\subsection{Variation of the remaining core passage} \label{sec:partial-core-variation} The initial state now moves in both radius and angle. To follow its phase to the outgoing section, we keep these two variations separate until the last crossing. The weighted variational equation from Section~\ref{sec:core-variation} still has uniformly bounded growth. Its initial condition acquires an angular correction, and the integral of its strong component determines the resulting change in the radial derivative. For a projective state $y=(M,U,V,W)$ with $V>0$, write \[ \mathcal C(y)=(r(y),U,VW),\qquad r(y)=\sqrt{(M-1)^2+V^2},\qquad z(y)=\operatorname{arsinh}\frac{1-M}{V}. \] Let $\mathcal R_\downarrow$ be the actual downward return defined in \eqref{eq:partial-core-oriented-return}. Denote its radius component by $\mathcal R_{\downarrow,R}$, and put \[ K_v=\exp(2\pi+48\varepsilon_0),\qquad C_p=4096K_v. \] Thus $K_v$ is the growth constant in \eqref{eq:core-variation-growth}. \begin{theorem} \label{thm:partial-core-radius-response} Let $y_0$ satisfy \eqref{eq:partial-core-initial-state}, and set $k=r(y_0)$, $u_0=U_0$, $P_0=V_0W_0$. For any ordinary initial direction $w\in\R^4$, define \[ v=(v_R,v_U,v_P)=D\mathcal C(y_0)w,\qquad \delta z=Dz(y_0)w. \] Then \begin{equation} \begin{split} |D\mathcal R_{\downarrow,R}(y_0)w-v_R| \leq C_p\biggl\{ |v_U|+\frac{u_0+P_0}{k}(|v_R|+|v_P|) +\left(ku_0+\frac{P_0^2}{k}\right)|\delta z| \biggr\}. \end{split} \label{eq:partial-core-radius-response-general} \end{equation} The estimate holds at interior entrances and at the incoming boundary $V_0=1$, including states with a zero transverse wall. \end{theorem} The first two terms measure changes in the initial radius and wall amplitudes. The last term accounts for moving the starting point along the core. Its strong-wall coefficient is $ku_0$, whereas its weak-wall coefficient is quadratic in $P_0$. We retain this distinction when passing to the reciprocal radius. \subsubsection{The moving initial angle} Let $\mathcal Y=(R,U,P)$ along the reference passage, and let $F(z,\mathcal Y)$ be the angular vector field in \eqref{eq:core-angle-system}. Write $\alpha=z(y_0)$ and $\beta=z(t_*)$ for its initial and final angles. For the initial direction $w$, the variation at fixed angle is $\xi=\delta\mathcal Y-\delta z\,F$, as in \eqref{eq:fixed-angle-variation}. At the initial point, \begin{equation} \xi(\alpha)=v-\delta z\,F(\alpha,k,u_0,P_0). \label{eq:partial-core-moving-initial} \end{equation} Here $v$ and $\delta z$ are derivatives of the actual initial coordinates. The identity follows by differentiating the smooth coordinate changes and the actual flow. In particular the variational equation \eqref{eq:core-variational-equation} applies to this $\xi$. In the original time it reads $\dd\xi/\dd t=ARd\,J\xi$; the angular integrals below use the positive weight $\dd z=ARd\,\dd t$ along the actual orbit. At a core point with $R=k$, the field satisfies \begin{equation} |F_R|\leq2\left(ku_0+\frac{P_0^2}{k}\right),\qquad |F_U|\leq2ku_0,\qquad |F_P|\leq\frac{4P_0}{k}. \label{eq:partial-core-initial-field} \end{equation} For the first bound, use the radial integrand estimate \eqref{eq:core-radius-integrand} at this point. Since $1\leq V\leq k$ and $W=P_0/V\leq P_0$, \[ UV\leq ku_0,\quad UW\leq u_0/60000,\quad U^2/k\leq u_0/60000,\quad W^2/k\leq P_0^2/k. \] Division by $d\geq999/1000$ gives the stated bound. The other two follow from the wall factors $F_U=UL_U$, $F_P=PL_P$ and \eqref{eq:core-log-rate-bounds}. Recall $S_k\xi=(\xi_R,k\xi_U,\xi_P)$. As $P_0\leq1$ and $k\geq6$, \eqref{eq:partial-core-initial-field} implies \[ \|S_kF(\alpha,k,u_0,P_0)\|_\infty \leq4(k^2u_0+P_0/k). \] Introduce the quantity \begin{equation} Q_0=|v_R|+k|v_U|+|v_P| +(k^2u_0+P_0/k)|\delta z|. \label{eq:partial-core-variation-size} \end{equation} The triangle inequality in \eqref{eq:partial-core-moving-initial} gives $\|S_k\xi(\alpha)\|_\infty\leq4Q_0$. Lemma~\ref{lem:core-variation-growth}, applied on every initial part of the passage, therefore gives \begin{equation} \|S_k\xi(z)\|_\infty\leq Q,\qquad Q=4K_vQ_0,\qquad \alpha\leq z\leq\beta. \label{eq:partial-core-variation-growth} \end{equation} The radius $k$ in this weighting is fixed at the reference initial state; it is not differentiated inside the linear equation. We first work with any bound $\|S_k\xi\|_\infty\leq Q$ on a part of the core. Set \[ x=k\xi_U,\qquad x_0=x(\alpha). \] The strong row of \eqref{eq:core-transverse-rows} and the decay of $U$ give \begin{equation} |x(z)|\leq e^{-k(z-\alpha)/4} \bigl(|x_0|+3ku_0Q(z-\alpha)\bigr). \label{eq:partial-core-strong-envelope} \end{equation} Indeed $x'=B_{UU}x+h$, with $B_{UU}\leq-k/4$ and $|h|\leq3kUQ$. Multiplication by $e^{k(z-\alpha)/4}$ leaves a nonpositive damping coefficient and forcing at most $3ku_0Q$. This is the scalar comparison used in \eqref{eq:core-strong-variation}; it uses only the angle difference and applies to the present initial point. Its integral is \begin{equation} \int_\alpha^\beta|x(z)|\dd z \leq\frac{4|x_0|+48u_0Q}{k}. \label{eq:partial-core-strong-integral} \end{equation} The constants follow by integrating $e^{-ks/4}$ and $s e^{-ks/4}$ over $s\geq0$. Put $\gamma=(101/99)^2$. The amplitude integrals needed for the radius row are \begin{equation} \int_\alpha^\beta U\dd z\leq\frac{4u_0}{k},\qquad \int_\alpha^\beta W^2\dd z\leq2\gamma P_0^2,\qquad \int_\alpha^\beta\frac{W}{kV}\dd z \leq\frac{2\gamma P_0}{k}. \label{eq:partial-core-variation-integrals} \end{equation} These are the estimates in \eqref{eq:core-amplitude-integrals}, with the initial product $P_0$ in place of the entrance weak wall. For clarity, $P\leq P_0$ and $R\geq99k/100$, while $V\geq1$ gives $\cosh\alpha,\cosh\beta\leq101k/100$. Thus the same endpoint estimate for $\int\cosh^2z\dd z$ yields $\int V^{-2}\dd z\leq2\gamma$. The last two integrals follow from $W=P/V$. \begin{lemma} \label{lem:partial-core-radial-variation} On every part of the passage on which $\|S_k\xi\|_\infty\leq Q$, \begin{equation} |\xi_R(\beta)-\xi_R(\alpha)| \leq256\left\{\frac{|x_0|}{k} +\frac{(u_0+P_0)Q}{k}\right\}. \label{eq:partial-core-radial-variation} \end{equation} \end{lemma} \begin{proof} The radius row estimate \eqref{eq:core-radial-variation-integrand} gives \[ \left|\frac{\dd\xi_R}{\dd z}\right| \leq \left(16U+\frac{8W^2}{k^2}+\frac{8W}{kV}\right)Q +3|x|. \] Here $2/\cosh z+1/k\leq3$. Integrating with \eqref{eq:partial-core-strong-integral} and \eqref{eq:partial-core-variation-integrals} gives \[ \frac{12|x_0|}{k} +\frac{208u_0Q}{k} +\frac{16\gamma P_0^2Q}{k^2} +\frac{16\gamma P_0Q}{k}. \] Since $P_0\leq k$ and $\gamma<2$, this is bounded by the right-hand side of \eqref{eq:partial-core-radial-variation}. \end{proof} The strong contribution here is $|x_0|/k$. For a complete core passage the incoming condition $\cosh\alpha=k$ supplies the further factor $1/k$ in \eqref{eq:core-radial-variation}. Keeping the present estimate allows the starting angle to vary through the core. \subsubsection{Projection onto the moving exit} At the exit set \[ r_*=R(\beta),\qquad c_*=-\tanh\beta,\qquad f_*=F_R(\beta,\mathcal Y(\beta)). \] The endpoint variation is tangent to $V=1$. The same elimination as in \eqref{eq:core-exit-projection} gives \begin{equation} D\mathcal R_{\downarrow,R}(y_0)w =s_*\xi_R(\beta),\qquad s_*=\frac{c_*}{c_*+f_*/r_*}. \label{eq:partial-core-exit-projection} \end{equation} This identity includes the derivative of the true exit time. Its contribution is a multiple of the flow vector, which is removed by the fixed-angle projection. At the downward exit, $c_*\leq-1/2$ and $f_*\leq0$. The exit factor bounds proved in Section~\ref{sec:core-variation}, together with $U(\beta)\leq u_0$ and $W(\beta)=P(\beta)\leq P_0$, give \begin{equation} 00$ and $K\geq1$ such that for every $03$ and $\ell>4$, all sufficiently small scales $a>0$ have the following property. Consider the positive-shear family \eqref{eq:radial-shear-family} with $n_3\leq a^\ell$, at the fixed time $T=(q/2)\log(1/a)$. At every $\rho\in[a,2a]$ for which its actual core state $Y(\rho)$ has $V\geq1$, the outgoing radius $f=\mathcal R_{\downarrow,R}\circ Y$ satisfies \eqref{eq:partial-core-family-radius-error} and \eqref{eq:partial-core-family-phase-error}. These are ordinary derivatives along the original shear curve, with $h,a,q,n_2,n_3$ fixed. \end{corollary} \begin{proof} Take the common small-$h$ threshold of Propositions~\ref{prop:small-section-coordinate-response} and \ref{prop:small-section-core-data}. Choose \[ K=2000+K_0+K_1 \] before $h,q,\ell$, and take the small-$a$ thresholds from those propositions, also requiring $a\leq1$. The radius bounds give \[ \frac2{Ka}\leq\frac1{16a}\leq k \leq\frac{32}{a}\leq\frac Ka,\qquad |k'|\leq\frac{1536}{a^2}\leq\frac K{a^2}. \] The wall and angle bounds give the other three inequalities in \eqref{eq:partial-core-radial-family-input}. Choose the entry bound in \eqref{eq:partial-core-entry-smallness} with $\eta=\varepsilon_0$. When $V\geq1$, the state satisfies \eqref{eq:partial-core-initial-state}. Proposition~\ref{prop:partial-core-radial-family} now applies, including at $V=1$. The flow identity at the end of Section~\ref{sec:small-section-partial-core} identifies this endpoint with the outgoing event on the original orbit, at time $T+\tau_\downarrow(Y(\rho))$. \end{proof} Thus the radial and angular derivatives supplied by the initial decay control the phase at the actual outgoing section. The remaining derivative estimates concern the two wall amplitudes. Their multiplier form will retain the small factors needed to pass from this exit to the next incoming section.