Revised PCA Background Models of 2006

C. Markwardt (U. Maryland/GSFC) and K. Jahoda (NASA/GSFC)
with contributions from A. Markowitz and R. Rothschild
Last Updated: 2007-09-18

Table of Contents

1. Summary
2. Recommendations to Guest Observers
3. Introduction
4. Brief Description of Model
5. Errors in the Previous Model
6. Changes to Model: Epochs 5A, 5B and 5C
7. Results
8. Unmodeled Variance
9. Systematic Offsets
10. Dead Time Correction
11. Conclusions and Future Outlook
12. Modification History

Summary

This report describes new background models for RXTE PCA observations for "Epoch 5," and the impact to guest observers. The new models have removed long term trends that became apparent over the years (with magnitudes of ~0.2 ct/s/PCU). Only models for "Epoch 5" -- May 2000 to the present -- were changed.

The unmodeled variance in the background rate, which fundamentally limits the sensitivity to source variations, is 0.04 ct/s/PCU or better, in the top layer 2-10 keV band, and 0.02 ct/s/PCU in the top layer 10-20 keV band. However, for purposes of determining absolute fluxes of sources at arbitrary points on the sky, the cosmic X-ray background is the limiting factor (see below).

Recommendations to Guest Observers

The new background models can be downloaded from the "PCA Digest page".

The most significant errors in the previous models were in the "bright source" (VLE) model. However, the "faint source" (L7) model was also improved. The PCA team recommends that for the maximum sensitivity and least systematic error, that observers re-estimate the background level for their observations after 2000.

In the future, the PCA team expects not to change the background models up to January 2004.

The systematic errors are minimized by integrating for at least 1600 seconds per time bin. For shorter time bins, the variance in the L7 rate, the nominal independent variable of the model, begins to become important.

Faint Sources (Absolute Flux Measurements) - For absolute X-ray flux measurements the variation of the cosmic X-ray background is the dominant source of error. (see below)

Introduction

The RXTE PCA is a non-imaging instrument, which means that for most work, both in spectroscopy and light curve analysis, the background must be subtracted based on an a priori model. Here, "background" is defined broadly to include anything that contributes non-source counts to the PCA instrument in orbit, including but not limited to:

• local particle environment;

• induced radioactivity of the spacecraft; and

• the cosmic X-ray background.

In general, these components vary as a function of time, so they must be somehow parameterized. The parameterized model is adjusted to fit a set of dedicated observations by the PCA of blank sky regions. Once a good fit is achieved, the same parameterization can be applied to other observations.

These release notes describe the improvements to the PCA background model since the last version was released in 2003.

Brief Description of Model

The model is based on blank-sky observations which are periodically made by RXTE. Some data editing has been done to remove outliers and probable stellar flares, but this amounts to less than 1% of the total data volume.

Models are parameterized by housekeeping rates. The two primary models are based on:

L7

the sum of seven different pair-wise coincidence rates;

VLE

the "very large event" counter, which registers events which deposit large amounts of change in the instrument.

The models are also parameterized by the particle dose from the previous SAA (scaled by an exponential decay time factor), and have a linear dependence on time.

Errors in the Previous Model

Over time, small discrepancies have been building in the background models.

For example, consider the light curve residuals shown in Figures 1 and 2. These light curves are drawn from background data, and show that the faint model is drifting slightly higher with time, and the bright (VLE) model is drifting significantly lower with time. Changes were undertaken to reduce the trends.

Changes to Model: Epochs 5A, 5B and 5C

In the past, PCA background models have been grouped into background "epochs," actually intervals of time, where significant changes in the PCA configuration had been made. Several of the changes have involved high voltage adjustments; one change involved the loss of propane in PCU0. Gain epoch 3, the previous longest interval, was split into two effective intervals to improve the model.

Since the time of the last PCA background release, background Epoch 5 has become the longest epoch of all, spanning almost 6 years. Thus, we have also split epoch 5 into three time segments, 5A, 5B and 5C. This is motivated by the changing space environment, and the changing instrument.

The space environment has changed as the RXTE spacecraft altitude has gradually decreased. The particle background are generally lower. Also, there is reason to believe that the nature of the particles has changed (perhaps having less energy), so the correlation between the housekeeping and background rates does not hold so perfectly. Splitting the data interval into multiple segments helps the model to track these changes more closely.

The instrument has also changed, in the sense that the gain has slowly drifted with time, which changes the energy to channel assignment. While this is an expected effect, it is non-trivial to account for it in the background analysis procedures, and causes strange effects around the positions of background spectral lines. It is far more straightforward to break the analysis into several time segments in order to reduce the gain drift effect.

The previous model was made long enough ago that the extrapolations were starting to diverge from reality. It was time to include more data.

Finally, it's worth pointing out that dividing the epoch 5 into different intervals will make future updates more straightforward, since it is likely that at least Epochs 5A and 5B will remain constant.

The new gain epochs are listed in Table 1

Results

He we present an analysis of light curves and spectra with the new models.

Long Term Light Curves

The results of the new models are shown in Figures 3 and 4. The trends have been removed. The spike near the end of 2003 is an analysis artifact. The scatter in the 2-10 keV residuals is primarily due to the cosmic X-ray background variance: each data point is the sum of different amounts of exposure of each background position. As the cosmic X-ray background is fainter at higher energies, it is less apparent in the 10-20 keV band (Figure 4. However, most users will be concerned with a single point on the sky -- their target -- which will show much smaller unmodeled variances (see below).

Spectra

Spectral results show some small changes.

The spectrum of the AGN IC4329A is shown in Figures 5 through 8. While small changes are present at low energies, the strongest differences are evident in the 25-30 keV range. The old model was contaminated by background lines from the calibration source near 29 keV. This line feature is much weaker after applying the new model.

A light curve of the same AGN over the time range 2003 Apr to 2005 Oct shows deviations of less than 0.04 ct/s/PCU between the old and new models.

A spectrum of Centaurus A is shown in Figure 9, using both the new and old background models. It shows that the results for this bright and hard source are barely changed. There are still very weak features at 29 and 60 keV, corresponding to calibration lines, of order 0.01 ct/s/keV. Such features might be expected to get somewhat stronger several years after the release of the model, and may take on a P Cygni type profile. This is due to a coupling between the temporal gain drift of the instrument, and the temporal background drift component of the model.

Unmodeled Variance

As with previous models we provide tables of unmodeled variances to users.

Tables 2 and 3 represent the Faint L7 unmodeled variances in the 2-10 and 10-20 keV bands, respectively. As usual, one should be aware that the cosmic variance from point to point (approximately 0.16 and 0.04 ct/s/PCU in the same bands), so there will always be a small and unknown offset in the count rates for a given point. However, even if the absolute count rate canot be established exactly, the unmodeled variance tables are useful for determining source flux variations.

Systematic Offsets

The PCA team is aware that the background model appears to have a constant offset which causes an incomplete subtraction. This is especially evident at high energies where the signal is low. The PCA team currently recommends applying a correction file. This is done within XSPEC with the corfile and recornrm commands. For each source spectrum, you should notice only the source free portion of the spectrum (say, above 85 keV). Then run the recornrm command, which adjusts the background to match in that energy range. Once this is done, you can notice whatever channels you desire and fit your model.

In summary, for each source spectrum,

  corfile background.pha  # Load the background as a correction vector
  ignore **-85.           # Notice only source-free part of spectrum
  recornrm                # Correct the background level
  notice whatever...      # Back to your regular fitting

Dead Time Correction

The PCA background model is based on observations of the blank sky. Therefore, the model already includes the dead time due to the background level. This means that for faint sources you typically should not apply any further corrections for dead time. For bright sources, the background model will have a different dead time than the source+background rate, since they are based on observations with different rates. You can correct for dead time for bright sources by one of two methods:

• correct both the source+background and background spectra for dead time;

• correct only the source+background spectrum for dead time using the excess count rate above the background level;

Typically the second step is easier to perform. Remember, that the deadtime correction must be based on the rate in the entire PCA energy band, not just the spectral band you are interested in.

Finally, you will likely need to use the 'corfile' correction described above to get a good background match during spectral fitting. Since this adjustment is comparable to performing dead time correction, it is likely that a precise dead time correction procedure is not necessary for all but the very brightest sources.

Conclusions and Future Outlook

The new models remove significant systematic trends in the long term behavior of the background model, and also reduce the artifacts seen around 25-30 keV.

The PCA team expects to not change the background models for epochs 5A and 5B, unless some major new problem or improvement is discovered. It is possible that the model for epoch 5C (affecting observations of January 1, 2004 and onward) could change as more data becomes available. We also expect to keep partitioning the background models into roughly two year intervals, so there is likely to be an epoch "5D" starting around the beginning of 2006, when there is enough data.

Modification History

• 2006 Aug 30 - add section on dead time; add table of contents

• 2007 Sep 18 - small formatting changes (no content changes)