Detection of the Cosmological 21 cm Signal in Auto-correlation at z ~ 1 with the Canadian Hydrogen Intensity Mapping Experiment

TL;DR

This work reports the first detection of the cosmological 21 cm auto-power spectrum at z ≈ 1 with CHIME, achieving a 12.5σ significance over 0.4 < k < 1.5 h Mpc⁻¹ using 94 nights of data in 608.2–707.8 MHz. A novel pipeline combining three RFI-mitigation methods, achromatic NS beamforming, and foreground filtering before time averaging enables a robust auto-correlation measurement, complemented by HyFoReS-based bandpass corrections and extensive validation. The analysis yields a 2D cylindrically and 1D spherically averaged P_{21}(k) with a consistent amplitude parameter A_{HI}² = 10⁶ Ω_{HI}²(b_{HI} + ⟨fμ²⟩)², showing agreement with cross-correlation results and IllustrisTNG expectations. Sub-band measurements around z ≈ 1.08 and z ≈ 1.24 demonstrate consistency and constrain HI clustering, while systematic tests across masks, baselines, and sky regions reinforce the robustness of the detection. The results establish CHIME as a direct probe of large-scale HI structure and set the stage for future BAO-scale measurements in auto-correlation at low and high redshift.

Abstract

We present the first detection of the cosmological 21 cm intensity mapping signal in auto-correlation at z ~ 1 with the Canadian Hydrogen Intensity Mapping Experiment (CHIME). Using 94 nights of observation, we have measured the 21 cm auto-power spectrum over a frequency range from 608.2 MHz to 707.8 MHz (z = 1.34 to 1.01) at 0.4 h Mpc^-1 < k < 1.5 h Mpc^-1, with a detection significance of 12.5 sigma. Our analysis employs significant improvements to the CHIME data processing pipeline compared to previous work, including novel radio frequency interference (RFI) detection and masking algorithms, achromatic beamforming techniques, and foreground filtering before time averaging to minimize spectral leakage. We establish the robustness and reliability of our detection through a comprehensive suite of validation tests. We also measure the 21 cm signal in two independent sub-bands centered at z ~ 1.08 and z ~ 1.24 with detection significance of 8.7 sigma and 9.2 sigma, respectively. We briefly discuss the theoretical interpretation of these measurements in terms of a power spectrum model, deferring the details to a companion paper. This auto-power spectrum detection demonstrates CHIME's capability to probe large-scale structure through 21 cm intensity mapping without reliance on external galaxy surveys.

Figures (29)

• Figure 1: Top: Normalized window function, $W^{\rm NS}$, versus NS baseline distance, shown at three representative frequencies near the bottom, middle, and top of the band. The left column shows the window function used in chimestacking which is derived from the inverse-variance weights, while the right column shows the scaled triangular window function adopted in this work. Bottom: Fourier transforms of the normalized window functions, illustrating the expected synthesized beams as a function of offset in the telescope $y$ coordinate. The inverse-variance weights produce a beam that narrows with increasing frequency and exhibits large frequency-dependent sidelobes, in part due to ripple-like structure from noise cross-talk. By contrast, the triangular window is scaled with frequency to maintain an identical synthesized beam across the band.
• Figure 2: Schematic illustration of how spectrally-smooth gain variations can couple to the RFI mask, resulting in leakage of foreground power into high-delay modes. The red and blue curves show the real component of a visibility as a function of frequency over a 10MHz sub-band for the same local ERA bin on two different days. They differ by an overall $1\%$ multiplicative factor that is approximately constant across this sub-band, representative of gain variations expected in CHIME. Gaps correspond to channels flagged by the RFI mask. The black curve shows the average of the two days, which is no longer spectrally smooth despite each day being smooth individually. This motivates our choice to apply foreground filtering on individual days prior to averaging.
• Figure 3: Bandpass leakage coefficients (fractional units) for the YY-polarization, 22m EW baseline as a function of frequency, measured from 10 bright point sources. Top panel: Leakage coefficients estimated from the sidereal-day stack of foreground-filtered hybrid visibilities using the point-source fitting procedure of \ref{['eq:dgcoeff']}. Bottom panel: Leakage coefficients estimated from the corresponding stack when HyFoReS is applied to the per-day foreground-filtered hybrid visibilities before sidereal-day averaging. HyFoReS reduces the amplitude of the bandpass leakage coefficients across most of the band.
• Figure 4: Number of $9.9404s\xspace$ integrations contributing to each local-ERA–frequency bin in the stack over all 94d of data. The center panel shows the two-dimensional distribution as a function of local ERA and frequency, restricted to the local-ERA range defining the field used in this work. Frequency channels that are entirely masked are shown in gray. The left panel shows the average over local ERA, while the bottom panel shows the average over frequency, excluding masked channels from the average. Summing the bottom panel over local-ERA bins yields a total integration time of 385h on this field.
• Figure 5: Top: Hybrid beamformed visibility as a function of frequency and time for the XX polarization and 22m EW baseline, evaluated at the telescope $y$ coordinate closest to that of 3C 295 (the brightest radio source in the field considered in this analysis) at transit. The time axis is expressed in degrees of 3C 295 hour angle, and the color scale shows the real component of the visibility after fringestopping. The top panel shows the average of unfiltered data over the full 94-day dataset, with features corresponding to the intrinsic spectrum of 3C 295 modulated by the primary beam response. Middle: Same quantity after applying the HyFoRes bandpass correction and a time-dependent high-pass filter along the frequency axis to the raw time series prior to averaging over sidereal days. This is the primary dataset used in our analysis, in which the foregrounds are suppressed by a factor of $\gtrsim 10^{4}$. Bottom: Difference between the even and odd sidereal-day stacks, which cancels foreground power that is stable across sidereal days. The close agreement with the middle panel demonstrates that foreground leakage is sub-dominant compared to thermal noise at this stage of the data processing. The frequency dependence of the noise level, including the prominent $\sim 30MHz$ ripple, reflects the primary beam response at the transit of Cygnus A, which is imprinted onto the data through the complex gain calibration procedure.