We fix \theta and write C_\tau=C^\bullet_{\theta,\tau} and U^k_\tau=C^k_\tau\cap\ker\iota_{X_\tau}. We suppress metric subscripts and use Chaubet–Dang’s convention v\wedge\alpha=(-1)^{k-1}\alpha\wedge v for horizontal (k-1)-forms. For a horizontal form u, the contact splitting and Lefschetz map are
C^k_\tau=U^{k-1}_\tau\wedge\alpha_\tau\oplus U^k_\tau,
\qquad \ell_\tau u=u\wedge d\alpha_\tau,
\qquad U^{-1}_\tau=U^5_\tau=0.
For k=0,1,2, f\in U^{k-1}, and g\in U^k, we define the contact chirality \Gamma, a degree-reversing involution, by
\Gamma(f\wedge\alpha+g)
=\ell^{2-k}g\wedge\alpha+\ell^{3-k}f.
In higher degrees it is the inverse. Thus
\Gamma(C^k_\tau)=C^{5-k}_\tau,\qquad\Gamma^2=\operatorname{Id}.
The powers of \ell denote the corresponding Lefschetz maps. The inverses used below are restricted to the ranges of these maps.
The generalized zero-resonance part is acyclic by [DR, Theorem 2.1]. On the sum of the nonzero generalized resonance spaces, L is invertible and
d_\theta(\iota_XL^{-1})+(\iota_XL^{-1})d_\theta=\operatorname{Id}.
Thus the whole cut C_\tau is acyclic. The spectral projector commutes with d_\theta, \iota_X, wedging by \alpha, and the Lefschetz maps. Thus \Gamma preserves the full resonance space.
Write \det E for the top exterior power of a finite-dimensional space. The determinant line of the complex is
\det C_\tau=\bigotimes_{k=0}^5(\det C^k_\tau)^{(-1)^k},
where exponent -1 denotes the dual line. Choose volume elements in degrees 0,1,2 and their \Gamma-partners in degrees 5,4,3. Their alternating tensor product, with the sign of [CD, Section 3.3], is choice-independent: paired rescalings cancel. By acyclicity, the canonical torsion isomorphism takes values in \det H^\bullet(C_\tau)=\mathbb C. It sends this tensor to the nonzero scalar \mathcal T_{\rm fin}(\tau). The differentiability of the spectral projectors established in [CD, Proposition 7.2 and Appendix B] implies that the finite-dimensional matrices depend differentiably on \tau. Here the projectors act from a stronger anisotropic space to a weaker one. We take a reference time \tau_* and write
\Pi_*=\Pi_{\theta,\tau_*},\qquad
\dot\Pi_*=(\partial_\tau\Pi_{\theta,\tau})|_{\tau=\tau_*}.
On the fixed space C=C_{\tau_*}, set
S_\tau=\Pi_{\theta,\tau}|_C,
\qquad \mathsf F_\tau=\Pi_*S_\tau:C\to C.
They are invertible near \tau_*. Since the projections commute with d_\theta, the differential in the fixed coordinates is unchanged S_\tau^{-1}d_\theta S_\tau=d_\theta|_C. We define the chirality in these coordinates by
\widetilde\Gamma_\tau
=S_\tau^{-1}\Gamma_\tau S_\tau
=\mathsf F_\tau^{-1}\Pi_*\Gamma_\tau S_\tau.
At \tau_*, differentiation of \Pi_{\theta,\tau}^2=\Pi_{\theta,\tau} gives
\Pi_*\dot\Pi_*\Pi_*=0,
\qquad \mathsf F_{\tau_*}=\operatorname{Id},\qquad\dot{\mathsf F}_{\tau_*}=0.
Using \Pi_*\Gamma_{\tau_*}=\Gamma_{\tau_*}\Pi_*, we obtain
\dot{\widetilde\Gamma}_{\tau_*}
=\Pi_*\dot\Gamma_{\tau_*}|_C
+\Pi_*\Gamma_{\tau_*}\dot\Pi_*|_C
=\Pi_*\dot\Gamma_{\tau_*}|_C.
Since the differential does not change in these coordinates, neither does the torsion isomorphism. Fix volumes \upsilon_k\in\det C^k, k=0,1,2. We differentiate the factors (\widetilde\Gamma_\tau\upsilon_k)^{(-1)^{5-k}} and use the derivative of \widetilde\Gamma^2=\operatorname{Id}. Pairing the terms in complementary degrees gives the half-supertrace identity [BK, Proposition 4.9], also used in [CD, Lemma 7.3]:
\begin{aligned}
\partial_\tau\log\mathcal T_{\rm fin}(\tau)
&=\sum_{k=0}^{2}(-1)^{k+1}
\operatorname{tr}_{C^k}(\widetilde\Gamma\dot{\widetilde\Gamma})=\tfrac12\operatorname{str}_C(\dot{\widetilde\Gamma}\widetilde\Gamma)
=\sum_{k=0}^{2}(-1)^{k+1}\operatorname{tr}_{C^k}(\Pi\Gamma\dot\Gamma).
\end{aligned}
We define a real function h on M, a horizontal two-form \delta, and maps \mathsf A_0,\mathsf A_1 on horizontal degrees zero and one by
\begin{aligned}
h&=\alpha(\dot X)=-\dot\alpha(X),
&\delta&=d\dot\alpha+(\iota_Xd\dot\alpha)\wedge\alpha,\\
\mathsf A_0u&=2(\ell^2)^{-1}(u\wedge\delta\wedge d\alpha),
&\mathsf A_1u&=\ell^{-1}(u\wedge\delta).
\end{aligned}
Here \Pi h denotes multiplication by h followed by projection to the cut. To differentiate the chirality, we use
\partial_\tau(d\alpha_\tau)^j
=j\,d\dot\alpha_\tau\wedge(d\alpha_\tau)^{j-1},
\qquad\dot\alpha(X)=-h.
In degree zero, for a fixed scalar u,
\begin{aligned}
\dot\Gamma u
&=2u\,d\dot\alpha\wedge d\alpha\wedge\alpha
+u(d\alpha)^2\wedge\dot\alpha \text{ and }
\Gamma\dot\Gamma u=\mathsf A_0u-hu.
\end{aligned}
The same differentiation in degrees one and two gives the diagonal blocks of \Gamma\dot\Gamma:
\begin{array}{c|cc}
k&U^{k-1}\wedge\alpha&U^k\\ \hline
0&0&\mathsf A_0-h\\
1&\mathsf A_0+h&\mathsf A_1-h\\
2&\mathsf A_1+h&-h
\end{array}
The off-diagonal blocks do not contribute to the trace after projection to the resonance space. Set
a_j=\operatorname{tr}_{U^j}(\Pi\mathsf A_j)\quad(j=0,1),\qquad
h_j=\operatorname{tr}_{U^j}(\Pi h)\quad(0\leq j\leq4),\qquad
\gamma_k=\operatorname{tr}_{C^k}(\Pi\Gamma\dot\Gamma).
The \mathsf A_j contributions cancel between adjacent full degrees:
\begin{aligned}
\gamma_0&=a_0-h_0,\quad
\gamma_1=a_0+h_0+a_1-h_1,\quad
\gamma_2=a_1+h_1-h_2,\\
-\gamma_0+\gamma_1-\gamma_2&=2h_0-2h_1+h_2.
\end{aligned}
For horizontal forms v and u, contraction gives
(\alpha\wedge\iota_{\dot X})(v\wedge\alpha)=h\,v\wedge\alpha,
\qquad (\alpha\wedge\iota_{\dot X})u
=\alpha\wedge\iota_{\dot X}u.
The second expression has no horizontal component. Thus the diagonal of this operator is h on U^{k-1}\wedge\alpha and zero on U^k. Hence
H=-h_0+h_1-h_2+h_3-h_4=-2h_0+2h_1-h_2.
The last equality follows from the cut Lefschetz isomorphisms and the fact that scalar multiplication commutes with them:
h_3=h_1,\qquad h_4=h_0.
Therefore
\partial_\tau\log\mathcal T_{\rm fin}(\tau)=-H.
For the cut of Section ?, the full ranks in (?) and flow dimension five give the sign exponent in [CD, (6.6)]:
n_{\rm cut}=\sum_{k=0}^2(-1)^k(3-k)\dim C^k
=3-6+6=3.
We write \mathcal T_{\rm dyn} for Chaubet–Dang’s dynamical torsion. The finite-cut formula is thus
\mathcal T_{\rm dyn}(\theta,\tau)
=(-1)^3G(0,\theta,\tau)\mathcal T_{\rm fin}(\tau).
At a crossing, the zero-complex sign exponent is m-1 and \operatorname{ord}_{s=0}Z=m-2. With \mathcal T_0 the refined torsion of the full generalized zero complex, [CD, (6.7)] gives \mathcal T_{\rm dyn}=-Z(0)\mathcal T_0 for m=2, and \mathcal T_{\rm dyn}=Z'(0)\mathcal T_0 for m=3, where Z(0)=0. Acyclicity alone does not remove this finite factor. Formula (6.8) of [CD] applies when zero is absent from the resonance spectrum. Deformation invariance [CD, Theorem 9] at the fixed acyclic character and (?) give
0=\partial_\tau\log\mathcal T_{\rm dyn}
=\partial_\tau\log G+\partial_\tau\log\mathcal T_{\rm fin}
=\partial_\tau\log G-H.
Integration proves (3.1), without choosing a logarithm.
The dimensions of the full resonance spaces remain constant at the nonacyclic trivial character. To extend H through \theta=0, set \mathcal K_\tau=\alpha_\tau\wedge\iota_{\dot X_\tau}. Using the forms f_j, functionals \varphi_i, and matrix F of (Proof), define in each full degree
K^{\rm var}_{ij}(z,\tau)=\varphi_i
(\Pi_{z,\tau}\mathcal K_\tau\Pi_{z,\tau}f_j).
The matrices F are uniformly invertible on \mathcal P. The entries of K^{\rm var} depend continuously on the parameters, as follows from this composition of operators:
\mathcal H_2\xrightarrow{\Pi}\mathcal H_1
\xrightarrow{\mathcal K_\tau}\mathcal H_1
\xrightarrow{\Pi}\mathcal H_0
\xrightarrow{\varphi_i}\mathbb C.
Writing F_k,K^{\rm var}_k for the matrices in degree k, we obtain the continuous extension
H(z,\tau)=\sum_{k=0}^5(-1)^k
\operatorname{tr}(F_k(z,\tau)^{-1}K^{\rm var}_k(z,\tau)).
Choose K_H\geq\sup_{\mathcal P}|H|. Equation (3.1) then yields
\left|\frac{G(0,\theta,\tau)}{G(0,\theta,0)}-1\right|
\leq e^{K_H\tau}-1.