Lognormal

Form

Suggested by, e.g., Larson (1973), the IMF takes the form of a lognormal distribution:

\[\xi(M) = \xi_0\frac{1}{M}\exp\left(\frac{-\left(\log M - \log M_c\right)^2}{2\sigma^2}\right).\]

Implemented in

This is implemented in the LognormalIMF class, where the user is free to pick whatever parameters they like or use the default (same as above).

Integrals

Behind the scenes we will use scipy.special.erf to evaluate these integrals. Defining

\[\mathtt{erf}(z) = \frac{2}{\sqrt{\pi}} \int^{z}_{0} e^{-t^2}\mathrm{d}t\]
and recalling \(\int^{b}_{a}f(x)\mathrm{d}x = \int^{b}_{0}f(x)\mathrm{d}x - \int^{a}_{0}f(x)\mathrm{d}x\) we can write
\[\begin{split}\begin{align*} \int^{b}_{a} e^{-t^2}\mathrm{t} &= \int^{b}_{0} e^{-t^2}\mathrm{t} - \int^{a}_{0} e^{-t^2}\mathrm{t} \\ &= \frac{\sqrt{\pi}}{2}\left(\mathtt{erf}(b) - \mathtt{erf}(a)\right) \\ \end{align*}\end{split}\]

Total number of stars

Use the substitution

\[\mu = \frac{\log M - \log M_c}{\sqrt{2}\sigma} = \frac{\ln(M/M_c)}{\sqrt{2}\sigma\ln10}\]
which has the derivative
\[\mathrm{d}\mu = \frac{1}{\sqrt{2}\sigma\ln10}\frac{\mathrm{d}M}{M}\]
to change variables
\[\xi(M)\mathrm{d}M = \xi_0\sqrt{2}\sigma\ln10\exp\left(-\mu^2\right)\mathrm{d}\mu\]

\[\begin{split}\begin{align*} \int^{M_\textrm{max}}_{M_\textrm{min}} \xi(M)\mathrm{d}M &= \int^{\mu_\textrm{max}}_{\mu_\textrm{min}} \xi(\mu)\mathrm{d}\mu \\ &= \xi_0\sqrt{2}\sigma\ln10\int^{\mu_\textrm{max}}_{\mu_\textrm{min}} \exp\left(-\mu^2\right)\mathrm{d}\mu \\ &= \xi_0\sigma\ln10\sqrt{\frac{\pi}{2}}\left(\mathtt{erf}(\mu_\textrm{max}) - \mathtt{erf}(\mu_\textrm{min})\right) \end{align*}\end{split}\]

Total mass of stars

We will have an extra factor of \(M\). Rearranging our substitution to \(M = M_c e^{\sqrt{2}\sigma\ln 10\mu}\):

\[\begin{split}\begin{align*} \int^{M_\textrm{max}}_{M_\textrm{min}} M\xi(M)\mathrm{d}M &= \int^{\mu_\textrm{max}}_{\mu_\textrm{min}} M_c e^{\sqrt{2}\sigma\ln 10\mu}\xi(\mu)\mathrm{d}\mu \\ &= \xi_0\sqrt{2}\sigma\ln10M_c\int^{\mu_\textrm{max}}_{\mu_\textrm{min}} e^{\sqrt{2}\sigma\ln 10\mu}\exp\left(-\mu^2\right)\mathrm{d}\mu \\ &= \xi_0\sqrt{2}\sigma\ln10M_c\int^{\mu_\textrm{max}}_{\mu_\textrm{min}} \exp\left(-\left[\mu^2 - \sqrt{2}\sigma\ln 10\mu\right]\right)\mathrm{d}\mu \end{align*}\end{split}\]

This can be solved by another substitution, which we use to complete the square

\[\begin{split}\begin{align*} \mu' &= \mu - \frac{\sigma\ln10}{\sqrt{2}} \\ \Rightarrow \mu'^2 -\frac{(\sigma\ln10)^2}{2} &= \mu^2 - \sqrt{2}\sigma\ln10\mu \\ \int^{M_\textrm{max}}_{M_\textrm{min}} M\xi(M)\mathrm{d}M &= \xi_0\sqrt{2}\sigma\ln10M_c\int^{\mu_\textrm{max}}_{\mu_\textrm{min}} \exp\left(-\left[\mu'^2 -\frac{(\sigma\ln10)^2}{2}\right]\right)\mathrm{d}\mu \\ &= \xi_0\sqrt{2}\sigma\ln10M_c e^\frac{(\sigma\ln10)^2}{2}\int^{\mu_\textrm{max}}_{\mu_\textrm{min}} \exp\left(-\mu'^2\right)\mathrm{d}\mu \\ &= \xi_0\sigma\ln10M_c e^\frac{(\sigma\ln10)^2}{2}\sqrt{\frac{\pi}{2}}\left(\mathtt{erf}(\mu'_\textrm{max}) - \mathtt{erf}(\mu'_\textrm{min})\right) \end{align*}\end{split}\]

See also