The ^{55}Fe X-ray Energy Response of Mercury Cadmium Telluride Near-Infrared Detector Arrays

A technique involving ^{55}Fe X-rays provides a straightforward method to measure the response of a detector. The detector’s response can lead directly to a calculation of the conversion gain (e^- ADU^{-1}), as well as aid detector design and performance studies. We calibrate the ^{55}Fe X-ray energy response and pair production energy of HgCdTe using 8 HST WFC3 1.7 \micron flight grade detectors. The results show that each K$\alpha$ X-ray generates 2273 \pm 137 electrons, which corresponds to a pair-production energy of 2.61 \pm 0.16 eV. The uncertainties are dominated by our knowledge of the conversion gain. In future studies, we plan to eliminate this uncertainty by directly measuring conversion gain at very low light levels.

💡 Research Summary

The paper presents a straightforward, physics‑based method for calibrating the absolute response of HgCdTe near‑infrared (NIR) detector arrays using the well‑known X‑ray emission from a ^55Fe radioactive source. ^55Fe decays by electron capture and emits Mn K‑shell X‑rays, the dominant line being the Kα photon at 5.9 keV. Because the energy required to create an electron‑hole pair in a semiconductor (the pair‑production energy) is a material constant, measuring how many electrons a single Kα photon generates provides a direct route to both the pair‑production energy and the detector’s conversion gain (electrons per analog‑to‑digital unit, e⁻ ADU⁻¹).

Experimental setup
Eight flight‑grade HgCdTe detector arrays, identical to those used in the Hubble Space Telescope Wide Field Camera 3 (WFC3) 1.7 µm channel, were mounted in a cryogenic test dewar at ~77 K. A calibrated ^55Fe source was positioned close to the detector backside so that emitted X‑rays entered the detector bulk with minimal attenuation. The detectors were read out in their standard read‑reset mode, and the resulting ADU values for each X‑ray event were recorded. Prior to the X‑ray measurements, each array’s conversion gain was estimated using the conventional photon‑transfer curve (PTC) method, which relates signal variance to mean signal in the high‑flux regime.

Data analysis
For each X‑ray hit, the charge cloud was integrated over the affected pixels, and the total ADU was converted to electrons using the pre‑measured gain. By accumulating a large number of events, the authors obtained a statistical distribution of electrons per Kα photon. The mean number of electrons was found to be 2273 with a standard deviation of ±137 electrons. Dividing the photon energy (5.9 keV) by this mean yields a pair‑production energy of 2.61 eV per electron‑hole pair, with an uncertainty of ±0.16 eV. The uncertainty is dominated by the systematic error in the PTC‑derived gain; the PTC method assumes linearity and Gaussian noise at relatively high illumination levels, which may not hold at the few‑electron regime relevant for X‑ray calibration.

Key insights

1. Material‑specific pair‑production energy – The measured 2.61 eV is significantly lower than the 3.65 eV commonly quoted for silicon CCDs, confirming that HgCdTe requires less energy to generate a charge pair. This lower energy contributes to the higher quantum efficiency of HgCdTe in the NIR.
2. Direct gain verification – By comparing the X‑ray‑derived electron count with the PTC‑derived gain, the authors demonstrate an independent check on the conversion gain. The agreement within uncertainties validates the PTC approach for these devices but also highlights its limitation in the ultra‑low‑signal regime.
3. Uncertainty budget – The dominant error term is the gain calibration; other contributions (statistical spread of X‑ray events, charge diffusion across pixels, Kβ contamination) are comparatively minor.

Future work
To reduce the gain‑related uncertainty, the authors plan to implement a direct gain measurement at the single‑electron level using ultra‑low‑flux illumination (e.g., a calibrated photon‑counting LED or a laser attenuated to the photon‑starved regime). This approach would bypass the assumptions inherent in the PTC method and provide a true low‑signal gain value. Additionally, they intend to refine the charge‑diffusion model to correct for any loss of electrons that spread beyond the integration window, further tightening the error bars on the pair‑production energy.

Implications
Accurate knowledge of both the conversion gain and the pair‑production energy is essential for precise photometric calibration, noise modeling, and detector design. For space‑based missions such as JWST, Roman, and future NIR survey telescopes, these parameters directly affect the ability to detect faint sources, characterize exoplanet atmospheres, and measure cosmological signals. The ^55Fe X‑ray technique offers a simple, repeatable, and material‑agnostic calibration tool that can be applied not only to HgCdTe but also to other semiconductor detector technologies (e.g., InGaAs, Si:As). By establishing a clear pathway to reduce systematic uncertainties, this work paves the way for next‑generation NIR instruments with truly photon‑limited performance.