D4000n Measurements in Galaxies

• D4000n is a spectral break index defined as the flux ratio across the 4000 Å break, serving as a proxy for average stellar age in galaxies.
• Recent developments using PAUS narrow-band photometry enable the measurement of D4000n over large galaxy samples with statistical precision previously limited by high SNR spectroscopy.
• A weighted-sum estimator, which accounts for filter width, overlap, and variance, provides near-spectroscopic performance in galaxy classification and robust error calibration.

The D4000$_{\rm n}$ (narrow) spectral break index is a quantitative measure of the strength of the $4000\,\text{\AA}$ break in galaxy spectra, employed extensively as a proxy for average stellar ages and in galaxy classification. Traditionally determined via spectroscopic data, recent advances—most notably methodologies validated within the Physics of the Accelerating Universe Survey (PAUS)—have extended the measurement of D4000$_{\rm n}$ to narrow-band (NB) photometric surveys. This development enables large-sample studies and statistical analyses previously limited by the need for high SNR continuum spectroscopy.

1. Definition and Spectroscopic Determination of D4000$_{\rm n}$

D4000$_{\rm n}$ characterizes the flux-density ratio across the $4000\,\text{\AA}$ break, exploiting its sensitivity to stellar population age. As defined in Balogh et al. (1999), the "narrow" implementation uses the following continuum bands:

• Blue: $3850$–$3950\,\text{\AA}$
• Red: $4000$–$4100\,\text{\AA}$

The index is expressed as

$$\mathrm{D4000_n} = \frac{\langle F_\nu \rangle_{\text{red}}}{\langle F_\nu \rangle_{\text{blue}}}$$

where $\langle F_\nu \rangle$ denotes the average flux density per unit frequency within each window.

In direct spectroscopy, the average is computed as

$$\langle F_\nu \rangle = \frac{1}{(\lambda_{\rm max}-\lambda_{\rm min})(1+z)} \int_{\lambda_{\rm min}(1+z)}^{\lambda_{\rm max}(1+z)} F_\nu(\lambda)\, d\lambda,$$

with observed-frame integration limits set by redshift $z$.

2. Narrow-Band Photometric Measurement with PAUS

The PAUS survey employs 40 contiguous, quasi top-hat filters with FWHM $\approx 130\,\text{\AA}$, spaced every $100\,\text{\AA}$ from 4550–8450 Å ($R\sim65$). At any redshift $z$, the D4000$_{\rm n}$ windows shift to the observed frame (e.g., $\sim$6500 Å at $z\sim0.7$), with each of the two continuum bands typically sampled by $3$–$5$ filters—some partially overlapping the defined boundaries.

For each galaxy, the set of PAUS filters contributing to the blue and red bands is determined from the object’s spectroscopic redshift (VIPERS: $0.562

3. Weighted-Sum Estimator for $\langle F_\nu \rangle$

Given that PAUS filters have finite widths and variable overlap with the target D4000$_{\rm n}$ intervals, Renard et al. (2021) introduce a statistically principled estimator for the average flux density: $$\langle F_\nu \rangle \approx \frac{ \sum_i \Delta\lambda_i\, r_i^2\, \sigma_i^{-2}\, F_{\nu,i} }{ \sum_i \Delta\lambda_i\, r_i^2\, \sigma_i^{-2} }$$ where:

• $F_{\nu,i}$ and $\sigma_i$ are the measured flux and its error in filter $i$
• $\Delta\lambda_i = \int R_i(\lambda) d\lambda$ is the effective width of filter $i$
• $r_i$ is the fraction of the filter transmission within the (redshifted) D4000$_{\rm n}$ window:

$$r_i = \frac{ \int_{\lambda_{\rm min}(1+z)}^{\lambda_{\rm max}(1+z)} R_i(\lambda)\, d\lambda }{ \int R_i(\lambda)\, d\lambda }$$

This estimator weights by both filter width and inverse-variance and down-weights incomplete filter overlaps, mitigating edge artifacts. Application of this weighted average to both D4000$_{\rm n}$ bands yields the direct photometric D4000$_{\rm n}$ index.

4. Error Propagation, Signal-to-Noise, and Completeness

The variance of $\langle F_\nu \rangle$ is

$$\sigma_{\langle F_\nu \rangle}^2 = \left( \sum_{i} (\Delta\lambda_i\, r_i^2 / \sigma_i)^2 \right)^{-1}$$

and errors on D4000$_{\rm n}$ itself follow from standard propagation for the ratio of two independent weighted means.

The signal-to-noise ratio (SNR) for D4000$_{\rm n}$ is thus

$$\mathrm{SNR}(\mathrm{D4000_n}) = \frac{\mathrm{D4000_n}}{\sigma(\mathrm{D4000_n})}$$

Measured SNR statistics for the full PAUS-VIPERS sample show mean $\langle\mathrm{SNR}\rangle\sim4$ for D4000$_{\rm n}$. For the "bright" subsample ($i_{\mathrm{AB}}<21$), $>99\,\%$ of galaxies achieve $\mathrm{SNR}>3$ in D4000$_{\rm n}$; above $i_{\mathrm{AB}}=23$, this completeness drops to $~12\,\%$. Thus, direct D4000$_{\rm n}$ measurements are statistically robust to $i_{\mathrm{AB}}\sim21$, with deeper samples requiring stacking or median statistics to retain statistical precision.

5. Performance Validation via Synthetic PAUS

Using spectroscopic VIPERS data re-sampled to PAUS-like narrow-band photometry (termed sPAUS) at known redshifts, the weighted estimator’s performance was benchmarked:

• Mean bias: $-0.23\,\%$ relative to spectroscopic D4000$_{\rm n}$
• Gaussian scatter: $\sigma_\mathrm{bias} = 3.19\,\%$
• 32 % within $|\mathrm{bias}| < 1\,\%$; 67 % within $2.5\,\%$; 93 % within $5\,\%$

A small oscillatory bias ($\sim\pm2$–$3\,\%$) as a function of redshift is attributable to the sampling configuration of rest-frame D4000$_{\rm n}$ bands by the fixed narrow-band filter set; these effects diminish at higher $z$ due to broader observed-frame coverage. In practical terms, the estimator is unbiased to within $1\,\%$, so no additional corrections are generally applied in real PAUS data.

6. Correlations with Galaxy Physical Properties

The analysis of 17 195 matched PAUS–VIPERS galaxies with SED-fit star-formation rates (SFRs) and stellar masses (from CIGALE), classified into red/blue types using unsupervised Siudek et al. (2018) methodology, allows assessment of D4000$_{\rm n}$–$M_*$ and D4000$_{\rm n}$–SFR correlations. Linear fits to median-binned data give:

	Blue: $\alpha$	Red: $\alpha$
PAUS direct D4000$_{\rm n}$	$0.09 \pm 0.01$	$0.25 \pm 0.01$
PAUS CIGALE D4000$_{\rm n}$	$0.12 \pm 0.01$	$0.32 \pm 0.02$
VIPERS D4000$_{\rm n}$	$0.11 \pm 0.01$	$0.33 \pm 0.02$

For blue galaxies, the direct PAUS result is consistent within $1\sigma$ with both spectroscopy and SED fitting. For red galaxies, PAUS-direct slopes are systematically underestimated, a trend linked to a slight negative bias in D4000$_{\rm n}$ that increases with D4000 (and thus mass/age). D4000$_{\rm n}$–SFR relations derived from direct NB photometry closely match those mapped by spectroscopy when analyses use median statistics.

7. Comparison to SED-Fitting-Based Reconstruction

CIGALE (Code Investigating GALaxy Emission) reconstructs D4000$_{\rm n}$ by fitting full UV–NIR SEDs incorporating flexible star-formation histories. This yields a per-object $\langle\mathrm{SNR}\rangle\sim20$ for D4000$_{\rm n}$, about five times higher than direct NB methods.

However, validation tests indicate CIGALE underestimates its formal errors by $\gtrsim50\,\%$ (as assessed by normalized differences with spectroscopic D4000$_{\rm n}$), so the reported SNRs are overestimated accordingly. Optimal cuts in D4000$_{\rm n}$ for galaxy classification yield:

	D4000$_{\rm n,cut}$ (bright/full)	Correctly classified
PAUS direct	1.39 / 1.53	76 % / 69 %
PAUS CIGALE	1.38 / 1.38	89 % / 90 %
VIPERS spec	1.42 / 1.40	84 % / 85 %

PAUS CIGALE outperforms spectroscopy in this galaxy type classification, an effect attributed to artificially enhanced bimodality in the model SED fits. Direct PAUS NB measurements achieve near-spectroscopic classification efficiency for the bright sample but decrease with increasing photometric noise at fainter magnitudes.

8. Sample Definition and Selection Effects

The studied PAUS sample is built by cross-matching the PAUS NB catalogue within the CFHTLS W1 field (14 deg$^2$) to VIPERS PDR-2 spectroscopic data. Applying VIPERS spec-z quality flags ($2\!\leq\!z_{\rm flg}\!<\!10$ or $22\!\leq\!z_{\rm flg}\!<\!30$), and limiting to $0.562$_{\rm n}$ bands, yields 17 375 objects. Exclusion of sources lacking valid D4000$_{\rm n}$ in either PAUS or CIGALE results in a final sample of 17 241 galaxies.

PAUS is complete to $i_{\mathrm{AB}}\lesssim23$, but the SNR for direct D4000$_{\rm n}$ measurements exceeds 3 for only $\sim66 \%$ of all galaxies and $\sim99 \%$ of the bright ($i_{\mathrm{AB}}<21$) sample. Beyond $i_{\mathrm{AB}}=23$, reliable (SNR$>$3) direct D4000$_{\rm n}$ estimation falls to $\sim12 \%$ completeness. Aggregating (stacking) galaxies within appropriate bins restores high-fidelity D4000$_{\rm n}$ scaling relations over the full survey magnitude range.

---

Renard et al. (2021) demonstrate that direct, weighted-sum estimators applied to PAUS narrow-band data yield D4000$_{\rm n}$ values with minimal bias and robust scatter, compatible with spectroscopic indicators when applied to median-stacked samples. While individual SNRs are lower than those achieved by SED-fitting approaches, error estimates are better calibrated and the results are essentially model-independent. The methodology supports precise, unbiased mapping of stellar population diagnostics and galaxy type classification across deep, wide photometric surveys (Renard et al., 2022).