NICER Responses (ARFs and RMFs)

Overview

When performing spectral analysis for any X-ray astronomy mission, the analyst needs calibration information about the detector response. This information is typically embodied in two types of files: the Ancillary Response Files (ARFs) and Response Matrix Files (RMFs). This document describes how to retrieve these files for NICER spectral analysis.

Read this thread if you want to: Retrieve ARFs and RMFs for NICER spectral analysis.

Last update: 2022-04-28

Important Note. If you are having issues or problems generating responses, please see the Troubleshooting section below.

Introduction

When performing spectral analysis for any X-ray astronomy mission, the analyst needs calibration information about the detector response. Response information is the link between astrophysical quantities, such as flux and spectral shape, and detector quantities such as count rate in each spectral bin. Thus, having the appropriate response information is key to making inferences about astrophysical observations.

It is typical for modern astrophysical observatories store response information as data files in FITS formats. There are standard formats for response information that most analysis software conforms to.

A commonly used spectral analysis software package called XSPEC provides an environment where astrophysical spectral models, response information, and observed spectra can be combined to estimate spectral parameters.

NICER's analysis system supports the standard response formats and XSPEC.

Response information comes in two primary types, Ancillary Response Files and Response Matrix Files.

Ancillary Response Files (ARFs). ARFs embody the total throughput of the detector, also known as effective area. Measured in units of cm2, the ARF essentially provides a way of translating an modeled flux, in units of photons cm-2 s-1, into counts. The ARF is a multiplicative product of the geometric area, which is energy independent and measured in units of cm2, and the quantum efficiency, which is the probability to convert a single incident photon into an observed detector count. The construction of an ARF by calibration scientists involves understanding both geometric effects, such as throughput of a detector system via ray tracing, physical effects such as transmission through windows or filters, and detector effects such as conversion efficiencies. The ARF also typically contains information about off-axis effects such as vignetting.

For NICER, the ARF contains all of these effects: transmission properties of the X-ray Concentrators, transmission throughput of filters and windows, and detector quantum efficiency. There are some caveats as described below.

Response Matrix Files (RMFs). The RMF contains information about energy redistribution. In this case, redistribution means that a photon of a given incident energy may have a measured pulse height that differs significantly from the nominal incident energy. This typically occurs because of imperfect charge collection within the detector, which causes some of the charge deposited by a photon to be lost or missed. Even if the "full" charge of an incident photon is collected, detector effects such as Fano-like and electronic noise will broaden the peak, leading to the detector resolution. The actual response matrix provides, for each incident photon energy, a probability of being measured at each possible pulse height.

For NICER, the RMF contains all known detector resolution and charge collection efficiency effects.

Please Note. NICER previously recommended to use a method to merge responses based upon precomputed files. This method is no longer recommended, but is described on the NICER Precomputed Responses page.

NICER ARF

As mentioned above, the NICER transmission properties of the X-ray Concentrators, transmission throughput of filters and windows, and detector quantum efficiency.

The NICER ARF is based on the following modeling efforts:

  • Ray tracing. The NICER-specific ray tracing program called CONSIM is used to estimate the X-ray throughput of the NICER X-ray Concentrator (XRC) optics. This simulation includes geometric optic terms, the geometrical point spread function, X-ray scattering in the Rayleigh-Rice domain, and reflectivity of gold concentrator surfaces.
  • Pre-launch measurements. X-ray throughput of filters and windows were measured at the BESSY synchrotron facility.
  • Other Detector Modeling. Overall FPM detector quantum efficiency is modeled based upon measurments taken at BESSY.
  • Post-launch calibration. On-orbit measurements of the Crab nebula were used to estimate the throughput and roughness of individual shells of each XRC.
  • Ray-trace simulations of vignetting. Vignetting is the primary cause of off-axis throughput changes.

Figure 1. NICER full-array ARF (sum of 52 operating modules). Key detector and optics features are identified.

Figure 1 shows a typical NICER ARF. This ARF is the sum of all 52 operating detectors. The peak effective area of about 1850 cm2 is located at a photon energy about 1.7 keV. Absorption edge features of carbon (0.285 keV), nitrogen (0.397 keV), oxygen (0.535 keV), aluminum (1.562 keV), and silicon (1.840 keV) are visible, and correspond to transmission edges of the windows and filters, and to some extend the dead layer of the silicon detector itself. Absorption edge features at about 2.2-3.5 keV and 11-14 keV are due to the reflectivity of gold M and L shells, respectively.

NICER uses a calculator task named 'nicerarf' to compute on- and off-axis ARF files, as described below. This technique became available as of HEASoft 6.29 (NICERDAS version 8). The computed ARF will be tailored to a specific observation because each observations conditions are slightly different.

NICER RMF

As mentioned above, the RMF contains all known detector resolution and charge collection efficiency effects.

The NICER RMF is based on the following items:

  • Modeling. The detector response model is based upon the model presented by Scholze & Procop (2009), with additional modifications for the NICER geometry.
  • Pre-launch calibration data. Detectors were calibrationed individually before integration into the full instrument. This allowed the team to estimate the intrinsic Fano-like detector resolution and electronic noise of each module. In addition, certain detectors were taken to the BESSY synchrotron facility for detailed calibration and validation against the response model.
  • Post-launch calibration data. Observations of the Crab nebula and background fields were used to estimate the trigger efficiency curves of detectors individually.
  • Effects of optical loading. Optical loading changes the resolution and trigger efficiency function parameters. These are now included in RMF generation.

Figure 2. NICER sample RMF for a single module, and for an incident photon energy of 3.75 keV. Key detector features are identified.

Figure 2 shows a sample response matrix for a single FPM module, and for a single incident photon energy of 3.75 keV. This would be the spectrum that NICER would record if the astrophysical target emitted a single narrow line at 3.75 keV. Visible in this plot are the main photopeak at 3.75 keV, along with its intrinsic broadening and low energy tail. The tail, along with the flat shelf, are indicative of imperfect charge collection processes. Imperfect charge collection can occur when some of the charge deposited by the X-ray either escapes the silicon, or enters a low-field "dead" portion of the silicon where it can recombine more rapidly than drift to the charge collection anode. A silicon escape peak appears at an energy of 3.75-1.84 = 1.91 keV. In addiiton, there are fluorescence peaks of oxygen, aluminum and silicon.
Figure 3. NICER sample RMF for a single module. Key detector features are identified.

Figure 3 shows a sample response matrix as a two dimensional array, covering all incident photon energies in one chart. This is a single-module RMF plotted on log intensity scale with the pulse height axis for X and the incident photon energy for Y. The same features that appeared in Figure 2 are also visible here, along with the energy-related trends. In addition, at high energies, a Compton-scattering related peak appears at low energies.

NICER uses a calculator task named 'nicerrmf' to compute RMF files, as described below. This technique became available as of HEASoft 6.29 (NICERDAS version 8). The computed RMF will be tailored to a specific observation because each observations conditions are slightly different.

How are NICER ARFs and RMFs Delivered?

NICER's primary delivery mechanism for calibration data is HEASARC's Calibration Database (CALDB) system.

If you need help to install NICER software or calibration please see the thread titled Setting Up a NICER Analysis Environment to download or update your NICER CALDB.

The CALDB area contains the products that are designed to be used by the tasks 'nicerarf' and 'nicerrmf', and not directly used by the analyst.

Generating an ARF for Your Observation

For spectral analysis with XSPEC or any other analysis program, you will need both an ARF and RMF for your observation. The NICER team now recommends using the tasks 'nicerarf' and 'nicerrmf' for generating responses tuned for your observation. These tasks became available in HEASoft 6.29 (NICERDAS 8). While you are urged to upgrade to the newest NICER software, if you are using older software where the nicerarf/nicerrmf tasks are not available, your only choice is to use NICER Precomputed Responses.

If you extract a new spectrum from a different observation, or from the same observation with a different time selection, you should generate a new response. You should not use a response generated for one spectrum with another spectrum.

Prerequisites

We assume the following:

  • The user has a working software and calibration setup
  • Spectrum named myspectrum.pha which is accumulated from your cleaned event file. The spectrum file should contain a GTI extension (which is standard)
  • The cleaned event file should have been created with HEASoft 6.29 (NICERDAS 8) or later, so that it contains "FPM Selection" information.
  • You know the RA and Dec of your target. Here we will assume RA=1.2345, Dec-67.89
  • The target is point-like.

To generate an ARF for your spectrum, use the following 'nicerarf' command. nicerarf myspectrum.pha 1.2345 -67.89 niNNNNNNNNNN.mkf \ niNNNNNNNNNN_0mpu7_cl.evt myspectrum.arf outwtfile=myspectrum_wt.lis where

  • myspectrum.pha is your NICER spectrum file, which should include a GTI extension
  • 1.2345 and -67.89 are the right ascension and declination of your target (in J2000 degrees). Of course you should enter the R.A. and Dec. of your target
  • niNNNNNNNNNN.mkf is the filter file (or attitude file)
  • niNNNNNNNNNN_0mpu7_cl.evt is your cleaned event file
    WARNING: if you have created a barycentered event file do not use it for ARF/RMF generation; use the original un-barycentered "cl" event file.
  • myspectrum.arf is the name of the output ARF file
  • myspectrum_wt.lis is the name of an auxiliary file which contains weights that can be used by the RMF generator

You should enter the correct RA and Dec of your target at the indicated command line positions. These values should be the celestial coordinates of the astrophysical target (and not the mean NICER pointing direction, for example). After completion, the ARF file is generated and saved with the indicated file name.

As noted above, if you have created a barycentered event file, do not use this file or any spectra generated from it. Use only unbarycentered event file and spectrum inputs to nicerarf. For more information, please see the NICER Response Common Issues thread for more information.

The outwtfile (=myspectrum_wt.lis in this example) is an output file created by nicerarf that contains weighting factors for each NICER detector, based on which detectors are on or off, as well as on-axis or off-axis. This information can be used by the RMF generator 'nicerrmf' to generate an appropriately weighted RMF file.

PLEASE NOTE: two bugs were uncovered in nicerarf and associated tools that were distributed in HEASoft 6.29b (NICERDAS 8b) and earlier. Please see the NICER Response Bug Fixes analysis thread for more discussion of these bugs and how to fix them.

What if I don't have "FPM Selection" Information? (NOT RECOMMENDED)

In some cases you may not have "FPM Selection" information in your cleaned event file. This may be because

  • You processed it with a version of NICER software before HEASoft 6.29 (NICERDAS 8)
  • You have manipulated the event file with non-NICER tools and lost the FPM Selection information
Since FPM Selection data has been available for more than a year with default processing, this technique is no longer recommended, but is here for reference only.

Please note that you can always run nicerl2 from HEASoft 6.29 or later and the FPM Selection information will be added automatically.

In this case you can still run nicerarf, but you are responsible to account for which detectors are selected or deselected. nicerarf can use the filter file (.mkf file) instead of the cleaned event file to determine which detectors are on or off.

To run 'nicerarf' without FPM Selection information, use the following nicerarf command. This is NOT RECOMMENDED because you can generate an event file with FPM Selection information. nicerarf myspectrum.pha 1.2345 -67.89 niNNNNNNNNNN.mkf \ niNNNNNNNNNN.mkf myspectrum.arf outwtfile=myspectrum_wt.lis # NOT RECOMMENDED where

  • myspectrum.pha is the spectrum, as above
  • 1.2345 and -67.89 are RA and Dec, as above
  • niNNNNNNNNNN.mkf is the filter file (or attitude file)
  • niNNNNNNNNNN.mkf is the filter file, here used for detector on-off information
  • myspectrum.arf is the name of the output ARF file
  • myspectrum_wt.lis is the name of an auxiliary file which contains weights that can be used by the RMF generator

As you can see, the only change is to substitute the filter file in place of the cleaned event file. The nicerarf task will use the filter file's FPM_ON column to determine which detectors are on or off in order to make a properly weighted ARF file.

The limitation of this approach is that nicerarf does not know if you have additionally screened the event file to remove certain detectors. ("FPM Selection" information in the event file automatically handles this) If you have screened out detectors from your event file, you will need to inform nicerarf of this manually using the detlist parameter. You can use the "-" notation to indicate excluded detectors. For example, if you excluded detectors 14 and 34 from the event file, you would nicerarf myspectrum.pha 1.2345 -67.89 niNNNNNNNNNN.mkf \ niNNNNNNNNNN.mkf myspectrum.arf outwtfile=myspectrum_wt.lis \ detlist=launch,-14,-34 # NOT RECOMMENDED (use FPM Selection technique above) where

  • detlist=launch,-14,-34 indicates that beginning with the launch configuration of detectors, you manually screened out detectors 14 and 34.
You can also set detlist to an affirmative list of detectors that you did include. See the help file for nicerarf for more information.

What If My Source Is Not Point-Like

The default command line invocation of nicararf listed above is for point-like targets. This may not be applicable for diffuse targets. nicerarf has support for certain kinds of diffuse targets.

The supported target surface brightness profiles are

  • Point-like. Point-like, or spatial extent smaller than ~10 arcsec.
  • Gaussian. Gaussian surface brightness profile (gaussian sigma larger than ~10 arcsec)
  • Radial. User-specified custom radial surface brightness profile
  • Flat. Flat sky brightness profile

Please be aware that there are some limitations to diffuse target ARF computations. First, the data grid used to tabulate spatial information is ~10 arcsec. Therefore, targets whose spatial extents are smaller than this size will not gain any benefit from specifying them as diffuse; it is better to enter them as point-like. For extremely broad surface brightness profiles, computation time may become overwhelming; users should consider if the "flat" diffuse profile may be more useful.

Please see the help file for nivignette for more information about specifying diffuse target information.

Generating an RMF for Your Observation

Prerequisites

We assume the following:

  • The user has a working software and calibration setup
  • Spectrum named myspectrum.pha which is accumulated from your cleaned event file. The spectrum file should contain a GTI extension (which is standard)
  • A weighting file myspectrum_wt.lis was generated by nicerarf (see above)

To generate an RMF for your spectrum, issue the following command: nicerrmf myspectrum.pha niNNNNNNNNNN.mkf myspectrum.rmf \ detlist=@myspectrum_wt.lis where

  • myspectrum.pha is your NICER spectrum file, which should include a GTI extension
  • niNNNNNNNNNN.mkf is the filter file
  • myspectrum.rmf is the name of the output RMF
  • myspectrum_wt.lis is the weighting file generated by nicerarf as described above

Upon completion, nicerrmf creates the RMF file (myspectrum.rmf) in this example. It is weighted according to the weighting coefficients supplied from the myspectrum_wt.lis file, as generated in the ARF stage described above.

What If I Want to Keep the Precomputed Responses and Not Use the Calculator?

What happens if you want to use the "old" method of using precomputed responses instead of the new response calculators?

NICER is committed to maintaining an online presence for the precomputed responses released in July 2020 for at least six months after the next big calibration release. This means that the NICER team will maintain the precomputed responses on-line until at least February, 2022. After this date, the precomputed responses may become unavailable.

However, the NICER team will not update the precomputed responses. This means that although they are available for download, they will not represent the state of the art, and most importantly cannot be used for off-axis or high optical loading conditions.

What If I Want to Generate an RSP File, or do ARF weighting?

In some circumstances the user may wish to generate an "RSP" file, which is the joint combination of the RMF and ARF into a single response file. This may ease data analysis downstream, since it requires less files.

To generate an RSP file, nicerarf should be run with the following additional options nicerarf ... savedetarf=YES outdetlisfile=myspectrum_arf.lis and then nicerrmf run with these options nicerrmf ... outmode=RSP detlist=@myspectrum_arf.lis

If one desires the RMF to be weighted by the ARF, but still remain a separate file, then use the following options to nicerrmf instead. nicerrmf ... outmode=RMF detlist=@myspectrum_arf.lis

Using ARF and RMF files in XSPEC

Once you have retrieved ARF and RMF files (either from CALDB, or generated as described below), you can load them into XSPEC.

First, be sure you have a spectrum. Load it into XSPEC with a command like this: data 1:1 myspectrum.pha Obviously your spectrum file name may be different, and there are many advanced usages available using the "data" command.

Next you can load the ARF and RMF with commands like this: resp 1:1 myspectrum.rmf arf 1:1 myspectrum.arf These should be the ARF and RMF paths you generated, as described above. The exact path and file names may be different from this example. Note that, in XSPEC, you should load the RMF file before loading the ARF.

At this point you should have successfully loaded an ARF and RMF, and you are ready for spectral analysis.

Troubleshooting

If you are having trouble generating NICER responses, we have additional documentation which may help you.

NICER Response Common Issues discusses common issues and user mistakes related to response generation, and how to correct them.

Common NICER Error Messages and How to Fix Them discusses many different NICER software error conditions including those related to response generation.

Modifications

  • 2020-10-16 - fixed typos
  • 2020-07-22 - initial draft
  • 2021-04-16 - add navigation bar
  • 2021-07-16 - add new nicerarf and nicerrmf tasks, move pre-computed response information to its own page
  • 2021-08-09 - add warning about not using barycentered event files
  • 2021-09-01 - add warnings about NICER response bugs
  • 2021-10-13 - add more background discussion about response features
  • 2021-11-30 - correct typo in XSPEC "response" command
  • 2022-04-25 - link to Troubleshooting topics; clarify no barycentered files