Prospects for disentangling dark matter with weak lensing

TL;DR

This work tackles the problem of disentangling ultra-light axion dark matter from baryonic feedback using weak lensing, focusing on a mixed cold+axion dark matter scenario. The authors develop a halo-model-based framework that combines a standard $ ext{LCDM}$ halo model with an axion halo component (featuring soliton cores) and a baryonic-feedback extension, parameterized by $m_{ m ax}$, $f_{ m ax}$, and $ heta_{ m AGN}$. They build an emulator for the axion nonlinear power spectrum to enable efficient MCMC analyses, forecast LSST Year-1 cosmic shear constraints, and investigate the potential for power-spectrum reconstruction as a model-agnostic diagnostic. The results show strong LSST Y1 sensitivity to axions across a broad range of masses, but reveal degeneracies with feedback at intermediate masses; with external feedback constraints or via reconstruction of $P_{ m m}(k)$, LSST can break these degeneracies and potentially detect a non-negligible axion component (e.g., $f_{ m ax} o 0.1$ for $m_{ m ax} o 10^{-25}$ eV at a few-sigma level). The study highlights the need for dedicated hydrodynamical simulations including axion components to robustly disentangle dark matter and baryonic physics in upcoming surveys and CMB lensing analyses.

Abstract

We investigate the degeneracy between the effects of ultra-light axion dark matter and baryonic feedback in suppressing the matter power spectrum. We forecast that galaxy shear data from the Rubin Observatory's Legacy Survey of Space and Time (LSST) could limit an axion of mass $m = 10^{-25}\,\mathrm{eV}$ to be $\lesssim 5\%$ of the dark matter, stronger than any current bound, if the interplay between axions and feedback is accurately modelled. Using a halo model emulator to construct power spectra for mixed cold and axion dark matter cosmologies, including baryonic effects, we find that galaxy shear is sensitive to axions from $10^{-27}\,\mathrm{eV}$ to $10^{-21}\,\mathrm{eV}$, with the capacity to set competitive bounds across much of this range. For axions with $m \sim 10^{-25}\,\mathrm{eV}$, the scales at which axions and feedback impact structure formation are similar, introducing a parameter degeneracy. We find that, with an external feedback constraint, we can break the degeneracy and constrain the axion transfer function, such that LSST could detect a $10^{-25}\,\mathrm{eV}$ axion comprising 10\% of the dark matter at $\sim 3 σ$ significance. Direct reconstruction of the non-linear matter power spectrum provides an alternative way of analysing weak lensing surveys, with the advantage of identifying the scale-dependent features in the data that the dark matter model imposes. We advocate for dedicated cosmological hydrodynamical simulations with an axion dark matter component so that upcoming galaxy and cosmic microwave background lensing surveys can disentangle the dark matter-baryon transfer function.

Figures (10)

• Figure 1: For four cosmological scenarios we show Left: the ratio of the power spectrum, $P_\mathrm{m}(k)$, compared to a $\Lambda$CDM dark matter-only case, $P_\mathrm{\Lambda\mathrm{CDM}}(k)$, as a function of wavenumber $k$ and Right: the corresponding forecasted result of a $\Lambda{\rm CDM}$ inference (ignoring the presence of axions and baryonic feedback) of LSST year 1 cosmic shear data simulated with the four power spectrum scenarios. The four cases we consider are: strong baryonic feedback from the BAHAMAS simulations $\Theta_{\rm{AGN}} = 8.0$BAHAMAS, a phenomenological $A_{\rm{mod}}$ suppression AAGPE2022CPAAGPE and two axion models vogt with masses, $m_\mathrm{ax}$, each comprising a fraction, $f_\mathrm{ax}$, of the total dark matter Rogers_S8tensionaxions. The marginalised posteriors in the $S_{8}-\Omega_{\rm{m}}$ plane are statistically indistinguishable despite the different power spectrum transfer functions; the inner and outer contours respectively indicate the 68% and 95% credible regions. This highlights how scenarios with distinct power spectra are degenerate in the $S_8-\Omega_{\rm{m}}$ plane, but would be distinguishable if we could reconstruct their matter power spectra and isolate the differences in shapes.
• Figure 2: Left: the effect of varying axion mass $m_\mathrm{ax}$ on the matter power spectrum $P_\mathrm{m} (k)$, where $m_\mathrm{ax} = 10^{-25}\,\mathrm{eV}$ is highlighted in orange. The axion fraction $f_\mathrm{ax}$ is fixed to 0.1 and feedback is not included, i.e. the model in Eq. \ref{['eq:power_cold_axion_no_feedback']}. Centre: the effect of varying axion fraction $f_\mathrm{ax}$ on $P_\mathrm{m} (k)$, where $f_\mathrm{ax} = 0.1$ is highlighted in orange. The axion mass $m_\mathrm{ax}$ is fixed to $10^{-25}\,\mathrm{eV}$ and feedback is not included, i.e. the model in Eq. \ref{['eq:power_cold_axion_no_feedback']}. Right: the effect of varying feedback parameter $\Theta_\mathrm{AGN}$ on $P_\mathrm{m} (k)$, where $\Theta_\mathrm{AGN} = 8.0$ is highlighted in red. We use the $\Lambda$CDM halo model without axions but with feedback, i.e. as presented in Sec. \ref{['subsection:halomodel_LCDM']} but with the window function in Eq. \ref{['equ:LCDMbaryonWFC']}. In each panel, the same combined axion and feedback model, i.e. Eq. \ref{['equ:total_pk']}, is highlighted in blue and we always show the ratio to the $\Lambda$CDM, feedback-free limit $P_{\Lambda\mathrm{CDM}}(k)$. A lighter axion mass generally suppresses the matter power spectrum to increasingly large scales and a larger fraction suppresses more at a given wavenumber, introducing more complex shapes in power suppression compared to feedback.
• Figure 3: Accuracy of the emulator for axionHMCODE mixed cold and axion non-linear power spectra. We show the 68%, 95% and 99% upper limits on the distribution of the absolute difference between emulator prediction and test power spectrum in ratio to the test value. We achieve sub-percent accuracy up to $k=10\,h\,\rm{Mpc}^{-1}$.
• Figure 4: Posterior distributions from a $\Lambda{\rm CDM}$ analysis of a simulated LSST Y1-like shear correlation function that includes the effects of both axions and feedback. One analysis accounts for baryon feedback by varying the $\Theta_\mathrm{AGN}$ parameter (green), while the other neglects to account for this effect (red) and neither analysis models the axion dark matter component. Left: the fake data are generated for $m_{\rm{ax}}=10^{-21}\,\mathrm{eV}$, $f_\mathrm{ax} = 0.1$ and $\Theta_{\rm{AGN}}=8.0$; right: $m_{\rm{ax}}=10^{-25}\,\mathrm{eV}$, $f_\mathrm{ax} = 0.1$ and $\Theta_{\rm{AGN}}=8.0$. We report marginalised posteriors of $\Omega_{\rm{m}}$, $S_{8}$ and $\Theta_{\rm{AGN}}$; the inner and outer contours respectively indicate the 68% and 95% credible regions. The true parameter values are shown as dashed black lines. We note that the true $S_{8}$ includes the effect of axions on the linear matter power spectrum and so can be lower than the equivalent $\Lambda{\rm CDM}$ value Rogers_S8tensionaxions. The heavier axion ($m_{\rm{ax}}=10^{-21}\,\mathrm{eV}$) slightly boosts power on small scales relative to no axions (though the feedback still overall suppresses the power relative to the feedback-free limit). Feedback strength is therefore marginally under-estimated in order to account for the unmodelled axion effect. The lighter axion ($m_{\rm{ax}}=10^{-25}\,\mathrm{eV}$) strongly suppresses power. As this axion effect is also unmodelled, cosmological and feedback parameters are strongly biased from the truth. Neither case recovers the true cosmology as axions are not modelled.
• Figure 5: Posterior distributions from an extended cosmological analysis of a simulated LSST Y1-like shear correlation function, where axion mass, axion fraction and baryonic feedback parameters are varied. Left: the fake data are generated for $m_{\rm{ax}}=10^{-27}\,\mathrm{eV}$, $f_\mathrm{ax} = 0.1$ and $\Theta_\mathrm{AGN} = 8.0$; centre: $m_{\rm{ax}}=10^{-25}\,\mathrm{eV}$, $f_\mathrm{ax} = 0.1$ and $\Theta_\mathrm{AGN} = 8.0$; right: $m_{\rm{ax}}=10^{-21}\,\mathrm{eV}$, $f_\mathrm{ax} = 0.1$ and $\Theta_\mathrm{AGN} = 8.0$. We report marginalised posteriors of $\log (m_\mathrm{ax}\,[\mathrm{eV}])$, $f_\mathrm{ax}$ and $\Theta_\mathrm{AGN}$; the inner and outer contours respectively indicate the 68% and 95% credible regions. The true parameter values are shown as dashed black lines. For the centre panel, we also report the posterior assuming the feedback strength is known (black contours). For the lightest axion ($m_\mathrm{ax} = 10^{-27}\,\mathrm{eV}$), we recover the truth and a non-zero preference for axions. The heaviest axion ($m_\mathrm{ax} = 10^{-21}\,\mathrm{eV}$) has a weaker effect on the matter power spectrum for $k<10\,h\,\rm{Mpc}^{-1}$ and so we place only an upper limit $f_{\mathrm{ax}} \lesssim 0.4$, with no preference for axions; feedback strength is recovered. For the intermediate-mass axion ($m_\mathrm{ax} = 10^{-25}\,\mathrm{eV}$), the scales at which baryonic feedback and axions suppress the matter power spectrum are the same. It is therefore hard to disentangle the two effects without external information on feedback, although a competitive limit on the axion fraction is still feasible.