THESAN-DARK-1: Dark Matter Companion Simulation

• THESAN-DARK-1 is the dark-matter-only extension of THESAN-1, mapping reionization histories to z=0 halo properties and Local Group analogues.
• The study quantitatively links reionization redshift with halo mass and local overdensity using varying Gaussian smoothing scales to capture small-scale patchiness.
• The analysis reveals that environmental factors, such as proximity to Virgo-like clusters, drive asynchronous reionization within Local Group pairs despite modest mass differences.

THESAN-DARK-1 is the dark-matter-only companion to the full radiation-hydrodynamic THESAN-1 simulation. The two are used together as a matched pair: THESAN-1 provides the baryonic physics and self-consistent ionizing radiative transfer needed to determine when each region or halo reionizes, while THESAN-DARK-1 extends the same initial conditions to $z=0$ so that the $z\gtrsim 5.5$ reionization histories can be tied to present-day halo properties, especially for Local Group analogues. In this role, THESAN-DARK-1 is used to quantify how the reionization redshift $z_{\rm reion}$ correlates with halo mass, local overdensity, and present-day pair properties, and to assess how nearby massive structures regulate the reionization history of Milky Way–Andromeda–like systems (Zhao et al., 22 Jul 2025).

1. Position within the THESAN simulation programme

THESAN is a suite of large-volume radiation-magneto-hydrodynamic simulations designed to model both large-scale reionization morphology and realistic galaxy formation. In the broader suite, the flagship radiation-hydrodynamic calculation evolves a cubic volume of side $L_{\rm box}=95.5\ \mathrm{cMpc}$ with the IllustrisTNG galaxy-formation model, on-the-fly radiative transfer, non-equilibrium thermochemistry, and realistic ionizing sources, using the moving-mesh code AREPO with the AREPO-RT extension (Jamieson et al., 2024). THESAN-DARK-1 does not supply this baryonic and radiative physics; instead, it provides the dark-matter backbone needed to continue the same initial conditions to $z=0$ and thus connect reionization-era structure to present-day halo demographics (Zhao et al., 22 Jul 2025).

The operational link between the two runs is a merger-tree analysis plus a cross-link at $z=5.5$. Haloes in THESAN-DARK-1 are mapped back onto their baryonic counterparts in THESAN-1, producing a catalogue of “bijectively matched” subhaloes. This is the step that enables a direct association between an early reionization history and a final halo mass, environment, or Local Group pair configuration. A plausible implication is that THESAN-DARK-1 functions as the $z=0$ interpretive layer of the THESAN reionization programme, while THESAN-1 remains the source of the reionization physics itself.

2. Core quantities: $z_{\rm reion}$, smoothing scale, and environmental metrics

The paper defines the reionization redshift $z_{\rm reion}$ as the latest time when the volume-smoothed neutral hydrogen fraction drops below a threshold, with the fiducial choice $x_{\rm HI}=0.5$; the appendix also tests $z\gtrsim 5.5$0 and $z\gtrsim 5.5$1 (Zhao et al., 22 Jul 2025). The environmental measure is a smoothed overdensity field, with Gaussian smoothing scales ranging from 125 ckpc to 1 cMpc. The analysis is built around the claim that small-scale structure and large-scale environment both matter, but in different ways.

This dependence on smoothing scale is not a numerical footnote but a physical part of the interpretation. Using a smaller smoothing length retains more of the patchiness of the ionizing field and local density structure, which generally yields earlier inferred reionization times and stronger gradients in $z\gtrsim 5.5$2. Larger smoothing scales wash out small-scale structure, suppress local variations, and make different environments look more similar. The appendix shows that changing $z\gtrsim 5.5$3 from 125 ckpc to 1 cMpc shifts the median $z\gtrsim 5.5$4–mass relation downward, especially at $z\gtrsim 5.5$5, and reduces the contrast in $z\gtrsim 5.5$6 across overdensity bins. The physical interpretation given in the paper is that reionization is highly patchy on sub-Mpc scales, so averaging over too large a volume “infills” bubbles and delays the apparent moment when the neutral fraction falls below threshold.

Within the broader THESAN literature, the reionization-redshift field is also used as a chronology of ionized bubble growth and mergers, reinforcing the idea that $z\gtrsim 5.5$7 is a structurally meaningful field rather than a purely descriptive timestamp (Jamieson et al., 2024).

3. Mass bias, overdensity bias, and the imprint of patchy reionization

A central result is that $z\gtrsim 5.5$8 depends on halo mass and local overdensity in the expected “biased reionization” sense, but with important nuance (Zhao et al., 22 Jul 2025). At fixed mass, haloes in denser regions reionize earlier. This is clearest at $z\gtrsim 5.5$9: haloes with $z_{\rm reion}$0 show a strong positive correlation between mass and reionization redshift, and the most massive systems, especially those with $z_{\rm reion}$1, are almost all early-reionizing.

The environmental trend is similarly strong. At $z_{\rm reion}$2, $z_{\rm reion}$3 rises steeply with overdensity $z_{\rm reion}$4, especially over $z_{\rm reion}$5–100. In the $z_{\rm reion}$6 plane, the distribution peaks around $z_{\rm reion}$7–5, which the authors interpret as a signature of inside-out reionization: moderately overdense regions host clustered sources that ionize themselves first, while underdense regions lag behind. By $z_{\rm reion}$8, the correlation is still present, but weaker, because nonlinear structure growth and halo mergers partially erase the original imprint.

These results place THESAN-DARK-1 within an explicitly environmental picture of reionization. Halo mass is not the sole predictor of reionization timing, and present-day mass alone does not retain the full information content of the earlier radiation field. The simulation instead supports a coupled bias in which local source clustering and ambient density jointly regulate when a region crosses its reionization threshold.

4. Virgo-like clusters and the finite range of protocluster influence

The most direct environmental experiment in the paper concerns nearby massive structures. The authors select 20 Virgo analogues with $z_{\rm reion}$9 and examine Local Group pairs within 20 cMpc of them (Zhao et al., 22 Jul 2025). They find that Virgo-like clusters accelerate reionization in their surroundings out to roughly $L_{\rm box}=95.5\ \mathrm{cMpc}$0–$L_{\rm box}=95.5\ \mathrm{cMpc}$1 cMpc.

Within that distance, the central cluster dominates the timing: reionization is earlier near the cluster center, with $L_{\rm box}=95.5\ \mathrm{cMpc}$2 at small radii, and the ionized region expands outward over time. Beyond $L_{\rm box}=95.5\ \mathrm{cMpc}$3–6 cMpc, local sources and the ambient background become more important, and by $L_{\rm box}=95.5\ \mathrm{cMpc}$4 cMpc the delay relative to the cluster center reaches $L_{\rm box}=95.5\ \mathrm{cMpc}$5–$L_{\rm box}=95.5\ \mathrm{cMpc}$6 Myr. The paper interprets this as a transition from source-dominated to environment-dominated reionization.

The significance of this result is twofold. First, it establishes a finite but substantial radius of influence for proto-cluster environments in the Local Universe. Second, it implies that present-day Local Group analogues cannot be interpreted in isolation: their reionization histories depend not only on their own progenitors, but also on whether they formed inside the expanding ionized domain of a nearby massive structure.

5. Local Group analogues and asynchronous pair reionization

The Local Group analysis is the main $L_{\rm box}=95.5\ \mathrm{cMpc}$7 application of THESAN-DARK-1. The paper identifies 224 LG analogues near Virgo-like clusters, with the pair criteria $L_{\rm box}=95.5\ \mathrm{cMpc}$8, $L_{\rm box}=95.5\ \mathrm{cMpc}$9, isolation within 2 cMpc of the midpoint, and mass ratio $z=0$0 (Zhao et al., 22 Jul 2025). The reionization timing of the two members depends on the smoothing scale:

$z=0$1	M31-like $z=0$2	MW-like $z=0$3
125 ckpc	$z=0$4	$z=0$5
250 ckpc	$z=0$6	$z=0$7
500 ckpc	$z=0$8	$z=0$9
1 cMpc	$z=5.5$0	$z=5.5$1

The more massive member tends to reionize earlier on average, but the difference is modest and sensitive to smoothing. The median redshift offset between the two members is

$z=5.5$2

again for $z=5.5$3. This means the typical asymmetry is only of order $z=5.5$4–0.6, but with substantial scatter; in some cases the less massive halo reionizes first. In time units, the offsets reach up to roughly $z=5.5$5 Myr between pair members.

The controlling variable is not primarily mass ratio. Pairwise timing offsets correlate more with present-day separation than with mass ratio. Pairs with small separations, especially $z=5.5$6 cMpc, often reionize nearly synchronously, with $z=5.5$7 Myr. More widely separated pairs show larger differences, sometimes approaching $z=5.5$8 Myr. By contrast, the mass ratio dependence is weak and non-monotonic: nearly symmetric pairs may show somewhat earlier and more synchronized reionization, but the scatter is large, and the paper concludes that mass ratio is not the primary driver. There is also only a weak dependence on total dynamical energy, with more bound systems tending to reionize earlier and more synchronously, but this trend is modest. The physical reading offered by the authors is that the shared large-scale environment of the pair, not simply the instantaneous pair properties, is the main control on whether two Local Group haloes reionize together or offset in time.

6. Physical interpretation, broader THESAN context, and astrophysical implications

The paper’s overall picture is an extreme inside-out reionization scenario (Zhao et al., 22 Jul 2025). Dense peaks produce the first galaxies and ionize their surroundings quickly; ionized bubbles then expand into adjacent lower-density regions; voids and diffuse outskirts reionize later, often because they are reached by external fronts rather than by local sources. In the maps discussed by the authors, the ionized core grows outward from the most massive halo, and the photoionization rate, neutral fraction, and temperature all evolve coherently near the center before spreading outward. This reading is consistent with the broader THESAN analysis of bubble growth, which finds a hierarchical and percolative history with a dominant bubble emerging by $z=5.5$9–10 (Jamieson et al., 2024), and with the galaxy–IGM studies that distinguish inside-out and outside-in reionization environments in late-time Ly$z=0$0 opacity statistics (Garaldi et al., 2024).

For the Local Group and its dwarf galaxies, the implication is that reionization timing is not universal even within a single group environment. Some dwarf satellites may have been exposed to ionizing radiation very early if they lay near a Virgo-like proto-cluster or within the denser part of a Local Group progenitor region, while others in lower-density or more isolated environments could have reionized substantially later. This matters for interpreting fossil star formation histories in Milky Way dwarfs: old stellar populations, early quenching, and “reionization fossils” need not reflect a single global reionization time, but rather a spatially varying local history. Within the wider THESAN framework, this emphasis on environmental suppression and satellite accretion is echoed by THESAN-ZOOM, where photoevaporation during reionization suppresses further star formation in minihaloes and the remnants of Pop III stars primarily reside in satellite galaxies of larger haloes at lower redshifts (Zier et al., 5 Mar 2025).

The paper is also explicit about its interpretive limits. The exact numerical values of $z=0$1 depend on the smoothing scale and on the chosen neutral-fraction threshold, so the conclusions are robust in a qualitative sense but not intended as unique, threshold-independent timestamps. What remains stable across these choices is the qualitative structure of the result: reionization in the Local Universe is strongly biased by mass and environment, accelerated by nearby Virgo-like clusters over $z=0$2–10 cMpc scales, and only loosely synchronized within Local Group pairs unless they share very similar environments and small separations.