deletions | additions       diff --git a/quenching_time.tex b/quenching_time.tex  index cd56871..94550e5 100644  --- a/quenching_time.tex  +++ b/quenching_time.tex 
...
First, motivated by the dearth of \emph{isolated} galaxies with $\mstar<10^9\msun$ that are quiescent at $z\approx0$ (see Introduction), our model assumes that all satellites with $\mstar(z=0)<10^9\msun$ were actively star-forming prior to first infall.  However, because most galaxies with $\mstar(z=0)<10^4\msun$ may have been quenched at high redshift by cosmic reionization \citep[e.g.,][]{Weisz2014a, Brown2014}, we do not model those masses.  At $\mstar(z=0)=10^{4-5}\msun$, satellites' star-formation histories show a mix of complete quenching by $z\gtrsim3$ (e.g., Bootes I, Leo IV) and signs of star formation at $z\lesssim1$ (e.g., And XI, And XII, And XVI) \citep{Weisz2014a, Weisz2014c, Brown2014}, so quenching at these masses may come from a mix of reionization and the host-halo environment.  %Leo T had recent star formation, suggesting that galaxies at least down to logM_star ~ 5 can form stars today if not for environment.  That said, the 100\% quiescent fraction for satellites at this $\mstar$ means that if both processes are responsible, both are highly efficient.  Furthermore, if the satellites that were quenched by reionization have a similar infall-time distribution to those that were quenched by the host-halo environment, our modeling approach remains valid.  Thus, we include this $\mstar$ in our results, but we label it distinctly to emphasize caution in interpretation. 
...
Several works have shown that this model successfully describes the dependence of satellite quiescent fractions on host-centric distance \citep[e.g.,][]{Wetzel2013, Wetzel2014, Wheeler2014} because infall time correlates with host-centric distance \citep[e.g.,][]{Wetzel2015}.  However, this correlation means that we must account for observed satellites' distances in computing their infall times.  %including incompleteness at large distances for fainter satellites, in computing their infall times.  Thus, in ELVIS we only select satellites out to the maximum host-centric distance that they are observed in each $\mstar$ bin.  In fact, this matters most at the highest $\mstar$ bin, where all observed satellites (M32, NGC 205, LMC/SMC) reside $<61\kpc$ from the MW or M31.  Figure~\ref{fig:quench_times} shows the inferred environmental quenching timescales (the time duration from first infall to being fully quiescent/gas-poor) versus $\mstar$ (top axis shows corresponding subhalo $\mpeak$).  Blue circles show the satellites in the MW and M31, and we shade the lowest $\mstar$ bin to highlight caution in interpretation because of reionization.  We derive error bars from the 68\% uncertainty in the observed quiescent fractions in Figure~\ref{fig:quiescent_fraction}.  %these uncertainties are typically larger than the host-to-host scatter in satellites' infall times in ELVIS.  As explored in \citet{Wetzel2015}, many satellites first fell into a another host halo (group), typically of $\mvir=10^{10-12}\msun$, before falling into the MW/M31 halos.  Because the importance of this environmental preprocessing in lower-mass groups remains unclear, we present quenching timescales both neglecting (left panel) and including (right panel) such group preprocessing.  %Thus, the left panel of Figure~\ref{fig:quench_times} uses time since infall into the MW/M31 halos, ignoring group preprocessing, while the right panel uses time since infall into \emph{any} host halo, including group preprocessing.  The latter results in longer quenching timescales, though it primarily shifts the upper 16\% of the distribution.  Both panels show shorter median quenching timescales for less massive satellites: $\sim5\gyr$ at $\mstar=10^{8-9}\msun$, $2-3\gyr$ at $\mstar=10^{7-8}\msun$, and $ < 1.5\gyr$ at $\mstar<10^7\msun$, depending on the inclusion of group preprocessing. 
...
We also compare these timescales with previous studies of more massive satellites of other hosts.  The red squares in Figure~\ref{fig:quench_times} show the timescales from \citet{Wheeler2014}, who used nearly identical methodology, combining the the galaxy catalog from \citet{Geha2012} with satellite infall times (including group preprocessing) from the Millennium II simulation \citep{BoylanKolchin2009}.  They examined satellites with $\mstar\approx10^{8.5}$ and $10^{9.5}\msun$ around hosts with $\mstar>2.5\times10^{10}\msun$, which they found likely spans $\mvir\approx10^{12.5-14}\msun$, much higher than the MW/M31.  %which could mean that the quenching timescales in \citet{Wheeler2014} are \emph{shorter} than for similar mass satellites of MW/M31-like hosts.  Similarly, the green curves in Figure~\ref{fig:quench_times} show the quenching timescales for more massive satellites in groups with $\mvir=10^{12-13}\msun$ from \citet{Wetzel2013}, who also used identical methodology, combining a galaxy group catalog from SDSS \citep{Tinker2011, Wetzel2012} with satellite infall times (including group preprocessing) measured in their cosmological simulation.  Altogether, Figure~\ref{fig:quench_times} indicates a complex dependence of the environmental quenching timescale on satellite $\mstar$.