 20s and covered an area of approximately 7 × 82 arc sec2 . The variations of the electron density for each position along the slit in the E-W direction is shown in Fig. 9. Here we can see a persistent pattern of variations all along the slit as well as in the E-W direction. The variations along the slit (N–S) are similar in size to those seen along the slit in the CH dataset, i.e. 4–5 arc sec seperated by 10–15 arc sec. In the E-W direction, the variations in Fig. 9 are remarkely similar to those in Fig. 5 despite the fact that it covers a region of ∼7 arc sec. This therefore tends to indicate that these are mostly temporal in nature rather than 1 arc sec size structures, although such small scale structures probably also exist within Fig. 9. The electron density values ranged between log Ne = +0.23 9.56+0.28 −0.25 and log Ne = 11.45−0.18 . In Fig. 9 the grey scales were allowed only to range between log Ne =9.8–11.2 (6.3 109 − 1.6 1011 ), in order to show more clearly the variations in den- sity. The value of the electron density averaged over the total area observed (i.e. 7×82 arc sec2 ) is approximately 10.52+0.11 −0.10 . The distribution of electron density values in this dataset can be seen in Fig. 8, these being significantly larger than those found in the CH. In Fig. 10 (upper panel) we can see a distinctive variation in the electron density corresponding to each position along the slit. These values are the result of averaging over the 1h7min observation period (∼ 7 arc sec, E-W direction). This variation in the electron density along the slit is due to individual features of ∼ 5–10 arc sec size. In Fig. 10 (lower panel) we show five different plots together, corresponding to five consecutive regions of 16 arc sec along the slit, in the N-S direction. In each of these plots we show the electron density, averaged over these 16 arc sec (N-S), for each scan position along the E-W direction. For each of these regions of 7 × 16 arc sec2 we tabulate the average electron density, limited in the E-W direction between 623 to 630 arc sec. The values ranged between log Ne =10.67±0.15 for the region limited in the N-S direction between −233 and −219 arc sec and log Ne =10.35±0.15 between −202 and −188 arc sec. These values are, within the errors (1σ), similar to the previously mentioned average electron density over the total area covered, 10.52+0.11 −0.10 . Moreover, the density variations are similar along the slit and the rastered E-W direction for each of these small regions. No spatial variations smaller than ∼ 3–4 arc sec (∼30 min) are present here. The long time-scale variations present in Fig. 10 (lower and upper panel) are probably due to arc sec scale features passing through during the sit-andstare nature of the dataset and are distinct from the shorter scale variations mentioned above, that can be seen in Fig. 11. These shorter scale variations (in the E–W direction) seem to be related to the similar temporal variations found in the CH dataset, with periods of approximately 8 and 16 min. In Fig. 11 we locate some representative sections along the slit image (plotted as A, B, C & D, see Fig. 12) whose density and line intensity variations in time are shown in more 1144 M.E. Pérez et al.: Temporal variability in the electron density at the solar transition region Fig. 9. The electron density (in cm−3 ) as derived from O iv for the AR dataset. Saturated areas are those where the electron density is higher than 1.6 1011 cm−3 and black areas are those with values lower than 6.3 109 cm−3 . Fig. 10. Top panel: Electron density (with 1σ error bars) values corresponding to each position along the slit. These values are the result of an average over the 1h7min observation period (∼ 7 arc sec, E-W direction). Bottom panel. Density values corresponding to each scan position along the E-W direction for the AR dataset. These values are the result of an average over 16 arc sec, with the exact position range in the N-S direction specified in the top, in each of the small frames. Dashed line: electron density averaged over the total covered area of 7×82 arc sec2 , log Ne = 10.52+0.11 −0.10 . detail. The intensity values plotted in this figure correspond to the O iv 1399 Å line. As an indication, the typical 1σ error in log Ne is indicated on each plot, at the positions where the minimum and maximum values for the errors were found. Regions A (from −173 to −177 arc sec South) and B (from −215 to −219 arc sec South) correspond to the higher density areas in Fig. 9. In region A we found the biggest density variations between ∼ 625 and ∼ 625.7 arc sec West, in particular at position −175 arc sec South (dotted line) we found a logarithmic varia+0.18 tion in density of between 10.54+0.13 −0.16 and 11.02−0.15 , while at M.E. Pérez et al.: Temporal variability in the electron density at the solar transition region 1145 Fig. 11. Some of the electron density and intensity values for the AR dataset. These are the values corresponding to locations A, B, C and D in the slit image represented in Fig. 12. The exact range in arc sec is given here in brackets for each of these locations. The typical 1σ error in log Ne is indicated on each plot, at the positions where the minimum and maximum values for the errors were found. position −176 arc sec South (dashed line) there was a variation +0.12 of between 10.46+0.13 −0.15 and 10.89−0.12 . Between ∼ 626.7 and ∼ 627.5 arc sec West, the variations were between 10.92+0.12 −0.11 and 11.44+0.30 −0.22 at −173 arc sec South (continuous line). In region B, between ∼ 623.6 and ∼ 624.5 arc sec West, we found a logarithmic variation in density of between 10.94+0.08 −0.08 and +0.23 11.45−0.18 at −216 (dotted line) and −217 (dashed line) arc sec South. For this same area, between ∼ 626.8 and ∼ 627.5 arc sec West, this variation was between the logaritmic values +0.06 10.34+0.07 −0.08 and 10.76−0.06 at −215 arc sec South (continuous line). On average we found a variation of a factor of 1.5 in Ne between consecutive positions E-W, that is over 0.44 arc sec, while the mean errors in the electron density were a factor of two less. In region C (from −227 to −230 arc sec South), on average we found a variation of a factor of 1.3 between consecutive positions E-W. For instance, we found from −227 to −230 arc sec South variations of a factor of two in Ne between ∼ 628.6 and ∼ 629.7 arc sec West, i.e. within 1 arc sec, thus suggesting that these are temporal in nature. In region D (from −242 to −246 arc sec South), we find three areas in the E-W direction with variations in the electron density greater than a factor of two; namely between ∼ 624.2 and ∼ 624.8 arc sec West, between ∼ 626.4 and ∼ 627.1 arc sec West, and between ∼ 628.5 and ∼ 629.2 arc sec West. Again, these are temporal in nature due to the small area covered in the E-W direction. When checking for some kind of periodicity we found that, while for some regions along the slit there was no appreciable periodicity, for others there was evidence for approximately 0.8, 1.1 and 1.6 arc sec periodicities which corresponds to ∼8, ∼11 and ∼16 min period. Our binning on the E-W direction was 4 min which corresponds to ∼0.4 arc sec. The longer periods appear mainly in the northern half of the image, that is the less intense part of the AR, although the electron densities are higher in this region. These were in areas of five arc sec (the running mean for this analysis) at around −195 and −177 arc sec South. From −221 to −230.5 arc sec South there was evidence for periodicities of 0.8 arc sec (8 min) and 1.1 arc sec (11 min) in the density, that extended to approximately −245 arc sec (see also Fig. 11). 1146 M.E. Pérez et al.: Temporal variability in the electron density at the solar transition region ered an area of ∼ 10 × 85 arc sec2 . The variations of the electron density values for each position is shown in Fig. 14. The allowed range of values for the grey scale is the same as in Fig. 13, log Ne =9.56–10.5. The density values ranged between +0.23 log Ne =9.56+0.28 −0.28 and log Ne =10.90−0.22 . The average electron density for the whole image was 10.06+0.25 −0.25 . This average was calculated only for the ∼32% density values larger than the low density limit for this dataset. In this dataset we found periodicities of ∼ 1.8 and ∼ 2.5 arc sec (∼12 & 16 min period) around 9 arc sec North and −20 arc sec South. A period of ∼ 2 arc sec was present in Ne between −31 and −35 arc sec South. Our binning on the E-W direction was 4 min which corresponds to ∼0.6 arc sec. Again, some of the variations in Figs. 13 & 14 are temporal in nature due to their sub arc sec variability while others could be spatial in origin. For both datasets (see Figs. 13 & 14) there appears to be evidence of super-granular cells, with the increases in electron density occurring along the network boundaries. For example in Fig. 13 there is one from –40 arc sec to 0 arc sec in the NS direction and another from 0–55 arc sec again in the N-S direction. 5. Discussion Fig. 12. O iv 1399 Å corresponding to the AR dataset. This image results from an average over the whole observational period. The overplotted regions A, B, C and D in this slit image represent areas with peculiar density variations (see Fig. 11 for further details). 4.3. ‘Quiet’ Sun (QS1 & QS2) Like the CH dataset, the low signal-to-noise in the ‘quiet’ Sun datasets make it very difficult to obtain a reliable estimate for some positions in our raster/temporal image. In fact, a large fraction of the area is in the low density limit. From Fig. 8, we can see that the overall distribution of electron densities is intermediate between that of the CH and AR. Both datasets were centered at (3, 0) arc sec, i.e. disk center. The first, QS1, started at 17:09:42UT and ended at 18:16:42UT on 10 July ’96. This dataset covered an area of ∼10 × 112 arc sec2 . The corresponding variations of the electron density values for each position along the slit with position along the E-W direction is shown in Fig. 13. The density values ranged +0.53 between log Ne =9.56+0.28 −0.25 and log Ne =11.09−0.33 . The average electron density for the whole image was 10.05+0.22 −0.23 . This average was calculated only for the 30% density values over the minimum density value, log Ne =9.56. In this dataset we found periodicities of ∼ 2 and ∼ 2.5 arc sec (∼13 & 16 min period) around 55 arc sec North. We found a period of ∼1.5 arc sec (∼10 min) in Ne in an area of five arc sec (running mean) around 49 arc sec North. Our binning on the E-W direction was 4 min which corresponds to ∼ 0.6 arc sec. The second dataset, that we called QS2, started at 22:32:46UT and ended at 00:00:09UT on 10 July ’96, and cov- Datasets taken in a coronal hole, a‘quiet’ Sun region at disk center plus an active region show variations in the electron density in the transition region over time periods of a few minutes. Such variations can be as large as a factor of two in ∼5 minutes, but unfortunately the time resolution of our datasets do not permit us to detect faster variability. Electron density enhancements due to spatial structures of 5–10 arc sec are also clearly visible. In each dataset there are a few locations where the electron density showed periodicities of between 8 and 16 min. There is a remarkable agreement between the scale and temporal variability in the coronal hole and ‘quiet’ Sun datasets, in agreement with a bright point study by Habbal et al. (1990) who found that these were indistinguishable. The above study also showed that bright point detection had two maxima, one at coronal temperatures and the other at 1–2 105 K, i.e. around the formation temperature of O iv. Numerous studies (e.g. the statistical analyses of the HRTS3 mission by Dere et al. 1983 or SOHO Chae et al., 1998, Pérez et al. 1999), have shown ultraviolet explosive events occurring in a burst-type manner in the solar transition region. Explosive events have been connected to magnetic reconnection occurring on time-scales of minutes over regions with sizes of few arc sec. The distribution of density increases along the network boundaries, as reported in our present work, is consistent with the predominant location of explosive events as already observed by Dere (1991) and recently by Chae et al. (1998). Moreover, these density enhancements are in good agreement with numerical simulations of explosive events by Sarro et al. (1999), who found increases of a factor of two or three at these temperatures. In the CH dataset we calculate a birthrate of 6 10−21 cm−2 s−1 for these density enhancements, in excellent agreement with that M.E. Pérez et al.: Temporal variability in the electron density at the solar transition region 1147 Fig. 13. The O iv electron density variations in the QS1 dataset. Fig. 14. The O iv electron density variations for the QS2 dataset. derived by Dere et al. (1983) for the explosive event birthrate as observed in C iv. Judge et al. (1998) found evidence in support of the ‘ nanoflare ’ picture of coronal heating, that would explain his observations of predominantly downward-propagating compressive waves in the solar transition region. Judge et al. do not rule out other heating mechanisms such as resonant absorption of Alfvén waves as described by Ofman et al. (1998). This mechanism would be consistent as well with a density structure showing filamentary and closely spaced density enhancements up to a factor of two, varying on a time-scale of minutes. Thus, in principle, this could also be the cause of the downward-propagating compressive waves. Nevertheless, other work by Peter & Judge (1999) and Teriaca et al. (1999) recently suggested that nano-flares predominantly occurring around the O vi formation temperature (3 105 K) could account for the redshift observed in the low and middle transition region and for the blue-shift seen in the upper transition region and coronal lines. This idea of nano-flares/explosive events occurring in the high transition region is also in agreement with the present detected electron density enhancements. From this point of view, the larger range of density values detected for the active region could be explained in terms of higher frequency of occurrence and/or energy releases in the active region with respect to the ‘quiet’ Sun or coronal holes. Nevertheless, a preliminary analysis of the line widths variations for the data presented here has not been conclusive. Further numerical work based on this type of model is required. 1148 M.E. Pérez et al.: Temporal variability in the electron density at the solar transition region Acknowledgements. Research at Armagh Observatory is grant-aided by the Department of Education for N. Ireland while partial support for software and hardware is provided by the STARLINK Project which is funded by the UK PPARC. This work was supported by PPARC grants GR/K43315 and GR/L57449. We would like to thank the SUMER and EIT teams at Goddard Space Flight Center for their help in obtaining the data. The SUMER project is financially supported by DLR, CNES, NASA, and PRODEX. SUMER is part of SOHO, the Solar and Heliospheric Observatory of ESA and NASA. MEP is supported via a studentship from Armagh Observatory. References Brage T., Judge P.G., Brekke P., 1996, ApJ 464, 1030 Chae J., Wang H., Lee C., Goode P.R., Schühle U., 1998, ApJ 497, L109 Cheng C.-C., 1980, ApJ 238, 743 Dankworth W., Trefftz E., 1978, A&A 65, 93 Dere K.P., Bartoe J.-D.F., Brueckner G.E., 1983, Sol. Phys. 123, 41 Dere K.P., 1991, J. Geophys. Res. 96, 9399 Doschek G.A., Feldman U., Laming J.M., et al., 1998, ApJ 507, 991 Dufton P.L., 1977, Comp. Phys. Comm. 13, 25 Dufton P.L., Berrington K.A., Burke P.G., Kingston A.E., 1978, A&A 62, 111 Dwivedi B.N., Gupta A.K., 1991, Sol. Phys. 138, 283 Feldman U., Doschek G.A., 1978, A&A 65, 215 Foster V.J., Keenan F.P., Reid R.H.G., 1996, A&A 308, 1009 Foster V.J., Reid R.H.G., Keenan F.P., 1997, MNRAS 288, 973 Griffiths N.W., Fisher G.H., Woods D.T., Siegmund O.H., 1999, ApJ 512, 992 Habbal S.R., Dowdy J.F., Withbore G.L., 1990, ApJ 352, 333 Haughan S.V.H., 1997, The Component Fitting System (CFIT) for IDL. CDS Software no. 47 Hayes M.A., Shine R.A., 1987, ApJ 312, 943 Horne J.H., Baliunas S.L., 1986, ApJ 302, 757 Judge P.G., Hansteen V., Wikstol O., et al., 1998, ApJ 502, 981 Landi E., Landini M., Dere K.P., Young P.R., Mason H.E., 1999, A&AS 135, 339 Mazzotta P., Mazzitelli G., Colafrancesco S., Vittorio N., 1998, A&AS 133, 403 Nakagawa Y., Stenflo J.O., 1979, A&A 72, 67 Nussbaumer H., Storey P.J., 1982, A&A 115, 205 Ofman L., Klimchuk J.A., Davila J.M., 1998, ApJ 493, 474 O’Shea E., Doyle J.G., Keenan F.P., 1998, A&A 338, 1102 Pérez M.E., Doyle J.G., Erdélyi R., Sarro L.M., 1999, A&A 342, 279 Peter H., Judge P.G., 1999, ApJ 522, 1148 Safronova M.S., Johnson W.R., Safronova U.I., 1996, Phys. Rev. A 54, 2850 Sarro L.M., Erdélyi R., Doyle J.G., Pérez M.E., 1999, A&A (in press) Seaton M.J., 1964, MNRAS 127, 191 Siegmund O.H.W., Gummin M.A., Stock J.M., et al., 1994, Proc. SPIE 2280, 89 Spadaro D., Leto P., Antiochos S.K., 1994, ApJ 427, 453 Teriaca L., Banerjee D., Doyle, J.G., 1999, A&A 349, 636 Wikstol O., Judge P.G., Haansten V.H., 1997, ApJ 483, 972 Wilhelm K., Lemaire P., Curdt W., et al., 1997, Sol. Phys. 170, 75 Zhang H.L., Graziani M., Pradhan A.K., 1994, A&A 283, 319