1 Introduction

Plumes are an intrinsic feature of polar coronal holes. They appear in white-light coronagraph and eclipse images as ray-like structures, extending up to many solar radii above the limb of the Sun (see e.g. Van de Hulst 1950; Saito 1965). From their morphology they appear to be tracing the open field lines emerging from the coronal holes. The characteristic comet-like shape at the base of the plumes is best seen in the polar holes where the geometry is most favourable. Here, they can be observed in extreme ultraviolet (EUV) images and easily identified from their shape and enhanced intensity relative to the coronal hole (e.g. Ahmad & Withbroe 1977; Widing & Feldman 1992; Walker et al. 1993). They are also seen at X-ray wavelengths (Ahmad & Webb 1978). By tracing plume structures from the lower corona up to 15 $R_\odot$, DeForest et al. (1997, 2001) have argued that the plumes seen in white-light and in the EUV are different parts of the same structure. In this paper, we only discuss the base of the plumes as observed in EUV, projected against the solar disk or very close ( $R < 1.1 \times R_{\odot}$) to the limb.

Although more difficult to identify at low latitude where the plume may be viewed more or less "end-on'', masking the plume-like shape, there have been some tentative reports of possible identifications in low-latitude holes viewed near the limb (Wang & Sheeley 1995; Woo 1996). Del Zanna & Bromage (1999b) reported the first detection by SOHO of a low-latitude plume visible on the disk, seen in the Elephant's Trunk equatorial coronal hole of 1996. This plume was identified by its spectroscopic characteristics, rather than its morphology. Here, we present observations of polar plumes, showing that they exhibit similar characteristics to the low-latitude plume seen in the Elephant's Trunk.

Before the advent of SOHO, it was not possible to distinguish clearly between the plasma characteristics of plumes and those of the surrounding coronal hole (inter-plume) regions because of the limitations of the observations. Density and temperature measurements in plumes were mostly based on coronagraph or Skylab data from just above the limb (e.g. Ahmad & Webb 1978). Using white light coronagraph observations and applying the Van de Hulst (1950) technique, the plume densities were estimated to be 3-5 times higher than the surrounding coronal hole. Widing & Feldman (1992) used a Mg VIII density diagnostic ratio on Skylab data but were not able to obtain consistent density values for plumes, due to insufficient spectral resolution. Temperatures were estimated from intensity gradients of EUV emission lines, assuming hydrostatic equilibrium and ionization equilibrium (Widing & Feldman 1992). Ahmad & Webb (1978) derived a mean temperature of $8 \times 10^5$ K (at 1.07 $R_\odot$) for some polar plumes observed by the Skylab S-054 experiment. At this time there was no concensus on the subject of the temperature of plumes; it was not clear whether they were cooler or hotter than the inter-plume regions. The confusion may be due in part to the fact that plumes are sometimes seen to have a very bright base, where the density and temperature are enhanced relative to those in the ambient coronal hole and the main body of the plume (Bromage et al. 1997, 2000; Young et al. 1999). In other cases (including the examples presented here), the plumes appear more diffuse overall, with a near-isothermal temperature which is lower than that in the surrounding coronal hole (e.g. Del Zanna & Bromage 1999b). This variation in the nature of the plumes probably represents two different stages in the evolution of the structure, with a magnetic reconnection event occurring initially at the base, increasing the density and temperature there. Subsequently a more diffuse structure develops which may fade over a timescale of about a day. Lamy et al. 1997) have shown a tendency for plumes to recur at the same location after disappearing for a time.

It is now firmly established that coronal holes are the source of the fast solar wind, as already pointed out by Noci (1973) and Krieger et al. (1973). The question now is how it is generated and whether plumes might be the source.

The SOHO spectroscopic instruments (CDS, SUMER and UVCS) have provided improved spatial and spectral resolution, allowing more detailed analyses to be made in the plume/interplume regions. Hassler et al. (1999) reported finding blue-shifts in SUMER (Solar Ultraviolet Measurements of Emitted Radiation) spectra from the coronal hole network junctions, while Giordano et al. (2000) and Patsourakos & Vial (2000) have found evidence from UVCS (Ultraviolet Coronagraph Spectrometer) and SUMER observations, respectively, that the fast solar wind is accelerated in the interplume regions. Similar results were recently found by Teriaca et al. (2002), also using UVCS and SUMER observations.

Solar wind models depend heavily on the density and temperature of the source region. The uncertainty in these quantities obtained from the early observations of plumes is reflected in the variety of different theoretical solar wind-plume models which have been proposed (e.g. Wang 1994; Velli et al. 1994; Del Zanna et al. 1997; Casalbuoni et al. 1999). More direct measurements of electron densities have now become possible, using spectral line ratios from SOHO data. Densities derived in this way are more accurate, although it should be kept in mind that these are averaged values (for the emitting plasma along the line of sight).

Wilhelm et al. (1998) presented off-limb (in the range 1.03-1.6 $R_\odot$) SUMER spectroscopic observations of plume and interplume regions. Densities of both plume and interplume regions were estimated using Si VIII line ratios, and found to start at $N\mbox{$\rm _{e}$ } \simeq 1 \times 10^{8}$ cm^-3 and then decrease with height, with plumes having shallower density gradients. A limitation of this study was that the plume and interplume observations were not simultaneous, but were separated in time by about six months. They also determined plume and interplume temperatures, using a Mg IX line ratio, finding temperature profiles in the plumes which initially appeared constant at around $T=7.8 \times 10^{5}$ K but decreased further out, while in interplume regions the temperature tended to increase steadily with height. SUMER and UVCS off-limb observations (cf. Hassler et al. 1997; Noci et al. 1997) have shown broader spectral line widths in the interplume lanes than in the plumes, indicating lower temperatures in the plumes. Such off-limb spectral observations may suffer from contamination by foreground or background emission, in particular, plume emission behind the coronal hole (interplume) regions can affect measurements of densities and temperatures. Similarly, in the "plume'' regions observed off-limb, there may be contamination from "interplume'' plasma, or from other plumes present along the line of sight.

DeForest et al. (1997) used EIT images to estimate plume temperatures. They found them to be in the range $1{-}1.5 \times 10^6$ K, while the interplume temperatures appeared higher (by no more than 30%). Problems with this technique are discussed in detail below (Sect. 4). Del Zanna & Bromage (1999b) found the Elephant's Trunk plume to be cooler than the surrounding coronal hole, with a temperature of about $7.8 \times 10^5$ K.

Another parameter which can be determined spectroscopically is the composition of the coronal plasma. This can provide a link between the fast solar wind and regions which might be its source. Coronal abundances have been found to differ from photospheric values in a way which appears to be related to the first ionization potential (FIP) of the various elements (see the reviews of Feldman 1992, and Raymond et al. 2001). The degree of this "FIP effect'' can vary depending on the coronal structure being observed. Early observations revealed that the slow solar wind had a FIP effect of about 4, with low-FIP ($\leq$10 eV) elements more abundant than the high-FIP ones. On the other hand, the fast solar wind exhibited abundances closer to photospheric values. A recent study (Von Steiger et al. 2000), based on a re-analysis of the Ulysses data has confirmed that the fast solar wind has photospheric abundances, but suggests an average FIP effect of about 2 for the slow solar wind.

The only pre-SOHO measurement of plume abundances is that of Widing & Feldman (1992), who studied the brightest polar plume observed by the Skylab S082A EUV spectroheliograph, in December 1973. They derived one of the largest FIP biases reported (a factor of 10). Since then, it has long been thought that plumes have a large FIP effect. If this is the case, then plumes cannot be the major contributors to the fast solar wind. However, a new analysis of the Skylab data is presented here (Sect. 2.2), which shows the FIP effect to be very small or even absent in the plume.

More recent studies using SOHO-CDS data indicate that plumes exhibit only a small FIP effect (Del Zanna & Bromage 1999a; Del Zanna 1999; Del Zanna & Bromage 1999b; Young et al. 1999). However, problems with the instrument calibration led to uncertainties of about a factor of two in the derived abundances. The CDS calibration has now been substantially improved (see Del Zanna et al. 2001a and Del Zanna 2002) thus allowing a more accurate determination of the FIP effect in plumes.

We have observed a large number of polar plumes with CDS, during solar minimum conditions in 1996, and in 1997. In this paper, we present CDS observations of two "typical'' polar plumes seen on the disk. These were chosen in order to focus attention on the base of plumes, and to avoid problems of contamination associated with off-limb measurements. The principal aims of this paper are:

(a) to identify the characteristic signature of a plume in CDS spectra;
(b) to demonstrate that plumes can also be characterised spectroscopically in terms of their density, temperature and composition;
(c) to show that polar plumes, identified from EIT images, exhibit the same spectral characteristics as the low-latitude plume seen in the Elephant's Trunk coronal hole of 1996;
(d) to examine the spectroscopic diagnostic techniques applied to the CDS spectra in order to determine the above quantities, high-lighting the problems encountered, in particular when deducing element abundances (see Sect. 2); and
(e) to explain how it is that polar plumes (which are cool objects) may be seen in the EIT 195 Å and 284 Å images which generally reveal plasma emitting at a temperature of more than $1.5 \times 10^6$ K. In Sect. 4 we present the first in-flight direct study of the emission that contributes to the EIT filters, based on a comparison between EIT and CDS near-simultaneous observations, and we discuss the validity of the EIT filter ratio estimates of temperature.