\subsection{What purity gives in a differential module}\label{sec:seedbridge} The following lemma is the precise passage from the formal Gaussian functional to an analytic period. Its hypotheses must be verified after each change of variables or localization. \begin{lemma}[The seed functional]\label{lem:seedfunctional} Let $M$ be a finite-dimensional differential module over $K=\C(t)$, let $s\in M$, and suppose a $K$-linear horizontal map $J:M\longrightarrow\C((t))$ satisfies $J(s)=1$. Then the differential submodule $U=K\langle\partial_t\rangle s$ has a horizontal quotient $L:U\longrightarrow K$ with $L(s)=1$. At an ordinary base point, $L$ extends locally to an analytic horizontal functional on $M$. For the period conclusion, assume explicitly that $M$ is the generic exponential de Rham system of regular functions $f,g$ on a smooth affine complex variety, where $f$ is the base projection and $f,g$ are algebraically independent. Remove the finite exceptional base set in the Hien--Roucairol comparison theorem. Then the extension is represented by a flat rapid-decay cycle, and its period evaluated on $s$ is identically one. \end{lemma} \begin{proof} On the generators $s,\nabla s,\nabla^2s,\ldots$, the values of $J$ are $1,0,0,\ldots$. Therefore its restriction takes values in $K$. Every rational relation between these generators is respected because $J$ was defined on $M$, so this restriction is well defined and horizontal. Its value at $s$ makes it surjective. Choose a point where $M$, $U$ and $L$ are ordinary, extend the fiber functional from $U$ to $M$, and solve the dual analytic ODE. Uniqueness preserves its restriction to $U$. Perfectness of the period pairing is Hien--Roucairol~\cite[Theorem~2.7]{HienRoucairol}; their Theorem~3.5 represents holomorphic solutions by rapid-decay periods. Together these statements give the asserted cycle. A scalar rational normalization is an invertible gauge away from its zeros and poles and must be retained in the period. \end{proof} There are three distinct verifications here. First, integration by parts must make $J$ descend to the exact displayed de Rham quotient. Second, every inverted spatial coordinate must be a valid formal Laurent unit on the selected branch. Third, the comparison theorem requires a smooth affine total space and regular functions after removing the finite set of poles in the parameter. The theorem gives a representation, not a bound at infinity. In particular, it gives no shrinking estimate on its own. There is also a useful purely formal way to obtain a contradiction. \begin{lemma}[A full formal frame and a positive seed]\label{lem:positiveformal} Assume the hypotheses of the first paragraph of Lemma~\ref{lem:seedfunctional}. Suppose that, after a finite ramification $r=t^{-1/N}$ at infinity, $M$ has a full formal frame of rank-one exponential factors. For every factor with zero exponential part, assume its horizontal functional takes $s$ into $r\C[[r]]$. Then $J(s)=1$ is impossible. \end{lemma} \begin{proof} Enlarge the ramification to include the rational powers in the frame. The module over $\C((r))$ is then a direct sum of rank-one modules, so the quotient on $U$ extends to a horizontal map from the whole module to $\C((r))$. A nonzero exponential factor has no such map: the logarithmic derivative of a Laurent series has at most a simple logarithmic pole and cannot equal the derivative of a nonzero negative-power exponential part. On the zero part, all horizontal functionals are constant linear combinations of the supplied frame. Their values on $s$ have strictly positive $r$-order and cannot sum to one. \end{proof} The word \emph{full} is essential. Counting some critical points or writing a few scalar asymptotic expansions does not supply this lemma's hypothesis. The finite and escaping pieces must together give an invertible period matrix of the full de Rham rank. Also, a fractional power becomes an integer after ramification; the proof uses its strictly positive order, not an unsupported monodromy inference.