[Paper Review] The nearest neighbor statistics for X-ray source counts I.The method

A. Sołtan|arXiv (Cornell University)|Dec 31, 2010

Statistical and numerical algorithms3 citations

TL;DR

This paper introduces the nearest neighbor statistics (NNST) method to constrain X-ray source counts below the conventional detection threshold by analyzing spatial clustering in photon distributions. Applied to a 465-ks Chandra AEGIS observation, NNST extends source count estimates down to 2×10⁻¹⁷ erg cm⁻² s⁻¹ in the soft band, matching deep-field results and demonstrating effectiveness for probing extremely faint source populations.

ABSTRACT

Most of the X-ray background (XRB) is generated by discrete X-ray sources. It is likely that still unresolved fraction of the XRB is composed from a population of the weak sources below the present detection thresholds and a truly diffuse component. It is a matter of discussion a nature of these weak sources. The goal of the present paper is to explore the effectiveness of the nearest neighbor statistics (NNST) of the photon distribution for the investigation of the number counts of the very weak sources. All the sources generating at least two counts each induce a nonrandom distribution of counts. This distribution is analyzed by means of the NNST. Using the basic probability equations, the relationships between the source number counts N(S) and the NNST are derived. It is shown that the method yields constraints on the N(S) relationship below the regular discrete source detection threshold. The NNST was applied to the medium deep Chandra pointing to assess the source counts N(S) at flux levels attainable only with the very deep exposures. The results are in good agreement with the direct source counts based in the Chandra Deep Fields (CDF). In the next paper of this series the NNST will be applied to the the CDF to assess the source counts below the present flux limits.

Motivation & Objective

• Investigate the effectiveness of nearest neighbor statistics (NNST) for detecting very weak X-ray sources below standard detection thresholds.
• Address the challenge of estimating number counts for sources too faint to be individually detected in deep X-ray surveys.
• Assess whether NNST can provide reliable constraints on the X-ray source count function $N(S)$ at flux levels inaccessible to traditional source extraction.
• Validate the NNST method by comparing its results with direct source counts from the Chandra Deep Fields (CDF), establishing its credibility for future deep-field applications.

Proposed method

• Model the observed photon distribution as a combination of Poisson-distributed background counts and clustered counts from discrete point sources.
• Use nearest neighbor statistics (NNST) to analyze spatial clustering of photons, where the distribution of distances to the nearest source reveals source density and flux distribution.
• Derive analytical relationships between the source number count function $N(S)$ and the NNST using basic probability theory and Poisson statistics.
• Account for instrumental effects such as PSF variations and exposure gradients by incorporating them into the normalization factor $N_o$, which depends on the flux normalization $\mathcal{N}$ and the conversion factor $\text{cf}$.
• Apply the method to a 465-ks Chandra AEGIS pointing, using randomized source distributions to estimate uncertainties in the NNST-derived $N(S)$ slope.
• Correct for background contamination and exposure variations by calibrating the count normalization $N_o$ using the observed count distribution and energy band conversion factors.

Experimental results

Research questions

• RQ1 Can nearest neighbor statistics (NNST) detect and constrain the number counts of X-ray sources below the formal detection threshold of 4–5σ?
• RQ2 How accurately can NNST estimate the differential source count slope $b$ in the $N(S) \propto S^{-b}$ relation at flux levels below 10⁻¹⁶ erg cm⁻² s⁻¹?
• RQ3 To what extent does the NNST method recover source counts consistent with direct measurements from the Chandra Deep Fields (CDF)?
• RQ4 What are the dominant systematic uncertainties affecting NNST in the hard X-ray band (2–8 keV), and how do they limit its effectiveness?
• RQ5 Can NNST be extended to even deeper exposures (e.g., 2 Ms) to probe flux levels below 10⁻¹⁷ erg cm⁻² s⁻¹?

Key findings

• The NNST method successfully constrains the X-ray source count function $N(S)$ down to $2 \times 10^{-17}$ erg cm⁻² s⁻¹ in the 0.5–2 keV band, a factor of 5–10 below the standard detection threshold for this exposure.
• The derived source count slope $b = 1.595 \pm 0.035$ in the soft band is consistent with direct measurements from the Chandra Deep Fields (CDF), validating the method’s accuracy.
• In the hard band (2–8 keV), the NNST method yields a slope of $b = 1.676 \pm 0.340$, but with large statistical uncertainties due to lower source contribution and higher particle background contamination.
• Sources contributing $2 \leq k \leq 20$ counts in the hard band account for only $1.19^{+0.96}_{-0.42}$% of total counts, indicating weak clustering and low signal-to-noise.
• The uncertainty in the count normalization $N_o$ is estimated at ~3.5% due to PSF and exposure variations, which is small compared to the intrinsic uncertainty in $\mathcal{N}$.
• The method is expected to extend source count constraints to $\sim 4 \times 10^{-18}$ erg cm⁻² s⁻¹ in the soft band and $\sim 2 \times 10^{-17}$ erg cm⁻² s⁻¹ in the hard band in future 2 Ms exposures, though accuracy in the hard band remains limited.