= Solution
\b[First velocity moment.] Assume a stationary, spherical, nonstreaming <collisionless stellar system> with constant <mass-to-light ratio> $\Upsilon$. Use a luminosity-weighted <galactic distribution function>, with <luminosity density> $\lambda$. Multiplying the <Collisionless Boltzmann equation> by $v_i$ and integrating over velocities, with its velocity-space boundary terms vanishing, gives
$$
\partial_j\big(\lambda\langle v_i v_j\rangle\big)=-\lambda\partial_i\Phi.
$$
Spherical symmetry makes the stress tensor
$$
T_{ij}=q_t\delta_{ij}+(q-q_t)\frac{x_ix_j}{r^2},\qquad q=\lambda\sigma_r^2,\qquad q_t=\lambda\sigma_t^2.
$$
Here $\sigma_t^2$ is one tangential component, as in the paper; the total tangential <variance> is $2\sigma_t^2$. Its radial divergence is $q'+2(q-q_t)/r$, since the divergence of the radial unit vector is $2/r$. With the <velocity-anisotropy parameter> $\beta=1-\sigma_t^2/\sigma_r^2$, and enclosed <mass> $M(r)=\Upsilon L(r)$, this gives the <Spherical Jeans equation>
$$
\boxed{q'+\frac{2\beta q}{r}=-\lambda\Phi'(r)=-\gamma\frac{\lambda L(r)}{r^2},\qquad\gamma=G\Upsilon.}
$$
This uses a constant ratio of total gravitating mass to luminosity, rather than an arbitrary independent dark halo.
\b[Projection and elimination of anisotropy.] Write $R$ for projected radius and $r$ for spherical radius. The <Abel transform> of the <luminosity density> gives
$$
\mu(R)=2\int_R^\infty\frac{\lambda(r)r\,dr}{\sqrt{r^2-R^2}},
$$
whose inverse is the displayed density inversion in the paper. At a point on this <line of sight>, $\sin^2\theta=R^2/r^2$. The line-of-sight second <stellar velocity moment> is $\sigma_r^2\cos^2\theta+\sigma_t^2\sin^2\theta=\sigma_r^2(1-\beta R^2/r^2)$. Therefore the measured quantity $J(R)=\mu(R)\sigma_p^2(R)$ satisfies
$$
J(R)=2\int_R^\infty\left(1-\beta(r)\frac{R^2}{r^2}\right)\frac{q(r)r\,dr}{\sqrt{r^2-R^2}}.
$$
The <Spherical Jeans equation> gives $\beta q=-rq'/2-\gamma\lambda L/(2r)$. Substitution yields
$$
J(R)=2\int_R^\infty\frac{q(r)r\,dr}{\sqrt{r^2-R^2}}+R^2\int_R^\infty\frac{q'(r)\,dr}{\sqrt{r^2-R^2}}+\gamma R^2\int_R^\infty\frac{\lambda(r)L(r)\,dr}{r^2\sqrt{r^2-R^2}}.
$$
The inner $\gamma L$ printed in the PDF's definition of $p$ must therefore be $\lambda L$: the literal expression has an extra gravitational factor and lacks the density. Both this derivation and dimensional consistency require the correction. Define the <projected spherical Jeans residual> by
$$
p(R)=J(R)-\gamma R^2\int_R^\infty\frac{\lambda(r)L(r)\,dr}{r^2\sqrt{r^2-R^2}}.
$$
To express it entirely in terms of the radial stress, put
$$
A(R)=\int_R^\infty\frac{q(r)r\,dr}{\sqrt{r^2-R^2}}=\int_0^\infty q\!\left(\sqrt{R^2+z^2}\right)dz.
$$
Differentiating the nonsingular $z$ representation gives $A'(R)=R\int_R^\infty q'(r)/\sqrt{r^2-R^2}\,dr$. Hence
$$
\boxed{p(R)=2A(R)+RA'(R)=\frac1R\frac{d}{dR}[R^2A(R)].}
$$
If $R^2A$ vanishes at zero and infinity, then $\int_0^\infty6\pi Rp(R)\,dR=6\pi[R^2A]_0^\infty=0$. These are the absence-of-boundary-stress conditions needed for the global result.
\b[The virial relation, term by term.] Interchange the order in the residual integral and use
$$
\int_0^r\frac{R^3\,dR}{\sqrt{r^2-R^2}}=r^3\int_0^{\pi/2}\sin^3\theta\,d\theta=\frac23r^3.
$$
It follows that
$$
\int_0^\infty6\pi Rp(R)\,dR=6\pi\int_0^\infty RJ(R)\,dR-4\pi\gamma\int_0^\infty r\lambda(r)L(r)\,dr.
$$
The first term is $2K/\Upsilon$: spherical averaging assigns one third of the integrated velocity-square to any fixed viewing direction, giving $K=3\pi\Upsilon\int RJ\,dR$. The gravitational potential energy is
$$
W=-\int_0^\infty\frac{GM(r)}r\,dM(r)=-4\pi G\Upsilon^2\int_0^\infty r\lambda(r)L(r)\,dr.
$$
Thus the integrated residual is exactly $(2K+W)/\Upsilon$, and its vanishing is the <virial theorem>. Solving for $\gamma$ gives the <global projected virial estimate of a mass-to-light ratio>,
$$
\boxed{\gamma=\frac{3\int_0^\infty R\mu(R)\sigma_p^2(R)\,dR}{2\int_0^\infty r\lambda(r)L(r)\,dr}.}
$$
It is independent of the radial anisotropy profile when the full aperture and the stated boundary conditions are used. A finite aperture retains a boundary term.
Back to article page