The studied sample consists of 23 events. Gross spectral and spatial characteristics are summarized in Tables 1 (click here) to 3 (click here), and the individual events are described in the Appendix.
Table 1: The U burst sample
The bursts extend from decimetric to metric waves, with turnover frequencies ranging from 100 to 380 MHz and starting frequencies from 170 to 460 MHz (Table 1 (click here)). The average total bandwidth is about half the centre frequency (Table 2 (click here)). These are not statistical properties of type U bursts observed by the Tremsdorf spectrograph, but are largely due to the requirement that the bursts should be observable by the Nançay radioheliograph (NRH). Two cases have the spectrum and the spatial configuration of type U(N) bursts.
The vast majority of the bursts is located near or above the limb in
projection onto the plane of the sky. This is at least partly due to
foreshortening and to the possibility to observe behind-limb events. If
isotropically emitting radio sources are placed uniformly in longitude
at an altitude of e.g. 
/3
above the photosphere, 63% of the observed sources will appear projected
above the limb. This is consistent with the distribution of the ascending
branch source A of our type U sample: 38% are at projected distances below
1 
, 62% at distances between 1 and 1.4 
(multiple
bursts from the
same day and the same site have been counted as a single event). The
distribution of our small sample is therefore consistent with the assumption
of isotropic emission from uniformly distributed type U burst sources on the
Sun, in agreement with earlier conclusions based on the location of H
flares associated with such radio bursts (Fokker 1970).
The last two columns of Table 1 (click here) list the H
and soft
X-ray emissions which accompany the U bursts. It is clear from a comparison
with Table 3 (click here) that even when the type U burst is associated with
active regions on the disk it does not necessarily occur with an
H
flare (SGD Comprehensive Reports; e.g. 4 Nov. and 25 Jan. 1994,
13 Nov. 1992).
Hard X-ray observations are not available during most of the events of interest. As a proxy (cf. Dennis & Zarro 1993), the time intervals where the derivative of the soft X-ray flux is positive are given in the last column of Table 1 (click here) (GOES satellites; data kindly provided by SDAC at Goddard Space Flight Center). The majority of the U/III burst groups (14/23) occurs together with weak X-ray brightenings, but a significant number (9/23) has no or probably no counterpart detected by GOES. Three of them occur on the disk. The 13 Nov. 1992 burst shows that this is at least in part a problem of sensitivity, since Yohkoh-SXT detects a jet in time coincidence with the U-burst (Aurass et al. 1994). Our selection of simple and spectrally well-defined events likely favours the tendency of weak flare association.
Most of the presented observations are consistent with type U bursts being
emitted by electron beams which are guided along large-scale
(
1 
) closed magnetic loop structures as suggested by Labrum
& Stewart (1970):
The most clear-cut examples of textbook type U behaviour are:
Table 2: Spectral properties of type U bursts
Table 3: Spatial properties of type U burst sources
During many events additional radio sources appear. In general the type U
bursts are observed within groups of type III or type J bursts. Only six of
the studied events are not accompanied by type III bursts between 10 s before
and 10 s after the U burst. Often (e.g.
Figs. 16 (click here), 17 (click here) and 19 (click here)) the sources of
type III bursts and of the type U bursts (A) are spatially separated even if
the bursts occurred in time coincidence (cf. also Labrum &
Stewart 1970). In those cases where type III bursts and U bursts occurred
within a few seconds, the type III sources were usually (11 out of 12 cases)
closer to the ascending branch than to the descending branch of the U burst:
the average distance in the plane of the sky
to the ascending U sources was found to be (
)
, while the distance to the descending branch source was
(
) 
, where 
is the distance between the two
U burst sources. Other events (e.g. Figs. 21 (click here), 22 (click here)
and 24 (click here)) are so complex that the association of
spectral and spatial features must be inferred by analogy with the
clear-cut cases. In some of the spectral records significantly different
starting frequencies of quasi-simultaneous type III bursts and U bursts are
seen (Figs. 9 (click here), 19 (click here)).
These facts suggest that electron beams are injected in the low corona into structures of different magnetic connectivity. This extends conclusions inferred from earlier imaging observations of type III bursts (Raoult & Pick 1980; Lantos et al. 1984; Pick & van den Oord 1990).
In a number of events (8 out of 23) a new burst was observed at the site of the ascending branch (A) slightly before or simultaneously with the brightening of the descending branch at site B. There are cases where this renewed brightening is stronger than the descending U branch. Two events may be used to illustrate this situation: In Fig. 4 (click here) the renewed brightening (11:39:24 UT) is identified as a negatively drifting (type III) burst. This shows repetitive production of electron beams. On the other hand in Fig. 24 (click here) the renewed brightening at site A appears while only a descending U branch is seen in the dynamic spectrum. The fact that this brightening contributes - at least at 164 MHz - twice the flux density of the simultaneous burst at the remote site (cf. Table 2 (click here)) conflicts with upward travelling beams from the low corona, and seems to require downward reflection of beams by a scattering process near the loop apex (Karlicky et al. 1996). In the other cases the renewed brightening is too weak to allow for a spectral identification.
Figure 1: The event 28 January 1994 11:29 UT. The radio source centroids
of the U burst are superposed as crosses onto an enlargement of a Yohkoh SXT
full disk image (see Fig. 12 (click here)).
A - the ascending branch; B - the descending branch. F - ``fundamental
mode" emission (at 164 MHz or 236.6 MHz); H - ``harmonic mode" emission (at
236.6 MHz)
Figure 2: Composite picture of the source positions of five type U
bursts observed during 3h 40m on 09 August 1990. The ascending branch is given
by the solid crosses, the descending branch by the hatched crosses. The
dimension of the crosses gives the half-width of the undeconvolved sources
measured with the east-west and the north-south branch of the NRH
Figure 3: 25 October 1994: Position and diameter of the type U burst sources
at 435 MHz are superposed on a partial frame Yohkoh SXT image (see text).
Lower right: inserted full disk SXT image showing the pre-eruptive active
region and the partial frame position at 10:07:39 UT. Compare also
Fig. 6 (click here)
Figures 1 (click here) and 12 (click here) (Sect. B.3 (click here)) show
the unique case in our sample where a type U burst seems to occur as a pair of
harmonically related emission features. The low-frequency lane consists of an
ascending branch below 170 MHz, followed by a diffuse descending branch
extending up to 300 MHz. The high-frequency lane (220-280 MHz) is much
fainter but gives clear evidence for ascending and descending branches. The
top frequency ratio of both lanes is 
, i.e. nearer to a 3:2 than
to the 2:1 ratio. There is no spectral indication of emission at frequencies
below 100 MHz which could be identified as the ``true fundamental" of this
event. Therefore we refer to the low frequency branch as fundamental (F) and
to the high frequency branch as harmonic (H) in the following text and tables,
but these are to be understood as purely descriptive terms.
At 236 MHz the sources of the ascending and descending H branch and the
descending F branch occupy three different positions which project upon an
arcade of loops seen by Yohkoh-SXT (bottom part of Fig. 12 (click here)).
Figure 1 (click here) gives an enlargement of the corresponding part of the
SXT image together with the centroid positions of the radio sources. The
arcade extends northward behind the limb and is visible in the SXT images
already several days before. It is dominated by a very bright loop in the
foreground. The two highest type U sources correspond with the spectral
F lanes at 164 MHz (
, 
in
Fig. 1 (click here)). The 236 MHz F source is situated east of the 164 MHz
F sources, at lower height. The H sources at 236 MHz
(
and 
) are situated lower than the 164 MHz F sources
and have a greater mutual distance. To summarize:
source !).
Given the contradiction of the observations with the F/H emission model, including ducting, we conclude that both lanes in the burst pattern are rather due to a simultaneous injection of different beam ensembles in two nested loops.
Figure 2 (click here) collects in a single graph the 164 MHz positions of
five U bursts observed during 3h 40min on 09 August 1990 (cf.
Figs. 23 (click here)-27 (click here)) superposed on the Meudon
H
spectroheliogram. The spectra of the bursts are quite different, but
despite their temporal separation the 164 MHz emissions have the same source
configuration. A comparison of the time histories of the three events in
Figs. 24 (click here), 26 (click here) and 27 (click here) even suggests
some degree of similarity in the succession of beam injections during different
events: two bursts near the ascending source site A are followed by the
brightening of the descending source at B and a simultaneous renewed
brightening at A. The observed similarities reveal stable beam injection
conditions and a stable magnetic structure which guides the beams.
The flare of 25 October 1994 (Fig. 3 (click here); cf. Figs. 6 (click here)-9 (click here)) illustrates a dynamic situation. It is related with evolving coronal structures partly seen by Yohkoh SXT (Manoharan et al. 1996; Aurass et al. 1996). Beginning with the soft X-ray maximum of the flare until several hours later type U bursts repeat at systematically varying sites. Figure 3 (click here) shows the 435 MHz radio image of the first burst of the event sequence (see also Fig. 6 (click here)). It is superposed onto the almost simultaneously recorded Yohkoh SXT partial frame image of a system of sheared flaring loops in AR 7792 (S10 W10) which are under ongoing relaxation. The U burst is probably associated with newly formed magnetic structures within the active region. Later in the U burst sequence, the site of beam injection changes and the source sites - seen at lower frequencies - are situated in loops bridging another part of the magnetic neutral line far outside the flaring region and having much larger span between sites A and B (cf. Figs. 7 (click here) to 9 (click here)). Comparing the bursts of Figs. 8 (click here) and 9 (click here) at the same frequency (236 MHz), we note again a remarkable repetition of features with a time lag of 1 h: a starting pair of a type III burst and an ascending type U branch is followed by the descending type U branch and a renewed ascending branch.
The two U burst sequences illustrate that the emission may occur in stable as well as in highly dynamic coronal structures. In both cases there is a hint to repetitive sequences of homologous beam injection. Evidence for repetitive U burst emission over several minutes had been given by Aschwanden et al. (1992). Such stable patterns up to several hours appear in contradiction with stochastic or chaotic models of energy release and transport (e.g. Vlahos 1993) and deserve closer investigation.
Figure 18 (click here) (Sect. B.8 (click here)) gives a type U burst with a smooth transition to a type V continuum. Despite this spectral peculiarity, the basic type U pattern of two separate sources A and B is present. The descending type U branch and the continuum are at the same position (B; dashed cross). The source is much bigger than in other events because the emission occurs on top of a complex source which persisted from a preceding burst.
Comparison with the Yohkoh-SXT image suggests that the radio sources are associated with a loop system, and that the site B is in one of the legs. The accuracy of the position measurement is less than in the other events, due to the geometry of the NRH and to residual uncertainties of the ionospheric corrections. Irrespective of this restriction the location of the type V source at a place different from that of the beam injection is similar to those observed by Dulk et al. (1980). In the present case the observations suggest the continuum emission to come from one side of the loops. Prolonged emission after the descending U branch is also found in Figs. 20 (click here) and 26 (click here), although with shorter durations than during the type V burst. In all three cases the emission comes from the vicinity of the preceding descending U branch.
