Carbonaceous dust grains seen in the first billion years of cosmic time

The study investigates when and how carbonaceous dust grains formed in the early Universe by searching for the ultraviolet 2,175 Å attenuation bump—an indicator of carbonaceous dust such as PAHs or graphitic grains—in galaxies during the first billion years (z ≈ 4–7). Dust is abundant in some z ≳ 7 galaxies, but its rapid emergence challenges standard dust-formation pathways (e.g., AGB stars) given the limited cosmic time available. Direct spectroscopic detection of the 2,175 Å feature at high redshift can break degeneracies in interpreting dust masses and constrain the origin, composition, and production timescales of early dust, thereby informing models of galaxy evolution, star formation histories, and interstellar medium enrichment at early epochs.

Previous observations have revealed large dust masses up to z ≈ 8, prompting reconsideration of dust production channels including AGB stars, supernovae, and rapid interstellar grain growth. The 2,175 Å feature is prominent in the Milky Way and LMC extinction curves but weak or absent in the SMC, and has been attributed to PAHs or nano-sized graphitic grains. Its presence and strength vary with metallicity and dust–star geometry; radiative transfer effects can attenuate the bump in integrated light. Prior to this work, spectroscopic detections in galaxies were largely confined to massive, dusty systems at z ≲ 3, with stacked photometric hints up to z ≲ 6. Correlations between bump strength and metallicity have been observed at lower redshift, and γ-ray burst sightlines have also shown the feature at z ≤ 2. Theoretical studies debate the efficiency and survival of dust produced in supernova ejecta and the contribution from WR stars, with mixed predictions about carbonaceous grain yields and destruction by reverse shocks.

• Observations: JWST/NIRSpec multi-object spectroscopy (MSA) with PRISM/CLEAR configuration covering 0.6–5.3 μm at R ≈ 100 as part of JADES. Three visits (21–25 Oct 2022) observed 253 sources with per-object exposure times of 9.3–28 h; JADES-GS-z6-0 received 27.9 h.
• Sensitivity: 1D spectra reached 3σ continuum sensitivity ~6–40 × 10^−22 erg s^−1 cm^−2 Å^−1 (27.2–29.1 AB mag) near 2 μm.
• Sample selection: From NIRCam-selected high-z candidates with spectroscopic confirmation using BAGPIPES and BEAGLE, select z > 4 galaxies with median S/N ≥ 3 per spectral pixel in rest-frame 1,268–2,580 Å. Final parent sample: 49 galaxies at 4.02 ≤ z ≤ 11.48.
• UV bump preselection: Fit rest-UV continuum with power laws in four windows (excluding 1,920–1,950 Å to avoid C III) using MultiNest. Identify candidates via a strong slope turnover across 2,175 Å (negative γ3−4 and γ4 < 0 within 1σ), yielding 10 galaxies (4.02 < z < 7.20) with bump signatures.
• Spectral processing: Use running median filtering (15 pixels) and bootstrap uncertainties. Flux calibration, slit-loss corrections (point-source assumption), and extraction over 5-pixel apertures; robustness checked with 3-pixel extractions.
• Bump measurement: Model excess attenuation over a smooth power-law continuum with a Drude profile, fitting amplitude and central wavelength. For the individual z=6.71 galaxy and for stacked spectra, quantify significance via residuals and χ² comparisons between power-law-only and power-law+Drude models.
• Stacking: Construct inverse-variance-weighted stacks (20 Å bins) of all 49 galaxies and of the 10 bump candidates after shifting to rest frame and normalizing to the predicted bump-free continuum at 2,175 Å.
• Ancillary measurements: Determine Balmer decrement (Hα/Hβ) for nebular extinction E(B−V), gas-phase and stellar metallicities, stellar masses, ages, and star formation rates; assess presence of C III] λ1907,1909 emission.

• First direct spectroscopic detection of the UV 2,175 Å attenuation bump at z > 3 in an individual galaxy at z = 6.71 (JADES-GS-z6-0). The feature deviates from a power-law continuum at 6σ significance.
• Drude fit results for JADES-GS-z6-0: bump amplitude A_max = 0.43 ± 0.07 mag; central wavelength λ_max = 2,263 ± 20 Å; substantial χ² improvement (power-law only χ² = 72.5 vs. power-law+Drude χ² = 5.0).
• Physical properties of JADES-GS-z6-0: Balmer decrement Hα/Hβ ≈ 3.7, implying nebular E(B−V)_neb = 0.25 ± 0.07 mag; gas-phase and stellar metallicities Z ≈ 0.2–0.3 Z⊙; evidence for significant dust obscuration and substantial metal enrichment relative to peers at similar mass and redshift; tentative color gradient in imaging.
• Parent sample (49 galaxies, 4.02 ≤ z ≤ 11.48) vs. bump subset (10 galaxies, 4.02 < z < 7.20):
  • All-sample stack shows no bump but clear C III] λ1907,1909 emission.
  • Bump-subset stack reveals a clear depression centered near 2,175 Å at 5.4σ significance, with mildly redder UV slopes and higher dust/metallicity than the parent sample.
  • Drude fit to subset stack: amplitude A_max = 0.10^{+0.03}{−0.01} mag (reported also as 0.10 ± 0.01 mag), λ_max = 2,236^{+30}{−20} Å (also quoted as ±20 Å).
• The individual galaxy’s bump parameters are comparatively strong and at slightly longer λ_max than typical Milky Way sightlines, suggesting potential differences in grain composition/mixture.
• Timescale implications: At z = 6.71 the Universe is ~800 Myr old. Standard AGB-driven carbonaceous dust formation (requiring ≥300 Myr for low-mass stars to reach AGB at low metallicity) would imply star formation onset at z ≳ 10, yet no clear evidence for such old stellar populations is found. Results point to faster channels (e.g., WR stars or supernovae) for early carbonaceous dust production.

The detection of a strong UV bump at z = 6.71 and a significant bump in a stacked subset of z ≈ 4–7 galaxies demonstrates that carbonaceous dust grains were already present within the first billion years. The strength and slightly redder central wavelength of the bump in JADES-GS-z6-0 compared with typical Milky Way sightlines may reflect differences in the dust grain population (e.g., more graphitic grains vs. PAHs), and/or simpler dust–star geometries at early times that preserve the bump in integrated spectra. The observed dust and metallicity levels indicate substantial chemical enrichment. Given the limited cosmic time and lack of strong evidence for >300 Myr-old stellar populations, the results favor rapid dust production mechanisms, such as formation in WR stellar winds and/or supernova ejecta, over standard AGB-dominated channels alone. These findings impose constraints on dust formation and survival models (e.g., grain growth, reverse shock destruction), highlighting the need to account for efficient early production of carbonaceous grains. The prevalence of the bump in a dustier, slightly more metal-enriched subset supports a link between metallicity, dust content, and bump visibility, consistent with trends at lower redshift.

This work provides the first direct spectroscopic evidence of the 2,175 Å attenuation bump in the reionization era, including a robust detection in a z = 6.71 galaxy and in a stacked sample of z ≈ 4–7 galaxies. The strengths and wavelengths of the features, together with dust and metallicity indicators, imply that carbonaceous dust formed rapidly in the early Universe, likely via short-timescale channels such as WR stars and/or supernova ejecta. These results refine our understanding of dust production timelines and carriers in young galaxies and challenge models relying solely on AGB contributions within the available cosmic time. Future work should expand spectroscopic samples, obtain higher-resolution spectra to better constrain bump profiles, disentangle dust–star geometries via spatially resolved spectroscopy, and use multiwavelength diagnostics (e.g., mid-IR PAH features, far-IR dust emission) to identify the dominant carriers and refine dust formation and destruction models in the early ISM.

• Sample selection requires sufficient S/N in rest-UV, potentially biasing toward brighter, dustier systems; the bump appears only in a subset (10/49), limiting generalizability.
• Integrated spectra are affected by dust–star geometry and radiative transfer, which can weaken or mask the bump; spatial variations (e.g., color gradients) may influence measured strength.
• Central wavelength and amplitude uncertainties remain, and the inferred carrier (PAHs vs. graphitic grains) is not uniquely identifiable from the UV bump alone.
• Stacked spectra average diverse properties and can dilute or bias individual features. Systematic uncertainties (e.g., slit-loss corrections) are expected to be smooth with wavelength but residual systematics could subtly affect continuum shape.
• Constraints on star formation histories and ages rely on SED modeling with inherent degeneracies, complicating definitive timescale inferences for dust formation channels.