niscorpeon - NICER SCORPEON background modeling tools


This task describes the NICER SCORPEON family of modeling tools, available with the NICERDAS analysis system.

The SCORPEON model is a background model for NICER data. The model is unique in that it embodies contributions from many physically-motivated components. These include sky-related components such as the cosmic X-ray background (CXB), local hot bubble (LHB) and galactic halo; and also non X-ray background components such as cosmic rays (COR), South Atlantic Anomaly (SAA), trapped electrons (TREL), precipitating electrons (PREL), low energy storm-related electrons (LEEL), and so on.

At the moment, SCORPEON is designed to estimate the background for spectra. It has two different kinds of capabilities.

SCORPEON's most powerful and sophisticated capability is the ability to include the background model as a parameterized model within XSPEC. The scientist is able to adjust (fit) the background to match the data along with the scientific parameters of interest. This allows a better fit to the data. Over- or under-subtraction of the data is no longer a problem.

Another benefit of this approach is that the uncertainty of the background, which is normally ignored, now feeds directly into the uncertainty of the scientific parameters of interest. This allows much more realistic confidence bounds on scientific parameters.

Finally, it is well known that astrophysical backgrounds like the CXB, Halo, and so on, vary from point to point on the sky and are not fixed to a single quantity. Using this modeling approach, one can get a better handle on the true values of these parameters.

However, the benefits of this approach do come at a cost. The SCORPEON model is more complicated to use within XSPEC. It requires that user become familiar with the details of potential NICER background contributors. For these reasons, SCORPEON also offers the option to output a background file instead of a background model.

The two SCORPEON software tasks available are,

If you are uncertain which SCORPEON background estimator task to use, niscorpspectmod or niscorpspect, here is a summary of which to use.

Background Model Versioning

The SCORPEON tasks support working with different versions of the SCORPEON model. This facilitates reproducibility of scientific analysis for older works that may have been analyzed with earlier model versions. It is also possible to load different versions of background model for comparison purposes. The available released background versions are:

You can choose an alternate model version number with the bkgver parameter.

Model Variants

Some of the SCORPEON models come in different flavors or variants. The variants provide additional or different functionality for the same basic component. You may select such variants using the bkgvariant=NAME parameter, for example bkgvariant=detailed.

The "detailed" variant provides addiitional detailed parameters for some model components. The actual spectral shape and normalization of the detailed variant is the same as the default variant, but the detailed variant provides additional parameters that are adjustable in the case of special circumstances. Currently, only the NXB model component has a detailed variant, which provides additional normalization adjustment factors for sub-components such as COR and SAA, as well as normalization adjustment factors for the strong background lines.

Currently the only supported variant is for the nxb component (non-X-ray background). This variant supplies additional parameters to allow a higher fidelity fit to bright backgrounds, or spectra with very high statistics. Most users will not need the detailed variant.

A planned "sf" variant provides background models tailored to NICER's SLOW+FAST data mode. This mode consists of only events that have both slow and fast pulse heights, which automatically implies energies above about 0.6 keV. A threshold at 0.572 keV is applied to the background model. The total background for SLOW+FAST events may be lower than the default mode.

How SCORPEON background parameters are estimated

The SCORPEON model is more than just a set of functions. The model is also an a priori estimate of the values of many background parameters which give a good starting value for spectral analysis. It is quite possible that further fitting may improve the fit quality even more, but the goal is that the a priori estimates give a "good enough" result for many situations. This setion describes how these a priori parameter estimates are derived.

Please note that these descriptions are necessarily model version-dependent and thus subject to change. The descriptions below are intended to reflect the "v22" SCORPEON model release.

SCORPEON Sky Background Components

For the astrophysical sky backgrounds, there are contributions from the CXB, galactic halo, and LHB.

The cosmic X-ray background (CXB) parameters are estimated from the following work,

Namely, the CXB norm is set by default to 10.91 phot/s/cm2/sr and the CXB power law index is fixed at Gamma = 1.45. The CXB norm is allowed to vary by +/- 30% based on cosmic variance in the NICER field of view (Moretti et al. 2009, "A new measurement of the cosmic X-ray background," A&A, 493, 501, doi 10.1051/0004-6361:200811197).

The galactic halo is an APEC plasma model estimated based on the work of

The halo temperature is fixed at 225 eV and the emission measure is based on a raytrace through the paper's empirical disk-like density model (non-polytrope version). The emission measure parameter is allowed to vary in a range of +/- 35%, since the halo is more clumpy than the empirical density model. The emission measure units are in cm-6 pc.

The local hot bubble (LHB) is an APEC plasma model estimated based on the work of

The LHB temperature is fixed at 99 eV. The emission measure normalization is estimated from maps provided by the authors, using three crude intensity contours. The emission measure parameter is allowed to vary in a range of +/- 35% because of the crudeness of the estimate as well as additional clumpiness that may be found in the true LHB. The emission measure units are in cm-6 pc.

The sky background parameters are:

The solar wind charge exchange (SWCX) model consistes of lines at the following energies,

By default these lines are considered narrow, and the centroid and sigma values are fixed. In addition the normalizations of these parameters is fixed to 0 by default. Normalizations of these values are in units of photons/s/sr. In large accumulations of megaseconds of background data, lines at these energies are detectable at the ~1 photon/s/sr level. Neutral oxygen is occasionally detected, especially when looking through the earth's polar cusp regions. So far a quantitative model is not available to estimate the normalizations of these lines, so they should be activated (thawed on an as-needed basis).

The swcx parameters are:

SCORPEON Non X-ray Background Components

There are several non X-ray background components. These components all use a response matrix which is diagonal and no ARF, which is appropriate for non X-ray charged particle interactions.

There is a constant(CON) component which reflects residual non-variable background count rates. This residual level presumably corresponds to the cosmic gamma-ray background (both primary and secondary). This con_norm component is normalized by the average number of detectors enabled, and so is typically ~52 for an observation where all detectors are enabled.

A cosmic-ray (COR) component corresponds to background due to cosmic-rays. These particles have a huge energy range (typically ~MeV through 10s of GeV), and are modulated by the geomagnetic field (COR_SAX) as well as the solar modulation potential (SOLARPHI). This background is typically very predictable because the interplanetary cosmic ray flux is nearly constant, and the COR_SAX and SOLARPHI modulating functions are well known. However, this parameter is allowed to vary by default. The cor_norm normalization is given in units of overshoot rate per second (per FPM), and is typically in the range of 5-30.

The model includes a component due to South Atlantic Anomaly (SAA). Although the default screening excludes SAA data, SCORPEON does include a way to model background for data taken while passing through this region. If the SCORPEON model detects an SAA passage, the saa_norm parmaeter will be estimated, but allowed to vary during fitting. The saa_norm normalization is given in units of overshoot rate per second (per FPM), and is typically zero outside of SAA and in the range 0-90 within SAA.

Trapped electrons (TREL) are electrons from the solar wind that become trapped by the earth's magnetic field when the right conditions occur. These electrons range in energy between ~keV and ~MeV. The fluxes of these electrons can vary by large amounts on hourly through monthly timescales, and are not predictable. In addition, trapped electrons cannot be readily distinguished from the other kinds of electron populations. Therefore this parameter is estimated, but always allowed to vary during fitting. The trel_norm normalization is given in units of overshoot rate per second (per FPM), and is typically in the range 0-30.

Precipitating electrons (PREL) are typically created during reconnection events in the outer geomagnetosphere. When they do occur, large fluxes of electrons can essentially rain down on NICER. These electrons can have relativistic energies, although the strongest count rates in the X-ray band are due to ~keV electrons passing through the optics and directly into the detectors. Precipitating electron flares are confined almost exclusively to the extreme polar horns (COR_SAX < 1.5) As mentioned above, precipitating electrons cannot be readily distinguished from the other kinds of electron populations. This parameter is typically set to a value of zero, but if passages through the polar horns are detected, the parameter is allowed to vary. The prel_norm normalization is given in units of overshoot rate per second (per FPM), and is typically in the range 0-10, but during extreme flares can be as high as 1000.

The nxb parameters are:

Noise Peak Model

The NICER noise peak is due to noise events that trigger the electronics. Due to how the NICER triggering algorithm functions, noise events typically appear just below threshold in a spectrum, which in NICER's case means centered at an energy of about 150 eV.

The noise peak core is extremely predictable. The core is a gaussian, whose centroid, sigma width and normalization are all very repeatable functions of undershoots. Thus, the noise peak core is estimated by the SCORPEON model using these relations. When analyzing data in the 0.25 keV range and above, the noise peak core does not typically interfere with the source spectrum very much (the centroid is typically ~130 eV and sigma is typically ~25 eV, so E=0.25 eV is about 5 sigma from the centroid).

However, on occasion, some detectors experience broadened noise peaks. In addition to the narrow core, there is a broader wing. Even though this wing is suppressed by a factor of ~10,000 compared to the core, for extremely strong noise peaks, this broad wing can intrude into the scientific band of interest. Even in this case, the wing can be modeled as a gaussian as well.

Thus, the noise peak is included as a single gaussian function. The centroid, sigma and normalization are set by the undershoot-based model. The normalization is allowed to vary by default. If the scientist experiences strong residuals in the 0.22-0.30 keV range, they are advised to allow the sigma to vary as well. In the cases of a very broad wing, the sigma can be as large as about 75 eV. The scientist should never thaw the noise peak centroid since there are not enough data points to constrain it from a NICER cleaned events spectrum.

The noise peak parameters are:

"Soft Landing" Procedures

During initial testing, it became clear that it is not possible to estimately the precise contribution of each background component.

The electron components (TREL, PREL and LEEL) are difficult to distinguish because they each produce wildly different amounts of in-band counts for the same amount of overshoots, by almost three orders of magnitude. The NICER system does not have a way to distinguish these components a priori, so naturally the estimates will be uncertain.

A similar comment can be made about the noise peak. While the core of the peak is well-modeled by a gaussian, whose parameters can be determined very precisely from the number of undershoots per detector, there is a "tail" to this gaussian which can extend upward to about 0.4 keV. This tail is particularly pronounced when undershoots exceeds about 100 ct/s/FPM (the FPM_UNDERONLY_COUNT column in the filter file reports this value). The properties of the tail are must less well behaved.

To accomodate these uncertainties, the SCORPEON model can employ some "soft landing" procedures that adjust the model to more closely match the spectrum being analyzed. These procedures are similar to the 3C50 model's procedures, which renormalize the background model to the real data. Doing this means of course that the model is not as "a priori," but also means that a better background model can be achieved.

SCORPEON's soft landing process uses out-of-band counts (below 0.26 keV and above 12 keV), so in principle should not be heavily biased by source counts, although this is a risk for extremely bright sources.

The soft landing process is enabled when bkgsoftlanding=YES, which is the default. Set bkgsoftlanding=NO to disable it.

When the process is enabled, the high energy counts between 12-15 keV from the user's spectrum are used to adjust the trel_norm to more closely match the data; if the prel_norm is enabled and COR_SAX falls below 1.5 GeV/c, there is a risk of precipitating electrons, so prel_norm is adjusted instead. The low energy counts between 0.22 and 0.26 keV are used to adjust the norm and sigma of the noise peak, if the count rate exceeds ~10 ct/s/keV.





Jul 2022