arXiv:0904.0878v1 [astro-ph.SR] 6 Apr 2009
Evidence of widespread hot plasma in a non-flaring coronal active region
Dipartimento di Scienze Fisiche & Astronomiche, Universit` a di Palermo, Sezione di Astronomia,
Piazza del Parlamento 1, 90134 Palermo, Italy
Harvard-Smithsonian Center for Astrophysics, Cambridge, MA 02138, USA
James A. Klimchuk
NASA Goddard Space Flight Center, Greenbelt, MD 20771, USA
Royal Observatory of Belgium, 3 Circular Avenue, B-1180 Brussels, Belgium
Nanoflares, short and intense heat pulses within spatially unresolved magnetic
strands, are now considered a leading candidate to solve the coronal heating prob-
lem. However, the frequent occurrence of nanoflares requires that flare-hot plasma be
present in the corona at all times. Its detection has proved elusive until now, in part
because the intensities are predicted to be very faint. Here we report on the analysis
of an active region observed with five filters by Hinode/XRT in November 2006. We
have used the filter ratio method to derive maps of temperature and emission mea-
sure both in soft and hard ratios. These maps are approximate in that the plasma is
assumed to be isothermal along each line-of-sight. Nonetheless, the hardest available
ratio reveals the clear presence of plasma around 10 MK. To obtain more detailed in-
formation about the plasma properties, we have performed Monte Carlo simulations
assuming a variety of non-isothermal emission measure distributions along the lines-
of-sight. We find that the observed filter ratios imply bi-modal distributions consisting
of a strong cool (logT ∼ 6.3−6.5) component and a weaker (few percent) and hotter
(6.6 < logT < 7.2) component. The data are consistent with bi-modal distribu-
tions along all lines of sight, i.e., throughout the active region. We also find that the
1INAF - Osservatorio Astronomico di Palermo “G.S. Vaiana”, Piazza del Parlamento 1, 90134 Palermo, Italy
– 2 –
isothermal temperature inferred from a filter ratio depends sensitively on the precise
temperature of the cool component. A slight shift of this component can cause the
hot component to be obscured in a hard ratio measurement. Consequently, tempera-
ture maps made in hard and soft ratios tend to be anti-correlated. We conclude that
this observation supports the presence of widespread nanoflaring activity in the active
Subject headings: Sun: corona - Sun: X-rays
The solar corona, once thought to be heated in a quasi-steady fashion, has proven to be much
more difficult to explain than early observations suggested. Developements over the last few
years point to a different picture in which coronal loops and perhaps also the diffuse corona are
heated impulsively by nanoflares occurring within unresolved strands (Parker 1988; Cargill 1994;
Cargill & Klimchuk 1997; Klimchuk 2006; Warren et al. 2003; Parenti et al. 2006). Nanoflares
can explain a number of observations that are clearly inconsistent with static models, such as the
density excess and superhydrostatic scale height of ∼ 1 MK loops. Furthermore, the proposed
mechanisms for coronal heating predict that the heating should be impulsive when viewed prop-
erly from the perspective of individual magnetic flux strands (Klimchuk 2006). Despite these
successes, the existence of nanoflares is still under debate. No “smoking gun” has yet been found.
The strongest evidence in support of nanoflares would be the detection of very hot plasma
(Cargill 1995). Hydrodynamic simulations show that impulsively heated strands should reach
temperatures in excess of 5 MK. Much higher temperatures are possible depending on the energy
of the nanoflare. Until now, there has been littleevidence of such very hot plasma outsideof proper
flares. There are likely two reasons for this. First, few studies have explicitly looked for such ex-
treme temperatures during quiescent (nonflaring) conditions. Second, and more importantly, the
intensity of the hot emission is expected to be very weak. The reason is as follows. Tempera-
ture rises abruptly when a nanoflare occurs, but density is much slower to respond. By the time
chromospheric evaporation has raised the density significantly, the coronal plasma has cooled to
temperatures far below the peak value. Low densities combined with rapid initial cooling means
that the hot phase contributes relatively little to the total emission measure of a heating and cooling
event. Hydrodynamic simulations predict that the emission measure of the hottest plasma is 2-3
orders less than that of the dominant coronal plasma (Klimchuk et al. 2008). The emitted radiation
is correspondingly faint. Ionization nonequilibrium effects may diminish the hot component even
more (Reale & Orlando 2008; Bradshaw & Cargill 2006).
– 3 –
Recent observations suggest that very hot plasma may indeed be present at low levels in
nonflaring active regions (Zhitnik et al. 2006; Urnov et al. 2007; Patsourakos & Klimchuk 2008;
McTiernan 2008; Schmelz et al. 2009). We have therefore undertaken a focused investigation to
determineas precisely as possibletheamountof very hotplasmain oneparticularactiveregionand
to evaluate whether it is consistent with nanoflare heating. We use wide-band multi-filter imaging
observations(Reale et al. 2007) from theX-Ray Telescope (XRT, Golub et al. (2007)) on board the
Hinode mission (Kosugi et al. (2007)). These observations have the advantage of a strong signal
compare to spectrometer observations of individual spectral lines. The disadvantage is that each
band is sensitive to a range of plasma temperatures. To distinguish hot from cool plasmas, we use
different filter combinations as described below.
The major problem is to screen out the dominant cooler emitting components which may
inhibit the detection of the minor hot component. Hinode/XRT is equipped with several wide-
band filters, which might fulfill this task. The filters with a band ranging down to relatively low
energies do not allow the detection of the hot plasma, because it is overwhelmed by the much
larger quantities of cooler plasma. At the other extreme, flare filters hardly detect flare-hot plasma
with very low emission measure. There is a third opportunity offered by the wide choice of filters:
filters of intermediate thickness which may be the right compromise between sensitivity level and
2. The observation and preliminary analysis
Through its nine filters in the soft X-ray band the XRT is sensitive to the emission of plasma
in the temperature range 6.1 < log T < 7.5. The CCD camera has a 1 arcsec pixel and
the Full-Width-at-Half-Maximum (FWHM) of the Point Spread Function (PSF) is ≈ 0.8 arcsec
(Golub et al. 2007). We consider the same data as those already illustrated in Reale et al. (2007).
The 512×512 arcsec2field of view includes active region AR10923 observed close to the Sun cen-
ter on 12 November 2006. The filters used were Al poly (F1), C poly (F2), Be thin (F3), Be med
(F4) and Al med (F5), with exposure times of 0.26 s, 0.36 s, 1.44 s, 8.19 s and 16.38 s, respec-
tively. The selected dataset covers one hour, starting at 13:00 UT, and the time interval between
one exposure and the next in the same filter is about five minutes (12 images in each filter). This
active region was monitored by the XRT all along its passage from the east to the west limb. It
flared several times during this period. However, during the selected hour no major flare and no
significant rearrangement of the region morphology occurred (the average of the RMS intensity
deviation of the individual pixels is ≈ 12%), so that we can average over the whole hour. We used
the up-to-date standard XRT software to preprocess the data, including corrections for the read-
out signal, flat-field, CCD bias. The observed images were co-aligned with a cross-correlation
– 4 –
6.06.5 7.07.58.0 8.5
Response p.u.EM x Exp.Time
Fig. 1.— Temperature response function of the 5 filters used in the XRT observation. To compen-
sate for the different exposure times, the responses are multiplied by the respective total exposure
Each XRT filter has a different response to plasma temperature, described by a function
GFj(T), where j indicates the filter order number. Thinner filters F1 and F2 are more sensitive to
the emission of cooler plasma than thicker filters F4 and F5. Images of the same region in different
filters provide information about the temperature of the emitting plasma. If the emitting plasma
is isothermal and optically thin along the line of sight, the ratio between signals in thick and thin
filters is a function of the temperature only. Reale et al. (2007) devised a special filter ratio from a
combination of all available filters (Combined Improved Filter Ratio, CIFR), which optimizes for
the signal-to-noise ratio and reveals a very fine thermal structure of the active region.
The analysis presented here considers also several standard ratios of singlefilters, and we take
an up-to-date filter temperature calibration with a special fine tuning to make the analysis entirely
self-consistent (see Appendix). Fig. 1 shows the temperature response functions of the filters. Five
available filters allow to define 10 different ratios of single filters. We will exclude the ratio F2/F1
from our analysis because the dependence on temperature is double-valued over a large range. We
are then left with 9 relevant ratios. The 6 ratios with F2 and F1 in the denominator, and also CIFR,
are able to diagnose plasmas down to logT ≈ 6, i.e. relatively cool plasma; the two ratios with
F3 in the denominator down to logT ≈ 6.2; and the ratio F4/F5 down to logT ≈ 6.3 − 6.4. An
important point is that the coolest plasma detectable by the last ratio is about twice as hot as the
– 5 –
coolest plasma detectable by the first two ratios. The F4/F5 ratio is a monotonic function of logT
in the range 6.4 ≤ logT ≤ 7.5. We have set minimum acceptance thresholds to compute the
ratios. All pixels with signal below 200 DN/s for F1, 100 DN/s for F2, 30 DN/s for F3, 2 DN/s for
F4 and 10 DN/s for CIFR have been masked out.
3. Analysis and results
3.1. Data analysis
Fig. 2 shows temperature and emission measure maps obtained with CIFR (left column) and
with the hard (F4/F5) filter ratio. To improvethe signal-to-noiseratio, the F4/F5 maps are obtained
after binning over boxes of 4×4 pixels. We have estimated the uncertainties on temperature and
emission measure in each pixel from the photon counts derived from the DN in each filter, with a
conversion factor obtained assuming an average spectrum at central temperature about logT = 6.7
(the conversion factor does not change much in the range 6.5 ≤ logT ≤ 7). Error propagation to
temperature and emission measure has been applied according to Klimchuk & Gary (1995). The
uncertainties are globally consistent with the average root mean square fluctuations of the tem-
perature and emission measure values of the pixels surrounding a given pixel (Reale & Ciaravella
2006). The average errors in temperature and emission measure obtained from CIFR in each pixel
are around 0.004 (∼ 1%) and 0.015 in the log, respectively, in regions of high signal (e.g. region
SH in Fig.4), and 0.008 and 0.03 in regions of low signal (e.g. region HH in Fig.4). The errors
from F4/F5 in each 4×4 pixel box are around 0.05 and 0.2, respectively, in regions of high signal,
and 0.1 and 0.25 in regions of low signal.
The difference in resolution and detail definition is immediately apparent: the softer ratio
maps show much better-defined structuring. While the soft ratio detects plasma mostly in the
range 6.3 < logT < 6.6, the hard ratio appears to detect hotter components to logT > 7. In
particular we find hot plasma at the boundary of dense structures and in an extensive region on the
center left of the active region, where the emission measure is instead low. We also notice that the
temperature maps do not look similar, while the emission measure maps to some extent do, e.g.
we find high emission measure in the same zones of the active region.
Therefore we find different thermal conditions using different filter ratios. In particular, rela-
tively hot regions in the CIFR and other soft filter ratios appear as relatively cool in the hard filter
ratio. This is not trivially expected and deserves further investigation, that we will show later.
Our interest is also to exploit the multi-band imaging observation to obtain quantitative in-
formation about the thermal distribution of the plasma both along the line of sight and across the
– 6 –
Important information comes from the so-called Emission Measure vs Temperature diagram
(EM(T), e.g. Orlando et al. (2000)). This diagram is a histogram of the distributionof the emission
measure in temperature bins. To build this histogram we consider the maps of temperature and
emission measure obtained with a given filter ratio and sum the emission measure of all pixels with
atemperatureinacertainbin. Fig.3showsEM(T)obtainedfromtherelevantfilterratios, including
CIFR; the bin size is the same as the temperature resolution of the filter response functions, i.e.
∆logT = 0.1. The distributions obtained with soft simple filter ratios and with CIFR mostly
overlap: they are strongly peaked at logT ≈ 6.4 and fall by about three decades in a temperature
range 6.2 < logT < 6.7. EM(T) obtained with both the intermediate filter ratios F4/F3 and F5/F3
are similar to those from the soft filters, but differ in several respects: they are slightly shifted to
higher temperatures (peaking at logT ≈ 6.5); they have a somewhat smaller amplitude; and they
decrease with temperature less slowly on the hot side (decreasing by 2.5 decades at logT = 6.8).
EM(T) obtained from the hard filter ratio F4/F5 peaks at the same temperature and with the same
amplitude, but its hot tail extends to logT > 7 still at about a few percent of the peak. The error in
each EM value is negligible in this scale.
Our question is now whether the hot component found with the hard filter ratio is real or
just, for instance, an artifact of higher uncertainties due to the shallow dependence of this ratio on
temperature and to the lower photon statistics.
In order to investigate this issue, we first recall from Fig. 2 that regions of enhanced tem-
perature occur at different locations in the soft and hard filter ratio maps. The maps appear anti-
correlated. Now we identify two 64x64 pixel (16x16 boxes in F4/F5 maps) subregions that are
reasonably homogeneous, as indicated in Fig. 4. The one to the left has enhanced temperatures in
the hard ratio map, and the one to the right has enhanced temperatures in the soft ratio maps. We
refer to these as the hard-hot (HH) and soft-hot (SH) regions, respectively.
The EM(T) distributions obtained separately for the two subregions are shown in Fig. 5. We
immediately see that they are very different from each other. They are also qualitatively different
from the distribution for the whole active region. Their amplitude is much smaller because they
include much less plasma.
EM(T) of the soft-hot region (right panel) is narrow in all filter ratios, including the hard ratio.
None of the distributions has a hot tail, with all dropping to essentially zero by logT = 6.8. The
peaks of the distributions are all clearly at a higher temperature (logT ≈ 6.5 − 6.6) than those of
the whole active region (logT ≈ 6.4 − 6.5).
EM(T) of the hard-hot region (left panel) is different in several aspects. First, EM(T) of the
hard ratio is considerably hotter than all the others: it is quite broad and its peak is at logT ≈ 6.9
with a hot tail extending beyond logT = 7. It appears to be detached from the other distributions,
– 7 –
and its amplitude is smaller by a factor of 50. The EM(T) of the other filter ratios are consistently
narrow and peaked at logT ≈ 6.4−6.5. Thisis slightlycoolerthan thepeaks inthe soft-hotregion.
The small but clear temperature difference is highly significant, as we will show. It explains the
anti-correlation that is evident in the hard and soft temperature maps of Figure 2. The key result is
that in regions where the hard ratio extends to high temperatures (e.g. left frame in Fig. 4) the soft
ratios peak at slightly lower temperatures than in regions where the hard ratio is confined to cooler
temperatures (right frame in Fig. 4).
It is important to remember that the plasma is multi-thermal along each line-of-sight, whereas
our EM(T) distributions were obtained by assuming that it is isothermal. Important questions are
therefore: can we combine the information from our measurements into a coherent picture? Are
filter ratio diagnostics reliable? Do they provide sensible information? Is hot plasma really hot or
are we being fooled by large errors in T associated with the shallow ratio vs T curve of the F4/F5?
3.2. MonteCarlo Simulations
There are multiple nonlinear effects that complicate the derivation of temperature maps and
emission measure distributions. These include the nonuniform thermal distribution along the line
of sight, its variation from pixel to pixel, the differing sensitivities of the filters, and the non-linear
weighting of the thermal components through the filter responses. In order to assess quantitatively
all such effects we perform MonteCarlo simulations including all realistic ingredients.
to the hard-hot and soft-hot regions of Figures 4 and 5, and then to generate EM(T) distributions
exactly as we did for the real data. We take the sub-regions to be 64×64 pixels in size, and we
assume that EM(T) along the different lines of sight within each sub-region are variations on the
same “parent” distribution EMlos(T). The parent distribution has a parameterized form. We vary
one or more of the parameters at each pixel by randomly sampling from a probability distribution
(either normal or log-normal). This gives us an input EM(T) at each pixel from which we compute
intensities corresponding to observations made with the different filters. As a final step, we modify
the intensities with Poisson noise to mimic photon counting statistics.
The EMlos(T) parent distributions are either single or double step functions (top-hat func-
tions). The amplitude (σEM) and central temperature (σ[logT]) of the step function can vary, but its
width (dlogT) is fixed for the entire sub-region simulation.
From the parent EMlos(T) we compute the corresponding emission value for each of the five
filters. We then randomize the emission values, according to Poisson statistics on photon counts.
The DN rate is first integrated on the exposure time to obtain the total DN counts. The DN counts
– 8 –
are then converted to photon counts, which are randomized and then converted back to DN counts
and DN rate. The 5 new DN rates are assigned to a pixel and the whole procedure is repeated
64×64 times to build a 64×64 pixel image.
The resulting image is analyzed as done for the real XRT images, i.e. derive temperature and
emission measure maps with the filter ratio method and build the related EM(T). We apply the
same rebinning, the same constraints and the same acceptance thresholds as those used for the real
data. We obtain an EM(T) distribution for each filter ratio.
We show end results of this procedure in Figs. 6,7,8,9 in a format similar to – and with the
same resolution as – that used in Figs. 3,5. The histogram in each panel is the total input EMlos(T)
for the sub-region, i.e., the modified (randomized) EMlos(T)’s for all the pixels summed together.
The connected points are the EM(T) obtained from the simulated observations of this input. For
the sake of clarity we do not show the EM(T) obtained with CIFR, which is invariably very similar
to those obtained with the soft filter ratios (the histogram is now the total input EMlos(T)).
Fig. 6 shows results for single component parent EMlos(T)’s with four different central tem-
peratures, increasing from logT = 6.2 in the top-left to logT = 6.5 in the bottom-right. Only the
emission measure amplitude and photon counting errors vary within each case. The amplitude is
normally distributed with a Gaussian half width σEM= 30%. Therefore the total input EM(T) has
thesameshape as theparent EMlos(T) in each pixel. TheDN rate levelis set so as to becomparable
to the one observed in the soft filters (F1 and F2) in the left region in Fig. 4.
For logT = 6.2 (upper-left) only the soft filter ratios yield meaningful temperature and EM
values, which closely resemble the input EM. There is a small spread due only to photon and
EM amplitude fluctuations, and not to temperature variations. For logT = 6.3 (upper-right), the
soft filter ratios and one of the medium ones still reproduce the input EMlos(T) quite well. The
other medium filter ratio yields a somewhat broader distribution peaked at lower temperature. Due
to random photon counting errors, the hard filter ratio produces a spurious faint hot component
centered at logT = 6.9. Although this hot component has some resemblance to that observed in
the left panel of Fig. 3, the agreement is in fact not good because the cool components are cooler
than the observed ones. For logT = 6.4 and logT = 6.5 we once again obtain single EM(T)
components in all filter ratios, mostly centered at the center of the parent EMlos(T). The spurious
hot component in the hard ratio disappears because of the improved photon counting statistics
afforded by the hotter input distribution.
Fig. 7 shows the effect of randomizing the central EM(T) temperature. This is done for two
single-component parent EMlos(T)’s at different central temperatures: logT = 6.2 and 6.4 in the
left and right panels, respectively. Pixel-to-pixel variations in the central temperature are taken
from log-normal distributions with a half-widths σ[logT]= 0.1 and 0.05, respectively. The effect
– 9 –
is clearly to broaden the soft EM(T)’s. A spurious hot component appears in the hard filter ratio,
but only for the cooler case, again due to poor photon counting statistics. An important conclusion
is that observed distributions like those in the left panel of Fig. 5 cannot be produced by single
component input distributions: high temperature values in the hard ratio can only occur when the
emitting plasma is all cooler than diagnosed in the soft ratios.
In Fig. 8 two examples show the effect of adding the second hot component to the input
The parent EMlos(T) of the hot component, another step function, is centered at logT = 6.8 and
has a width of dlogT = 0.15 on the left and 0.2 on the right. The input distribution is broadened
further by randomizing the central temperature with σ[logT] = 0.05 and 0.1, respectively. The
figure shows that the presence of the second hot and weaker component naturally leads both to
an increase of the temperature of the EM(T) from the soft filter ratios and to the detection of a
single peaked hot component from the hard filter ratio. The medium filter ratios F5/F3 and F4/F3
both produce a distribution with a peak around logT = 6.4 − 6.5 but with a hot tail extending to
logT ≈ 7. The distributions in the soft and hard filter ratios look very similar to the ones observed
from the hard-hot subregion on the left in Fig. 3. The central temperatures, amplitudes, and widths
all agree well. The hot tails predicted for the medium filter ratios is not observed, however.
hard-hot subregion, at least for the types of input distributions that we have so far explored. In
particular, they produce a hot component in the hard ratio at the same time that they produce a cool
component with sufficiently high temperature (peaking at logT ≈ 6.4) in the soft ratios. These
are important constraints that must both be satisfied. The total emission measure summed over
temperature also agrees well with the observations, i.e., to within 20% for all filters except for F3
We now turn our attention to the soft-hot subregion, whose EM(T) is shown on the right of
Fig. 5. In contrast with the hard-hot subregion, all ofthe filter ratios, includingthe hard ratio, givea
single quite narrow EM(T) distribution peaking at logT ≈ 6.5. The left panel of Fig. 9 shows that
this distribution can be recovered from a narrow input EMlos(T) with the same peak temperature.
However, the right panel shows that it can also be recovered if the input EMlos(T) contains a
weaker hot component. The hot component is not detected even by the hard ratio because the
much stronger cool component is hot enough to dominate the observed signal. This would not be
the case if the cool component were slightly cooler, because the sensitivity of the hard filters is
a steep function of temperature in this range. The hard filters are nearly blind to plasma cooler
than logT = 6.4, and if this plasma were invisible, the hard ratio would be dominated by the hot
component, even though it is inherently weaker (Fig. 1).
Thus, the data are consistent with a picture in which a weak hot component is present every-
– 10 –
where, and the only significant difference between hard-hot and soft-hot type subregions is that the
cool component is slightly cooler in the hard-hot subregions.
In this work we have analyzed an active region observed by Hinode/XRT in November 2006,
i.e. in the first phase of the Hinode mission. The observation is made with five filters which can
be combined into ten different but not entirely independent filter ratios to provide temperature di-
agnostics. We have optimized the signal-to-noise ratio by averaging over one hour of observation,
i.e. 12 images per filter, in the absence of flaring activity and any other considerable variation.
The new achievement is to include in the analysis the ratio of the filters of intermediate thickness
(Be med, F4, and Al med, F5) – the thickest ones available in this observation – which screen
out the emission of plasma below logT = 6.3 and are sensitive therefore to possible weak hot-
ter plasma components. We have derived temperature and emission measure maps for all filter
ratios (except for the softest one) for signal above a high acceptance level. By combining the val-
ues obtained at individual pixels, we have built emission measure distributions as a function of
temperature which show evidence of a hot plasma tail up to logT > 7. From inspection of the
temperature maps we realize that there is a clear correlation between regions detected as hot in the
soft filter ratios and regions detected as cool in the hard filter ratio and viceversa (i.e. an anticorre-
lation). To investigate this point further we have first extracted the emission measure distribution
from two regions of the different kind. The distribution of the former type is characterized by a
single narrow component with the peak at logT ≈ 6.5 in all filter ratios. The distribution of the
latter type is instead split into a cool component peaking at logT ≈ 6.4 detected by the soft filter
ratios and a hot component at logT ≈ 6.9 detected by the hard filter ratio. In order to interpret
this result correctly we have performed MonteCarlo simulations to build realistic synthetic XRT
images in all 5 filters starting from model parent emission measure distributions along the line of
sight. We find that both kinds of emission measure distributions derived from the data can be co-
herently explained with a single type of line-of-sight parent distribution consisting of a high cooler
component at logT ≤ 6.5 and lower hotter component extending to logT ∼ 7. The difference
between obtaining a single or split distribution is determined by the peak temperature of the cool
component. If it is below logT = 6.3 the hard filter ratio is able to detect the hot component,
otherwise the cool component invariably obscures the other one.
We are unable to obtain a perfect match between simulations and data results for a variety
of reasons. First, although we have performed a fine tuning of the hard filters calibration through
detailed feedback on the data there is still room for some uncertainty. In particular, it is possible
– 11 –
we have applied the minimum possible correction to obtain consistent results. This assumption is
conservative and we may obtain even hotter emission measure component after applying possible
less conservative corrections. Small calibration tuning on the F3 (Be thin) filter response may
improve the agreement of results from ratios involving this filter. Second, we have assumed very
simplified forms of the parent emission measure distribution along the line of sight. The results
are clear as such and do not require more refined forms. In any case we prefer here not to push the
temperature resolution of our emission measure analysis at any step beyond the one offered by the
filter calibration, i.e. 0.1 in logT. This includes also the randomization of the parameters. Finally,
our analysis does not account for the inherent organization of the plasma in coherent magnetic
loops. We do not pretend here to enter into the fine details of the emission measure distribution.
Going back to the morphology of the active region, we propose that hot plasma is widespread
in the active region but that its detection depends somehow inversely on the local level of activity.
In particular, the hot plasma may be difficult in high pressure loops, where the dominant cool com-
ponent is relatively hot, and easier to detect in fainter structures, where the dominant component
is relatively cool. This scenario is consistent with what we find when we overlay the temperature
maps obtained with the soft and hard filter ratios with an emission map obtain with the 171˚ Afilter
onboard TRACE, which is sensitive to plasma at ∼ 1 MK (Fig. 10). We immediately realize that
the central green loops (in the soft-hot subregion) are anchored to moss structures detected by
TRACE which are typically interpreted as the footpoints of high pressure loops (e.g. Peres et al.
(1994); Fletcher & de Pontieu (1999)). Longer loops detected by TRACE are clearly cool struc-
tures which occupy volumes complementary to the hotter structures detected with XRT. Hot (blue)
regions are generally faint localized at the boundaries.
Our analysis has therefore revealed clear evidence for small quantities of very hot plasma that
is widely distributedthroughoutthe activeregion. The diagnosedtemperatures are characteristic of
flares (∼ 10 MK) though no flares occurred within about 2 hrs of the observations. This provides
strong evidence for the nanoflare picture of coronal heating. The emission measure of the hot
components (logT ≥ 6.6) is about 3% of the cool one in both parent EMlos(T)’s of Figs. 8 (right)
and 9 (right). We mention that similar results are obtained from the analysis of the same active
region observed 2 days before (10 November), in a period when the region was even more quiet.
Wehaverun nanoflaresimulationswithourEnthalpy-BasedThermal EvolutionofLoops(EBTEL)
code (Klimchuk et al. 2008) and have no difficulty producing emission measure distributions with
this property. A similar EM distribution has been also obtained in Reale & Orlando (2008) with
a simulation of a loop heated by nanoflares localized at the footpoints and lasting 3 min, taking
deviations from ionization equilibrium into account.
The standard nanoflare model envisions that many unresolved strands at different stages of
heating and cooling should coexist within close proximity of each other. It therefore predicts a
– 12 –
broad emission measure distribution along each line of sight. (The distribution could be rather
narrow for a bright coronal loop that is heated by a storm of nanoflares confined to a short time
window). Should nanoflares be ruled out in those parts of the active region where we find no
detectable signal in the hard ratio? They cannot be ruled out for the following reason.
The hard filters are much more sensitive to hot plasma than to cool plasma. For example, the
sensitivity of F5 is 10 times greater at 10 MK than at 3 MK. As noted above, however, nanoflare
simulationspredict an emission measure that is at least 100 times smaller at the hotter temperature.
The observed signal will therefore be dominated by the cooler plasma. Even with the hardest filter,
there is a tendency for small but significant amounts of hot plasma to be masked by more dominant
The hot component is a minor component of the total emission measure distribution, much
smaller (about 3%) than the cooler ∼ 3 MK dominant component. The small emission measure
of the hot component explains why it has been so elusive so far: it is overwhelmed by the cooler
component along the line of sight and can be detected only after cutting the cool component off.
This screening is efficiently but not completely performed through the single thick filters: the
EM(T) distribution obtained with the ratio of the filters shows that most but not all of the dominant
3 MK plasma is masked out. The final cut is allowed by the temperature diagnostics provided by
the filter ratio.
In conclusion, the thick filters of Hinode/XRT detect hot flare-like plasma in an active region
outside of proper flares. This plasma is widespread and steady, and its amount is consistent with
We thank G. Peres and P. Grigis for useful suggestions. Hinode is a Japanese mission devel-
oped and launched by ISAS/JAXA, with NAOJ as domestic partner and NASA and STFC (UK)
as international partners. It is operated by these agencies in co-operation with ESA and NSC
(Norway). FR acknowledges support from Italian Ministero dell’Universit` ae Ricerca and Agenzia
Spaziale Italiana (ASI), contract I/015/07/0. PT is supported by NASA contract NNM07AA02C
to SAO. JAK was supported by the NASA Living With a Star program. SP acknowledges the sup-
port from the Belgian Federal Science Policy Office through the ESA-PRODEX programme. This
work was partially supported by the International Space Science Institute in the framework of an
international working team.
– 13 –
A.Fine tuning of filter calibration
In this paper we use the recently updated XRT filter thickness calibration (Narukage et al.
2009). The high quality of the XRT observations here analyzed allows to test this new instrument
calibration. While changes in transmissivity with respect to the old calibration are in general
limited for all filters used here (of the order of up to about 15%), the effect on the temperature and
emission measure diagnostics can be significant. This is particularly true in the case of the ratio of
hard filters (F4/F5) which spans a rather narrow range compared to other filter ratios.
By comparing the distributions of measured filter ratio values with ratios expected from the
filter temperature responses for all filter combinations, we can investigatepossible inconsistencies.
On our data we find that with the new calibration the ratio of the two thickest filters yields values
in a range not entirely compatible with the expected range, as shown in Fig. 11. None of the other
filter ratios present similar obvious mismatches. The left panel of Fig. 11 shows the distribution of
measured F4/F5 filter ratio values, whiletheright panel shows theexpected filter ratio as a function
of temperature according to the official calibration (dashed line) and after the corrections that we
devised here (solid line, the cool branch marked by the dotted line is excluded for the temperature
diagnostics); the horizontal line represents the peak of the observed ratio distribution. This shows
that for a large number of pixels the measured ratio of hard filters lies outside the range of values
allowed by the current release of the filter calibration, pointing to an obvious problem with the
temperature response of one or both involved filters.
We therefore devised a procedure to estimate empirically a correction factor needed to bring
the observations back into agreement with the expected values and investigated the plausibility of
this correction. We find this correction factor with an iterative method: we use the distribution
of emission measure as derived from the thick filter analysis as input to synthesize the expected
emission in each of the two filters, and then apply correction factors to the filter temperature re-
sponses in order to reproduce the observed distributions of emission observed in both filters. From
the inspection of the changes of the filter responses between the original and the newly released
calibration we find that for most filters the changes in assumed filter thickness roughly correspond
to a constant scaling factor (at least in a range of temperatures where the response is highest), and
therefore we consider a sensible choice to modify the response by applying a simple scaling factor.
At this stage the correction can be applied indifferently to any of the hard filters. Operatively, we
chose to act on the F4 filter by increasing its response in steps of 5%. We found that an increase
by about 10% allows to match the observed values. We note however that this is the minimum
correction factor needed.
As further step we examined all other filter ratios to extract additional information potentially
useful to determine which filter is most likely responsible for the mismatch with observations,
and therefore it is most sensible to correct. Fig. 12 shows the expected filter ratios for eight filter
– 14 –
combinations, using the new calibration after our correction. The horizontal lines in each plot
delimit ranges of the observed filter ratio distributions within 10% of the peak, and the vertical
lines indicate the corresponding temperature range. As expected, the ratios of thin filters give
consistent results. Before correction, the observed ratios of F5/F3 (Al med/Be thin) corresponded
to a wider temperature range with respect to F4/F3 (Be med/Be thin), in particular covering higher
temperatures. This is not expected considering that F4 should be relatively more sensitiveto hotter
plasma than F5, as indicated by their ratio as a function of temperature (see Fig. 11). We interpret
this as a possible suggestion that the F5 filter response is the one needing the most significant
correction. We find that the correction factor of ∼ 10% found with the procedure described above
yields more consistent results. Therefore in our analysis we assume an increase of about 10% to
the response of the F5 filter while the response curves of all other filters remain as provided by the
current XRT calibration. We note that this correction factor is quite reasonable as it corresponds
to a variation in filter thickness of about 3%, which is compatible with the uncertainties in filter
thickness as derived through laboratory measuments.
Finally, we apply a minor correction to the filter responses to correct an artifact of the too
coarse temperature resolution for the definition of the temperature responses. As shown in Fig. 11
(dashed line) the ratio of the medium filters presents an inflection around logT ∼ 7.0 due to the
insufficient temperature resolution around the peak of the filter responses. This causes an artificial
dip in the derived distribution of emission measure. We find that a small correction (<2%) is
sufficient to correct for this problem, as shown in Fig. 11. Further minor corrections are applied
in the low temperature range to obtain further minor improvements. The final correction factor
between the current calibration release and our procedure is between 5 and 10%.
Bradshaw, S. J., & Cargill, P. J. 2006, A&A, 458, 987
Cargill, P. J. 1994, ApJ, 422, 381
Cargill, P. J. 1995, in Infrared tools for solar astrophysics: What’s next?, ed. J. R. Kuhn & M. J.
Cargill, P. J., & Klimchuk, J. A. 1997, ApJ, 478, 799
Fletcher, L., & de Pontieu, B. 1999, ApJ, 520, L135
Golub, L. et al. 2007, Sol. Phys., 243, 63
Klimchuk, J. A. 2006, Sol. Phys., 234, 41
– 15 –
Klimchuk, J. A., & Gary, D. E. 1995, ApJ, 448, 925
Klimchuk, J. A., Patsourakos, S., & Cargill, P. J. 2008, ApJ, 682, 1351
Kosugi, T. et al. 2007, Sol. Phys., 243, 3
McTiernan, J. 2008, AGU Fall Meeting Abstracts, A1511+
Narukage, N. et al. 2009, ., in prep, .
Orlando, S., Peres, G., & Reale, F. 2000, ApJ, 528, 524
Parenti, S., Buchlin, E., Cargill, P. J., Galtier, S., & Vial, J.-C. 2006, ApJ, 651, 1219
Parker, E. N. 1988, ApJ, 330, 474
Patsourakos, S., & Klimchuk, J. A. 2008, Eos Trans. AGU, 89, 23
Peres, G., Reale, F., & Golub, L. 1994, ApJ, 422, 412
Reale, F., & Ciaravella, A. 2006, A&A, 449, 1177
Reale, F., & Orlando, S. 2008, ApJ, 684, 715
Reale, F. et al. 2007, Science, 318, 1582
Schmelz, J. T., Saar, S. H., DeLuca, E. E., Golub, L., Kashyap, V. L., Weber, M. A., & Klimchuk,
J. A. 2009, ArXiv e-prints
Urnov, A. M., Shestov, S. V., Bogachev, S. A., Goryaev, F. F., Zhitnik, I. A., & Kuzin, S. V. 2007,
Astronomy Letters, 33, 396
Warren, H. P., Winebarger, A. R., & Mariska, J. T. 2003, ApJ, 593, 1174
Zhitnik, I. A., Kuzin, S. V., Urnov, A. M., Bogachev, S. A., Goryaev, F. F., & Shestov, S. V. 2006,
Solar System Research, 40, 272
This preprint was prepared with the AAS LATEX macros v5.2.