Projects | Andrew Robertson

Roman Galaxy Redshift Survey

Wed, 26 Nov 2025 00:00:00 +0000

I am a member of the ’s , where I lead the development of mock galaxy catalogues used as input for simulating grism observations. These catalogues are generated using the semi-analytic model, run on the simulations, and incorporate a new calibration described . You can hear me talk about this work .

Overview

The Roman Galaxy Redshift Survey (GRS) will measure precise redshifts for millions of galaxies over a large fraction of the sky. These measurements will enable detailed studies of large-scale structure, including (BAO) and (RSD), providing powerful constraints on the expansion history of the Universe and the growth of cosmic structure.

Roman achieves this using slitless spectroscopy with a : instead of targeting individual galaxies, every object in the field is dispersed into a spectrum. This makes the survey extremely efficient, but also introduces significant challenges for data analysis.

Why grism spectroscopy is powerful—and difficult

Simulated Roman data, with a direct image (top) and grism image (bottom). Image credit: IPAC/STScI/Goddard

Grism observations provide a unique combination of depth, area, and spectral information:

They obtain spectra for all objects simultaneously, avoiding target selection biases
They enable precise redshift measurements over large cosmological volumes
They combine deep imaging and spectroscopy of the same galaxies, enabling consistent measurements of both their shapes and their spectra

However, these advantages come with complications:

Spectra from nearby objects can overlap and contaminate one another
Light from each galaxy is spread out across the detector, so its position, shape, and orientation all affect how its spectrum appears
The ability to measure a redshift depends on factors such as galaxy size, orientation, emission-line strength, and proximity to other objects

As a result, interpreting grism data is not straightforward. To be confident in our conclusions, we test analysis pipelines on simulated datasets where the true answer is known. These simulations also allow us to explore how different survey strategies affect the quality of the final measurements.

Approach: realistic mock universes

My work focuses on building realistic mock datasets that connect theoretical models of galaxy formation to the data products that Roman will observe.

Galaxy catalogues
I generate large mock catalogues using Galacticus, calibrated to reproduce key observables such as the stellar-to-halo mass relation and emission-line luminosity functions.

Spectral modelling
These catalogues include emission-line predictions (e.g. Hα, [OIII]) and continuum spectra, enabling realistic modelling of the signals detected by the grism.

Mock observations
The catalogues are passed through grism simulation pipelines to produce synthetic images that include instrumental effects, noise, and spectral overlap.

Survey design and validation
These mocks are used to test survey strategies, optimise observing configurations, and validate analysis pipelines. They provide a controlled way to assess how well we can recover cosmological information from the data.

Why this matters

The statistical power of Roman will be unprecedented, but so is the complexity of the data. Extracting robust cosmological constraints requires understanding the full chain from galaxy formation to observed spectra.

Mock datasets play a central role in this process. They allow us to:

Quantify selection effects and biases
Test redshift recovery and catalogue construction
Validate clustering measurements and cosmological inference

A major focus of my work is ensuring that these tests are realistic enough that any conclusions drawn from Roman data are reliable.

Looking ahead

Roman will transform our ability to map the large-scale structure of the Universe.

By combining wide-area spectroscopy with high-resolution imaging, it will enable joint analyses of galaxy clustering, weak lensing, and other probes. The challenge will be to ensure that our modelling and analysis methods keep pace with the quality of the data.

Strong Gravitational Lensing

Wed, 18 Mar 2026 00:00:00 +0000

Overview

Strong gravitational lensing occurs when a massive object—such as a galaxy or cluster—bends light from a more distant source, producing multiple images, arcs, or even complete Einstein rings. Because gravity depends only on mass (not whether it is luminous or dark), lensing provides a direct way to map the distribution of dark matter.

It is also one of the most sensitive probes of small-scale structure: even low-mass subhaloes can leave detectable imprints in the detailed morphology of lensed images.

My work focuses on using strong lensing as a test of dark matter physics, while ensuring that the inferences drawn are robust to modelling assumptions and observational limitations.

Galaxy-scale strong lensing

Galaxy-scale lenses provide one of the cleanest ways to detect dark matter substructure. In these systems, small perturbations to lensed arcs or Einstein rings can reveal the presence of otherwise invisible subhaloes.

A handful of low-mass substurctures have now been detected using both optical imaging and VLBI (radio) observations. These measurements are often interpreted as constraints on the abundance and internal structure of dark matter subhaloes, and therefore on the nature of dark matter itself.

However, the interpretation is not always straightforward. A key focus of my work is testing how robust these inferences are using realistic mock datasets. By generating synthetic observations and analysing them with the same tools used on real data, we can assess when apparent detections genuinely require new dark matter physics.

Example: SDSS J0946+1006

The lens system SDSS J0946+1006 has been widely discussed as evidence for a very dense dark matter subhalo.

In , we showed that the data can also be explained if the perturber hosts a faint galaxy. When this is included in the modelling, there is no need to invoke an unusually compact dark matter structure.

This system illustrates how sensitive lensing inferences can be to seemingly small modelling choices, and why careful validation is essential.

Cluster-scale strong lensing

Strong lensing galaxy cluster Abell 370. Image credit: NASA, ESA, and J. Lotz and the HFF Team (STScI)

On larger scales, galaxy clusters produce complex networks of arcs and multiple images, allowing detailed reconstruction of their mass distributions.

Several studies have suggested potential discrepancies between observed cluster substructure and predictions from CDM simulations. These include claims of an excess of massive subhaloes or differences in their spatial distribution.

While these results are intriguing, their interpretation remains uncertain. In particular, comparisons between simulations and observations can be affected by selection effects, projection, and modelling assumptions. In , I argued that some reported tensions may be less significant than initially claimed.

Even so, cluster-scale lensing remains a promising avenue for testing dark matter models, particularly as both simulations and observational datasets improve.

Merging clusters

Merging clusters provide a complementary test of dark matter physics.

In these systems, the different components—stars (galaxies), gas, and dark matter—can become spatially separated during the merger. By comparing their relative distributions, it is possible to constrain properties such as the self-interaction cross-section of dark matter.

The Bullet Cluster is the most famous example, and has provided some of the strongest constraints on self-interacting dark matter.

I am currently involved in efforts to interpret new observations of merging clusters, including a recently-completed JWST programmes and a recently-accepted JWST proposal to . This work combines simulations and lensing analyses to understand how reliably we can extract constraints on dark matter physics from these systems.

Self-Interacting Dark Matter (SIDM)

Wed, 18 Mar 2026 00:00:00 +0000

Overview

Dark matter makes up most of the matter in the Universe, but its fundamental nature remains unknown. The standard model—cold dark matter (CDM)—successfully explains structure on large scales. In simulations that include only dark matter, it predicts that haloes develop steep, centrally concentrated cusps.

Early observations of dwarf and low-surface-brightness galaxies instead appeared to favour constant-density cores, in tension with these predictions. This “ ” was one of the original motivations for considering alternatives such as self-interacting dark matter (SIDM), in which dark matter particles scatter off one another and naturally redistribute energy to produce lower-density centres.

Subsequent work complicated this picture. On the theoretical side, it became clear that baryonic processes—particularly energy injection from supernovae—can also transform cusps into cores within CDM. On the observational side, the situation evolved from a simple “cores vs cusps” dichotomy to a broader diversity of inner density profiles across galaxies, even at fixed mass.

At the same time, constraints from systems such as the showed that large, velocity-independent self-interaction cross-sections are ruled out. While modest cross-sections remain viable—and can produce cores in dwarf galaxies—much larger cross-sections lead to more dramatic evolution. In this regime, heat conduction can drive a process known as core collapse, producing dense central regions. This naturally generates a wide diversity of inner density profiles, but the cross-sections required are incompatible with constraints from galaxy clusters.

This has led to increasing interest in velocity-dependent SIDM models, which are well motivated in particle physics. In these models, interactions are strong in low-velocity systems (like dwarf galaxies), but suppressed in high-velocity environments (like galaxy clusters), allowing them to evade existing constraints while retaining observable signatures.

For a broader overview of these ideas, see this , or this , which provides an excellent introduction.

Simulating an SIDM Universe

During my PhD, I developed and implemented methods to include dark matter self-interactions in large-scale simulations, and ran some of the first cosmological hydrodynamical simulations with SIDM—variants of the and simulation programmes. These allowed us to study how self-interactions affect galaxies and clusters in realistic environments, where dark matter, gas, and stars all evolve together. While these simulations were not released as a public dataset, I am very happy to share them—please get in touch if they would be useful.

A key focus of this work has been connecting simulations to observable quantities. For example, I have generated mock observations to study , where projection effects and measurement noise make it difficult to robustly distinguish between dark matter models.

Although I no longer run large SIDM simulations myself, I remain closely involved in this area. In particular, I have helped supervise new simulation efforts as part of the collaboration, using the publicly released . This code addresses many of the numerical challenges involved in modelling processes such as core collapse, which are central to the SIDM models of most current interest.

I am also supervising the development of a suite of simulations of merging galaxy clusters—selected from large cosmological simulations—which will be used to extract constraints on the nature of dark matter from recent .

Strong lensing tests of the nature of dark matter

The cleanest tests of dark matter physics come from systems that are dominated by dark matter itself. The challenge is that such systems are, by definition, difficult to observe directly. However, we can detect them through their gravitational effects. Two of the main approaches currently being explored are stellar streams and strong gravitational lensing (see my for more details).

Strong lensing is particularly powerful because it can probe small-scale structure directly, through the perturbations caused by dark matter subhaloes. There have been a number of intriguing results on galaxy scales—from both optical imaging and ( ) from radio interferometry—as well as hints of unusual behaviour in cluster-scale systems.

At the same time, these measurements are highly sensitive to modelling assumptions. A useful example is , which has been interpreted as hosting an unusually dense dark matter subhalo. In , we showed that the data can also be explained if the perturber hosts a faint galaxy, without requiring extreme dark matter properties. This highlights how easily astrophysical effects can masquerade as new physics.

A key goal going forward is to test these analyses using realistic mock data generated from both CDM and SIDM simulations, and to understand how robust the inferred constraints really are.