\section{The prediction: a visibility measurement gives \texorpdfstring{$C$}{C}} \label{sec:prediction} To leading order every non-interferometric bound of Table~\ref{tab:bounds} measures $C/R_0^3$ and cannot separate $C$ from the free $R_0$. An interferometer can, because $\EG$ is dominated by wavelengths of the size of the body, which a smearing $R_0\ll a$ does not affect. The proposition makes this quantitative for a sphere. \begin{proposition}[Rate at all separations] \label{prop:saturated} For a uniform sphere of mass $M$, radius $a$ and density $\rho$, with interatomic spacing $\lesssim R_0\ll a$, in a superposition with separation $d$, \begin{equation} \Gamma=\frac{C}{\hbar}\,\frac{GM^2}{a}\,f\!\Bigl(\frac da\Bigr)\bigl[1+\varepsilon\bigr], \label{eq:saturated} \end{equation} with $f$ of Eq.~\eqref{eq:EG-sphere} and $|\varepsilon|\leq(3/\sqrt\pi)R_0/a$ at every $d$. To leading order, $\varepsilon=-\frac94\frac{R_0^2}{a^2}\frac{s(1-s^2/12)}{f(s)}$ for $R_0\ll d<2a$ (hence $-\frac92R_0^2/(ad)$ for $R_0\ll d\ll a$), $\varepsilon\to-(3/\sqrt\pi)R_0/a$ for $d\ll R_0$, and $\varepsilon=-3R_0^2/(a^2f)$ for $d\geq2a$. For $d\geq2a$ the rate saturates, \begin{equation} \Gamma_{\rm sat}\equiv\lim_{d\to\infty}\Gamma =\frac{6C}{5\hbar}\,G\Bigl(\frac{4\pi\rho}{3}\Bigr)^{1/3}M^{5/3}. \label{eq:gamma-sat} \end{equation} To this accuracy the rate does not depend on $R_0$. \end{proposition} \begin{proof} Equation~\eqref{eq:EG-sphere} and Appendix~\ref{app:energy} give $f$. Smearing multiplies the kernel by $e^{-k^2R_0^2}=1-k^2R_0^2+\dots$. The $k^2R_0^2$ term is the contact term $-4\pi R_0^2\,\delta^3(\mathbf{x}-\mathbf{y})$, so $\delta\EG=-2\pi GR_0^2\int(\Delta\rho)^2d^3x=-2\pi GR_0^2\rho^2V_\triangle(d)$, with $V_\triangle$ the volume of the symmetric difference of the two balls, $2\pi a^2d(1-d^2/12a^2)$ for $d<2a$ and $8\pi a^3/3$ beyond. Dividing by $\EG=GM^2f/a$ gives the three leading forms. When $d\lesssim R_0$ the density difference is a dipole layer thinner than $R_0$, the expansion in $R_0^2$ fails, and the exact correction saturates at $-(3/\sqrt\pi)R_0/a$; $|\varepsilon|$ grows monotonically as $d$ decreases, which gives the bound at every $d$. The expansion is not analytic in $R_0$: the step edges of the balls add terms of order $(R_0/a)^3$, $+(6/\sqrt\pi)(R_0/a)^3\,GM^2/a$ for disjoint balls. Between disjoint balls the cross term changes only by terms exponentially small in $(d-2a)/R_0$; at contact, $d=2a$, the smeared tails overlap only within a region of volume of order $aR_0^2$ around the contact point, so the change is of higher order in $R_0/a$ than the self-term correction. With $a=(3M/4\pi\rho)^{1/3}$ the limit $d\to\infty$ gives $\Gamma_{\rm sat}$. \end{proof} \noindent The proposition means that an interferometer does not see $R_0$ as long as $R_0\ll a$: it measures $C$. Once the spheres no longer overlap the rate is set by the self-energy $\frac65GM^2/a$ minus the cross term $GM^2/d$: $\EG(d)/\EG(\infty)=1-5a/6d$ is $0.583$ at $d=2a$ and within $10\%$ of saturation for $d\geq8.33\,a$. The degeneracy between $C$ and $R_0$ reopens only for $R_0\sim a$. In the allowed region~\eqref{eq:region} $R_0$ exceeds the interatomic spacing of about $2.5$~\AA{} for $C>0.17$, so the continuum density applies. Point nuclei add $G\sum_im_i^2[1/(\sqrt\pi R_0)-\operatorname{erf}(d/2R_0)/d]$ to $\EG$; for silica ($\sum m_i^2/\sum m_i=21.6$~u) this is at most $4\times10^{-3}$ of $\EG$ at $R_0=4.5$~\AA{} and below $10^{-4}$ for $d\geq10$~nm. \begin{figure}[tb] \centering \includegraphics{fig-rate.pdf} \caption{Decoherence time $\tau=\hbar/(C\EG)$ at $C=1$ against sphere radius, for silica and for $\rho=5000$~kg\,m$^{-3}$, at $d\gg a$ (solid), $d=2a$ (dashed) and $d=0.01\,a$ (dotted), for $R_0\ll a$, where it depends on $R_0$ by at most $1.7\,R_0/a$. The squares mark $1~\mu$g of silica at $d=1$~mm and at $d=1~\mu$m, the circle the MAQRO-class sphere at $d\gg a$. The horizontal line is a free-evolution time of $100$~s. The time scales as $1/C$.} \label{fig:rate} \end{figure} \paragraph{Benchmarks.} For $1~\mu$g of silica ($\rho=2200$~kg\,m$^{-3}$, $a=47.7~\mu$m) at $d=1$~mm, $\EG=24.2\,GM^2/d$ and the decoherence time is \begin{equation} \tau=\frac{\hbar}{C\EG}=\frac{65.4~\mathrm{ps}}{C}. \label{eq:tau-silica} \end{equation} In the overlapping regime the time grows as $1/d^2$, $C\tau=2a^3\hbar/(GM^2d^2)$ for $d\ll a$, which gives $\tau=0.346~\mu\mathrm{s}/C$ at $d=1~\mu$m. Appendix~\ref{app:energy} lists the values at other separations and the size of the smearing corrections. Proposed nanosphere superpositions have $d\lesssim a$~\cite{RomeroIsart2011}, so the overlapping branch of $f$ is the one they test. A space platform of the MAQRO class~\cite{Kaltenbaek2016,Kaltenbaek2022}, with free evolution of order $100$~s, reaches $d\sim a$: a sphere of $a=150$~nm and $\rho=5000$~kg\,m$^{-3}$ has $\tau=39.5~\mathrm{s}/C$ for $d\gg a$. A silica sphere needs $a\geq173~\mathrm{nm}\;C^{-1/5}$ for $\Gamma\times100~\mathrm{s}\geq1$ at $d\gg a$. Figure~\ref{fig:rate} shows the decoherence time against radius. \paragraph{What a measurement determines.} At any separation $C=\hbar\Gamma_{\rm grav}/\EG$, with $\EG$ known from $M$, $a$ and $d$, where $\Gamma_{\rm grav}$ is the measured visibility decay rate after the known decoherence channels (residual-gas collisions, black-body emission, scattering and absorption~\cite{RomeroIsart2011,GallisFleming1990}, trap and recoil noise) have been subtracted. If the residual is the survivor's and $R_0\ll a$, one measurement determines $C$. The saturated regime $d\geq2a$ needs no knowledge of $d$; the overlapping regime needs $d$ calibrated, since $\tau\propto d^{-2}$, and in exchange gives the steepest signal. Scanning $d$ across $d\sim a$ over-determines $C$, since $f$ has no free parameter. A rate below the prediction means either a smaller $C$ or $R_0\gtrsim a$; the two are separated by the radius dependence, $\Gamma\propto a^5\rho^2$ at fixed $C$ in the saturated regime against a suppression once $R_0$ approaches $a$. A null residual excludes the survivor at the tested $(C,R_0)$.