Do galaxy mergers increase star formation and turbulence at cosmic noon?

A. Marchal; E. Wisnioski; I. Kanowski; J. T. Mendel; N. M. F\"orster Schreiber; T. Tsukui

arxiv: 2604.15002 · v1 · submitted 2026-04-16 · 🌌 astro-ph.GA

Do galaxy mergers increase star formation and turbulence at cosmic noon?

I. Kanowski , J. T. Mendel , E. Wisnioski , N. M. F\"orster Schreiber , A. Marchal , T. Tsukui This is my paper

Pith reviewed 2026-05-10 10:36 UTC · model grok-4.3

classification 🌌 astro-ph.GA

keywords galaxy mergersstar formation ratevelocity dispersioncosmic noongalaxy interactionsH-alpha kinematicsKMOS3D

0 comments

The pith

Galaxy interactions at cosmic noon raise star formation rates by 0.1 dex with no detectable rise in velocity dispersion.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper tests whether mergers and interactions boost both star formation and gas turbulence in galaxies at redshifts 1 to 2, using integral-field spectroscopy from the KMOS3D survey. It classifies galaxies via an interaction strength parameter Q_P and compares likely interacting systems to control galaxies matched in stellar mass and cosmic time. The analysis finds a modest enhancement in H-alpha flux and star-formation rate, matching the size of the effect seen in nearby galaxy pairs, yet velocity dispersion shows no significant difference. This distinction matters because cosmic noon is the era when most stars formed, so merger-driven processes help shape how galaxies assemble and why their internal motions behave as they do.

Core claim

Likely interacting galaxies exhibit increased H-alpha fluxes and star-formation rates at the level of approximately 0.1 dex relative to isolated controls, while showing no significant increase in the level of velocity dispersion.

What carries the argument

The interaction strength parameter Q_P that separates likely interacting from isolated galaxies, paired with spatially non-parametric deconvolution of H-alpha kinematics to extract velocity dispersion.

If this is right

Interacting galaxies at z approximately 1-2 show a star-formation enhancement comparable in size to that observed in local pairs.
Velocity dispersion does not rise, consistent with physical saturation in already turbulent gas or with current spectral-resolution limits.
Merger effects on star formation appear continuous across cosmic time rather than qualitatively different at high redshift.
The null result on turbulence implies that any merger-driven stirring is either brief or masked by the high baseline turbulence at cosmic noon.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

Galaxy-formation simulations should check whether their sub-grid turbulence prescriptions reproduce the observed decoupling between star-formation boost and dispersion at high redshift.
Deeper or higher-resolution integral-field observations could reveal small dispersion increases currently below detection threshold.
Extending the analysis to other tracers of turbulence, such as molecular-gas line widths, would test whether the result is H-alpha specific.
The modest SFR offset supports models in which mergers contribute to the scatter in the star-forming main sequence but do not dominate it at cosmic noon.

Load-bearing premise

The Q_P parameter cleanly distinguishes interacting from isolated galaxies and that matching controls solely on stellar mass and lookback time removes all other influences on measured star-formation rate and velocity dispersion.

What would settle it

A statistically significant increase in velocity dispersion for the interacting sample when using an independent interaction classifier or spectra with finer velocity resolution would falsify the no-increase result.

Figures

Figures reproduced from arXiv: 2604.15002 by A. Marchal, E. Wisnioski, I. Kanowski, J. T. Mendel, N. M. F\"orster Schreiber, T. Tsukui.

**Figure 1.** Figure 1: The 3D-HST parent sample, with updated redshift information and quality cuts applied (small grey points), overplotted with the full KMOS3D sample (large grey circles), the KMOS3D sample with quality cuts applied as described in Section 2.3 (large black circles), and the final kinematic sample described in Section 3.5 (large open circles). The solid line describes the fit to the 90 percent mass completeness… view at source ↗

**Figure 2.** Figure 2: A comparison of 𝑄𝑃 and 𝑟𝑄𝑃 for all galaxies in the spectroscopic redshift sample, coloured by the effective radius of each galaxy. The colourbar spans the 1 − 99th percentile of the galaxy effective radii. Closed black circles denote KMOS3D galaxies which pass the quality cuts given in Section 2.3 and open circles show the final kinematic sample described in Section 3.5. Vertical lines denote the control 𝑄… view at source ↗

**Figure 3.** Figure 3: This figure shows the galaxy U4_34173. The left-most panel shows the galaxy in a 3-colour JWST image (f090w, f115w, f150w), with the KMOS field of view overplotted as a box. The remaining panels show the results of the ROHSA-SNAPD modelling, where each map is the median fit from 100 noise resamples. As the KMOS3D cubes were cropped to remove noisy edge pixels, as discussed in Section 3.2, we pad the edges … view at source ↗

**Figure 4.** Figure 4: A comparison of interaction strength 𝑄𝑃 with the jackknifed Δ log offset in galaxy H𝛼 flux. Symbols denote the velocity regularisation value, 𝜆𝜇, used in the fit. Vertical solid and dashed lines show the control and likely interacting 𝑄𝑃 bounds respectively. Small offsets in 𝑄𝑃 are applied to the different 𝜆𝜇 results for visual clarity. deconvolved velocity dispersion offsets in the 0.5 < 𝑧 < 1.5 bin are w… view at source ↗

**Figure 5.** Figure 5: The ΔAv offset from the control for both the kinematic sample (closed points) and the high-quality star-forming 3D-HST spectroscopic sample (open points). The layout of the figure is similar to [PITH_FULL_IMAGE:figures/full_fig_p009_5.png] view at source ↗

**Figure 6.** Figure 6: The two panels compare interaction strength with the Δ log offset from the control of galaxy H𝛼 SFR (left panel) and deconvolved velocity dispersion (right panel). The velocity dispersion offsets are calculated using only well-resolved galaxies, as described in Section 3.5. The layout of each panel is similar to [PITH_FULL_IMAGE:figures/full_fig_p010_6.png] view at source ↗

**Figure 7.** Figure 7: The Kaplan-Meier estimator results of galaxy velocity dispersions for velocity regularisation values of 𝜆𝜇 = 2, 4 and 8. Only galaxies with redshifts in the range 𝑧 = 0.5 − 1.5 are included. The median, 16th and 84th percentiles of the Kaplain-Meier estimator are shown by errorbars with closed and open markers for the likely interacting and control samples respectively. 2024). However, the lack of increase… view at source ↗

**Figure 9.** Figure 9: The Δ offset of the percent difference between the convolved ROHSA-SNAP and observed velocity dispersion maps for the kinematic sample. The layout of the figure is similar to [PITH_FULL_IMAGE:figures/full_fig_p011_9.png] view at source ↗

read the original abstract

Mergers and interactions can significantly affect the morphological and dynamical properties of galaxies, however the impact of mergers on turbulence at $z > 1$ has not been observationally constrained. In this work we use the interaction strength parameter $Q_P$ to identify likely interacting and isolated galaxies at cosmic noon ($z \sim 1-2$) within the KMOS\textsuperscript{3D} integral field spectroscopy survey, utilising redshifts from the 3D-HST, CANDELS and UVCANDELS surveys. For $186$ galaxies, we measure deconvolved H$\alpha$ kinematics, including velocity dispersion, using a spatially non-parametric approach to account for observational effects in the dynamically diverse range of galaxies. We compare offsets in H$\alpha$ flux, star formation rate (SFR), dust attenuations, and velocity dispersion of likely interacting galaxies to isolated control galaxies matched in mass and lookback time. We find increased H$\alpha$ fluxes and SFRs in the likely interacting sample at the level of $\sim 0.1$ dex, a similar enhancement to studies of local pairs. In contrast, we find no significant increase in the level of velocity dispersion in interacting galaxies compared to their controls. The lack of increase in dispersion may reflect a combination of physical and observational factors, including limits to increasing turbulent motions in an already turbulent medium and spectral resolution limits.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Desk Editor's Note private letter to a colleague

Mergers at z~1-2 show a 0.1 dex SFR boost but no extra velocity dispersion in this KMOS3D sample.

read the letter

The main thing to know is that this paper measures a modest rise in Hα flux and star formation rate for likely interacting galaxies at cosmic noon, but finds no corresponding increase in gas velocity dispersion relative to controls. The null result on turbulence is the part that has not been checked observationally before at these redshifts. They draw 186 galaxies from the KMOS3D integral-field survey, flag interactors with the Q_P parameter, match controls on stellar mass and lookback time, and extract deconvolved Hα kinematics with a non-parametric method that avoids assuming disk or merger models for every object. The SFR offset lines up with local pair studies, and the authors note that the dispersion result could reflect either physical limits in an already turbulent medium or observational ceilings from spectral resolution and beam smearing. The non-parametric kinematics step is a practical choice for a mixed sample, and the survey data themselves are from a established program, so the measurements rest on real observations rather than new modeling. The soft spots sit in the sample construction and the strength of the null. Q_P is a standard metric, yet the paper does not report a purity or completeness test for the threshold used, so some contamination between the interacting and isolated groups is possible. Matching on only two variables leaves open differences in environment, gas fraction, or dust that could contribute to the SFR offset or hide a small dispersion change. The abstract lists resolution limits as one explanation, but without the actual error budgets, significance values, or a quantitative assessment of how much beam smearing affects the dispersion comparison, it is difficult to judge whether the measurement had the power to detect an increase if one were present. The split sample is modest, which further limits how firmly the null can be read. This work is aimed at people modeling galaxy growth and dynamical heating during the peak epoch of star formation. A reader who needs an updated observational anchor on merger effects at z>1 will get a usable data point from it. The paper deserves a serious referee because the underlying dataset is solid and the turbulence question is open, even though the analysis will need closer checks on the matching statistics and measurement uncertainties before the conclusions can be taken as firm. I would send it out for review.

Referee Report

3 major / 3 minor

Summary. The manuscript examines the effects of galaxy mergers and interactions on star formation and turbulence at cosmic noon (z ~ 1-2) using the KMOS3D integral field spectroscopy survey. It applies the interaction strength parameter Q_P to classify 186 galaxies (with redshifts from 3D-HST, CANDELS, and UVCANDELS) as likely interacting or isolated, measures deconvolved Hα kinematics including velocity dispersion via a spatially non-parametric method, and compares Hα flux, SFR, dust attenuation, and velocity dispersion to control samples matched in stellar mass and lookback time. The central results are a ~0.1 dex enhancement in Hα fluxes and SFRs for the interacting sample (similar to local studies) but no significant increase in velocity dispersion, with the null result attributed to limits in an already turbulent medium and spectral resolution constraints.

Significance. If the findings are robust, the work provides valuable observational constraints on merger-driven processes at high redshift, where baseline turbulence is elevated. The modest SFR boost aligns with local pair studies, implying continuity in merger-induced star formation across cosmic time. The absence of a dispersion increase highlights possible saturation of turbulent motions and observational challenges with IFS data, contributing to models of galaxy dynamical evolution. A strength is the non-parametric kinematic extraction, which accommodates the morphological diversity of the sample without assuming idealized models.

major comments (3)

[sample classification and Q_P section] In the section on sample classification using Q_P: the choice of interaction threshold and its ability to cleanly separate interacting from isolated systems is central to attributing the ~0.1 dex SFR offset and the null dispersion result. Without reported purity/completeness estimates, cross-validation against visual inspections, or mock observations, residual misclassification could dilute signals and undermine the contrast between the two outcomes.
[control matching and sample comparison section] In the control sample construction and matching procedure: matching solely on stellar mass and lookback time leaves open the possibility of residual confounders (e.g., environment, gas fraction, or dust properties) that affect both Hα flux/SFR and deconvolved dispersion at z ~ 1-2. The paper should demonstrate that the 0.1 dex enhancement persists under expanded matching or include quantitative tests for unmatched variables.
[kinematics extraction and dispersion results] In the kinematics measurement and results section on velocity dispersion: given the abstract's reference to spectral resolution limits and beam-smearing, a detailed error budget, resolution tests, and assessment of how the non-parametric deconvolution performs on the high-turbulence baseline population is needed to confirm that the null result is not an artifact of measurement uncertainties.

minor comments (3)

[abstract] The abstract would benefit from explicitly stating the adopted Q_P threshold value and the exact number of galaxies in each subsample for immediate clarity.
[throughout manuscript] Ensure uniform notation for velocity dispersion (e.g., consistently σ or σ_v) and Hα quantities across text, figures, and tables.
[discussion section] Consider adding a brief comparison table or figure panel showing the local-universe pair studies referenced for the ~0.1 dex SFR enhancement to facilitate direct visual comparison.

Simulated Author's Rebuttal

3 responses · 0 unresolved

We thank the referee for their constructive and insightful comments on our manuscript. We have addressed each major concern by outlining specific revisions that will strengthen the robustness of our sample classification, control matching, and kinematic analysis. Our point-by-point responses follow.

read point-by-point responses

Referee: In the section on sample classification using Q_P: the choice of interaction threshold and its ability to cleanly separate interacting from isolated systems is central to attributing the ~0.1 dex SFR offset and the null dispersion result. Without reported purity/completeness estimates, cross-validation against visual inspections, or mock observations, residual misclassification could dilute signals and undermine the contrast between the two outcomes.

Authors: We agree that validating the Q_P classification is important. The threshold was adopted from established literature on high-redshift interactions. In the revised manuscript we will add a sensitivity analysis varying the Q_P threshold to show that both the 0.1 dex SFR enhancement and null dispersion result are stable. We will also perform a cross-check against visual classifications for the subset of galaxies with high-resolution CANDELS imaging. Full mock-based purity and completeness estimates are not feasible within the scope of this observational study, but we will discuss how any misclassification would dilute rather than fabricate the observed signals. revision: partial
Referee: In the control sample construction and matching procedure: matching solely on stellar mass and lookback time leaves open the possibility of residual confounders (e.g., environment, gas fraction, or dust properties) that affect both Hα flux/SFR and deconvolved dispersion at z ~ 1-2. The paper should demonstrate that the 0.1 dex enhancement persists under expanded matching or include quantitative tests for unmatched variables.

Authors: We thank the referee for this suggestion. In the revised manuscript we will expand the matching to incorporate local environment density from CANDELS catalogs and dust attenuation where available. We will demonstrate that the ~0.1 dex SFR offset remains significant under these additional constraints and will include quantitative comparisons of unmatched variables between the interacting and control samples to assess any residual differences. revision: yes
Referee: In the kinematics measurement and results section on velocity dispersion: given the abstract's reference to spectral resolution limits and beam-smearing, a detailed error budget, resolution tests, and assessment of how the non-parametric deconvolution performs on the high-turbulence baseline population is needed to confirm that the null result is not an artifact of measurement uncertainties.

Authors: We appreciate this recommendation. The revised methods section will include a comprehensive error budget for the deconvolved Hα velocity dispersion, explicitly accounting for spectral resolution, beam smearing, and noise contributions. We will also present resolution tests performed on mock IFS data cubes with input dispersions matching the turbulent z ~ 1-2 baseline population. These tests will validate the accuracy of our spatially non-parametric deconvolution method and confirm that the null result is not an artifact of observational limitations. revision: yes

Circularity Check

0 steps flagged

No circularity: purely observational sample comparison

full rationale

The paper performs a direct observational comparison of measured Hα fluxes, SFRs, dust attenuations, and deconvolved velocity dispersions between galaxies classified as likely interacting (via the external Q_P parameter) and mass+lookback-time matched controls drawn from the KMOS3D survey. No derivations, model equations, fitted parameters, or predictions are presented that reduce to the input data by construction. The central claims are statistical offsets (~0.1 dex SFR enhancement, null result for dispersion) obtained from non-parametric kinematic measurements; these do not rely on self-definitional loops, fitted inputs renamed as predictions, or load-bearing self-citations. The analysis is self-contained against external benchmarks (local pair studies) and does not invoke uniqueness theorems or ansatzes from prior author work.

Axiom & Free-Parameter Ledger

1 free parameters · 1 axioms · 0 invented entities

The central claims rest on the assumption that Q_P traces true interaction strength and that the control matching isolates the merger effect; no free parameters or invented entities are described in the abstract.

free parameters (1)

Q_P interaction threshold
The specific cutoff value used to classify galaxies as interacting is a key selection parameter whose value is not stated in the abstract.

axioms (1)

domain assumption Q_P parameter reliably identifies interacting galaxies with low contamination
Invoked to define the likely interacting sample versus controls.

pith-pipeline@v0.9.0 · 5574 in / 1268 out tokens · 28711 ms · 2026-05-10T10:36:39.102897+00:00 · methodology

discussion (0)

Reference graph

Works this paper leans on

5 extracted references · 5 canonical work pages

[1]

Rep., 114, 321 Baron D., Netzer H., Lutz D., Davies R

Argudo-Fernández M., et al., 2013, A&A, 560, A9 Astropy Collaboration et al., 2013, A&A, 558, A33 Astropy Collaboration et al., 2018, AJ, 156, 123 Astropy Collaboration et al., 2022, ApJ, 935, 167 Athanassoula E., 1984, Phys. Rep., 114, 321 Baron D., Netzer H., Lutz D., Davies R. I., Prochaska J. X., 2024, ApJ, 968, 23 Barton E. J., Geller M. J., Kenyon S...

work page doi:10.26182/r3bp-1226 2013
[2]

The mass ratio of each galaxy was calculatedbycomparingtothemassofthegalaxywhichcontributed the largest individual interaction strength,𝑄𝑖 𝑃, to the galaxy. It is also possible to study galaxies with lower𝑄𝑃 values and study trends in merger impacts with interaction strength, in a sim- ilar way to previous works which have studied trends in projected sepa...

work page 2013
[3]

A2 and could account for the larger offset

In the later, galaxies with AGN have been removed, this is not possible in the 3D-HST sample used in Fig. A2 and could account for the larger offset. A3 Galaxy effective radii Figure 2 demonstrates that galaxies with larger effective radii tend to have larger𝑄𝑃 values. This is because, as described Equation 3, a galaxy with a larger effective radius would...

work page 2014
[4]

Performing size-matching test on the final kinematic sample itself was not possible as the control sample size was too small to match to a reasonable number of controls. MNRAS000, 1–15 (2026) Do mergers increase SF and turbulence?15 0 100 Counts 6 4 2 QP 0.2 0.1 0.0 0.1 0.2 0.3 0.4 0.5 Jackknifed median log SFRH (3D-HST) Likely interaction Unlikely intera...

work page 2026
[5]

APPENDIX B: UPDATES TOROHSA-SNAP B1 Liikelihood We have included the option to use a Cauchy likelihood in the cost function ofROHSA-SNAPD 6, instead of the default Gaussian likeli- hood used in Kanowski et al. (2025). As Kanowski et al. (2025), appliedROHSA-SNAPDto mock observations generated from simu- lateddatawithaddedGaussiannoise,aGaussianlikelihoodw...

work page 2025

[1] [1]

Rep., 114, 321 Baron D., Netzer H., Lutz D., Davies R

Argudo-Fernández M., et al., 2013, A&A, 560, A9 Astropy Collaboration et al., 2013, A&A, 558, A33 Astropy Collaboration et al., 2018, AJ, 156, 123 Astropy Collaboration et al., 2022, ApJ, 935, 167 Athanassoula E., 1984, Phys. Rep., 114, 321 Baron D., Netzer H., Lutz D., Davies R. I., Prochaska J. X., 2024, ApJ, 968, 23 Barton E. J., Geller M. J., Kenyon S...

work page doi:10.26182/r3bp-1226 2013

[2] [2]

The mass ratio of each galaxy was calculatedbycomparingtothemassofthegalaxywhichcontributed the largest individual interaction strength,𝑄𝑖 𝑃, to the galaxy. It is also possible to study galaxies with lower𝑄𝑃 values and study trends in merger impacts with interaction strength, in a sim- ilar way to previous works which have studied trends in projected sepa...

work page 2013

[3] [3]

A2 and could account for the larger offset

In the later, galaxies with AGN have been removed, this is not possible in the 3D-HST sample used in Fig. A2 and could account for the larger offset. A3 Galaxy effective radii Figure 2 demonstrates that galaxies with larger effective radii tend to have larger𝑄𝑃 values. This is because, as described Equation 3, a galaxy with a larger effective radius would...

work page 2014

[4] [4]

Performing size-matching test on the final kinematic sample itself was not possible as the control sample size was too small to match to a reasonable number of controls. MNRAS000, 1–15 (2026) Do mergers increase SF and turbulence?15 0 100 Counts 6 4 2 QP 0.2 0.1 0.0 0.1 0.2 0.3 0.4 0.5 Jackknifed median log SFRH (3D-HST) Likely interaction Unlikely intera...

work page 2026

[5] [5]

APPENDIX B: UPDATES TOROHSA-SNAP B1 Liikelihood We have included the option to use a Cauchy likelihood in the cost function ofROHSA-SNAPD 6, instead of the default Gaussian likeli- hood used in Kanowski et al. (2025). As Kanowski et al. (2025), appliedROHSA-SNAPDto mock observations generated from simu- lateddatawithaddedGaussiannoise,aGaussianlikelihoodw...

work page 2025