DESI Spare Fiber Program Overview

• DESI Spare Fiber Program is a dual-purpose initiative using idle fibers for transient follow-up and low-redshift galaxy mapping, enhancing overall survey yield.
• It employs dynamic fiber assignment and advanced selection algorithms to secure early-time spectra for extragalactic transients with high classification reliability.
• The program attains high completeness in dwarf galaxy redshift measurements, advancing studies in galaxy–halo connections and local peculiar velocities.

The DESI Spare Fiber Program encompasses a set of secondary surveys operating in parallel with the Dark Energy Spectroscopic Instrument (DESI) primary cosmology observations. Its core premise is the strategic utilization of unused ("spare") fibers per DESI pointing to execute efficient spectroscopic follow-up for extragalactic transients and for the systematic mapping of low-redshift galaxies, maximizing scientific yield without adverse impact on DESI’s primary mission. The program includes both transient follow-up ("DESI Transient ToO") and the low-redshift dwarf galaxy survey ("DESI LOW-Z"), using dynamic fiber assignment and advanced selection algorithms to enable early-time transient classification and high-completeness dwarf galaxy redshift measurements (Hall et al., 18 Jan 2026, Darragh-Ford et al., 2022).

1. Scientific Motivation and Objectives

Spare fiber programs are motivated by the dual opportunity presented by idle fiber resources and the pressing need for spectroscopic data in areas under-served by primary survey priorities. Key objectives are:

• DESI Transient ToO Program: Acquire early-time spectra of extragalactic transients (including thermonuclear and core-collapse SNe, tidal disruption events) identified by wide-field imaging surveys (ZTF, DESIRT, and future LSST), providing robust redshift and classification for events that typically remain unclassified (>80% of TNS alerts). The program extends statistical studies to fainter transients ($r \sim 22$–22.5) and acts as a pathfinder for large-scale time-domain follow-up in the Rubin Observatory era.
• DESI LOW-Z Survey: Map the local ($z < 0.03$) Universe with high completeness, emphasizing faint dwarf galaxies ($M_* < 10^9 M_\odot$, $19 < r < 21$), thereby supporting peculiar-velocity studies, galaxy–halo connection constraints, and multi-messenger source identification (Darragh-Ford et al., 2022).

Both efforts leverage spare fibers (typically 5–50% of all available) per DESI tile, with zero measurable impact on primary cosmological survey S/N, tiling, or completeness (Hall et al., 18 Jan 2026, Darragh-Ford et al., 2022).

2. DESI Instrumentation and Fiber Allocation

DESI comprises 5000 robotically actuated fibers (1.5″ diameter), feeding ten three-arm spectrographs (wavelength coverage 3600–9824 Å, $R\sim2000$–5500) on the 4-m Mayall telescope. Each pointing (“tile”) covers $\sim$8 deg$^2$, observed in 1–3 exposures (“dark” program: 900–1200 s total, “bright” program: 150–300 s total). Main survey assignment (“fiberassign”) reserves 3000–4000 fibers for cosmology targets (LRGs, ELGs, QSOs, BGS), leaving hundreds per exposure unassigned.

Spare fiber allocation is governed by a priority function

$$\max_{\text{assignment }A} \sum_{i\in\text{fibers}} P_{i}(t_i)$$

subject to geometric collision ($<62″$ separation) and survey-specific priority constraints, where $P_{i}$ encodes assignment priority (main survey $>$ ToO $>$ filler). Transient ToO and LOW-Z programs receive lower priority than main targets but are favored over generic fillers (Hall et al., 18 Jan 2026, Darragh-Ford et al., 2022).

3. Transient ToO Survey Strategy

The transient program exploits daily targets-of-opportunity (ToOs) submission, drawing on public TNS alerts (PSN class), filtered ZTF alerts (BTSbot selection), and DESIRT candidates with $r<22$, prioritizing observations within days of trigger to sample transient evolution. The program also classifies transients from serendipitous matches of archival DESI data (DR1/DR2) with TNS alerts within $±120$ d and $<0.75″$ positional offset.

• Exposure regimes: “Dark” program (moon $<60\%$): 1000 s total, $r$-band depth $\sim$22.5 mag; “Bright” program: 180 s total, usable to $r\lesssim21.5$ mag.
• Temporal response: Median observed delay $\Delta t_{\mathrm{obs}-\mathrm{trigger}} \sim 8$ d, 50% within 8 d.
• Classification methodology: Photon-based spectral classifier (NGSF + human vetting) requiring $\geq20\%$ transient flux fraction for secure typing.

Coordination with DESIRT provides external imaging (DECam: 100 deg$^2$ fields, $r\sim23.5$ mag, 3 d cadence), supporting candidate selection and providing light-curve context for spectroscopic data (Hall et al., 18 Jan 2026).

4. LOW-Z Galaxy Survey Methods

The LOW-Z survey employs a three-tiered selection strategy, integrating catalog-level photometric cuts with CNN-based prioritization:

• Photometric cleaning: Non-PSF/non-Gaia duplicate objects, dereddened magnitudes ($m_o$), enforced quality flags across $g, r, z$ bands.
• Catalog-level cuts: Surface brightness and color cuts optimized for $z<0.03$ completeness,

$$\mu_{r_o,\rm eff} + \sigma_\mu - 0.7(r_o - 14) > 16.8\ [\mathrm{mag\,arcsec^{-2}}]$$

$(g - r)_o - \sigma_{gr} + 0.06(r_o - 14) < 0.99$

• CNN prioritization: 34-layer ResNet architecture trained on SAGA and expanded LOW-Z redshift catalogs, operating on $3\times144\times144$ pixel cutouts, focal-loss objective to optimize low-$z$ yield.

Tiered target assignment discriminates high-priority ($\sim$20 deg$^{-2}$, CNN Tier 1), intermediate ($\sim$80 deg$^{-2}$, photometric Tier 2), and overlapping ($\sim$200 deg$^{-2}$, Tier 3) candidates, with observed fiber allocation fraction $\sim$50% (Darragh-Ford et al., 2022).

5. Program Performance Metrics

Key performance is characterized as follows:

• Transients:
  • Limiting magnitude: $r\sim22.5$ mag (1000 s, dark), SNR $\sim$5 per res. element,

  $$\mathrm{SNR} \simeq \frac{F_\star t^{1/2}}{[F_\star + F_{\rm sky} + (\mathrm{RN}/t^{1/2})^2]^{1/2}}$$ - Typing success: ToO submission acquires $\sim$13% of PSN ToOs ($\sim$1 SN night$^{-1}$); inclusion of ZTF+DESIRT yields up to $\sim$3 SN night$^{-1}$. Classification reliability exceeds 80% for flux fractions $>$20%. - Impact: $<$5% of fiber usage, non-measurable effect on main survey S/N/completeness (Hall et al., 18 Jan 2026).

• LOW-Z:

  • Catalog-level cuts: $C>95\%$ completeness at $E\sim1\%$ efficiency; CNN selection achieves $C\sim85$–90%, $E\sim20\%$ (≥10× efficiency improvement).
  • Spectroscopic yield: Early survey (Apr–Jul 2021), 143,486 unique targets, 22,679 galaxies with $M_*<10^9\,M_\odot$, 2,019 galaxies at $z<0.03$.
  • Comparative density: Outperforms SDSS-DR8 and GAMA for dwarf counts and completeness (Table below).

Survey	$z<0.03$ density [deg$^{-2}$]	$r$ limit
LOW-Z	3.7	21
SDSS-DR8	0.5	17.8
GAMA	17	19.65

(Darragh-Ford et al., 2022)

6. Scientific Outcomes

During the first six months (Dec 2024–Jun 2025):

• Transients: Spare-fiber ToO program generated spectra for 184 targets, resulting in 59 classifiable SNe (44 SNe Ia, 11 SNe II, 4 SNe Ib/c); additional 198 classifiable transients from 649 archival serendipitous spectra (123 SNe Ia, 22 SNe II, 3 SNe Ib/c, 2 TDEs). Notable early TDE detection (2022emf) occurred 7 d prior to first ZTF imaging alert. Transients were classified to median $z\sim0.15$, and down to $r\sim22.5$, probing fainter flux than ZTF’s $5\sigma$ detection limit (Hall et al., 18 Jan 2026).
• LOW-Z: Over 22,000 dwarf galaxy redshifts obtained, with observed density rivaling or exceeding past surveys, mapping the faint end of the local galaxy mass function. Five-year projections forecast mapping $>10^5$ low-$z$ galaxies and $>10^6$ faint dwarfs ($M_*<10^9\,M_\odot$) over 14,000 deg$^2$, supporting precision peculiar-velocity, density-field, and host catalog studies (Darragh-Ford et al., 2022).

7. Future Directions and Implications

DESI Spare Fiber Programs empirically validate the use of idle spectroscopic multiplex for impactful secondary science without primary survey compromise. This suggests extensibility to future facilities, particularly Rubin–DESI coordination for time-domain probes. The demonstrated selection algorithms enable tuning for increased density (to $r<22$ at $\sim$800 deg$^{-2}$, $M_*>10^7\,M_\odot$), promising comprehensive local universe tomography and robust transient host identification. A plausible implication is that similar approaches will underpin next-generation multi-messenger source discovery and low-mass galaxy studies (Hall et al., 18 Jan 2026, Darragh-Ford et al., 2022).