MIRAC-5: Advancing Mid-IR Imaging

• MIRAC-5 is a ground-based, adaptive-optics mid-IR imager designed for high-resolution, high-contrast thermal observations of warm exoplanets and circumstellar disks.
• It features a next-generation GeoSnap detector, advanced cryogenic optics, and a robust chopping/nodding strategy to effectively manage noise and thermal background.
• Upcoming upgrades, including a coronagraphic module with MAPS integration, promise enhanced capabilities for detailed atmospheric diagnostics and detection of faint companions.

Searching arXiv for MIRAC-5 and related commissioning/coronagraph papers to ground the article in the cited literature. arxiv.search query: "MIRAC-5 Mid-Infrared Array Camera" Mid-Infrared Array Camera-5 (MIRAC-5) is the fifth incarnation of the Mid-Infrared Array Camera family, a ground-based, adaptive-optics-capable mid-infrared imager developed for high-spatial-resolution and high-contrast observations in the thermal infrared. Introduced as a new 3–13 micron camera built around a Teledyne GeoSnap detector and later commissioned on the 6.5-m MMT as a system sensitive from 2 to 13 μm, MIRAC-5 is intended for direct studies of warm exoplanets, wide-orbit companions, protoplanets, circumstellar disks, and other sources whose luminosity and composition are most accessible in the mid-IR. Its architecture combines a long-wave HgCdTe array, cryogenic optics, internal chopping, BLINC relay optics, and MAPS adaptive optics, with later development focused on an annular groove phase mask coronagraph for contrast-limited observations (Bowens et al., 2022).

1. Instrument lineage and scientific role

MIRAC-5 belongs to an instrument line that began in 1988 and has been repeatedly updated to track detector and adaptive-optics developments. It is described as a joint University of Arizona / University of Michigan effort and as the successor to MIRAC-3 and MIRAC-4. Relative to MIRAC-4, the new system replaces a problematic 256×256 Si:As BIB detector, which had image artifacts and limited astronomical utility, with a much larger 1024×1024 MCT array, thereby increasing field of view, improving detector quality, and broadening scientific reach (Bowens et al., 2022).

The instrument is designed to occupy a specific observational niche. Mid-IR observations probe cooler objects, penetrate dust more effectively than shorter wavelengths, and reduce contrast demands relative to reflected-light imaging. At the same time, large ground-based telescopes equipped with adaptive optics can deliver finer angular resolution and, in some regimes, stronger contrast performance than space-based systems. MIRAC-5 is therefore positioned as complementary to the James Webb Space Telescope rather than redundant, and commissioning work frames it as filling an important role before ELT-class mid-IR instruments become available (Bowens et al., 2024).

The science emphasis follows directly from this niche. MIRAC-5 targets direct imaging of warm astronomical systems that emit strongly in the mid-IR, especially protoplanetary disks, wide-orbit brown dwarfs, planetary companions in the contrast-limited regime, warm exoplanets and their atmospheric molecules, and dusty environments in which thermal emission traces temperature and composition (Bowens et al., 2024).

2. Detector, cryogenics, and optical configuration

At the core of MIRAC-5 is a Teledyne Imaging Sensors GeoSnap long-wave HgCdTe detector. The baseline GeoSnap format is 2048×2048, but the MIRAC-5 implementation uses a single 1024×1024 quadrant, and the initial device does not have the anti-reflection coating of the full baseline version. Even so, the reported quantum efficiency is about 65% across 3–13 microns, while the commissioned detector is described as sensitive from 2 to 13 μm at the half-peak QE cutoffs; the coated baseline device is expected to exceed 90% QE at 10 microns (Bowens et al., 2022).

The detector parameters reported for MIRAC-5 establish the system as a fast, high-background thermal-IR imager. The pixel pitch is 18 μm, the well depth is 1.2 million electrons, the readout integrated circuit can support up to 120 Hz full-frame operation, and standard imaging operation spans 0.1 to 85 Hz. The ROIC is CTIA-based and is intended to suppress persistence. Reported noise figures depend on context: the detector summary quotes 133 e− read noise, the original instrument description gives about 140 e−/pixel RMS single-frame read noise in low-gain mode, and on-sky pair-subtraction analysis gives 165 e−/pixel per frame. Dark-current figures are likewise context-dependent, with about 6600 e−/s at 38 K in the early detector description, lab characterization above $10^4$ e−/s/pix, and substantially larger on-telescope instrumental darks because warm internal background is included (Bowens et al., 2022).

The cryogenic and optical system is built around AO-fed operation. The MIRAC-5 cryostat uses a Cryomech PT405 cryocooler, with the GeoSnap detector expected to run at 35–40 K; commissioning describes pulse-tube cooling at 35–45 K with temperature precision better than 20 mK. The older BLINC cryostat remains part of the system and is LN$_2$-cooled at 77 K, serving as the reimaging optics. The telescope beam is reimaged from f/15 to f/29.8 at the detector, using an off-axis ellipsoid to place the telescope focal plane on the detector and the entrance pupil on a cold stop. The instrument retains an adjustable front focal distance via a “trombone” slide, while the back focal distance adjustment used in MIRAC-4 was removed because GeoSnap cabling is stiffer and bulkier (Bowens et al., 2022).

The plate scale is reported as 0.019 arcsec/pixel in the original design description and 0.0192 arcsec/pixel in commissioning, producing a field of view of about 19 arcsec. On the MMT, that sampling strongly oversamples the 10 μm diffraction limit of about 0.317 arcsec, a deliberate choice for high-contrast work (Bowens et al., 2024).

Parameter	Reported value	Context
Detector format	1024 × 1024 pixels	Single GeoSnap quadrant
Pixel pitch	18 μm	GeoSnap implementation
Well depth	1.2 × 10$^6$ e−	Low-gain / detector summary
Frame rate	up to 120 Hz; normal 0.1–85 Hz	Full-frame support / normal imaging
Plate scale	0.019–0.0192 arcsec/pix	MMT configuration
Field of view	about 19 arcsec	MMT configuration

3. Modulation strategy, chopping, and noise control

A defining operational feature of MIRAC-5 is its use of scene modulation to manage detector $1/f$ noise and thermal background. The GeoSnap detector is explicitly noted not to be affected by excess low-frequency noise (ELFN), unlike Aquarius-like arrays, but it does exhibit $1/f$ noise. This motivates the observing strategy: chopping or nodding modulates the astrophysical scene and removes slow drifts, while high frame rate allows rapid sampling before the background saturates the well (Bowens et al., 2022).

MIRAC-5 operates in both staring mode and chop/nod mode. The commissioning results state that staring mode becomes inefficient for long integrations or faint targets because of $1/f$ noise, with a systematic noise floor of about 0.41 ADU/pix after sufficiently many frames. The practical conclusion is band-dependent: N-band and N′ generally require chopping plus nodding, whereas L′ and M′ can often be handled with nodding only once the background is sufficiently low (Bowens et al., 2024).

The internal chopper is located in BLINC at a pupil plane and is driven by a rotational voice-coil actuator. In the design description it provides a 1-degree beam angle corresponding at the MMT to a 7.6 arcsec throw, or about $24\ \lambda/D$ at 10 microns; commissioning describes a source motion of about 7.4 arcsec, or 410 pixels, along the horizontal axis. The waveform is engineered for fast acceleration and deceleration with minimal ringing and low power consumption, and an accelerometer-based system compensates for orientation-dependent gravitational torque (Bowens et al., 2022).

The detector-noise regime is quantified in the commissioning work. In an 85 Hz frame, proper chopping and nodding reduce the $1/f$ contribution to less than 10% of the Poisson noise from the observed background. The exposure-time calculator therefore incorporates both a single-frame SNR model with source, atmosphere, telescope-plus-instrument background, dark current, read noise, and $1/f$ noise terms, and an empirical law

$$\sigma_{1/f}^2 = g^2 k_f \left(\frac{\nu_{\rm chopper}}{\nu_{\rm detector}}\right)^\alpha,$$

with $_2$0 (ADU/pix)$_2$1 and $_2$2. The same framework includes explicit chop and nod overheads, so observing efficiency depends on detector frame rate and chop/nod frequency (Bowens et al., 2024).

A plausible implication is that MIRAC-5 is deliberately engineered to operate in the regime where detector noise is no longer the only limiting factor. The original instrument paper makes this point qualitatively: at 85 Hz and around half well, shot noise is already comparable to or larger than detector noise, so atmosphere, telescope background, and source brightness become central determinants of sensitivity (Bowens et al., 2022).

4. Commissioning on the MMT

Commissioning observations on the MMT were conducted over six engineering runs, improving performance and defining operating procedures. The campaign verified detector readout and noise behavior, bad-pixel operability around 94.5%, throughput, effective background, sky variability, image quality, and the exposure-time calculator. Throughput measurements were obtained using Arcturus (Alpha Boo) on 22 May 2024 with nodding, sky frames, and darks, and the quoted throughputs include telescope, atmosphere, instrument, and detector QE (Bowens et al., 2024).

The central commissioning result is that total throughput is approximately 10% in the current configuration, with a planned dichroic upgrade expected to raise that to about 20% in many bands. The temporary dichroic is described as low-transmission, roughly 40–50%, whereas the planned replacement is expected to exceed 90% transmission. Reported current and future throughputs include 13% and 22% in L′, 5.5% and 16% in M′, 7.2% and 15% in N-band, 8.8% and 17% in the ammonia filter, and 8.0% and 17% in N′ (Bowens et al., 2024).

The thermal background is high, as expected for ground-based mid-IR work. Effective background levels are reported as $_2$3 e−/pix/s in L′, $_2$4 in M′, $_2$5 in N-band, and $_2$6 in N′, with lower future values anticipated after the dichroic upgrade. The commissioning analysis concludes that telescope plus instrument emission dominates over sky variability on the observed nights; a 1 K telescope temperature change produces only about 2% change in thermal emission, while water-vapor changes can affect sky emission substantially (Bowens et al., 2024).

Image quality already approaches the diffraction limit in N′. Measured FWHM values are 0.31″ in L′, 0.23″ in M′, and 0.38″ in N′, compared with ideal diffraction-limited values of 0.12″, 0.15″, and 0.37″ for a 6.5-m telescope using the quoted approximation $_2$7. N′ is therefore essentially diffraction-limited in the core, whereas L′ and M′ remain broadened by atmospheric effects. The N′ data imply a Strehl ratio of about 0.75 from encircled-energy comparison with an ideal Airy profile, and full MAPS commissioning is expected to deliver diffraction-limited performance in all bands (Bowens et al., 2024).

The background-limited sensitivities reported for an 8-hour exposure at $_2$8, after the planned dichroic upgrade, are 18.0 in L′, 15.6 in M′, and 12.6 in N′. Current values are 17.4, 14.1, and 11.3, respectively (Bowens et al., 2024).

5. Scientific applications: thermal-IR companions, disks, and ammonia

The principal scientific drivers for MIRAC-5 are characterization of gas giant exoplanets, imaging of forming protoplanets, and observations of dusty environments such as circumplanetary disks and protoplanetary disks. The motivating logic is explicit: objects with effective temperatures of about 300–1000 K radiate most of their energy in the mid-IR, so the 3–13 μm regime is naturally favorable for thermal-emission studies of warm companions and young planets in formation (Miller et al., 5 Aug 2025).

A major early goal is the first continuum mid-infrared detections for several known wide-orbit gas giants. The more specific molecular program focuses on atmospheric ammonia. MIRAC-5 includes a dedicated ammonia filter centered at 10.59 microns with a 0.64 μm FWHM, spanning 10.27–10.91 μm, designed to target NH$_2$9 absorption. The target case emphasized in the original instrument paper is GJ 504b, a warm companion with $^6$0 K (Bowens et al., 2022).

The ammonia strategy was developed using HELIOS and petitRADTRANS/PETIT-style atmospheric models for targets including HR 8799c, Kappa And b, and GJ 504b. The optimal NH$^6$1 feature was found in the 10.4 and 10.8 μm absorption bands, and the chosen filter passband was selected to maximize signal-to-noise. For GJ 504b, the estimated exposure time for a 3$^6$2 ammonia detection is about 8 hours of total observing time using the N′ filter together with the ammonia filter. The same study reports that if the target is warmer, around 600 K, the required exposure time can rise substantially, in some cases to about 15 hours. The filter tolerances are relatively forgiving: shifting the band edges by $^6$3 μm changes the S/N by only a few percent (Bowens et al., 2022).

This science program is significant because it moves MIRAC-5 beyond continuum photometry into atmospheric diagnostics. The commissioning paper also highlights NH$^6$4 as a dedicated use case, and the broader instrument concept connects thermal-IR continuum and molecular-band measurements with exoplanet luminosity, temperature, and composition (Bowens et al., 2024).

6. Coronagraphic upgrade, MAPS integration, and comparison with JWST

The next major stage in MIRAC-5’s development is a coronagraphic upgrade on the MMT in combination with the MMT Adaptive optics exoPlanet characterization System (MAPS). In the coronagraph paper, MIRAC-5 is described as a HgCdTe 1024×1024 detector camera sensitive from roughly 2–13 μm with an 85 Hz frame rate, 1.2 Me− well depth, 18 μm pixels, and about 19″ field of view, now being transformed from a high-background imaging instrument into a coronagraphic high-contrast system for exoplanet and disk science (Miller et al., 5 Aug 2025).

MAPS provides the adaptive-optics backbone. It is described as a high-order system using 336 voice-coil actuators operating at 1–2 kHz, with about 100–220 corrected modes and $^6$5 ms delay response; commissioning also describes a 336-actuator adaptive secondary mirror, visible or infrared pyramid wavefront sensors, and a 1 kHz control loop. In the upgraded optical train, a ZnSe dichroic sends wavelengths below 2 μm to MAPS wavefront sensors and transmits 2–15 μm light to MIRAC-5, which remains housed in the cryogenic BLINC/MIRAC-5 relay (Miller et al., 5 Aug 2025).

The coronagraph itself is a classical vortex implementation using an annular groove phase mask optimized for the N2 band, spanning 10–12.5 μm. The AGPM is mounted at the f/15 focal plane and cooled to about 77 K, with a downstream Lyot stop in a relayed pupil plane. Preliminary cryogenic laboratory tests measured on-axis stellar rejection factors of $^6$6 at 8.0 μm, $^6$7 at 9.0 μm, $^6$8 at 10.5 μm, $^6$9 at 11.5 μm, and $1/f$0 at 12.5 μm. The rejection near 10.5 μm is the principal reason this device was selected (Miller et al., 5 Aug 2025).

Lyot-stop design is treated as a throughput-versus-contrast optimization problem because the MMT pupil contains a central obscuration and spider vanes. The optimization used HEEPS with PROPER optical propagation, VIP post-processing, and AO residual phase screens from COMPASS. The preferred Lyot stop undersizes the MMT effective pupil to 98% of the effective diameter and oversizes the spider masks to about 3.2% of the effective diameter; the existing pupil mask already has an inner diameter of 18% of the effective external diameter, which simulations suggest is near the correct regime. The resulting off-axis throughput is about 74.8% (Miller et al., 5 Aug 2025).

The expected performance is explicitly framed in terms of raw and post-processed contrast. HEEPS simulations indicate raw contrast of order $1/f$1 at about 1″ in N2 under ideal alignment. After classical median-subtraction angular differential imaging, the projected sensitivity is about $1/f$2 at 5$1/f$3 around 1″, with a few hours needed for a bright target such as Procyon; more advanced post-processing could improve this by up to a factor of 10. For the simulated Procyon case, the sequence becomes background-limited beyond about 1.5″ (Miller et al., 5 Aug 2025).

The comparison with JWST is deliberately qualified. MIRAC-5+MAPS is not presented as a replacement for JWST/MIRI, whose space environment eliminates atmospheric thermal noise and turbulence. Instead, the argument is that on sufficiently bright stars and at separations near 1″, the ground-based system can be competitive in the contrast-limited regime. The coronagraph paper quotes roughly $1/f$4 raw contrast at 1″ for JWST/MIRI F1140C from on-sky results, and notes that JWST/MIRI PCA post-processing can reach about $1/f$5 inside 1″ at 3$1/f$6; commissioning similarly states that once AO is commissioned and a coronagraph installed, MIRAC-5 will have contrast-limited performance comparable to JWST in some regimes (Miller et al., 5 Aug 2025).

Because the AGPM has a small inner working angle, pointing control is critical. The planned low-order control loop is QACITS, which infers tip-tilt offsets from asymmetry in the coronagraphic image and feeds them back to keep the star centered on the phase mask during long integrations. This suggests that the full MIRAC-5 architecture is intended to combine high-order AO correction from MAPS, pupil-plane control, vortex suppression, and thermal-IR throughput into a single platform optimized for bright nearby systems, especially at separations of order $1/f$7 to $1/f$8, where $1/f$9 in the N-band for the 6.5-m MMT (Miller et al., 5 Aug 2025).