Intermediate Band Imaging Survey (IBIS)

• IBIS is a medium-band imaging survey using ~250 Å filters on DECam to select high‑redshift LAEs and LBGs across z ≈ 2.3–3.5.
• It combines deep, multi-filter observations (e.g., five adjacent bands in the XMM-LSS field reaching 5σ ~25.5 mag) to isolate continuum breaks and emission-line excesses.
• IBIS underpins DESI‑II target selection and cosmological analyses, forecasting sub‑3% angular BAO precision over vast areas using robust clustering methodologies.

Searching arXiv for papers on the Intermediate Band Imaging Survey (IBIS). arXiv search: "Intermediate Band Imaging Survey" The Intermediate Band Imaging Survey (IBIS) is a wide-field medium-band survey on DECam at the Blanco telescope in Chile, designed for the selection of high-redshift tracers, especially Lyman-$\alpha$ emitters (LAEs), in support of spectroscopic programs such as DESI-II. Its observing strategy combines intermediate-width filters, with $\Delta\lambda \approx 250\,\text{\AA}$, and broad-band imaging from Subaru, HSC, and Rubin/LSST to isolate spectral breaks and emission-line excesses across the high-redshift interval $2.3BAO analyses at $z\gtrsim 2$ (Ebina et al., 30 Sep 2025, Feder et al., 6 Dec 2025).

1. Survey architecture and observational definition

IBIS is described as a medium-band imaging survey on DECam at Blanco. In the XMM-LSS field used for clustering validation, it provides five adjacent medium bands spanning roughly $4000<\lambda<5300\,$Å, each about $260$ Å wide, and reaches a $5\sigma$ depth of $\sim 25.5$ mag in the deep field. More generally, the survey concept uses intermediate-width filters with $\Delta\lambda \approx 250\,\text{\AA}$, together with broad-band Subaru and Rubin/LSST imaging, to isolate both continuum breaks and line excesses (Ebina et al., 30 Sep 2025).

Published descriptions reflect more than one survey configuration. The clustering analysis describes IBIS as planned to cover about $3000\,\deg^2$, whereas the angular BAO forecasts evaluate a fiducial $\Delta\lambda \approx 250\,\text{\AA}$0 LAE survey and then extend the same methodology to larger Stage-V configurations. This suggests that the literature distinguishes between the currently characterized survey and forecast survey realizations used for cosmological performance studies (Ebina et al., 30 Sep 2025, Feder et al., 6 Dec 2025).

Quantity	Reported value	Context
Instrument platform	DECam at Blanco	Survey implementation
Medium-band width	$\Delta\lambda \approx 250\,\text{\AA}$1	General survey concept
XMM-LSS bands	5 adjacent bands, $\Delta\lambda \approx 250\,\text{\AA}$2Å	Clustering validation field
Deep-field depth	$\Delta\lambda \approx 250\,\text{\AA}$3 mag	XMM-LSS field
Stated footprint	$\Delta\lambda \approx 250\,\text{\AA}$4	Planned survey in clustering study
Fiducial BAO footprint	$\Delta\lambda \approx 250\,\text{\AA}$5	Forecast configuration

The central observational rationale is that medium-band imaging occupies an intermediate regime between broadband photometry and spectroscopy. It yields photometric redshift information precise enough for tomographic angular BAO, while remaining wide-field enough to construct dense target samples for later spectroscopic use. At the same time, the photo-$\Delta\lambda \approx 250\,\text{\AA}$6 precision is stated to be insufficient for full 3D anisotropic BAO reconstruction, which defines both the power and the limitation of the IBIS design (Feder et al., 6 Dec 2025).

2. Target selection and tracer populations

IBIS is intended to construct dense samples of high-redshift galaxies for large-scale-structure work, with emphasis on LAEs and, in practice, a broader population that includes Lyman break galaxies (LBGs) with emission lines. The target-selection strategy in the clustering analysis is explicitly exploratory and combines several medium-band color criteria: one class selects a flux peak in a central medium band relative to the two adjacent bands, another uses a synthetic $\Delta\lambda \approx 250\,\text{\AA}$7-band to identify Ly$\Delta\lambda \approx 250\,\text{\AA}$8 excess, and a third compares one band to a synthetic continuum built from the other four medium bands. These selections define redshift slices of width $\Delta\lambda \approx 250\,\text{\AA}$9 across $2.3 (Ebina et al., 30 Sep 2025).

Spectroscopic validation is performed with DESI ancillary spectroscopy, which is used primarily to determine the true redshift distribution $2.348% receive secure DESI redshifts, sufficient to characterize the redshift distribution. The measured interloper fractions are described as generally modest, and for the higher-$2.3Ebina et al., 30 Sep 2025).

A central result of the validation work is that the IBIS-selected sample is not a pure LAE sample. Instead, the modeling indicates that it is composed of an overlapping mixture of LAEs and LBGs with emission lines. In the $2.346.1% of the medium-band-selected galaxies are also recovered in a traditional $2.3LBG catalog. The sample therefore occupies the continuum between LAE and LBG selection rather than defining a disjoint tracer class. This is important for interpreting both its clustering amplitude and its utility for future spectroscopic target selection (Ebina et al., 30 Sep 2025).

The BAO forecast work adopts a more idealized tracer definition. There, the fiducial IBIS forecast is based on a pilot-survey-informed LAE sample spanning $2.3, divided into five redshift bins defined by medium-band filters: roughly $2.3$z\gtrsim 2$0, $z\gtrsim 2$1, $z\gtrsim 2$2, and $z\gtrsim 2$3, with mean redshifts $z\gtrsim 2$4 and $z\gtrsim 2$5. The forecast uses LAEs rather than LBGs because LBGs have broader redshift distributions that would smear the BAO feature too strongly after projection (Feder et al., 6 Dec 2025).

3. Clustering measurements and halo interpretation

Because DESI fiber assignment in the ancillary program is not random and introduces complicated angular selection effects, the current IBIS analysis does not attempt a conventional full 3D clustering measurement. Instead, it measures angular clustering within the narrow medium-band shells and projects the signal into $z\gtrsim 2$6. The basic relation is written as

$z\gtrsim 2$7

with the narrow-shell approximation

$z\gtrsim 2$8

and the estimator

$z\gtrsim 2$9

This formalism allows the target sample to be used directly while avoiding the full complexity of DESI fiber assignment (Ebina et al., 30 Sep 2025).

The observed clustering is detected clearly in the $4000<\lambda<5300\,$0 sample and in the total $4000<\lambda<5300\,$1 sample, while the wide $4000<\lambda<5300\,$2 sample is only marginally detected. Using a power-law real-space correlation function,

$4000<\lambda<5300\,$3

the analysis infers $4000<\lambda<5300\,$4–$4000<\lambda<5300\,$5 Mpc, with values around $4000<\lambda<5300\,$6–$4000<\lambda<5300\,$7 Mpc, and a corresponding linear bias $4000<\lambda<5300\,$8–$4000<\lambda<5300\,$9. These values are reported as consistent with earlier LAE and faint LBG measurements (Ebina et al., 30 Sep 2025).

The physical interpretation is developed with both HOD modeling and one-loop Eulerian perturbation theory. The HOD model adopts the standard five-parameter form

$260$0

$260$1

The best-fitting models are reported to give good fits, generally within about $260$2, with satellite fractions $260$3–$260$4. The inferred halo occupation suggests that the $260$5 sample resembles typical LAEs, whereas the $260$6 sample leans more toward LBG-like halo masses around $260$7 (Ebina et al., 30 Sep 2025).

The perturbative description uses velocileptors and writes the galaxy field as

$260$8

with counterterms and stochastic terms added in the power spectrum. The fitted linear biases are around $260$9 at $5\sigma$0 and $5\sigma$1–$5\sigma$2 at $5\sigma$3. The analysis reports only modest scale dependence at $5\sigma$4, somewhat stronger scale dependence at $5\sigma$5, and good agreement between auto-bias and cross-bias, implying that the IBIS-selected galaxies behave as well-defined biased tracers rather than as a pathological selection (Ebina et al., 30 Sep 2025).

4. Angular BAO formalism and forecast performance

The principal cosmological use case developed for IBIS is angular BAO. Because medium-band photo-$5\sigma$6 uncertainties remove most radial information, the forecast is built from angular power spectra $5\sigma$7 between redshift bins $5\sigma$8 and $5\sigma$9, rather than from full anisotropic 3D clustering. The BAO signal is parameterized by the dilation parameter $\sim 25.5$0, with

$\sim 25.5$1

$\sim 25.5$2

and

$\sim 25.5$3

Smooth broadband residuals are marginalized using cubic spline nuisance terms,

$\sim 25.5$4

with knot spacing $\sim 25.5$5, while the Gaussian covariance includes shot noise

$\sim 25.5$6

The analysis is performed in the flat-sky/plane-parallel approximation rather than the Limber approximation because the bin widths are only a few times the BAO scale (Feder et al., 6 Dec 2025).

The fiducial tracer model is anchored to the pilot survey analysis of Ebina et al. The LAE bias is assumed to rise smoothly from $\sim 25.5$7 to $\sim 25.5$8 across the redshift range. The comoving number density is varied over $\sim 25.5$9, corresponding to projected surface densities of about 625–1250 $\Delta\lambda \approx 250\,\text{\AA}$0 for the full sample, or 125–250 $\Delta\lambda \approx 250\,\text{\AA}$1 per redshift bin. Interloper-free baseline forecasts are supplemented with a more realistic case in which $\Delta\lambda \approx 250\,\text{\AA}$2 (Feder et al., 6 Dec 2025).

The headline forecast is that a $\Delta\lambda \approx 250\,\text{\AA}$3 IBIS-like LAE survey can constrain $\Delta\lambda \approx 250\,\text{\AA}$4 at $\Delta\lambda \approx 250\,\text{\AA}$5 with $\Delta\lambda \approx 250\,\text{\AA}$6 precision for a tracer density of $\Delta\lambda \approx 250\,\text{\AA}$7 and $\Delta\lambda \approx 250\,\text{\AA}$8. In the no-interloper baseline, the forecasts show that for $\Delta\lambda \approx 250\,\text{\AA}$9 and $3000\,\deg^2$0, the BAO position can already be recovered to better than $3000\,\deg^2$1, while for the $3000\,\deg^2$2 footprint the precision falls below $3000\,\deg^2$3, and at the highest assumed density $3000\,\deg^2$4 it reaches $3000\,\deg^2$5. Marginalizing over the spline broadband degrades $3000\,\deg^2$6 by only about $3000\,\deg^2$7 overall, and the study finds that including multipoles up to $3000\,\deg^2$8 captures most of the BAO information (Feder et al., 6 Dec 2025).

A key methodological conclusion is that these BAO forecasts are shot-noise dominated. The precision scales approximately as

$3000\,\deg^2$9

and is therefore more sensitive to increases in area and tracer density than to moderate changes in bias. Interlopers are modeled as a suppression of the signal,

$\Delta\lambda \approx 250\,\text{\AA}$00

so contamination primarily worsens the ratio of shot noise to recovered clustering signal rather than inducing a qualitatively different regime (Feder et al., 6 Dec 2025).

5. Role in DESI-II and Stage-V survey strategy

IBIS is framed not only as a source of imaging-based target selection, but also as a precursor cosmology survey. For DESI-II, which is expected to target LAEs and LBGs over $\Delta\lambda \approx 250\,\text{\AA}$01, IBIS is identified as a central imaging component alongside Subaru and Rubin imaging. In this role, early angular BAO measurements from IBIS can calibrate selection functions, identify the redshift bins and tracers with the best leverage, and expose potential systematics before the full spectroscopic analyses are carried out (Feder et al., 6 Dec 2025).

The same logic is extended to Spec-S5 and later Stage-V configurations. The forecast study considers a Spec-S5-like medium-band program with seven filters spanning roughly $\Delta\lambda \approx 250\,\text{\AA}$02 Å, corresponding to $\Delta\lambda \approx 250\,\text{\AA}$03, and a target LAE density of $\Delta\lambda \approx 250\,\text{\AA}$04 over $\Delta\lambda \approx 250\,\text{\AA}$05. Under those assumptions, the combined seven-bin angular BAO measurement reaches about $\Delta\lambda \approx 250\,\text{\AA}$06 precision on $\Delta\lambda \approx 250\,\text{\AA}$07. The paper also considers a longer-wavelength HSC-like extension to $\Delta\lambda \approx 250\,\text{\AA}$08 and finds that combined constraints can beat $\Delta\lambda \approx 250\,\text{\AA}$09 for survey areas above $\Delta\lambda \approx 250\,\text{\AA}$10 if tracer densities remain sufficient (Feder et al., 6 Dec 2025).

Within this framework, IBIS serves as a demonstrator for a broader observational strategy: medium-band imaging can be used not merely to feed spectroscopy, but to extract useful cosmological information before the spectroscopic program is complete. A plausible implication is that IBIS-like imaging can reduce survey-design uncertainty by testing tracer density, interloper control, binning choices, and BAO observability directly on the imaging sample itself (Feder et al., 6 Dec 2025).

6. Limitations, systematics, and simulation requirements

The most explicit limitation of the IBIS approach is radial resolution. The forecast study states that medium-band photo-$\Delta\lambda \approx 250\,\text{\AA}$11 errors are not good enough for full 3D anisotropic BAO reconstruction, and that BAO reconstruction is not attempted for IBIS because the photometric redshift errors remove radial modes and limit displacement-field estimation. The angular analysis therefore trades away radial information in exchange for robustness in projected clustering (Feder et al., 6 Dec 2025).

On the sample-construction side, the medium-band approach accepts somewhat higher contamination than traditional narrow-band LAE selection because spectroscopy can clean the sample later. The clustering study shows that interloper control is therefore a practical, not merely formal, part of the survey design: improving the higher-$\Delta\lambda \approx 250\,\text{\AA}$12 selection from $\Delta\lambda \approx 250\,\text{\AA}$13 to $\Delta\lambda \approx 250\,\text{\AA}$14 materially sharpens the characterization of $\Delta\lambda \approx 250\,\text{\AA}$15 and the subsequent clustering interpretation, even at the cost of roughly halving the sample (Ebina et al., 30 Sep 2025).

The simulation requirements implied by IBIS-selected tracers are unusually demanding. Using AbacusSummit-based mocks, the clustering paper argues that the current simulations barely have enough mass resolution to host the halos implied by the data. To resolve the halos hosting most of the selected galaxies under the standard HOD form, halos down to about $\Delta\lambda \approx 250\,\text{\AA}$16 are needed, which translates into particle masses of order $\Delta\lambda \approx 250\,\text{\AA}$17 or lower for the $\Delta\lambda \approx 250\,\text{\AA}$18 sample. A realistic survey-area mock over $\Delta\lambda \approx 250\,\text{\AA}$19 at $\Delta\lambda \approx 250\,\text{\AA}$20 would require a volume larger than $\Delta\lambda \approx 250\,\text{\AA}$21 Gpc on a side and roughly $\Delta\lambda \approx 250\,\text{\AA}$22 particles. This places mock generation and covariance control among the principal technical challenges for using IBIS-selected galaxies in precision cosmology (Ebina et al., 30 Sep 2025).

Taken together, these constraints define the present status of the Intermediate Band Imaging Survey. IBIS already functions as an empirically validated medium-band selection program for high-redshift galaxies, with measured clustering consistent with $\Delta\lambda \approx 250\,\text{\AA}$23–$\Delta\lambda \approx 250\,\text{\AA}$24 Mpc and $\Delta\lambda \approx 250\,\text{\AA}$25–$\Delta\lambda \approx 250\,\text{\AA}$26, while forecast work indicates that an IBIS-like survey can deliver sub-$\Delta\lambda \approx 250\,\text{\AA}$27 angular BAO precision at $\Delta\lambda \approx 250\,\text{\AA}$28 over a $\Delta\lambda \approx 250\,\text{\AA}$29 footprint. Its scientific role is therefore dual: it is both a target-selection infrastructure for future spectroscopy and a high-redshift angular-clustering experiment in its own right (Ebina et al., 30 Sep 2025, Feder et al., 6 Dec 2025).