Multi-Instrument DEM (RHESSI – GOES) Calculations

J.McTiernan

5th General RHESSI Workshop

8-June-2005

Why RHESSI and GOES?

• RHESSI and GOES are what we have for all flares

• And we want to do *all* flares…. We have 14000+

• There have been questions about GOES in the past (R.Schwartz, private communication). It would be nice to see if it’s possible to get RHESSI and GOES on the same plot. In 1998 we could not get GOES on the same plot with SXT, BCS.

• Not happy with RHESSI-only DEMs. They do not look like what should be expected from previous Yohkoh SXT-BCS DEMs

RHESSI flare… dude… this looks isothermal…

XDEM routines:

• Designed to easily incorporate multiple instruments:• Needs: Spectral Response (T response is calculated “on the fly”

using CHIANTI_KEV (CHIANTI 4.2)) • For Example, RHESSI DRM Or• GOES_TRANSFER function Or• TRACE_EUV_RESPONSE Or• SXIG12_LAMBDA_RESPONSE• Not just the T response: Using the spectral response allows for

inclusion of non-thermal emission.

RHESSI – GOES Temperature Response

• T response is very broad-band

• D(response)/DT is small

• Will probably not have really good temperature resolution. Maybe a few MK.

DEM calculation uses PIXON method:

• Same as HXT – RHESSI pixon imaging algorithm. But hacked into a 1-d spectral code. This can be applied to any problem with response#spectrum = data.

• ‘Fuzzy’ pixons, where high resolution image is smoothed at each pixel by parabolic-shaped pixon. Pixons are chosen to have the smallest possible number of pixons (or the largest amount of smoothing) consistent with the data.

• The smoothed image is the answer. In this case the ‘image’ is the DEM.

Test Case:

• Decay phase of flare

• 19-sep-2002 05:35 to 05:50

• 1 minute time intervals

• Why the decay? To avoid nonthermal emission.

Quicklook Spectrum:19-sep-2002

• 1 thermal component, about 20 MK

• 1 PL with cutoff, index of 10.2, very soft, probably thermal

What about the DEM?

• 2 GOES channels

• Approx. 25 RHESSI channels, 2/3 keV energy resolution, between 3 and 20 keV – includes the Fe line.

• Use power law DEM to start, (green dashed line in plots).

• 0.5 MK bins in T, from 3 to 40 MK

• Here is the result….

19-sep-2002 05:35 +1 MIN

19-sep-2002 05:36 +1 MIN

19-sep-2002 05:37 +1 MIN

19-sep-2002 05:38 +1 MIN

19-sep-2002 05:38 +1 MIN

19-sep-2002 05:39 +1 MIN

19-sep-2002 05:40 +1 MIN

19-sep-2002 05:41 +1 MIN

19-sep-2002 05:42 +1 MIN

19-sep-2002 05:43 +1 MIN

19-sep-2002 05:44 +1 MIN

19-sep-2002 05:45 +1 MIN

19-sep-2002 05:46 +1 MIN

19-sep-2002 05:47 +1 MIN

19-sep-2002 05:48 +1 MIN

19-sep-2002 05:49 +1 MIN

DEM results:

• 3 components, the stuff between the peaks is not really significant (error bars are restricted to be < 0.99 of DEM, so that they’d be plotted).

• Lowest T peak drops below significance early

• Position of peaks does not seem to vary.

Look At Fit and Residuals:

• Looks good on log scale, but points are off (note that sigma/counts < 0.10). The method wants a smoother spectrum.

• Some biases in residuals

Test with Power Law Input DEM

• Does the program try to insert peaks – at 3, 12 and 25 MK always? Doesn’t look like it here.

• Still problems with residuals, the high energy rate is overestimated. Would like to see scatter…

• Power law is almost recovered, but not quite.

• I think that these problems are the result of the broad T responses.

Case B:

• Pixon reconstruction of Electron spectrum from photon spectrum of 23-July-2002. (Courtesy of M. Piana)

• Biased residuals, the high energy part of the photon spectrum is overestimated

Case B:

• Bremsstrahlung Xsection is a broad response function

Case C:

• Reconstructed photon spectrum from counts spectrum, 20-jul-2002 flare

• This looks much better for most of the spectrum, not so hot at end points, probably due to smoothing.

Case C:

• Here we have a narrow response.

Conclusions:

• Click to add text

Conclusions:

• RHESSI-GOES DEM isn’t that bad – but there are annoying systematic problems that are probably due to broad responses.

• Future Work:• Try fractal pixons – rather than fuzzy pixons – this is binning

rather than smoothing, and should help with problems at edges. (Tom Metcalf has had some success with case B using power law shaped pixons.)

• Maybe use larger RHESSI energy bins, it is not clear whether having so many RHESSI data points helps when you have the very broad band responses.

• Add SXI, TRACE• Add nonthermal emission