SPIDER: Balloon-Borne CMB Polarimeter

• SPIDER is a balloon-borne polarimeter designed to map the millimeter-wave sky by measuring degree-scale polarization of the CMB using a multi-frequency, cryogenic instrument setup.
• The instrument employs six independent refracting telescopes with stepped cryogenic sapphire half-wave plates and antenna-coupled TES bolometers to achieve high sensitivity and suppress systematic errors.
• SPIDER’s Antarctic long-duration balloon flights have demonstrated its ability to constrain the tensor-to-scalar ratio and effectively separate foreground contamination from polarized Galactic dust.

SPIDER is a balloon-borne polarimeter designed to map the polarization of the millimeter-wave sky at large angular scales and to probe the degree-scale $B$-mode polarization of the cosmic microwave background (CMB). In its instrument form, SPIDER consists of six monochromatic refracting telescopes, large-format superconducting detector arrays, and cryogenic sapphire half-wave plates, operated from Antarctic long-duration balloon flights to obtain large-area, multi-frequency polarization maps (Filippini et al., 2011, Gualtieri et al., 2017). The name “SPIDER” also appears in arXiv literature as the title of unrelated work in optimization, segmentation, multimodal generation, neuroscience, graph trace reconstruction, ASIC design, and dexterous retargeting (Fort et al., 2021, Zhao et al., 2024, Lai et al., 2024, Zhang et al., 21 Jun 2026, Sun et al., 2022, Alvado et al., 19 Dec 2025, Pan et al., 12 Nov 2025). In cosmology and instrumentation, however, SPIDER denotes the suborbital CMB experiment.

1. Scientific objectives and observational niche

SPIDER’s principal scientific goal is a sensitive search for the $B$-mode polarization signature of primordial gravitational waves, parameterized by the tensor-to-scalar ratio $r$, down to $r\lesssim0.03$ $(3\sigma)$ (Filippini et al., 2011). Secondary objectives include measurement of lensing $B$-modes and characterization of polarized Galactic dust (Filippini et al., 2011). The experiment is optimized for degree-scale measurements because the peak of the inflationary $B$-mode angular power spectrum appears at multipoles $\ell\sim80$ (Rahlin et al., 2014), and later instrument summaries emphasized that degree-scale resolution is well matched to the peak of the primordial $B$-mode signal at $\ell\lesssim200$ (Gualtieri et al., 2017).

The survey geometry evolved across design, pre-flight, and flight descriptions. Early mission descriptions stated that a first flight would map $B$0 of the sky during a 20–30-day Antarctic balloon campaign (Filippini et al., 2011). Pre-flight integration work described an effective survey of $B$1 of the sky after integration-time weighting (Rahlin et al., 2014). The 2015 flight review summarized the realized program as $B$2 of the sky observed over a 16 day Antarctic long-duration balloon flight (Gualtieri et al., 2017). These documents collectively place SPIDER in the regime of wide-area, low-background, degree-scale CMB polarimetry rather than small-patch, high-resolution mapping.

A central scientific constraint arises from foregrounds. Forecast and survey papers concluded that polarized Galactic dust is as bright or brighter than the cosmological signal at all SPIDER frequencies, including 90 GHz, 150 GHz, and 280 GHz, in the selected southern field (Fraisse et al., 2011). This foreground-limited regime motivated SPIDER’s multi-frequency architecture and later addition of high-frequency receivers for dust characterization.

2. Instrument architecture and cryogenic platform

SPIDER’s payload comprises six identical, monochromatic refracting telescopes housed within a common liquid-helium cryostat (Filippini et al., 2011). Each telescope is a telecentric, two-lens system with polyethylene, anti-reflection-coated lenses cooled to 4 K, feeding a Lyot stop at 4 K and a detector focal plane at 300 mK (Filippini et al., 2011). A later instrument review described the flight system as six independent, single-frequency refracting telescopes in a single 1.2 m-diameter liquid-helium cryostat, with anti-reflection-coated polyethylene lenses, a cold black sleeve baffle between the secondary lens and Lyot stop, predominantly reflective metal-mesh filters, a single nylon absorber at 4 K, and an AR-coated UHMWPE vacuum window (Gualtieri et al., 2017).

The cryogenic chain is layered. One description gave a main tank of 1284 L of liquid He held near 1 atm and a 20 L superfluid tank pumped to $B$3 torr at float, providing a 1.6 K stage to cool the half-wave plate and internal baffles (Gualtieri et al., 2017). Pre-flight integration work described a 1300 L main tank, a pumped superfluid tank at 1.5 K fed by a capillary array delivering $B$4 mW of cooling power, and vapor-cooled shields at $B$5 K and 40 K (Rahlin et al., 2014). Each telescope has an independent $B$6He adsorption refrigerator or closed-cycle $B$7He refrigerator that brings the focal-plane array to $B$8 mK (Gualtieri et al., 2017, Rahlin et al., 2014).

The gondola and support systems were designed for long-duration Antarctic operation. SPIDER uses a lightweight carbon-fiber structure suspended beneath a $B$9 km balloon, with solar panels of order 2 kW and sun-shields for continuous operation in 24 hr Antarctic daylight (Gualtieri et al., 2017). Science data are stored redundantly, and power, housekeeping, and control are distributed across redundant flight computers and per-receiver control electronics (Rahlin et al., 2014).

3. Polarization modulation and detector systems

Polarization modulation in SPIDER is achieved with a stepped, cryogenic sapphire half-wave plate at the entrance pupil of each telescope (Filippini et al., 2011). For a half-wave plate angle $r$0, the incoming Stokes parameters are modulated as

$r$1

This strategy enables each detector to measure all polarization angles and suppresses differential gain and beam systematics (Filippini et al., 2011). A later thesis on the half-wave plate system stated that the polarization modulation of the optical stacks was modeled using a physical optics calculation and Mueller matrices, and that lab tests and integrated cryostat tests showed consistency with the model (Bryan, 2014).

The focal planes use antenna-coupled transition-edge-sensor bolometers. In the design configuration, $r$2 TES detectors were distributed among three observing bands centered at 90, 150, and 280 GHz (Filippini et al., 2011). For the January 2015 flight, SPIDER deployed a total of 2400 antenna-coupled TESs at 90 GHz and 150 GHz (Gualtieri et al., 2017). Each spatial pixel integrates two orthogonal polarization-sensitive slot-antenna arrays, summed coherently through superconducting microstrip trees and on-chip band-defining filters, with absorbed power dissipated in a gold termination resistor on a SiN island thermally linked to a Ti TES with $r$3 mK (Filippini et al., 2011). The 2015 flight review added that a higher-$r$4 Al TES with $r$5 K was wired in series for laboratory tests under higher loading (Gualtieri et al., 2017).

The later 280 GHz receivers adopted a different optical coupling architecture. These channels used conical, corrugated feedhorns coupled to a monolithic detector array fabricated on a 150 mm silicon wafer, with polarimetric sensing via waveguide probe-coupling to a multiplexed array of TES bolometers (Hubmayr et al., 2016). The SPIDER receiver carries three focal plane units at 280 GHz, containing 765 spatial pixels and 1,530 polarization sensitive bolometers in total (Hubmayr et al., 2016). Electromagnetic simulations for the waveguide–OMT–backshort stack showed band-averaged single-polarization coupling efficiency of about 94%, reflection of about 3%, and leakage plus radiative loss of about 3%; the measured white noise equivalent power was $r$6, consistent with the phonon noise prediction (Hubmayr et al., 2016).

4. Scan strategy, pointing control, and sky coverage

SPIDER’s observing strategy combines azimuth scanning, stepped elevation, and stepped half-wave-plate rotation. The Antarctic flight scans in azimuth while keeping the boresight safely away from the Sun, and elevation is stepped to provide cross-linked coverage (Filippini et al., 2011, Fraisse et al., 2011). Pre-flight integration work described a sinusoidal azimuth scan at speeds up to $r$7 with accelerations $r$8, while stepping the half-wave plate by $r$9 every 12 sidereal hours (Rahlin et al., 2014). Over a full day this yields nearly uniform polarization-angle coverage with matrix condition number $r\lesssim0.03$0 on roughly $r\lesssim0.03$1 of the sky, corresponding to an effective survey of $r\lesssim0.03$2 when weighted by integration time (Rahlin et al., 2014).

The azimuth drive uses a reaction wheel and a motorized pivot (Shariff et al., 2014). A 13 kHz PI control loop runs on a digital signal processor with fibre optic rate gyros as feedback, achieving azimuthal speed control with $r\lesssim0.03$3 RMS error (Shariff et al., 2014). Laboratory tests demonstrated sinusoidal azimuth scans of a 5000 lb payload at peak acceleration $r\lesssim0.03$4 and peak speed $r\lesssim0.03$5 while achieving sub-arcminute pointing control accuracy (Shariff et al., 2014). Elevation is controlled by stepper-motor-driven linear actuators, with the lab system consistently holding elevation to $r\lesssim0.03$6 (Shariff et al., 2014).

Pointing reconstruction combines fast inertial sensors with absolute references. Instrument summaries listed star cameras, dual-frequency GPS, sun sensors, and fiber-optic gyros, with instantaneous pointing accuracy $r\lesssim0.03$7 rms and post-flight reconstruction to $r\lesssim0.03$8 rms (Gualtieri et al., 2017). Pre-flight integration work quoted post-flight boresight pointing knowledge of $r\lesssim0.03$9 using boresight and star-tracking cameras together with a 3-axis fiber gyroscope (Rahlin et al., 2014).

5. Flight campaigns and measured performance

The first long-duration balloon flight in January 2015 deployed SPIDER at 90 GHz and 150 GHz with 2400 TESs (Gualtieri et al., 2017). After hardware failures and data-quality cuts, 1863 optically-coupled detectors remained: 675 at 90 GHz and 1188 at 150 GHz (Gualtieri et al., 2017). In-band optical loading at float was $(3\sigma)$0 pW at 90 GHz and $(3\sigma)$1 pW at 150 GHz, comparable to Planck-HFI at L2 (Gualtieri et al., 2017). The measured single-detector NET was $(3\sigma)$2 at 90 GHz and $(3\sigma)$3 at 150 GHz, giving instrument NET of about $(3\sigma)$4 and $(3\sigma)$5 after yield cuts (Gualtieri et al., 2017). During azimuth scans of $(3\sigma)$6 at $(3\sigma)$7 peak speed, platform vibrations remained below $(3\sigma)$8 and pointing reconstruction residuals were $(3\sigma)$9 (Gualtieri et al., 2017). The cosmic-ray glitch rate was about one glitch per detector every 3 minutes, with typical glitches shorter than 100 ms and coincidence rate across the focal plane below 0.03% (Gualtieri et al., 2017).

Pre-flight characterization of the 94 GHz and 150 GHz instrument showed white noise from 2 to 8 Hz, $B$0 knees well below 50 mHz, end-to-end optical efficiency of 35–40% with peak channels near 50%, beam full-widths at half maximum of $B$1 for 94/150 GHz, differential pointing, ellipticity, and width all at the $B$2 level, cross-polar response $B$3, and half-wave-plate encoder precision of $B$4 (Rahlin et al., 2014). Under flight loading, the same study reported $B$5 at 94 GHz and $B$6 at 150 GHz, with array sensitivities of about $B$7–$B$8 and projected map depth of $B$9–$B$0 over a 16-day, 85%-duty flight (Rahlin et al., 2014).

The second flight, in the 2022–23 season, added three 280 GHz receivers optimized for polarized dust mapping (Shaw et al., 2024). Pre-flight Fourier-transform spectroscopy measured center frequencies of $B$1 GHz for Y3 and $B$2 GHz for Y5, with Y4b assumed near 275 GHz, and an effective bandwidth of about 63 GHz (Shaw et al., 2024). Median in-flight optical loading per detector was 0.6 pW for Y3, 1.6 pW for Y4b, and 1.9 pW for Y5 (Shaw et al., 2024). Median single-detector $B$3 from stop-mode data was 624, 398, and $B$4 for Y3, Y4b, and Y5, respectively; cumulative array sensitivity for Y5 was about $B$5, and the combined 280 GHz array sensitivity was about $B$6, with scanning-mode data giving a consistent $B$7 (Shaw et al., 2024).

6. Systematics, foreground control, and scientific significance

SPIDER’s systematic-error program was driven by the requirement that spurious $B$8-modes remain below a cosmologically interesting signal. Simulation studies concluded that the expected level of systematic error in the instrument is significantly below the amplitude of a signal with $B$9 (Fraisse et al., 2011). In a detailed systematic budget, relative gain uncertainty, differential pointing, differential beam size, differential ellipticity, absolute and relative polarization-angle uncertainty, pointing jitter, polarized sidelobes, optical ghosting, half-wave-plate differential transmission, and magnetic pickup were each assigned target control levels, and adding all effects in quadrature gave a total false $\ell\sim80$0-mode amplitude $\ell\sim80$1 of the $\ell\sim80$2 signal at $\ell\sim80$3 (Fraisse et al., 2011).

Systematics control is distributed across optics, modulation, and readout. Pair differencing of co-located orthogonal TESs cancels common-mode signals, stepped half-wave plates modulate $\ell\sim80$4 at $\ell\sim80$5, and multi-layer high-$\ell\sim80$6 and superconducting shields attenuate external magnetic fields from the Earth by more than $\ell\sim80$7 in one design summary and by $\ell\sim80$8 in pre-flight integrated hardware descriptions (Filippini et al., 2011, Rahlin et al., 2014). Sidelobe simulations and measurements, aided by reflective forebaffles, an absorptive Eccosorb-lined optics sleeve, and internal baffle rings, showed sidelobe levels better than $\ell\sim80$9 dB beyond $B$0 (Rahlin et al., 2014).

Foreground separation is structurally central to SPIDER. Forecast papers concluded that at $B$1 the dust $B$2-mode power exceeds that of $B$3 by factors of about 9 at 150 GHz and about 500 at 280 GHz, while dust also dominates at 90 GHz by at least a factor of 5 (Fraisse et al., 2011). This is why the 280 GHz channels were added: they constrain $B$4-mode contamination from Galactic dust emission and provide expanded spectral leverage on the dust spectral energy distribution (Hubmayr et al., 2016). Early forecasts stated that SPIDER could reach $B$5 at 99% confidence after two $B$6-day flights even when foreground contamination is taken into account, while the no-foreground limit of $B$7 could be reached after one 20-day flight (Fraisse et al., 2011). Later instrument summaries forecast $B$8 across $B$9 after two flights, allowing a $\ell\lesssim200$0 detection of $\ell\lesssim200$1 (Gualtieri et al., 2017).

Within CMB instrumentation, SPIDER therefore occupies a specific role: a near-space, balloon-borne platform that combines cold refracting optics, stepped cryogenic half-wave plates, multiplexed TES focal planes, and multi-frequency foreground monitoring in a configuration intended both for primordial $\ell\lesssim200$2-mode searches and as a proving ground for detector and system architectures relevant to future satellite missions (Filippini et al., 2011).