5. Discussion and conclusions
We have analyzed the light curves of the individual pixels of 15 UVSP data sets and specifically looked for evidence of chromospheric oscillations. Only in a limited number of pixels (29%), did we find evidence of periodicities. For those pixels, which show periodicities, the maximum power is predominantly found between 2 and 5 mHz with a distinct maximum around 3-3.5 mHz. The highest count rates are found for frequencies of 2 mHz and are associated with flare-like brightenings.
The distribution of peak powers shown in Fig. 8a has a strong resemblance with the power distribution of the photospheric five minute oscillations as derived from Doppler shifts of C I and Fe I lines (see e.g. Fig. 1 in Rutten, 1995). This is a bit surprising because there is no a priori reason to expect that when at different spatial locations the peak power in the UVSP bandpass is measured, it would correspond to such a distribution. Alternatively one can argue that the photospheric five minute oscillations are not a stationary phenomenon but consist of distinct wave packets of finite duration. The power spectra of C I and Fe I lines can then be interpreted as the probability distributions for finding power at a given frequency. The measurements described in this paper should then be interpreted as a random sampling of this probability distribution. The distribution shown in Fig. 8a is clearly not representative of a distribution one would expect for the chromospheric three minute oscillations as observed in Ca II. Such a distribution is broad (2-10 mHz) and centered around 5.5 mHz.
A key question is where in the chromosphere the continuum in the UVSP bandpass has its origin. In the Vernazza et al. (1981) models it is formed in the lower chromosphere at . From a black body with this temperature, and an area of , the number of photons is photons/s/pixel in the UVSP bandpass. If, on the contrary, the emission is optically thin we can use the Vernazza et al. chromospheric models to calculate the emission measure of the plasma in the temperature range 11,000-39,000 K. Multiplying this number with the area of an UVSP pixel gives for a pixel an emission measure of . From the emissivity curve (Fig. 2) we estimate the number of photons in the optically thin interpretation to be photons/s/pixel which is a factor five less than the number of photons in the optically thick interpretation. We checked that the contribution by any overlying coronal plasma is negligible. This implies that the bulk of the observed emission is optically thick emission from the lower chromosphere with a small () optically thin contribution from the overlying chromospheric layers. We note however that if the true chromospheric densities would be a factor three higher than in the Vernazza et al models, then the optically thin component in the temperature range 11,000 - 39,000 K would dominate the observed emission.
Recently Carlsson and Stein presented in a number of papers (1992, 1994, 1995, 1997) a very dynamical picture of the chromosphere. These authors demonstrated that the formation of Ca II H&K grains is related to the presence of shocks. Because in a shock the temperature and the density increase, it is interesting to see whether the oscillations observed in the UVSP bandpass are related to such shocks. In Carlsson and Stein (1994) it is shown that the chromosphere has a multi-temperature structure with locally very high (shocks) and very low temperatures. The mean temperature is however relatively low (4000-5000 K). The high-temperature spikes in the shocks raise the observed brightness temperature above the true mean temperature. In Fig. 5 of Carlsson and Stein (1995, or Fig. 11 of Rutten, 1995) these authors do not show the shock temperatures above K but higher up in the chromosphere the shocks will further steepen and reach temperatures at which the emissivity curve for the UVSP continuum region, shown in Fig. 2, has its maximum. In their 1997 paper Carlsson and Stein demonstrate that the Ca II grains are associated with either isolated or merging shocks. Apart of the relevance of this work for grain formation, the simulations by Carlsson and Stein also provide information about wave propagation in the chromosphere. Therefore we discuss in the following various aspects of these simulations. Carlsson and Stein start with an atmosphere in radiative equilibrium which is perturbed by a piston at the base. The velocity profile of the piston is chosen so that (slightly higher up in the atmosphere) the observed velocity profile of an Fe I line is mimicked. Most of the photospheric (piston) power is in the frequency range 2-6 mHz. Carlsson and Stein distinguish three frequency ranges: 0-4.7 mHz (low:L), 4.7-7.1 mHz (medium:M) and above 7.1 mHz (high:H). Simulations with the piston only driven in the L range do not produce shocks and grains. The reason is that these frequencies are below the acoustic cut-off frequency so that the waves are evanescent. Simulations with the piston only driven in the M or H frequency band do produce shocks and grains, because the waves can propagate and steepen, but not in agreement with the observed grain patterns. A good correspondence with the observed patterns is found when the piston is driven over the full frequency range. In that case the following picture arises: only waves with frequencies 4.7 mHz can propagate and steepen into shock. The largest photospheric power at these frequencies is found close to the cut-off frequency so that most shock associated grains have this frequency. It is the combination of a broad spectrum of waves, the acoustic cut-off frequency and the photospheric power distribution which results in the specific time behaviour of the grains and shocks. Our results in Fig. 8a do not show a maximum for the observed peak powers near 4.7 mHz as one would expect from the Carlsson and Stein simulations. On the contrary, the distribution is exactly located in the frequency range where the waves are not propagating. This strongly suggests that we are observing evanescent waves driven by the photospheric five minute oscillations.
The above explanation does not hold for the dataset showing an excessively large number of pixels with peak power at 2 mHz (Fig. 8c) because these are clearly related to flare-like variations. Porter et al. (1995) studied the correlation between (micro-)flares observed with the UVSP (C IV at K) and the 3.5-5.5 keV emission as detected with the Hard X-ray Imaging Spectrometer (HXIS) on board the SMM. These authors find that small flare-like events detected in C IV have impulsive soft X-ray counterparts. It is possible that in our case (coronal) flare activity results in an increase of the emission measure at higher temperatures, say , so that the increase of the emission in the UVSP bandpass is related to optically thin emission from the transition region and the lower corona (see Fig. 2). This, and the origin of the 5 min. UVSP oscillations, can be tested by simultaneous observations of selected regions with the SUMER and CDS instruments onboard SOHO, thus covering spectral lines emitted from plasma temperatures ranging from 15,000 to 250,000K.
It is, however, also possible that the high count rates are still flare-related but have their origin in the UV continuum formed near the temperature minimum. Orwig and Woodgate (1987) found a temporal correlation between hard X-ray bursts (), ultraviolet line emission (O V at 1371 Å) and the UV continuum near 1600Å. This led Machado and Mauas (1987) to propose a model in which irradiation by flare-related XUV emission (mainly C IV at 1549Å) results in photo-ionization of Si I. This photo-ionization affects the continuum between 1350-1680Å which is formed above the temperature minimum (Doyle and Phillips, 1992).
Both of the above presented explanations (TR/coronal emission or XUV irradiation) are based on coronal energy release and do not explain the 2 mHz (500 s.) periodicity. It is however simpler to relate a time scale of 500 s. to the corona than to the chromosphere. The 500 s. period could be related to the Alfvén bounce time in a short coronal loop, or to the periodicity of the sausage mode, so that a modulation of the energy release is present. The combination of 2 mHz oscillations and high count rates is only found in one of the datasets and is therefore a rather specific case.
The relative absence of significant power peaks above 5 mHz is striking. Of course there is power above 5 mHz in our power spectra but apparently the bulk of the power is in evanescent waves. This can be explained if, apparently, evanescent waves oscillate more in phase, leading to a more (spatially) coherent oscillation and a `clearer' modulation of the lightcurve. Above 5 mHz the waves are propagating and their phase relation is lost so that, in the field of view of a pixel, the compressions and rarefactions do not result in a strong modulation of the light curve.
Finally we note that our statistical treatment of the power spectra is very general and can equally be applied to SUMER and CDS data from SOHO.
© European Southern Observatory (ESO) 1997
Online publication: April 8, 1998