• Definitions / Notation used
• Semidirect product law and an explicit action on $\mathrm{Conn}(P_H)$
• Assumptions vs Consequences
  • Definitional
  • Ansatz (this instantiation’s transport principle)
  • Consequences
• Why this matters
• Key takeaway
• Technical takeaway

In the previous article, the key fact was that $\mathrm{Conn}(P_H)$ is affine, modeled on $N=\Omega^1(Y,\mathrm{ad})$, so “shifting a connection by an adjoint 1-form” is a canonical move. What we still owe is the precise group that combines that move with ordinary $H$-gauge rotations, and a concrete statement of how it acts. The result is an inhomogeneous group that looks exactly like what you’d expect from any affine geometry: a semidirect product. In GU language, this is the transport group, and it’s the smallest structure that keeps track of how displacements and rotations compose when you move around the affine space of connections.

§Definitions / Notation used

• Same ambient setup: $X=X^4$, $Y=Y^{14}$ with Spin$(7,7)$ structure; immersion $\iota: X \hookrightarrow Y;$ tangent/normal split; metrics and Hodge stars as before.
• Gauge group: $H$ acts on connections by $A\mapsto A^\epsilon=\epsilon^{-1}A\epsilon+\epsilon^{-1}d\epsilon.$
• Translation module: $N:=\Omega^1(Y,\mathrm{ad}(P_H))$.
• Inhomogeneous group: $G := H \ltimes N$. Elements are pairs $\omega=(\epsilon,\phi)$ with $\epsilon\in H$, $\phi\in N$.

§Semidirect product law and an explicit action on $\mathrm{Conn}(P_H)$

The semidirect product structure is determined by one requirement: doing “a rotation then a translation” should compose consistently with doing “another rotation then another translation,” while respecting that rotations act on adjoint-valued objects by conjugation. Concretely, define multiplication in $G$ by

$$ (\epsilon_1,\phi_1)\cdot(\epsilon_2,\phi_2) := (\epsilon_1\epsilon_2, \mathrm{Ad}(\epsilon_2^{-1})(\phi_1)+\phi_2) (\epsilon_1\epsilon_2, \epsilon_2^{-1}\phi_1\epsilon_2+\phi_2). $$

This is the explicit semidirect product multiplication rule required in the draft’s group construction. The only “twist” is that the first translation $\phi_1$ is rotated by the second rotation $\epsilon_2$ before you add the second translation $\phi_2$.

Now give $G$ something to do. There is a natural right action of $G$ on the affine space of connections $\mathcal{A}=\mathrm{Conn}(P_H)$:

$$ \mathcal{A}\times G\to\mathcal{A}, (A,(\epsilon,\phi))\mapsto A\cdot(\epsilon,\phi):=A^\epsilon+\phi, $$

i.e.

$$ A\cdot(\epsilon,\phi)=\epsilon^{-1}A\epsilon+\epsilon^{-1}d\epsilon+\phi. $$

This is explicit about “what acts on what”: an element $(\epsilon,\phi)\in G$ acts on a connection $A\in\mathrm{Conn}(P_H)$ to produce another connection in $\mathrm{Conn}(P_H)$. It’s also the cleanest statement of “transport”: apply a gauge rotation, then translate in the affine direction $\phi\in N$.

Two quick sanity checks are worth stating (and they’re the reason the semidirect product rule above is the right one).

1. If you set $\phi=0$, you recover ordinary gauge transformations: $A\cdot(\epsilon,0)=A^\epsilon$.

2. If you set $\epsilon=e$ (identity), you recover pure affine shifts: $A\cdot(e,\phi)=A+\phi$.

   And because the group law was chosen to track how $\epsilon$ rotates $\phi$, one verifies (by a short computation) that

   $$ (A\cdot g_1) \cdot g_2 = A\cdot(g_1 g_2), $$

   so this really is a group action, not just a suggestive formula.

Notice what we didn’t do: we did not declare $\phi$ to be “the gauge field” or “torsion” or anything physical yet. At this stage, $\phi$ is simply the translation coordinate needed to parametrize motions in the affine space of connections. The physics begins when we pick a distinguished origin $A_0$ (soon) and form tensorial combinations measured relative to it — but the algebraic stage-setting is entirely here.

§Assumptions vs Consequences

§Definitional

• $N=\Omega^1(Y,\mathrm{ad}(P_H))$ is the model vector space for $\mathrm{Conn}(P_H)$.
• $H$ acts on $N$ by adjoint conjugation, $\phi \mapsto \epsilon^{-1}\phi\epsilon$.

§Ansatz (this instantiation’s transport principle)

• Promote the affine translations by $N$ to part of the gauge/transport symmetry by passing from $H$ to $G=H\ltimes N$.

§Consequences

• $G$ has the explicit semidirect product multiplication rule above, forced by the adjoint action.
• $G$ acts on the affine space of connections by $A\mapsto A^\epsilon+\phi$, giving a literal “rotate + shift” transport move.
• Once a background $A_0$ is chosen, $G$ will let us express a general connection as a transported image of $A_0$ and extract a tensorial $N$-valued field from that transport (next part, without yet defining it here).

§Why this matters

This $G$-action is the minimal bookkeeping that keeps the affine geometry of $\mathrm{Conn}(P_H)$ compatible with gauge covariance: it’s the framework in which you can pick an affine origin $A_0$, treat displacements $\delta A=A-A_0$ as honest $N$-valued tensors, and then build the torsion-first variables and Lagrangian pairings downstream (including the Shiab-based curvature contraction) without ever slipping into the forbidden move “the connection is a tensor”; $G$ is what makes the upcoming $A_0$-based construction genuinely covariant rather than a gauge-fixed hack.

§Key takeaway

$G=H\ltimes N$ is the group you get when you take seriously the fact that connection space is affine: you can rotate a connection (by $H$), and shift it (by $N$), and those operations compose with a conjugation twist.

§Technical takeaway

$$ (\epsilon_1,\phi_1)(\epsilon_2,\phi_2)=(\epsilon_1\epsilon_2, \epsilon_2^{-1}\phi_1\epsilon_2+\phi_2), A\cdot(\epsilon,\phi)=\epsilon^{-1}A\epsilon+\epsilon^{-1}d\epsilon+\phi. $$