INITIAL RESULTS FROM NuSTAR OBSERVATIONS OF THE NORMA ARM

The NuSTAR data consist of nine pointings whose details are summarized in Table 1. These nine pointings are comprised of two focal plane modules A and B (FPMA and FPMB) each having a field-of-view (FOV) of 13' × 13'. To increase sensitivity, adjacent pointings were tiled with significant overlap (∼50%) resulting in sky region covered by the survey of around 0.2 deg² (04 × 04), centered at (J2000.0) R.A. =16^h33^m47^s and decl. =−47°32'14''. In Galactic coordinates, this is l = 3367776 and b = 01825.

Data analysis relied on HEASoft 6.14 and the NuSTAR Data Analysis Software (NuSTARDAS 1.2.0¹⁴) with the latest calibration database files (CALDB: 2013 August 30). Raw event lists from FPMA and FPMB were reprocessed using nupipeline¹⁵ in five energy bands: 3–10 keV, 3–79 keV, 10–40 keV, 10–79 keV, and 40–79 keV.

2.1.1. Image Cleaning

2.1.2. Systematic Offset of Detected Sources

We ran wavdetect on individual event lists, and on the mosaic images, in order to create lists of detected sources in each energy band. In all cases, we assumed: a point-spread function (PSF) with a constant full-width at half-maximum (FWHM) of 18'' (Harrison et al. 2013); scale sizes of 1, 2, 4, and 8 pixels; and a threshold of 10⁻⁵. This threshold implies around 1 spurious source per observation. For the mosaics, we used the cleaned (non-ghost-ray removed) images assuming a background map that mimics the observed low-frequency (i.e., large scale) ghost ray patterns with wavelet scales with characteristic sizes of 8–32 pixels (Slezak et al. 1994; Starck & Murtagh 1994; Vikhlinin et al. 1997). Each pixel is 2'' wide, so the wavelet scales are 16''–64'', i.e., larger than the high-frequency scales expected for point sources. The lack of high-frequency scales in the background map leads to a poor modeling of the sharp edges and dark dips of the ghost ray pattern. This results in a large number of source detections that align with artifacts in the image, and we conclude that they are likely spurious.

We visually inspected the event lists and the mosaic images (in each band) searching for NuSTAR-detected sources that were coincident with Chandra sources (see Section  2.2). There are three NuSTAR-detected sources that have probable Chandra counterparts. The NuSTAR-derived positions show a systematic offset (i.e., with a similar direction and magnitude) with respect to the Chandra positions. In physical coordinates, this offset is +1.98 pixels (39) and +4.75 pixels (95) in the x and y directions, respectively, found by averaging the offsets of the Chandra sources. This is consistent with the expected performance from NuSTAR (Harrison et al. 2013). Therefore, we registered the mosaic images to the Chandra sources by subtracting these offset values from the reference pixel. We reran wavdetect to determine a final source position, positional uncertainty (quoted at 90% confidence), and detection significance in the 3–79 keV band. These values are reported in Table 2.

Table 2. NuSTAR-detected Sources and Their Angular Separation from Likely Chandra Counterparts

Name	R.A.	Decl.	90% Confidence Radius	Detection Significance	Offset
	(deg)	(deg)		(σ)
CXOU J163329.5−473332	248.37254	−47.55894	79	8.3	19
CXOU J163350.9−474638	248.46158	−47.77642	130	15.0	41
CXOU J163355.1−473804	248.48046	−47.63520	70	8.7	27

Notes. Results for two other sources detected by NuSTAR, 4U 1630−47 and NuSTAR J163433−473841, are presented in separate papers (King et al. 2014; Tomsick et al. 2014).

Download table as:  ASCIITypeset image

2.1.3. Spectral and Timing Analyses

In 2011, Chandra observed a ∼20 × 08 section of the Norma Arm, a subset of which is the ∼04 × 04 NuSTAR survey region described in this paper. Of the ∼1100 X-ray sources detected by Fornasini et al. (2014, submitted), we excluded all objects outside the 0.2 deg² NuSTAR survey region, and then rejected those whose ratio between net source counts in the 2–10 keV and 0.5–2 keV energy bands was less than 0.8. This yields a catalog of 22 relatively hard sources that are suitable low-energy X-ray counterpart candidates to sources detected at higher energies with NuSTAR.

Observations used in this study are ObsID 12532 and ObsID 12533. Reprocessing and reduction of this data relied on CIAO version 4.5. Spectra were extracted from each event list in the 0.3–10 keV band for a source region centered on the Chandra position (a circle of radius =10''), and for a source-free background region (a rectangle with dimensions: 200'' × 100'') on the same detector chip. Spectral data were grouped to contain a minimum of 20 source+background counts per bin.

NuSTAR detects a source at the 8.3σ level whose position (Table 2) is 19 away from, and compatible with, that of CXOU J163329.5−473332. The source appears in NuSTAR ObsID 5 and ObsID 6. However, it falls in the chip gap and among the ghost rays during ObsID 5, and so only data from ObsID 6 was used for spectral and timing analyses (Figure 2).

Zoom In Zoom Out Reset image size

Chandra spectral data from ObsID 12533 were fit with an absorbed power law yields N_H $=(12_{-9}^{+14})\times 10^{22}$ cm⁻² and $\Gamma = 1.2_{-1.8}^{+2.2}$ ($\chi _{\nu }^{2}/\mathrm{dof} = 0.6/3$). There were 125 ± 12 net source counts in 0.3–10 keV, distributed as 14 ± 4 cts (0.3–3 keV) and 111 ± 11 cts (3–10 keV). Using Cash (1979) statistics and Pearson (1900) χ² test statistics on unbinned Chandra data give consistent results.

An absorbed power law was then fit to the NuSTAR data only. With N_H fixed to the best-fit value from Chandra, we measure a photon index $\Gamma = 2.6_{-1.2}^{+1.3}$ that is consistent with the one from Chandra. The source emitted 120 ± 22 net counts in the NuSTAR energy band (3–24 keV), and nearly all of them (105 ± 20) were recorded below 10 keV.

We fit the combined spectra from Chandra and NuSTAR using a cross-instrumental constant fixed at unity for the Chandra data, and allowed to vary for the NuSTAR data. The best-fitting model parameters for the power-law model are N_H $= (17_{-7}^{+10})\times 10^{22}$ cm⁻² and $\Gamma = 2.0_{-1.1}^{+1.2}$ (Figure 3 and Table 3). A fit of equivalent quality ($\chi _{\nu }^{2}/\mathrm{dof} = 1.3/7$) is obtained with an absorbed blackbody model of temperature $kT = 2.0_{-0.6}^{+1.1}$ keV, and a lower column density N_H $=(7_{-3}^{+6})\times 10^{22}$ cm⁻². The instrumental constant ($0.9_{-0.5}^{+0.7}$ for the power law, $0.9_{-0.5}^{+0.6}$ for the blackbody), which is consistent with pre-flight expectations for NuSTAR (Harrison et al. 2013) and with joint Chandra–NuSTAR spectral fits of other Galactic objects (e.g., Gotthelf et al. 2014), indicates little variability in source flux between observations taken nearly 2 yr apart. Fitting the joint spectral data with a bremsstrahlung model leads to an unconstrained plasma temperature (≈12 keV).

Zoom In Zoom Out Reset image size

Table 3. Parameters from Absorbed Power Law (PL), Blackbody (BB), and Bremsstrahlung (FF) Models Fit to the Joint Chandra and NuSTAR Spectral Data for Sources in the Norma Arm Survey

Name	Model	Ca	NHb	Γ or kTc	Norm.d	$\chi _{\nu }^{2}/$dofe	Sf	Hg	HRh	Obs. Fluxi	Unabs. Fluxj
CXOU J163329.5−473332	PL	$0.9_{-0.5}^{+0.7}$	$17_{-7}^{+10}$	$2.0_{-1.2}^{+1.1}$	2.1	1.2/7	105 ± 20	15 ± 8	−0.7 ± 0.3	12.5 ± 5.2	26.2 ± 13.6
	BB	$0.9_{-0.5}^{+0.6}$	$7_{-3}^{+6}$	$2.0_{-0.6}^{+1.1}$	0.06	1.3/7				10.9 ± 2.7	13.2 ± 3.6
	FF	$0.9_{-0.5}^{+0.7}$	$15_{-5}^{+7}$	≈12⋆	1.3	1.2/7				12.0 ± 2.1⋆	21.2 ± 5.9⋆
CXOU J163350.9−474638	PL	$3.8_{-0.7}^{+0.9}$	$21_{-4}^{+6}$	3.7 ± 0.5	41.3	1.1/22	375 ± 28	25 ± 10	$-0.9_{-0.1}^{+0.2}$	29.1 ± 9.3	362 ± 238
	BB	$3.6_{-0.7}^{+0.9}$	$9_{-2}^{+3}$	$1.2_{-0.1}^{+0.2}$	0.06	1.3/22				25.8 ± 2.2	39.6 ± 10.9
	FF	$3.7_{-0.7}^{+0.9}$	$15_{-3}^{+4}$	$3.3_{-0.7}^{+1.0}$	4.0	1.1/22				27.6 ± 5.3	77.2 ± 31.3
CXOU J163355.1−473804†	PL	1.0 ± 0.3	6 ± 1	1.5 ± 0.3	1.5	2.0/35	256 ± 26	52 ± 12	−0.7 ± 0.2	32.8 ± 3.2	43.2 ± 7.1
	BB	1.0 ± 0.2	$1.2_{-0.4}^{+0.5}$	$2.1_{-0.2}^{+0.3}$	0.1	1.3/35				28.1 ± 3.7	29.1 ± 2.8
	FF	1.1 ± 0.3	$5.2_{-0.9}^{+1.1}$	$21_{-9}^{+32}$	1.9	1.8/35				32.7 ± 10.3	41.0 ± 18.6

Notes. ^aInstrumental constant fixed to 1 for the Chandra data and allowed to vary for the NuSTAR data. ^bColumn density in units of 10²² cm⁻². ^cPhoton index of the power law (PL) model, or plasma temperature (in keV) for the blackbody (BB) and bremsstrahlung (FF) models. ^dModel normalization (× 10⁻⁴). ^eReduced χ² over degrees of freedom (dof). ^fNet source counts from both NuSTAR modules combined in the soft (S) band: 3–10 keV. ^gNet source counts from both NuSTAR modules combined in the hard (H) band: 10–24 keV. ^hHardness ratio defined as (H − S)/(H + S). ⁱObserved flux (i.e., not corrected for absorption) in units of 10⁻¹³ erg cm⁻² s⁻¹ in the 0.3–24 keV band. ^jAbsorption-corrected flux in units of 10⁻¹³ erg cm⁻² s⁻¹ in the 0.3–24 keV band. ^⋆The fluxes are derived by fixing the plasma temperature to 12 keV. ^†The best fitting model for CXOU J163355.1−473804 requires a Gaussian component at 6.7 keV.

Download table as:  ASCIITypeset image

Figure 4 presents the light curve (3–24 keV, 100 s resolution) of CXOU J163329.5−473332. We searched for periods in the range of 0.004 s (i.e., twice the time resolution of the light curve data used in this fine timing analysis) to 18959 s (i.e., the observation duration), and we did not detect a significant pulsation signal in the soft (3–8 keV), hard (8–24 keV), or broad energy band (3–24 keV).

Zoom In Zoom Out Reset image size

The field of CXOU J163329.5−473332 was observed by XMM-Newton and this object was detected as part of the XMM-Newton Serendipitous Survey Catalog (Watson et al. 2009), although it appears faint and far off-axis (≳10 arcmin). We analyzed observation ID 0654190201 (rev. 2051), which was taken in 2011 February, with a total exposure time of 22 ks. The parameters from the XMM-Newton spectral fit of this source, i.e., the observed flux, column density, photon index, and blackbody temperature, are consistent with those derived from fits to the Chandra, NuSTAR, and combined Chandra–NuSTAR spectra.

We observed the near-infrared counterpart to CXOU J163329.5−473332 with the NEWFIRM telescope and its magnitudes are J = 15.29 ± 0.07 mag, H = 11.92 ± 0.10 mag, and K_s = 10.13 ± 0.06 mag (Rahoui et al. 2014, in press). The infrared spectrum displays strong CO lines in absorption (at 16198 Å and 22957 Å), a number of weak emission lines, and no Br-γ line. This spectrum is typical of an early MIII-type star feeding a small accretion disk (Rahoui et al. 2014, in press).

In ObsID 1 (Figure 2), NuSTAR detects a source at a significance of 15.0σ (Table 2) whose position is 41 from, and compatible with, the Chandra position of CXOU J163350.9−474638.

The Chandra spectral data (ObsID 12532) were fit with an absorbed power law to give N_H $= (13_{-5}^{+8})\times 10^{22}$ cm⁻² and $\Gamma = 2.0_{-1.1}^{+1.3}$ ($\chi _{\nu }^{2}/\mathrm{dof} = 0.8/6$). A thermal blackbody model ($kT = 1.4_{-0.4}^{+0.7}$ keV) fit to the data yields a similar column (N_H $= (8_{-3}^{+5})\times 10^{22}$ cm⁻²) and fit quality ($\chi _{\nu }^{2}/\mathrm{dof} = 1.1/6$). There are 190 ± 14 net source counts in the 0.3–10 keV range, with 30 ± 6 counts having energies below 3 keV, and the rest (159 ± 13) are above 3 keV.

We then fit an absorbed power law to the NuSTAR data alone while fixing N_H to the best-fit value from Chandra. The fit quality is decent ($\chi _{\nu }^{2}/\mathrm{dof} = 1.2/14$) and the photon index (Γ = 3.3 ± 0.3) is consistent with the value measured with Chandra. An absorbed blackbody provides an acceptable fit ($\chi _{\nu }^{2}/\mathrm{dof} = 1.4/22$) with a temperature kT = 1.1 ± 0.1 keV, similar to that measured with Chandra. The source emitted 400 ± 30 net counts in 3–24 keV, with most of them (375 ± 28) below 10 keV.

The spectra from Chandra and NuSTAR were jointly fit with an absorbed power law yielding N_H $=(21_{-4}^{+6})\times 10^{22}$ cm⁻² and Γ = 3.7 ± 0.5 (Figure 3 and Table 3). Although the fit quality is good ($\chi _{\nu }^{2}/\mathrm{dof} = 1.1/22$), the cross-instrumental constant is $3.8_{-0.7}^{+0.9}$, which indicates significant variability on year-long timescales.

Adding an exponential cutoff constrains the break energy (E_cut ⩽ 13 keV). However, this component is not required by the data since it returns a similar $\chi _{\nu }^{2}$ with 2 less dof. The measured N_H is larger in the joint fit than in the Chandra data alone, and fixing the column density to the Chandra value leads to a poorer fit ($\chi _{\nu }^{2}/\mathrm{dof} = 1.7/23$).

Thermal models also provide good fits to the data. A blackbody model ($\chi _{\nu }^{2}/\mathrm{dof} = 1.3/22$) gives a lower column density than for the power law (N_H $=(9_{-2}^{+3})\times 10^{22}$ cm⁻²), and has a temperature of $kT = 1.2_{-0.1}^{+0.2}$ keV. A bremsstrahlung model ($\chi _{\nu }^{2}/\mathrm{dof} = 1.1/22$) has an absorbing column consistent with the power law model (N_H $=(15_{-3}^{+4})\times 10^{22}$ cm⁻²), and a plasma temperature of $kT = 3.3_{-0.7}^{+1.0}$ keV (a value that is not constrained with the Chandra data alone).

The 3–24 keV light curve binned at 100 s is presented in Figure 4, and it shows CXOU J163350.9−474638 to be a relatively soft source that displays low variability on short timescales. The background is mostly due to 4U 1630−47 whose ghost-ray halo covers the extraction regions used to produce the light curves. The apparent decrease in background counts is not significant. There are no periodicities detected in the range of 0.004 s to 18407 s in any energy range. An upper limit of ∼30% (at 90% confidence) is derived for the fractional rms expected for a periodic signal.

This source appears in two Chandra observations (ObsID 12532 and ObsID 12533); the spectral data from these observations were summed to give 546 ± 24 net source counts (0.3–10 keV), divided into 168 ± 13 and 377 ± 20 net counts in the 0.3–3 keV and 3–10 keV bands, respectively. A power law model fit to the binned spectral data provides an adequate fit ($\chi _{\nu }^{2}/\mathrm{dof} = 1.3/24$) with N_H =(3 ± 1) × 10²² cm⁻² and a flat photon index: $\Gamma = 0.7_{-0.3}^{+0.4}$. A blackbody of temperature $kT = 1.9_{-0.3}^{+0.4}$ keV improves the fit slightly ($\chi _{\nu }^{2}/\mathrm{dof} = 1.2/24$).

The likely hard X-ray counterpart to CXOU J163355.1−473804 is detected at the 8.7σ level (3–79 keV) in the NuSTAR mosaic image. Ghost-ray photons contaminate the region around the source in ObsID 4, and so spectral and timing analysis relied only on data from ObsID 5. The 321 ± 28 net source counts (3–24 keV) were distributed as 256 ± 26 net counts in 3–10 keV, and 52 ± 12 in 10–24 keV.

We fit the NuSTAR data with power law and blackbody models holding N_H fixed to the best-fit value from Chandra, and derived a steeper photon index (Γ = 1.9 ± 0.3) or a plasma temperature consistent with that of Chandra (kT = 2.2 ± 0.3 keV), with both models giving poor fits ($\chi _{\nu }^{2}/\mathrm{dof} = 2.3/9$ and $\chi _{\nu }^{2}/\mathrm{dof} = 1.8/9$, respectively).

Jointly fitting the Chandra and NuSTAR data gives a poor fit ($\chi _{\nu }^{2}/\mathrm{dof} = 2.0/35$) when using only an absorbed power law: N_H =(6 ± 1) × 10²² cm⁻² and Γ = 1.5 ± 0.3 (Figure 3 and Table 3). The fit quality improves ($\chi _{\nu }^{2}/\mathrm{dof} = 1.3/35$) with a blackbody model having $kT = 2.1_{-0.2}^{+0.3}$ keV and a lower column density (N_H $= (1.2_{-0.4}^{+0.5})\times 10^{22}$ cm⁻²), or with a power law and exponential cutoff ($\chi _{\nu }^{2}/\mathrm{dof} = 1.4/30$) where N_H =(3 ± 1) × 10²² cm⁻², Γ = 0.6 ± 0.4, and the cutoff energy is $5_{-1}^{+3}$ keV. In both cases, the instrumental constant is 1.1 ± 0.2 suggesting little variability over yearlong timescales.

Residuals remain around 6.7 keV where emission from the fluorescence of ionized iron is expected. Indeed, the best spectral fits are obtained when a Gaussian component (σ = 0) is added to either the cutoff power law or the blackbody model. In order to analyze this line, we rebinned the NuSTAR spectra to have at least 20 source+ background counts and a minimum significance of 2σ. For the power law ($\chi _{\nu }^{2}/\mathrm{dof} = 1.1/48$), N_H =(2 ± 1) × 10²² cm⁻², with $\Gamma = 0.0_{-1.0}^{+0.6}$ and an exponential cutoff at $4_{-1}^{+3}$ keV. The line centroid is $6.72_{-0.08}^{+0.04}$ keV with an equivalent width of ∼500 eV (unconstrained).

For the blackbody model ($\chi _{\nu }^{2}/\mathrm{dof} = 1.2/50$), the line centroid is $6.7_{-0.2}^{+0.1}$ keV with an equivalent width of $414_{-312}^{+370}$ eV. The column density and blackbody temperature are N_H $= (1.2_{-0.5}^{+0.6})\times 10^{22}$ cm⁻² and $kT = 2.0_{-0.2}^{+0.3}$ keV, respectively. The radius of the emitting region implied by the blackbody model is very small (0.03–0.16 km assuming source distances of 2–10 kpc). Either the source is very distant (⩾20 kpc) or the blackbody is not the right model.

We replaced the blackbody continuum with a bremsstrahlung ($\chi _{\nu }^{2}/\mathrm{dof} = 1.3/50$) and obtained N_H =(5 ± 1) × 10²² cm⁻², $kT = 16_{-5}^{+14}$ keV, a line energy of $6.74_{-0.06}^{+0.05}$ keV, and an equivalent width of $911_{-365}^{+553}$ eV. We also modeled the continuum with apec ($\chi _{\nu }^{2}/\mathrm{dof} = 1.3/51$) and the resulting iron abundance is at least 40% greater than Solar (N_Fe ⩾ 1.4) with a plasma temperature of $kT = 12_{-3}^{+7}$ keV.

The NuSTAR light curve (3–24 keV) for CXOU J163355.1−473804 is presented in Figure 4. No coherent pulsations were detected for search periods ranging from 2 ms to ∼21 ks.

The likely infrared counterpart to CXOU J163355.1−473804 was observed with NEWFIRM giving magnitudes of J = 16.43 ± 0.07 mag, H = 15.45 ± 0.10 mag, and K_s = 14.99 ± 0.09 mag (Rahoui et al. 2014, in press). With a weak CO line at 16198 Å, a strong CO line at 22957 Å, and weak Br-γ emission, the infrared spectrum is typical of a late GIII-type star.

The Chandra position for CXOU J163329.5−473332 is encompassed by the 21 uncertainty radius of an INTEGRAL-detected source named IGR J16336−4733 (Krivonos et al. 2010), which was also detected in a short observation by Swift (Landi et al. 2011). The flux recorded by Swift-XRT (2–10 keV) and by NuSTAR (3–10 keV) translate to X-ray luminosities of 1.9 × 10³⁴[d/10 kpc]² erg s⁻¹, and 7.9 × 10³³[d/10 kpc]² erg s⁻¹, respectively. The available X-ray data of CXOU J163329.5−473332 show it to be a faint, absorbed (N_H ≳ 10²³ cm⁻²), and relatively hard X-ray source (the bulk of its photons are emitted in 3–10 keV).

Thus, CXOU J163329.5−473332 could be a faint low-mass X-ray binary (LMXB: e.g., Degenaar & Wijnands 2009) or a CV (Kuulkers et al. 2006) of the intermediate polar (e.g., Patterson 1994) variety due to the hard X-ray detection. The detection of CXOU J163329.5−473332 out to ∼20 keV with a moderately steep photon index ($2.4_{-0.8}^{+0.9}$) and low X-ray luminosity is consistent with both classifications. Another possibility is a binary system in which the compact object is a non-accreting magnetar (e.g., Thompson & Duncan 1996).

These NuSTAR observations of CXOU J163350.9−474638 extend the source spectrum beyond 10 keV. However, the source demonstrates significant variability in intensity (by at least a factor of four) over the 2 yr separating the Chandra and NuSTAR observations, which makes it difficult to draw firm conclusions from joint-fitting of the broadband X-ray spectral energy distribution.

Nevertheless, it is possible to compare the spectral parameters derived from single-instrument fits. The photon index is steeper in the NuSTAR data (by ∼50%) compared with the value measured with Chandra. This is not uniquely due to the fact that NuSTAR covers higher X-ray energies, since ∼90% of the photons recorded by NuSTAR were below 10 keV, i.e., in an energy range covered by Chandra. On the other hand, thermal models also fit the data well, and the blackbody temperature (kT = 1.2 ± 0.2 keV) and column density (N_H $= (9_{-4}^{+5})\times 10^{22}$ cm⁻²) are in agreement for both Chandra and NuSTAR spectra.

There are no catalogued IR/optical objects from Vizier¹⁶ or in the Vista Variables in the Via Lactae Survey (Minniti et al. 2010) compatible with the Chandra position. Thus, CXOU J163350.9−474638 lacks a stellar counterpart that would rule out a CV or XRB located nearby, while the steep power law disfavors an AGN. Given its thermal spectrum, its long-term variability, and the absence of multi-wavelength counterparts, we conclude that CXOU J163350.9−474638 could be a LMXB situated a large distance away, or perhaps an isolated, magnetized NS (i.e., a magnetar).

Prior to the NuSTAR survey, Chandra found CXOU J163355.1−473804 to be a relatively bright X-ray source with a hard spectral continuum. As the brightest of the three objects in this study, this permitted us to measure the source's broadband X-ray spectrum with relatively high precision. The spectrum combining Chandra and NuSTAR data is consistent with a cutoff power law of $\Gamma = 0.0_{-1.0}^{+0.6}$ and $E_{\mathrm{cut}}=4_{-2}^{+3}$ keV. Thermal models such as a blackbody with $kT = 2.0_{-0.2}^{+0.3}$ keV or a bremsstrahlung with $kT = 16_{-5}^{+14}$ keV also describe the data well, although the implied size of the emission region is not consistent with the blackbody model. The column density required by the best-fitting models (N_H ≲ 3 × 10²² cm⁻²) is lower than measured for the two other sources in the study, indicating that the source is either less intrinsically absorbed than the others, or more likely, that it is closer to us.

With NuSTAR, we are able to confirm the detection of an iron line that is hinted at in the Chandra data. The line energy of 6.7 keV suggests thermal Kα emission from highly ionized, helium-like iron (Fe xxv) in the optically thin plasma around an accreting white dwarf, i.e., a CV (e.g., Hellier & Mukai 2004; Pandel et al. 2005; Kuulkers et al. 2006). For example, EX Hya and V405 Aur are CVs that show a 6.7 keV line with equivalent widths ∼400–900 eV, i.e., consistent with the equivalent width measured in CXOU J163355.1−473804 (Hellier et al. 1998).

The identification of the infrared counterpart as a cool, GIII star supports the CV classification. Another factor favoring a CV nature for CXOU J163355.1−473804 is the apparent lack of change in intensity or spectrum during the 2 yr separating the Chandra and NuSTAR surveys, with no indication from all-sky X-ray monitors that the system underwent a major outburst (L_X ≳ 10³⁶ erg s⁻¹) in that time (or at any time in the past few decades).

Its lower absorbing column compared with the other sources in the survey suggests that CXOU J163355.1−473804 is at a distance of 2 or 3 kpc at most, i.e., in the Crux Arm, or in the nearest arc of the Norma Arm. At an assumed distance of 3 kpc, the absorption-corrected flux (0.3–79 keV) of the bremsstrahlung model translates to an X-ray luminosity of 5 × 10³³ erg s⁻¹. This is consistent with the persistent X-ray luminosity expected from a CV (e.g., Muno et al. 2004; Kuulkers et al. 2006).

Of the 22 hard Chandra sources in the survey region, 3 were detected by NuSTAR, and they ranked first, second, and fourth in order of the number of hard X-ray (⩾3 keV) counts recorded by Chandra. The third brightest source in the hard Chandra band is CXOU J163358.9−474214. This source was not detected in the NuSTAR event lists and mosaic images, despite the fact that it was located in a relatively ghost-ray free and stray-light free part of the image in ObsID 1. This indicates a variable nature for this object (significant variability was also observed with Chandra), and we establish a 3σ upper limit of 7 × 10⁻¹³ erg cm⁻² s⁻¹ on the absorbed source flux in the 3–10 keV range, i.e., higher than the average flux registered by Chandra in a similar energy band (2–10 keV: Fornasini et al. 2014, submitted). The Chandra error circle for this source contains a counterpart candidate seen in the near-IR by Two Micron All Sky Survey, and in the mid-IR by Spitzer and Wide Field Infrared Survey Explorer. The X-ray variability and the possible association with an IR-emitting source suggest a LMXB or a CV.

All other Chandra sources in the NuSTAR survey region had less than 35 cts in the hard Chandra band, which means they are too faint to be detected by NuSTAR given the exposure depth of this survey.

Sections of the Norma field have been observed by XMM-Newton and source candidates found therein are listed in the XMM-Newton Serendipitous Survey Catalog (Watson et al. 2009). Of the ∼150 sources in the field, 22 of these are both relatively bright (flux in the 0.2–12 keV band ⩾5 × 10⁻¹³ erg cm⁻² s⁻¹) and hard (hardness ratio between the 2–4.5 keV and 4.5–12 keV bands ⩾ 0.0). Only one of them coincides spatially with the error circle of a NuSTAR source: CXOU J163329.5−473332. It is one of the hardest sources (ranked 6th hardest out of 22), but it is also among the faintest (ranked 19th in flux out of 22).

Besides the analysis of X-ray sources, one of the primary goals of this pilot study is to optimize the strategy for future observations. Our experiences with this mini-survey showed us that some of our strategic choices were sound and some can be improved.

Based on the results of the Chandra survey, we knew that the mini-survey region contained several sources that NuSTAR could detect. As was done here, observers should select regions in such a way that they encompass the largest number of hard Chandra sources (or, when available, XMM-Newton sources) that are relatively bright, but not so bright that their ghost rays and stray light contaminate adjacent observations. With the exposures available in this survey (10–100 ks), NuSTAR was able to detect three out of four X-ray sources that had more than ∼100 cts in the hard Chandra band (⩾3 keV). The non-detection of the fourth source still gives the useful result that the source is variable. While this Chandra hard-band count rate could be used as a rule-of-thumb for a source's detectability in a typical mini-survey such as this, it is no guarantee since it does not account for X-ray sources that are variable, or that were in the soft state during the NuSTAR survey.

Another factor that led to the selection of this region was that we expected it to contain a relatively low level of stray light given the satellite's roll angle at the time the observations were performed. Even if stray light were to affect one or both of the modules, substitute coverage is available from the overlapping module and/or adjacent observation(s).

The value of exposure redundancy, not only thanks to the two modules but also by tiling observations with significant overlap (∼50% shifts), can not be overstated for eliminating or reducing imaging artifacts. This is an important factor that greatly facilitated the analysis of the faint sources in this study. Further improvements in this direction can be made by dividing up the 25 ks exposures into two or three 10–15 ks exposures tiled with slightly more overlap (roughly 2/3) between adjacent pointings. While data with more overlap will take more time to analyze (i.e., the spectra from separate observations will need to be merged to obtain meaningful statistics) the tradeoff is increased exposure redundancy in case pixels need to be discarded due to ghost rays or stray light.

Observers who wish to use NuSTAR for galactic surveys can prevent or reduce the effects of stray light and ghost rays in two ways: (1) by using opportunistic observations gathered only when known transients are off or emitting at low levels according to wide-field X-ray monitors such as MAXI, Swift-BAT, and INTEGRAL-ISGRI; and (2) by increasing the exposure redundancy. While we underestimated its effects during the planning of this survey, we now know more about the brightness and extent of the ghost-ray pattern from objects such as 4U 1630−47, which will help guide the selection of future surveys.

An open question is whether NuSTAR should continue to survey "regions" rather than using the observing time to place the most promising targets from these regions on axis. However, it is important to note that a targeted approach might have missed the discovery of the new X-ray transient NuSTAR J163433−473841.

While there are technical challenges, there are also tremendous scientific benefits from surveying the Galaxy with NuSTAR. Understanding the disk-wind connection in 4U 1630−47, the serendipitous discovery of NuSTAR J163433−473841, and insights into the faint members of the galactic X-ray population are primary among these. Surveys allow NuSTAR to offer a complete picture of the Inner Milky Way, which will add to our knowledge of the content of our host galaxy and unlock new mysteries.