Birth-environment archaeology (B5)

The environment-dependent IMF Marks et al., 2012Jeřábková et al., 2018 predicts that metal-poor, massive clusters form top-heavy — a shallower high-mass slope $\alpha_3$. This demo asks the inverse question honestly: given a present-day stellar mass spectrum, what about the cluster’s birth conditions can you actually recover? The answer is a sharp two-part contrast.

The forward model¶

BirthEnvironment carries three birth parameters — metallicity $[\mathrm{Fe/H}]$, embedded-cluster mass $\log M_{\rm ecl}$, and star-formation efficiency $\mathrm{sfe}$ — and env_to_imf_params (the Jerabkova generalized relation) maps them to the high-mass IMF slope $\alpha_3$, holding the low-mass slopes at their canonical Kroupa Kroupa, 2001 values. The truth here is a metal-poor, massive cluster:

— strongly top-heavy versus the canonical $\alpha_3=2.3$. We sample $N=10^5$ stars from this IMF and ask what they reveal.

Inputs and assumptions¶

The fit recovers one parameter, the high-mass IMF slope $\alpha_3$. The three birth-environment axes are the scientific quantities of interest, but — as the next sections show — they are not individually recoverable; they are held at truth and enter only through $\alpha_3$.

What you CAN read off the masses: $\alpha_3$¶

A direct maximum-likelihood fit of the high-mass slope (the per-star IMF log-likelihood $\sum_i \log p(m_i\mid\alpha_3)$, differentiable in $\alpha_3$) recovers it cleanly:

The uncertainty scales as the textbook $\sigma(\alpha_3)\propto N^{-1/2}$, and we validate the Cramér–Rao bound empirically: refitting $\alpha_3$ on independent $N=10^4$-star draws gives a measured scatter 0.0144 against the analytic CRLB 0.0170 (ratio 0.85, within the 12-sample noise). At a realistic $N=10^4$ complete sample the top-heavy slope is a $\sim 40\sigma$ detection — $\alpha_3$ is easy.

What you CANNOT: the birth environment¶

The map $(\,[\mathrm{Fe/H}],\,\log M_{\rm ecl},\,\mathrm{sfe})\to\alpha_3$ is three-to-one. The masses constrain exactly one combination — the gradient direction

so the environment-space Fisher information $\,\mathcal F_{\rm env} = (\nabla_{\rm env}\alpha_3)(\nabla_{\rm env}\alpha_3)^\top/\sigma_{\alpha_3}^2\,$ is rank 1. Its eigenvalues come out

— two of them machine-precision zero (condition number $\sim 10^{304}$). There are two flat directions: a continuum of metallicities, cluster masses, and efficiencies all produce the same $\alpha_3$ and hence the same mass spectrum. Recovering any single birth parameter requires an external constraint on the other two (e.g. an independent $\log M_{\rm ecl}$ from the cluster luminosity, or a spectroscopic $[\mathrm{Fe/H}]$).

This is the honest scientific message: the IMF slope is an observable; the birth environment is an inference with two irreducible degeneracies.

Interpretation — why the degeneracy is fundamental¶

It is the structure of the map, not measurement noise. The two zero eigenvalues in (3) come from the three-to-one map itself, so they do not shrink with better data. More stars do tighten $\sigma_{\alpha_3}$ — the one well-measured eigenvalue (the $1.4\times10^{4}$) grows $\propto N$ — but the two machine-precision zeros stay zero at any sample size. This is qualitatively different from a “we need a bigger survey” degeneracy: no amount of data recovers the birth environment from the masses alone. That is what makes it a floor rather than a forecast.

Reading the one direction you do constrain. The gradient (2) is the single combination of birth parameters that the mass spectrum measures, and its structure is physical. The signs match the environment-dependent IMF: $\partial\alpha_3/\partial[\mathrm{Fe/H}] > 0$ (metal-poor $\Rightarrow$ smaller $\alpha_3$ $\Rightarrow$ top-heavier) and $\partial\alpha_3/\partial\log M_{\rm ecl} < 0$ (more massive $\Rightarrow$ top-heavier). And per a natural variation of each parameter ($\Delta[\mathrm{Fe/H}]\sim1$ dex, $\Delta\log M_{\rm ecl}\sim1$ dex, $\Delta\mathrm{sfe}\sim0.3$), the embedded-cluster mass and the star-formation efficiency are the strong levers and the metallicity the weak one — so in this regime the high-mass IMF slope encodes cluster mass and efficiency far more than metallicity.

The implication for IMF archaeology. This is a caution for the IGIMF / top-heavy-IMF program. Inferring a cluster’s birth conditions from its present-day mass function reads a one-dimensional quantity ($\alpha_3$) and projects it back onto a three-dimensional birth-parameter space — a non-invertible projection. The honest statement is “the IMF is top-heavy, consistent with $\alpha_3 = X$”, not “therefore it formed at metallicity $Y$, embedded-cluster mass $Z$, and efficiency $W$.” Breaking the degeneracy requires supplying two of the three independently — e.g. a spectroscopic $[\mathrm{Fe/H}]$ and a dynamical $M_{\rm ecl}$ — at which point the third follows from $\alpha_3$.

Figure¶

Caveats¶

How to run¶

env -u VIRTUAL_ENV uv run --no-sync python scripts/demo_birth_environment.py

References¶

The environment-dependent IMF relations are Marks et al. (2012) and Jeřábková et al. (2018); the canonical low-mass slopes are Kroupa (2001). The BirthEnvironment + env_to_imf_params API is documented on the environment-dependent IMF theory page.