A D-Brane-Induced Quaternionic and PT-Symmetric Spacetime with Falsifiable Signatures

1. Introduction

The outstanding problem of quantum gravity is to reconcile the background–independent dynamics of general relativity with the microscopic degrees of freedom provided by quantum field theory. Two frameworks have achieved partial success from opposite directions. On the one hand, Dp–brane effective actions in Type II string theory show how gauge and gravitational modes merge through the Dirac–Born–Infeld (DBI) plus Chern–Simons (CS) terms [1]. On the other hand, the spectral-action programme in non-commutative geometry (NCG) derives all bosonic interactions from the high–frequency spectrum of a suitable Dirac operator D [2,3]. Despite conceptual affinities—both replace a fundamental space–time metric by algebraic data—the two approaches have remained technically disjoint. In particular, no rigorous derivation exists that starts from a standard string-theory action without ad-hoc deformations and ends with a four–dimensional spectral triple that is both consistent at the quantum level and falsifiable in principle.

Goal and strategy. We show that a single, space-filling D3–brane placed in a slowly varying Neveu–Schwarz two-form background generates—after the Seiberg–Witten (SW) scaling limit—a pair of linearly independent $SU\left(2\right)$ tensors that survive as deformations of the open-string metric. These tensors are naturally assembled into a quaternion-valued, PT-symmetric metric of the form

$${\mathbb{G}}_{\mu \nu }={\eta }_{\mu \nu }+{ϵ}_{0}cos\left(\omega t\right)e_1{T}_{\mu \nu }+{ϵ}_1{\displaystyle \frac{e_2}{r}}{R}_{\mu \nu },$$

where $e_{1,2}$ are fixed imaginary quaternions and $\{{ϵ}_0,{ϵ}_1\}$ are the only free parameters.1 With a minimal parity-time prescription $\mathcal{P}:x^i\to -x^i$, $\mathcal{T}:t\to -t$, $i\to -i$—while the internal Pauli matrices remain inert—the corresponding Dirac operator is pseudo-Hermitian. Heat-kernel techniques then imply that the spectral action reproduces the two activator profiles ${ϵ}_{0}cos\omega t$ and ${ϵ}_1/r$ without further assumptions. A single local counterterm, the linear-quaternion slice of the Seeley–DeWitt invariant $a_4$, renders the theory one-loop finite and anomaly free.

Main results.

(i)

First-principle derivation. Starting from the DBI+CS action and the SW limit we obtain a four-dimensional quaternionic metric whose PT symmetry is inherited—rather than imposed—by world-sheet parity.

(ii)

Renormalisability. All linear-quaternion anomalies cancel against a unique counter-term ${\Pi }_{1^{\mathrm{st}}Q}\left[a_4\right]$, leaving the scalar sector identical to conventional Einstein–Hilbert gravity at low energies.

(iii)

Phenomenological window. Current CMB and gravitational-wave data already limit $|{ϵ}_0|\lesssim {10}^{-7}$ and $|{ϵ}_1|\lesssim {10}^1$. Near-future atom interferometers and third-generation detectors will improve these bounds by one to two orders of magnitude, providing a decisive test of the model.

Outline.Section 2 derives the open-string metric from the DBI+CS action; Section 3 implements the SW scaling and identifies the surviving $SU\left(2\right)$ directions. Quaternionic geometry, PT symmetry and the pseudo-Hermitian Dirac operator are established in Section 4. Section 6 and Section 7 develop the heat-kernel expansion, the stochastic influence functional, and the one-loop renormalisation. Observable consequences are summarised in Section 9; concluding remarks and open problems appear in Section 10.

Throughout we use the mostly-minus signature $(+,-,-,-)$, set $\hslash =c=1$, and employ $({\alpha }^{\prime },g_s)$ for the string slope and coupling. Repeated Greek indices are summed unless stated otherwise; a glossary of symbols is collected in Appendix G.

2. D3–Brane DBI + CS Action

We consider a single, space–filling D3–brane propagating in ten–dimensional type-IIB string theory. Working in the static gauge $X^{\mu }\left(\xi \right)={\xi }^{\mu }$$(\mu =0,1,2,3)$ and setting the world-volume field strength $F_{\mu \nu }=0$, the bosonic action factorises into Dirac–Born–Infeld (DBI) and Chern–Simons (CS) terms2:

$$\begin{array}{cc}\hfill S_{\mathrm{D}3}& =S_{\mathrm{DBI}}+S_{\mathrm{CS}},\hfill \end{array}$$

(1)

$$\begin{array}{cc}\hfill S_{\mathrm{DBI}}& =-T_3\int d^4\xi \sqrt{-det\left(g_{\mu \nu }+2\pi {\alpha }^{\prime }{B}_{\mu \nu }\right)},\hfill \end{array}$$

(2)

$$\begin{array}{cc}\hfill S_{\mathrm{CS}}& ={\mu }_3\int \sum _q{C}_q\wedge e^{2\pi {\alpha }^{\prime }B}.\hfill \end{array}$$

(3)

2.1. Long–Wavelength Two–Form Background

To isolate the minimal Lorentz-breaking content one demands that no more than two independent antisymmetric tensors survive on the brane. A convenient ansatz is a slowly varying electric component plus a static magnetic monopole [4]:

$$B_{\mu \nu }=\underset{\mathrm{electric}\mathrm{wave}}{\underbrace{\left({\overline{B}}_{0i}+\delta B_{0i}cos\omega t\right)dt\wedge d{x}^i}}+\underset{\mathrm{magnetic}\mathrm{monopole}}{\underbrace{\beta {\displaystyle \frac{{ϵ}_{ijk}{x}^k}{r^3}}d{x}^i\wedge d{x}^j}},r^2={\delta }_{ij}{x}^i{x}^j.$$

(4)

The field strength $H=dB$ vanishes away from the origin, so the bulk equations remain intact. Two dimensionless parameters

$${ϵ}_0=2\pi {\alpha }^{\prime }\delta B_{0i},{ϵ}_1=2\pi {\alpha }^{\prime }\beta$$

(5)

will play the rôle of geometric activators in later sections.

2.2. Linearised Open–String Metric

For $\left(2\pi {\alpha }^{\prime }B\right)\ll 1$ the determinant in (1) can be expanded to quadratic order. The well-known result, often called the Seiberg–Witten open–string metric, is

$${\mathcal{G}}_{\mu \nu }=g_{\mu \nu }-{\left(2\pi {\alpha }^{\prime }\right)}^2{B}_{\mu \rho }{g}^{\rho \sigma }{B}_{\sigma \nu }+\mathcal{O}\left(B^3\right).$$

(6)

Substituting the Minkowski background $g_{\mu \nu }={\eta }_{\mu \nu }=\mathrm{diag}(+,-,-,-)$ and the profile (4), one obtains at linear order (cf. Appendix A for algebra)

$$\begin{array}{|c|}\hline {\mathcal{G}}_{\mu \nu }={\eta }_{\mu \nu }+{ϵ}_{0}cos\left(\omega t\right)e_1{T}_{\mu \nu }+{ϵ}_1{\displaystyle \frac{e_2}{r}}{R}_{\mu \nu }+\mathcal{O}\left(B^2\right)\\ \hline\end{array},\left\{T_{\mu \nu }={\delta }_{\mu }^0{\delta }_{\nu }^0,R_{\mu \nu }={\delta }_{\mu }^i{\delta }_{\nu }^i\right\}.$$

(7)

Here $e_{1,2}\equiv i{\sigma }_{1,2}/2$ are fixed imaginary quaternions. No other internal direction survives: the electric and magnetic pieces are orthogonal in $\mathbb{H}$, a fact that will seed the non-commutative geometry of Section 3.

2.3. Physical Reading

Equation (7) reveals two distinct, linearly independent deformations of flat Minkowski space:

• a time-like “spring” ${ϵ}_{0}cos\left(\omega t\right)e_1{T}_{\mu \nu }$,
• a radial “vortex” ${ϵ}_1{r}^{-1}{e}_2{R}_{\mu \nu }$.

Because $e_1{e}_2=-e_2{e}_1$, these deformations do not commute: they are the low-energy remnant of the non-commutative coordinates $[x^{\mu },x^{\nu }]\propto {\theta }^{\mu \nu }$ that emerge in the SW limit. Moreover, both terms are invariant under the minimal parity–time operation $\left(\mathcal{P}\mathcal{T}\right)x^{\mu }=(-t,-\mathbf{x})$ when accompanied by $i\to -i$, so the spectrum of the corresponding Dirac operator is guaranteed to be pseudo-Hermitian—a prerequisite for the spectral action employed later.

Section 3 carries these ingredients into the Seiberg–Witten scaling limit and identifies the precise non-commutativity tensor ${\theta }^{\mu \nu }$ that underlies the quaternionic algebra of our model.

3. Seiberg–Witten Limit and Open–String Data

The linearised metric (7) still depends on the string scale ${\alpha }^{\prime }$. In order to obtain a finite low–energy description we now implement the Seiberg–Witten (SW) scaling limit of the world–volume theory [4].

3.1. Seiberg–Witten Relations

For vanishing world–volume field strength $(F_{\mu \nu }=0)$ the closed and open variables are related by

$$\begin{array}{cccc}\hfill {\displaystyle \frac{1}{G+2\pi {\alpha }^{\prime }\mathrm{\Phi }}}& ={\left({\displaystyle \frac{1}{g+2\pi {\alpha }^{\prime }B}}\right)}_{\mathrm{sym}},\hfill & \hfill {\theta }^{\mu \nu }& =-{\left(2\pi {\alpha }^{\prime }\right)}^2{\left({\displaystyle \frac{1}{g+2\pi {\alpha }^{\prime }B}}\right)}_{\mathrm{anti}}^{\mu \nu },\hfill \end{array}$$

(8a)

$$\begin{array}{cc}\hfill G_s& =g_s\sqrt{{\displaystyle \frac{det(G+2\pi {\alpha }^{\prime }\mathrm{\Phi })}{det(g+2\pi {\alpha }^{\prime }B)}}},\hfill \end{array}$$

(8b)

where ${\mathrm{\Phi }}_{\mu \nu }$ is an arbitrary two–form corresponding to a field redefinition. We adopt the canonical gauge $\mathrm{\Phi }=0$, in which $G_{\mu \nu }$ coincides with the metric (7) computed from the DBI determinant.

3.2. Scaling Prescription

The SW limit is defined by

$${\alpha }^{\prime }⟶0,B_{\mu \nu }\sim {\alpha }^{\prime -1},\left\{G_{\mu \nu },{\theta }^{\mu \nu },G_s\right\}\mathrm{fixed}.$$

(9)

Intuitively, the electric and magnetic components of B are tuned large enough to compensate the vanishing string length, while $g_{\mu \nu }$ and $g_s$ are scaled to keep the open quantities finite. Substituting the background (4) into (8) and expanding to first order in the small parameters (5) yields

$$\begin{array}{c}\hfill \begin{array}{|c|}\hline G_{\mu \nu }={\eta }_{\mu \nu }+{ϵ}_{0}cos\left(\omega t\right)e_1{T}_{\mu \nu }+{ϵ}_1{\displaystyle \frac{e_2}{r}}{R}_{\mu \nu }+\mathcal{O}\left(B^2\right)\\ \hline\end{array},\end{array}$$

(10)

$$\begin{array}{c}\hfill \begin{array}{|c|}\hline {\theta }^{0i}=-\left(2\pi {\alpha }^{\prime }\right){ϵ}_0{e}_1{\delta }_0^0{\delta }_i^i,{\theta }^{ij}=-\left(2\pi {\alpha }^{\prime }\right){ϵ}_1{\displaystyle \frac{{ϵ}^{ijk}{x}^k}{r^3}}{e}_2\\ \hline\end{array},\end{array}$$

(11)

$$\begin{array}{c}\hfill \begin{array}{|c|}\hline G_s=g_s\left[1+\mathcal{O}\left(B^2\right)\right]\\ \hline\end{array}.\end{array}$$

(12)

Only two orthogonal imaginary quaternions, $e_1$ and $e_2$, survive; they span an $SU\left(2\right)$ algebra that will seed the quaternionic Clifford structure in Section 4.

3.3. Non-Commutative ★-Product and Gauge Map

At fixed $(G_{\mu \nu },{\theta }^{\mu \nu })$ any world–volume field $\phi$ multiplies according to the Moyal product

$$f★g=fexp\left(\frac{i}{2}{\overleftarrow{\partial }}_{\mu }{\theta }^{\mu \nu }\left(x\right){\overrightarrow{\partial }}_{\nu }\right)g,|\partial \theta |\ll \left|\theta \right|,$$

(13)

where the mild x–dependence of ${\theta }^{\mu \nu }$ (a monopole tail in r and a sinusoid in t) can be treated perturbatively. The usual Seiberg–Witten map ${\widehat{A}}_{\mu }\left[A\right]=A_{\mu }-\frac{1}{4}{\theta }^{\rho \sigma }\{A_{\rho },{\partial }_{\sigma }{A}_{\mu }+F_{\sigma \mu }\}+\dots$ then leads, up to $\mathcal{O}\left(\theta \right)$, to a non-commutative Yang–Mills action

$$S_{\mathrm{NCYM}}=-{\displaystyle \frac{1}{4g_{\mathrm{YM}}^2}}\int d^{4}x{G}^{\mu \rho }{G}^{\nu \sigma }{\widehat{F}}_{\mu \nu }★{\widehat{F}}_{\rho \sigma },g_{\mathrm{YM}}^2=2\pi G_s{\left(2\pi {\alpha }^{\prime }\right)}^{-1}.$$

(14)

Equations (10)–(14) constitute the complete set of open–string data that will enter the spectral geometry of the following sections.

3.4. Quaternionic Seed

Because ${\theta }^{\mu \nu }$ decomposes into two orthogonal blocks,

$${\theta }^{\mu \nu }={\theta }_{\left(1\right)}^{\mu \nu }{e}_1+{\theta }_{\left(2\right)}^{\mu \nu }{e}_2,e_a{e}_b=-{\delta }_{ab}+{ϵ}_{abc}{e}_c,$$

(15)

the associative algebra $\mathcal{A}=C^{\infty }\left({\mathbb{R}}^{1,3}\right)\otimes \mathbb{H}$ endowed with the product $(f\otimes h)★(g\otimes k)=(f★g)\otimes \left(hk\right)$ is quaternionic. The pair $(\mathcal{A},D_0)$, with $D_0=i{\gamma }^{\mu }({\partial }_{\mu }+{\Omega }_{\mu }\left[G\right])$, will act as the seed of a PT-symmetric spectral triple in Section 4.

Take-away. After the SW limit the D3–brane retains exactly two $SU\left(2\right)$ directions. They appear (i) in the open metric as activator profiles ${ϵ}_{0}cos\omega t$ and ${ϵ}_1/r$, and (ii) in the non-commutativity tensor ${\theta }^{\mu \nu }$ as an anti-commuting pair $e_{1,2}$. These features establish the algebraic backbone for the quaternionic, PT-symmetric geometry constructed below.

4. Quaternion–Valued Metric and Clifford Extension

The Seiberg–Witten analysis of Section 3 isolates two orthogonal, non-commuting internal directions $e_1,e_2\in \mathbb{H}$. In this section we promote that $\mathrm{SU}\left(2\right)$ doublet to a bona-fide quaternionic geometry, construct the enlarged Clifford bundle, prove $\mathcal{PT}$ symmetry and pseudo-Hermiticity of the Dirac operator, and—crucial for later sections—show how the leading Seeley–DeWitt coefficient automatically triggers the activator profiles ${ϵ}_{0}cos\omega t$ and ${ϵ}_1/r$. Throughout we keep terms up to $\mathcal{O}\left({ϵ}_{0,1}\right)$; higher orders will enter only in Section 6.

4.1. Quaternionic Metric: Minimal Ansatz

Let $\{e_1,e_2,e_3\}$ be the imaginary quaternion units ($e_a{e}_b=-{\delta }_{ab}+{ϵ}_{abc}{e}_c$, $e_a^{†}=-e_a$). Guided by the open metric (10) we introduce a quaternion-valued deformation of Minkowski space:

$$\begin{array}{|c|}\hline {\mathbb{G}}_{\mu \nu }\left(x\right)={\eta }_{\mu \nu }+{ϵ}_{0}cos\left(\omega t\right)e_1{T}_{\mu \nu }+{ϵ}_1{\displaystyle \frac{e_2}{r}}{R}_{\mu \nu }+\mathcal{O}\left({ϵ}^2\right)\\ \hline\end{array},{ϵ}_{0,1}\ll 1,$$

(16)

where $T_{\mu \nu }:={\delta }_{\mu }^0{\delta }_{\nu }^0$ projects on the lapse and $R_{\mu \nu }:={\delta }_{\mu }^i{\delta }_{\nu }^i$ on the spatial radius. Equation (16) is the minimal ansatz that

(i)

preserves Lorentz signature to $\mathcal{O}\left({ϵ}_{0,1}\right)$;

(ii)

retains exactly the two $\mathrm{SU}\left(2\right)$ directions singled out by the B-field; and

(iii)

reduces to the usual open metric when $e_{1,2}\mapsto \pm 1$.

Physical picture.

The term ${ϵ}_{0}cos\left(\omega t\right)e_1{T}_{\mu \nu }$ is a time-like spring aligned with $e_1$, whereas ${ϵ}_1{r}^{-1}{e}_2{R}_{\mu \nu }$ is a spatial vortex aligned with $e_2$. Because $e_1{e}_2=-e_2{e}_1$, spring and vortex do not commute—a geometric echo of the non-commutativity tensor ${\theta }^{\mu \nu }$ in Section 3.

4.2. Quaternionic Clifford Bundle

Let $S\to {\mathbb{R}}^{1,3}$ be the usual Dirac spinor bundle and set

$${\mathrm{\Gamma }}_{\mu }:={\gamma }_{\mu }\otimes {\mathbf{1}}_{\mathbb{H}},{\mathrm{\Gamma }}_{\left(a\right)}:={\gamma }_{*}\otimes e_a(a=1,2),$$

(17)

with ${\gamma }_{*}=i{\gamma }^0{\gamma }^1{\gamma }^2{\gamma }^3$. One checks $\{{\mathrm{\Gamma }}_A,{\mathrm{\Gamma }}_B\}=2{\mathcal{G}}_{AB}$, where ${\mathcal{G}}_{AB}={\eta }_{\mu \nu }\oplus (-{\delta }_{ab})$; thus the total Clifford algebra is $\mathrm{Cliff}(1,5)$ and the structure group factorises as $\mathrm{Spin}(1,3)\times \mathrm{SU}{\left(2\right)}_H$.

4.3. $\mathcal{PT}$ Symmetry and Pseudo-Hermiticity

Adopt the minimal parity–time rule of Section 1: $\mathcal{P}:x^i\mapsto -x^i$, $\mathcal{T}:t\mapsto -t,i\mapsto -i$, while the internal ${\sigma }_a$ (hence $e_a$) stay inert. Since $e_1,e_2$ are $\mathcal{PT}$–even,

$$\left(\mathcal{PT}\right){\mathbb{G}}_{\mu \nu }\left(x\right){\left(\mathcal{PT}\right)}^{-1}={\mathbb{G}}_{\mu \nu }\left(x\right)\forall x,$$

(18)

already at $\mathcal{O}\left(ϵ\right)$. With ${\nabla }^{\mathbb{G}}$ the Levi-Civita connection of ${\mathbb{G}}_{\mu \nu }$, define the quaternionic Dirac operator

$$D:=i{\mathrm{\Gamma }}^{\mu }\left({\partial }_{\mu }+\frac{1}{4}{\Omega }_{\mu }^{AB}{\mathrm{\Gamma }}_A{\mathrm{\Gamma }}_B\right),\eta :={\mathrm{\Gamma }}^0.$$

(19)

One immediately finds $D^{†}=\eta D{\eta }^{-1}$ and $\left(\mathcal{PT}\right)D{\left(\mathcal{PT}\right)}^{-1}=D$, so D is simultaneously pseudo-Hermitian and $\mathcal{PT}$-symmetric; its spectrum is real or comes in conjugate pairs, validating the spectral expansion in Section 6. Detailed proofs are relegated to Appendix B.

4.4. Linear SDW Trigger for the Activators

A key claim of this work is that the spring/vortex profiles (16) are not imposed by hand but emerge as stationary points of the spectral action. At leading order the relevant object is the linearised Seeley–DeWitt coefficient $a_2$ (details in Appendix C):

$$a_2\left(x\right)=-{\displaystyle \frac{4}{{\left(4\pi \right)}^2}}\left[{ϵ}_{0}cos\left(\omega t\right)e_1+{ϵ}_1{\displaystyle \frac{e_2}{r}}\right]+\mathcal{O}\left({ϵ}^2\right).$$

(20)

Inserting (20) into the cut-off spectral action $S_{\chi }={\sum }_n{f}_{2n}{\Lambda }^{4-2n}{a}_{2n}$ and integrating over the internal trace yields the activation Lagrangian

$${\mathcal{L}}_{\mathrm{act}}={\displaystyle \frac{{\kappa }_t}{2}}{ϵ}_0^2{cos}^2\left(\omega t\right)+{\displaystyle \frac{{\kappa }_r}{2}}{ϵ}_1^2{\displaystyle \frac{1}{r^2}},{\kappa }_{t,r}:={\displaystyle \frac{{\Lambda }^4{f}_0}{8{\pi }^2}}+{\displaystyle \frac{{\Lambda }^2{f}_2\{{\omega }^2,1\}}{12{\pi }^2}}.$$

(21)

Treating ${ϵ}_0\left(t\right)$ and ${ϵ}_1\left(r\right)$ as collective coordinates and varying $\int d^{4}x{\mathcal{L}}_{\mathrm{act}}$ gives

$$\begin{array}{cccc}\hfill {\ddot{ϵ}}_0\left(t\right)+{\omega }^2{ϵ}_0\left(t\right)& =0,\hfill & \hfill \Delta \left[{ϵ}_1\left(r\right)/r\right]& =0,\hfill \end{array}$$

(22)

whose normalisable solutions are precisely ${ϵ}_{0}cos\omega t$ and ${ϵ}_1/r$. Thus the linear SDW density $a_2$ self-consistently selects the activator profiles; higher SDW terms only dress the effective couplings ${\kappa }_{t,r}$.

Summary

• A Lorentzian metric can accommodate the SW deformations by upgrading two components to imaginary quaternions $e_1,e_2$.
• The enlarged Clifford algebra is $\mathrm{Cliff}(1,5)$; the Dirac operator is both pseudo-Hermitian and $\mathcal{PT}$-invariant.
• Linear perturbation of the SDW coefficient $a_2$ forces the activator Equation (22), whose solutions reproduce the spring ${ϵ}_{0}cos\omega t$ and vortex ${ϵ}_1/r$ profiles assumed in (16).
• These results lay the algebraic and dynamical foundation for the heat-kernel expansion of Section 6 and the renormalisation analysis of Section 7.

5. PT Symmetry and the Pseudo–Hermitian Dirac Operator

The quaternion–valued metric (16) equips the space–time manifold with a fixed internal $SU\left(2\right)$ frame $\{e_1,e_2\}$. To ensure that the ensuing quantum theory is physically well–defined one must (i) specify a consistent PT transformation acting simultaneously on space–time and on the quaternionic algebra, and (ii) verify that the Dirac operator constructed from ${\mathbb{G}}_{\mu \nu }$ is pseudo–Hermitian with respect to a positive Krein form. This section provides the required checks.

5.1. Minimal $\mathcal{PT}$ Rule

The guiding principle is to keep the rule minimal: space–time coordinates transform as in ordinary relativistic quantum mechanics, while the internal quaternions remain untouched. More precisely,

$$\mathcal{P}:(t,\mathbf{x})\mapsto (t,-\mathbf{x}),\mathcal{T}:(t,\mathbf{x},i)\mapsto (-t,\mathbf{x},-i),\mathcal{PT}=\mathcal{T}\mathcal{P}.$$

(23)

Because the Pauli matrices ${\sigma }_a$ and hence the quaternions $e_a=i{\sigma }_a/2$ live in an internal space, they are PT–even. Table 1 summarises the action on the basic building blocks.

Metric invariance.

Applying (23) to the quaternionic metric (16) gives

$$\left(\mathcal{PT}\right){\mathbb{G}}_{\mu \nu }\left(x\right){\left(\mathcal{PT}\right)}^{-1}={\mathbb{G}}_{\mu \nu }\left(x\right),\forall x\in {\mathbb{R}}^{1,3},$$

(24)

because $cos\left(\omega t\right)$ is even under $t\mapsto -t$, $r^{-1}$ is even under $\mathbf{x}\mapsto -\mathbf{x}$, and $e_{1,2}$ are PT–even. Hence the full geometry respects $\mathcal{PT}$ symmetry exactly to first order in ${ϵ}_{0,1}$.

5.2. Krein Structure and Pseudo–Hermiticity

Let ${\nabla }^{\mathbb{G}}$ be the Levi–Civita connection of ${\mathbb{G}}_{\mu \nu }$. The quaternionic Dirac operator is

$$D=i{\mathrm{\Gamma }}^{\mu }\left({\partial }_{\mu }+\frac{1}{4}{\Omega }_{\mu }^{AB}{\mathrm{\Gamma }}_A{\mathrm{\Gamma }}_B\right),{\mathrm{\Gamma }}_{\mu }={\gamma }_{\mu }\otimes {\mathbf{1}}_{\mathbb{H}},{\mathrm{\Gamma }}_{\left(a\right)}={\gamma }_{*}\otimes e_a.$$

(25)

Choosing the time–like Krein operator3

$$\eta ={\mathrm{\Gamma }}^0={\gamma }^0\otimes {\mathbf{1}}_{\mathbb{H}},$$

(26)

one verifies to $\mathcal{O}\left({ϵ}_{0,1}\right)$ that

$$D^{†}=\eta D{\eta }^{-1}⟹D\mathrm{is}\mathit{pseudo}-\mathit{Hermitian}.$$

(27)

Combined with the equality $\left(\mathcal{PT}\right)D{\left(\mathcal{PT}\right)}^{-1}=D$, Equation (27) ensures that the spectrum of D is either real or arranged in complex conjugate pairs, a property that underpins the heat–kernel expansion used in Section 6.

5.3. Spectral Triple in a Krein Space

Gathering the pieces, we obtain a Krein–space spectral triple

$$\begin{array}{|c|}\hline \left(\mathcal{A},\mathcal{H},D;\eta ,J,\gamma \right)\equiv \left(C^{\infty }\left({\mathbb{R}}^{1,3}\right)\otimes \mathbb{H},L^2({\mathbb{R}}^{1,3},S)\otimes \mathbb{H},D;{\gamma }^0\otimes \mathbf{1},J_C\otimes \mathbf{1},{\gamma }_{*}\otimes \mathbf{1}\right)\\ \hline\end{array},$$

(28)

where $J_C$ is the usual charge–conjugation on Dirac spinors and ${\gamma }_{*}=i{\gamma }^0{\gamma }^1{\gamma }^2{\gamma }^3$. All Connes axioms are satisfied up to $\mathcal{O}\left({ϵ}_{0,1}\right)$; the KO dimension is $d_{\mathrm{KO}}=6$ because the internal quaternion factor contributes two extra negative directions to the Clifford algebra $\mathrm{Cliff}(1,3)\otimes \mathbb{H}\cong \mathrm{Cliff}(1,5)$.

Take–Away

• The minimal rule (23) renders both the quaternionic metric and the Dirac operator strictly PT–invariant.
• With Krein form $\eta ={\gamma }^0$ the Dirac operator is pseudo–Hermitian; its spectrum is spectrally stable.
• The algebraic data $(\mathcal{A},\mathcal{H},D;\eta ,J,\gamma )$ define a consistent spectral triple, providing the backbone for the heat–kernel expansion and the renormalised spectral action employed in the next two sections.

6. Heat–Kernel Expansion up to $a_2$

With the pseudo–Hermitian spectral triple of Section 5 in place, we are ready to evaluate the spectral action in the ultraviolet cut–off scheme of Chamseddine–Connes [3]. Up to and including the Seeley–DeWitt density $a_2$ one has

$$\begin{array}{|c|}\hline {\mathcal{S}}_{\chi }(D,\Lambda )=Tr\chi \left(\frac{D^2}{{\Lambda }^2}\right)=f_0{\Lambda }^4{a}_0+f_2{\Lambda }^2{a}_2+\mathcal{O}\left({\Lambda }^0\right)\\ \hline\end{array},$$

(29)

where $f_n\equiv {\int }_0^{\infty }\chi \left(u\right)u^{\frac{n}{2}-1}du$ are the Mellin moments of the positive, rapidly decaying test function $\chi \left(u\right)$.

6.1. General Formulae

For a Laplace–type operator $D^2=-{\mathbb{G}}^{\mu \nu }{\nabla }_{\mu }{\nabla }_{\nu }+\frac{1}{4}R+\mathrm{\Phi }$ on the vector bundle $S\otimes \mathbb{H}\to {\mathbb{R}}^{1,3}$ one has [8,9]

$$\begin{array}{ccc}\hfill a_0\left(x\right)& ={\displaystyle \frac{1}{16{\pi }^2}}{Tr}_{\mathrm{int}}\mathbf{1},\hfill & \hfill \end{array}$$

(30)

$$\begin{array}{cc}\hfill a_2\left(x\right)& ={\displaystyle \frac{1}{16{\pi }^2}}{Tr}_{\mathrm{int}}\left(\frac{1}{6}R+\mathrm{\Phi }\right).\hfill \end{array}$$

(31)

Here R is the Ricci scalar of the quaternionic metric ${\mathbb{G}}_{\mu \nu }$ and $\mathrm{\Phi }$ is the bundle endomorphism generated by the commutator of the quaternionic gamma matrices with the spin connection: $\mathrm{\Phi }=\frac{1}{4}[{\mathrm{\Gamma }}_{\mu },{\mathrm{\Gamma }}_{\nu }]{\mathbb{G}}^{\mu \alpha }{\mathbb{G}}^{\nu \beta }{\Omega }_{\alpha \beta }$. The “internal” trace ${Tr}_{\mathrm{int}}$ runs over both Dirac and quaternion indices.

6.2. Linearised Evaluation on ${\mathbb{G}}_{\mu \nu }$

Insert the metric deformation

$${\mathbb{G}}_{\mu \nu }={\eta }_{\mu \nu }+{ϵ}_{0}cos\left(\omega t\right)e_1{T}_{\mu \nu }+{ϵ}_1{\displaystyle \frac{e_2}{r}}{R}_{\mu \nu }+\mathcal{O}\left({ϵ}^2\right),$$

and keep terms up to $\mathcal{O}\left({ϵ}^2\right)$. Because ${Tr}_{\mathrm{int}}{e}_a=0$ and ${Tr}_{\mathrm{int}}{e}_a{e}_b=-4{\delta }_{ab}$, all linear contributions vanish identically. Writing $d^{4}x=dt{d}^3\mathbf{x}$ and suppressing volume factors we obtain4

$$\begin{array}{cc}\hfill a_0& ={\displaystyle \frac{4}{16{\pi }^2}}\left[1+\frac{1}{2}{ϵ}_0^2{cos}^2\left(\omega t\right)+\frac{1}{2}{ϵ}_1^2{\displaystyle \frac{1}{r^2}}+\mathcal{O}\left({ϵ}^3\right)\right],\hfill \end{array}$$

(32)

$$\begin{array}{cc}\hfill a_2& =-{\displaystyle \frac{4}{48{\pi }^2}}\left[{ϵ}_0^2{\omega }^2{cos}^2\left(\omega t\right)+{ϵ}_1^2{\displaystyle \frac{1}{r^2}}+\mathcal{O}\left({ϵ}^3\right)\right].\hfill \end{array}$$

(33)

Crucially, the only space–time dependences that survive are ${cos}^2\left(\omega t\right)$ and $r^{-2}$, i.e. the squares of the spring and vortex profiles introduced at $\mathcal{O}\left(ϵ\right)$.

6.3. Activation Lagrangian

Substituting (32)–(33) into (29) and dropping the cosmological constant term ($\propto f_0{\Lambda }^4$) give the activation sector

$$\begin{array}{|c|}\hline {\mathcal{L}}_{\mathrm{act}}={\displaystyle \frac{{\kappa }_t}{2}}{ϵ}_0^2{cos}^2\left(\omega t\right)+{\displaystyle \frac{{\kappa }_r}{2}}{ϵ}_1^2{\displaystyle \frac{1}{r^2}},\\ \hline\end{array}\begin{array}{cc}\hfill {\kappa }_t& ={\displaystyle \frac{3{\Lambda }^4{f}_0}{64{\pi }^2}}+{\displaystyle \frac{{\Lambda }^2{f}_2{\omega }^2}{96{\pi }^2}},\hfill \\ \hfill {\kappa }_r& ={\displaystyle \frac{3{\Lambda }^4{f}_0}{64{\pi }^2}}+{\displaystyle \frac{{\Lambda }^2{f}_2}{96{\pi }^2}}.\hfill \end{array}$$

(34)

No quaternionic generators appear because the internal trace has removed them. The Lagrangian (34) acts as an effective potential stabilising the time–like spring and the spatial vortex. Its variation will generate the quaternionic geometric flow of Section 6.

Take–Away

• Up to $a_2$ the heat–kernel picks out ${cos}^2\left(\omega t\right)$ and $r^{-2}$ as the unique, leading order space–time dependences.
• The activation Lagrangian (34) contains only ${ϵ}_0^2$ and ${ϵ}_1^2$; all linear terms cancel by the internal trace.
• Equation (34) supplies the dynamical seed for the quaternionic geometric flow studied in the next section and ultimately for the phenomenological couplings summarised in Section 9.

7. One–Loop Renormalisation and Anomaly Cancellation

The heat–kernel calculation of Section 6 showed that the linear–quaternion slice of the Seeley–DeWitt density $a_4$ sources an apparent non–conservation of the $SU{\left(2\right)}_H$ currents, cf. Equation (40). In this section we prove that the corresponding anomaly is local and can be removed by a single counter–term, rendering the theory finite and gauge–invariant at one loop.

7.1. Spectral Regularisation and Divergent Structure

Throughout we keep the ultraviolet cut–off $\Lambda$ explicit, following the spectral–action prescription

$${\mathcal{S}}_{\chi }(D,\Lambda )=\sum _{n\ge 0}{f}_n{\Lambda }^{4-n}{a}_n\left[D^2\right],f_n:={\int }_0^{\infty }\chi \left(u\right)u^{\frac{n}{2}-1}du.$$

(35)

Loop corrections introduce an additional functional determinant,

$${\mathrm{\Gamma }}_{1\mathrm{L}}=\frac{1}{2}logdet\left(D^2/{\mu }^2\right)={\displaystyle \frac{1}{{\left(4\pi \right)}^2}}\sum _{k\ge 0}{b}_k{\mu }^{4-2k}{a}_{2k}\left[D^2\right],$$

(36)

where $\mu$ is the renormalisation scale and $b_k$ are scheme–dependent constants (here after minimal subtraction). Expanding around $d=4-ϵ$ dimensions yields the divergent piece

$$\mathrm{Div}{\mathrm{\Gamma }}_{1\mathrm{L}}={\displaystyle \frac{1}{ϵ}}\left({\beta }_0{\Lambda }^4{a}_0+{\beta }_2{\Lambda }^2{a}_2+{\beta }_4{a}_4\right),ϵ:=4-d.$$

(37)

Because ${Tr}_{\mathrm{int}}{e}_a=0$, the linear–quaternion projector ${\Pi }_{1\mathrm{Q}}[·]$ annihilates $a_0$ and $a_2$, while

$$\begin{array}{|c|}\hline {\Pi }_{1\mathrm{Q}}\left[a_4\right]=S_a{e}_a,\\ \hline\end{array}$$

(38)

with $S_a$ the local source defined in Equation (38). Hence only the coefficient ${\beta }_4$ is relevant for anomaly cancellation.

7.2. Counter–Term and Current Restoration

Introduce a local counter–term

$${\mathcal{S}}_{\mathrm{ct}}=-{\displaystyle \frac{1}{ϵ}}\left({\beta }_0{\Lambda }^4{a}_0+{\beta }_2{\Lambda }^2{a}_2\right)-{\displaystyle \frac{{\beta }_4}{ϵ}}{\Pi }_{1\mathrm{Q}}\left[a_4\right].$$

(39)

The scalar part (first parentheses) renormalises Newton’s constant and the cosmological term and plays no rôle in the $SU{\left(2\right)}_H$ Ward identities. The last term modifies the broken current equation (40) to

$${\nabla }_{\mu }{J}_a^{\mu }=S_a(1-{\beta }_4)+\mathcal{O}\left({\hslash }^2\right).$$

(40)

Choosing

$$\begin{array}{|c|}\hline {\beta }_4=1\\ \hline\end{array}$$

(41)

exactly cancels the linear–quaternion anomaly, i.e. ${\nabla }_{\mu }{J}_a^{\mu }=0$ to order ℏ. No further symmetry–breaking counter–terms are required.

PT Invariance.

Both $S_a$ and ${\Pi }_{1\mathrm{Q}}\left[a_4\right]$ are PT–even; therefore ${\mathcal{S}}_{\mathrm{ct}}$ preserves the global symmetry that guarantees the pseudo–Hermiticity of D.

7.3. Renormalisation–Group Flow

Denote by $\left\{g_i\right\}$ the scalar couplings in ${\mathcal{S}}_{\chi }$ (Newton, cosmological, and possible higher–derivative terms). The beta functions read $\mu \frac{d{g}_i}{d\mu }={\beta }_i\left(g\right)$. Because the $SU{\left(2\right)}_H$ anomaly has been removed, the running of $\left\{g_i\right\}$ is decoupled from the quaternionic sector:

$$\mu {\displaystyle \frac{d}{d\mu }}{\Pi }_{1\mathrm{Q}}\left[a_4\right]=0+\mathcal{O}\left({\hslash }^2\right).$$

(42)

At one loop, therefore, the spring/vortex parameters $({ϵ}_0,{ϵ}_1)$ renormalise solely through the classical matching to the underlying D3–brane data; they are not dressed by logarithmic divergences.

7.4. Higher Loops and Locality

Power counting shows that the linear–quaternion projection of $a_6,a_8,\dots$ is suppressed by additional factors ${\Lambda }^{-2}$ and can first appear at two loops. Moreover, every such contribution is local; if needed it can be cancelled by higher–dimensional counter–terms that respect PT symmetry. We thus conjecture that the single subtraction (39) is sufficient to all orders in perturbation theory.5

Summary

• One–loop divergences organise into the SDW basis $a_{0,2,4}$. Only $a_4$ carries a linear–quaternion piece.
• A single local counter–term $-{\Pi }_{1\mathrm{Q}}\left[a_4\right]$ removes the $SU{\left(2\right)}_H$ anomaly without spoiling PT symmetry.
• Scalar couplings run as in ordinary spectral gravity; the activator parameters $({ϵ}_0,{ϵ}_1)$ remain scale–invariant at one loop.
• Higher–loop anomalies, if any, are suppressed by extra powers of ${\Lambda }^{-2}$ and can be cancelled by local terms, preserving the predictivity of the two–parameter framework.

8. Path–Integral Origin of the Stochastic $\delta D$

The geometric flow of Section 6 required a noise term $\delta D$ in the quaternionic Dirac operator. In this section we derive that term from the world–sheet path integral of a single Type–IIB D3–brane, closely following the influence–functional strategy of Feynman and Vernon [16] while respecting the $\mathcal{PT}$ symmetry fixed in Section 5.

8.1. Microscopic Generating Functional

The full (Euclidean) partition function reads

$$\mathcal{Z}=\int \mathcal{D}X\mathcal{D}B\mathcal{D}\psi \mathcal{D}\overline{\psi }exp\left[-{\mathcal{S}}_{\mathrm{DBI}+\mathrm{CS}}[X,B]-{\mathcal{S}}_{\mathrm{F}}[\psi ,\overline{\psi };G(B,X)]\right],$$

(43)

where

• $X^M\left(\sigma \right)$ : brane embedding in static gauge,
• $B_{MN}$ : bulk NS–NS two–form,
• $\psi$ : open–string fermion in the bundle $S\otimes \mathbb{H}$,
• $G_{\mu \nu }(B,X)$ : open–string metric of Section 2 (quaternionic, $\mathcal{PT}$–even).

The fermionic term is

$${\mathcal{S}}_{\mathrm{F}}={\displaystyle \frac{1}{g_s}}\int d^{4}x\sqrt{\left|G\right|}\overline{\psi }\left(D_0+{ϵ}_{0}cos\omega t{e}_1+{ϵ}_1\frac{e_2}{r}\right)\psi ,$$

(44)

with $D_0$ the torsion–free Dirac operator for the background metric ${\eta }_{\mu \nu }$.

8.2. Mode Splitting and Coarse–Graining

Choose a coarse–graining scale ${\mu }_c\ll {\Lambda }_{\mathrm{UV}}$ and decompose each bosonic field into slow ($\overline{·}$) and fast ($\widetilde{·}$) parts:

$$\begin{array}{cccccc}\hfill B& =\overline{B}+\tilde{B},\hfill & \hfill X& =\overline{X}+\tilde{X},\hfill & \hfill \left|\partial \overline{B}\right|,\left|\partial \overline{X}\right|& <{\mu }_c<\left|\partial \tilde{B}\right|,\left|\partial \tilde{X}\right|.\hfill \end{array}$$

(45)

The functional measure factorises, $\mathcal{D}B=\mathcal{D}\overline{B}\mathcal{D}\tilde{B}$, etc., so that (43) becomes

$$\mathcal{Z}=\int \mathcal{D}\overline{B}\mathcal{D}\overline{X}{e}^{-{\mathcal{S}}_{\mathrm{DBI}+\mathrm{CS}}[\overline{B},\overline{X}]}\underset{{\displaystyle \equiv exp[-{\mathcal{S}}_{\mathrm{infl}}^{\mathrm{FV}}(\psi ,\overline{\psi };\overline{B},\overline{X})]}}{\underbrace{\int \mathcal{D}\psi \mathcal{D}\overline{\psi }{e}^{-{\mathcal{S}}_{\mathrm{F}}[\psi ;\overline{B},\overline{X}]}{〈e^{-\Delta {\mathcal{S}}_{\mathrm{int}}[\psi ,\overline{\psi };\tilde{B},\tilde{X}]}〉}_{\tilde{B},\tilde{X}}}}.$$

(46)

The average ${\langle \dots \rangle }_{\tilde{B},\tilde{X}}$ is Gaussian to leading order because the fast modes see the slow geometry as a fixed background.

8.3. Influence Functional and Gaussian Noise

Expanding ${\mathcal{S}}_{\mathrm{infl}}^{\mathrm{FV}}$ to quadratic order in the fermion bilinear gives

$${\mathcal{S}}_{\mathrm{infl}}^{\mathrm{FV}}=\frac{1}{2}\int \int d^{4}x{d}^{4}y\overline{\psi }\left(x\right)\Sigma (x-y)\psi \left(y\right)+\mathrm{local}\mathrm{terms},$$

(47)

where translation invariance of the bath implies $\Sigma (x-y)=\Sigma (y-x)$. At leading order one finds

$$\Sigma \left(z\right)={\sigma }^2{K}_{{\mu }_c}\left(z\right)\left(e_1\otimes e_1+e_2\otimes e_2\right),K_{{\mu }_c}\left(z\right)={\displaystyle \frac{exp[-{\mu }_c^2{z}^2]}{{\mu }_c^4}},$$

(48)

with ${\sigma }^2\propto g_s^2{\left({\alpha }^{\prime }\right)}^2{\mu }_c^2/{\Lambda }_{\mathrm{UV}}^2$. The anti–Hermitian part of $\Sigma$ defines a Gaussian noise [17], leading to the stochastic shift

$$D=D_0+{ϵ}_{0}cos\omega t{e}_1+{ϵ}_1{\displaystyle \frac{e_2}{r}}+\delta D,$$

(49)

with statistical moments

$${\langle \delta D\left(x\right)\rangle }_{\mathrm{stoch}}=0,{\langle \delta D\left(x\right)\delta D\left(y\right)\rangle }_{\mathrm{stoch}}={\sigma }^2{K}_{{\mu }_c}(x-y)\left(e_1,e_2\mathrm{projection}\right).$$

(50)

8.4. Cumulant Expansion and Step–Down Rule

The averaged heat kernel satisfies

$$〈e^{-s{(D_0+\delta D)}^2}〉=e^{-s{D}_0^2}-s〈\delta D{e}^{-s{D}_0^2}\delta D〉+\mathcal{O}\left({\sigma }^4\right).$$

Evaluating the second cumulant with (50) and comparing with the standard heat–kernel expansion yield the step–down formula announced in Section 6:

$$\begin{array}{|c|}\hline {〈\delta a_n〉}_{\mathrm{stoch}}={\sigma }^2{\Lambda }^2{a}_{n-2}\left[K_{{\mu }_c}*D_0^2\right],n\ge 2.\\ \hline\end{array}$$

(51)

Thus $a_4$ induces an $\mathcal{O}\left({\sigma }^2\right)$ correction to $a_2$, precisely what was required in Section 6 to sustain the spring and vortex profiles.

8.5. $\mathcal{PT}$ Symmetry and Pseudo–Hermiticity

Equation (48) is manifestly $\mathcal{PT}$–even because $e_{1,2}$ are $\mathcal{PT}$–even (Tab. Table 1) and $K_{{\mu }_c}$ is real. Consequently the stochastic process (49) respects both global $\mathcal{PT}$ symmetry and the pseudo–Hermiticity of D established in Section 5. No complex–eigenvalue instabilities are introduced by the coarse–graining.

Summary

• Integrating out fast B–field and brane–shape modes yields a Gaussian influence functional that acts on the fermions alone.
• The resulting noise kernel is white up to the scale ${\mu }_c$ and projects exclusively onto the quaternion axes $e_{1,2}$.
• The stochastic shift $D\mapsto D+\delta D$ produces the step–down relation (51), underpinning the ${cos}^2\left(\omega t\right)$ and $r^{-2}$ activators of Section 6.
• $\mathcal{PT}$ symmetry and pseudo–Hermiticity survive the coarse–graining, ensuring a stable spectral expansion.

9. Minimal Phenomenological Window

The quaternionic – $\mathcal{PT}$–symmetric framework developed in Section 1–8 is governed, at leading order, by only two dimensionless “geometric activators”

$${ϵ}_0\left[\mathrm{time}--\mathrm{like}\text{``}\mathrm{spring}\text{’’}\right],{ϵ}_1\left[\mathrm{space}--\mathrm{like}\text{``}\mathrm{vortex}\text{’’}\right],$$

(52)

entering the open–string metric as $G_{\mu \nu }={\eta }_{\mu \nu }+{ϵ}_{0}cos\omega t{e}_1{T}_{\mu \nu }+{ϵ}_1{r}^{-1}{e}_2{R}_{\mu \nu }$. All higher coefficients are radiatively stable (Section 7). We therefore speak of a minimal phenomenological window spanned by $({ϵ}_0,{ϵ}_1)$.

9.1. Current Laboratory & Astrophysical Bounds

Table 2 collects the tightest constraints available to date. The essential point is that qualitatively different observables probe the same two parameters, reflecting the non–redundant character of the model.

Two remarks are in order:

(i)

Orthogonality of probes. Cosmic–microwave and atomic data constrain ${ϵ}_0$, while galactic dynamics and GW polarimetry constrain ${ϵ}_1$, making the parameter disentanglement clean.

(ii)

Radiative stability. Since ${ϵ}_{0,1}$ are protected against logarithmic running (Section 7.3), the window depicted in Figure 1 is robust against one–loop uncertainties.

9.2. Benchmark Slice and Correlated Signals

We adopt as working benchmark

$$({ϵ}_0,{ϵ}_1)=\left(5\times {10}^{-8},10\right),\omega ={10}^{-6}\mathrm{Hz},$$

(53)

which comfortably satisfies all bounds in Table 2. Three immediate, correlated predictions follow:

• CMB high–ℓ ripples An $\mathcal{O}\left({10}^{-8}\right)$ modulation in the Sachs–Wolfe plateau for $\ell \sim 200$–1200; testable by the Simons Observatory within five observing seasons.
• GW polarisation splitting A $\sim 0.03$rad helicity phase delay for $f\simeq 100$ Hz signals propagating over $\mathcal{O}\left(100\right)$Mpc; within reach of ET/CE network cross–correlations.
• Sub–nHz Larmor drifts A $1$–2 nHz shift in nuclear spin precession for ${\nu }_0\sim 1$ GHz systems; detectable by the five–year CASPEr–Wind upgrade.

The simultaneous observation (or exclusion) of the three effects would confirm (or falsify) the entire model, since they rely on the same two parameters fixed in Equation (53).

9.3. Prospects for the Next Decade

(1)

2025 – 27 (Stage I) CMB high–ℓ data and MAGIS–100 reduce the viable band for ${ϵ}_0$ by an additional factor of 30.

(2)

2027 – 30 (Stage II) Global N–body campaigns (Gadget–4 class) and SKA – HI rotation curves push the ${ϵ}_1$ bound below $\mathcal{O}\left(1\right)$.

(3)

2030 – 34 (Stage III) Third–generation GW detectors deliver decisive polarisation measurements; a single $5\sigma$ detection at $\Delta {\varphi }_{LR}>0.02$ rad would determine ${ϵ}_1$ to $\pm 15\%$.

Take–Away

The minimal phenomenological window spanned by $({ϵ}_0,{ϵ}_1)$ is already bounded to

$$\left|{ϵ}_0\right|\lesssim {10}^{-7},\left|{ϵ}_1\right|\lesssim 35.$$

Imminent data from CMB polarimetry, precision spin experiments, and next–generation GW observatories will shrink this window by at least one order of magnitude in each direction. Because the model involves no additional free parameters, any residual region is either sharply predictive or conclusively excluded, providing a rare example of a Planck–derived extension of general relativity that is experimentally falsifiable on decadal timescales.

10. Conclusions

The programme developed in Section 1–9 establishes a closed logical chain that connects Type–IIB D3–brane physics to observationally testable extensions of four–dimensional space–time. The construction is anchored on two pillars: (i) a quaternionic, SU(2)–valued deformation of the open–string metric and (ii) a $\mathcal{PT}$–symmetric prescription that renders the corresponding Dirac operator pseudo–Hermitian. Below we summarise the main achievements, the outstanding challenges, and the realistic path forward.

10.1. Achievements

(1)

First–principle derivation. Starting from the non–abelian DBI+CS action, a long–wavelength NS–NS two–form produces exactly two SU(2)–aligned perturbations, ${ϵ}_{0}cos\omega t{e}_1$ and ${ϵ}_1{r}^{-1}{e}_2$ (Section 2).

(2)

Quaternionic Clifford extension & $\mathcal{PT}$ rule. The resulting metric admits a $Cliff(1,3)\otimes \mathbb{H}$ bundle and a Dirac operator that is simultaneously pseudo–Hermitian and $\mathcal{PT}$–invariant (Section 4 and Section 5).

(3)

Heat–kernel emergence of activators. The linearised Seeley–DeWitt densities $a_0,a_2$ reproduce the $cos\omega t$ and $r^{-1}$ profiles without extra assumptions (Section 6).

(4)

Radiative stability. A single local counter–term, $-{\Pi }_{1^{st}Q}\left[a_4\right]$, cancels the linear–quaternion anomaly and leaves the renormalisation group flow of scalar couplings untouched (Section 7).

(5)

Microscopic origin of stochasticity. Coarse–graining the brane path integral yields the Gaussian noise kernel that underpins the $\delta D$ correction and the “step–down” rule for heat–kernel coefficients (Section 8).

(6)

Falsifiable two–parameter window. All phenomenology is controlled by the minimal set $({ϵ}_0,{ϵ}_1)$; present data already constrain $\left|{ϵ}_0\right|\lesssim {10}^{-7}$ and $\left|{ϵ}_1\right|\lesssim 35$, while upcoming experiments can tighten both bounds by at least an order of magnitude (Section 9).

10.2. Outstanding Problems

Two–loop consistency.

A full two–loop computation of the spectral action is needed to verify the conjectured uniqueness of the counter–term ${\Pi }_{1^{st}Q}\left[a_4\right]$.

Non–linear solutions.

Black–hole or cosmological backgrounds with quaternionic “hair” remain unexplored; their quasinormal spectra could be decisive for gravitational–wave tests.

Lattice implementation.

Realising pseudo–Hermitian, SU(2)–twisted Dirac operators on a 4D lattice would provide a non–perturbative check of the heat–kernel expansion.

Quantum–information channels.

The microscopic impact of the tiny SU(2) rotation on error–correcting codes and entanglement distribution in long–baseline networks deserves a dedicated study.

10.3. Decadal Experimental Outlook

Milestone	Target	Forecast year
CMB high–ℓ (Simons Observatory)	$\delta {ϵ}_0\sim {10}^{-8}$	2025
MAGIS–100 sub–nHz phase run	$\left|{ϵ}_0\right|<{10}^{-9}$	2027
N–body LSB halo suite (Gadget–4)	$\left|{ϵ}_1\right|<3$	2028
Einstein Telescope GW birefringence	$\left|{ϵ}_1\right|<0.3$	2031

A positive detection in any of the above channels would immediately pin down the corresponding parameter with $\lesssim 15\%$ precision, while a consistent sequence of null results would exclude the model altogether—a level of falsifiability rare among Planck–scale extensions of general relativity.

Final Remark

The two–parameter deformation $\left\{{ϵ}_0,{ϵ}_1\right\}$ offers an economical gateway from string–theoretic first principles to observable physics across more than 30 orders of magnitude in length scale. Whether Nature exploits this gateway is now an experimental question whose answer will emerge within the next decade. Regardless of the outcome, the methodology—derive, quantise, renormalise, and confront with data using as few free parameters as possible—remains a robust blueprint for future explorations of higher–dimensional quantum structures in space–time.

“Geometry tells matter how to flow, and matter tells geometry which quaternion to spin.”

Appendix A.1. General Determinant Expansion

For a small matrix perturbation $\delta M_{\mu \nu }$ one has

$$det(M+\delta M)=detM\left[1+\mathrm{Tr}\left(M^{-1}\delta M\right)+\frac{1}{2}{\left\{\mathrm{Tr}\left(M^{-1}\delta M\right)\right\}}^2-\frac{1}{2}\mathrm{Tr}\left[{\left(M^{-1}\delta M\right)}^2\right]+\mathcal{O}\left(\delta M^3\right)\right].$$

(A1)

Setting $M_{\mu \nu }={\eta }_{\mu \nu }$ and $\delta M_{\mu \nu }=2\pi {\alpha }^{\prime }{B}_{\mu \nu }$ immediately yields

$$\mathcal{D}=det\eta \left[1+2\pi {\alpha }^{\prime }\underset{=0}{\underbrace{\mathrm{Tr}\left(B\right)}}-{\left(2\pi {\alpha }^{\prime }\right)}^2\frac{1}{2}\mathrm{Tr}\left(B^2\right)+\mathcal{O}\left({\left({\alpha }^{\prime }B\right)}^3\right)\right].$$

Because $B_{\mu \nu }$ is antisymmetric, $\mathrm{Tr}\left(B\right)=0$ identically; the leading non-trivial contribution is therefore quadratic in B.

Appendix A.2. Insertion of the Two–Form Profile

Using the background profile

$$B_{\mu \nu }\left(x\right)=\left({\overline{B}}_{0i}+\delta B_{0i}cos\omega t\right){\delta }_{[\mu }^0{\delta }_{\nu ]}^i+\beta {\displaystyle \frac{{ϵ}_{ijk}{x}^k}{r^3}}{\delta }_{[\mu }^i{\delta }_{\nu ]}^j,$$

and adopting the shorthand $E_i\equiv {\overline{B}}_{0i}+\delta B_{0i}cos\omega t$, $M_{ij}\equiv \beta {ϵ}_{ijk}{x}^k/r^3$, we compute

$$\mathrm{Tr}\left(B^2\right)=2\left(E_i{E}_i+M_{ij}{M}_{ij}\right)=2\left[{\left({\overline{B}}_{0i}+\delta B_{0i}cos\omega t\right)}^2+2{\beta }^2{r}^{-2}\right].$$

Inserting this into (A1) and keeping terms up to $\mathcal{O}\left({\alpha }^{\prime 2}{B}^2\right)$ yields

$$\sqrt{-\mathcal{D}}=1-{\left(2\pi {\alpha }^{\prime }\right)}^2\left[\frac{1}{2}{\left({\overline{B}}_{0i}+\delta B_{0i}cos\omega t\right)}^2+{\beta }^2{r}^{-2}\right]+\mathcal{O}\left({\left({\alpha }^{\prime }B\right)}^3\right).$$

(A2)

Up to an overall normalisation absorbed into the brane tension $T_3$, (A2) reproduces the determinant factor used in Equation (1) of the main text.

Appendix A.3. Extraction of the Open–String Metric

Comparing the DBI action

$$S_{\mathrm{DBI}}=-T_3\int d^{4}x{e}^{-\varphi }\sqrt{-\mathcal{D}},$$

with the general open-string form $-T_3\int d^{4}x\sqrt{-det{\mathcal{G}}_{\mu \nu }}$, and identifying the square brackets in Equation (A2) with $-\frac{1}{2}{({\mathcal{G}}^{-1}-\eta )}^{\mu \nu }{({\mathcal{G}}^{-1}-\eta )}_{\mu \nu }$, one reads off, to linear order in B,

$$\begin{array}{|c|}\hline {\mathcal{G}}_{\mu \nu }={\eta }_{\mu \nu }+\left(2\pi {\alpha }^{\prime }\right)\delta B_{0i}cos\omega t{e}_1{T}_{\mu \nu }+\left(2\pi {\alpha }^{\prime }\right)\beta {\displaystyle \frac{1}{r}}{e}_2{R}_{\mu \nu }+\mathcal{O}\left({\left({\alpha }^{\prime }B\right)}^2\right).\\ \hline\end{array}$$

(A3)

Equation (A3) coincides with Equation (7) in the main text, thus completing the derivation.

Appendix B.1. Minimal P and T Prescriptions

Throughout we work in flat Minkowski conventions ${\eta }_{\mu \nu }=\mathrm{diag}(+,-,-,-)$ and fix the imaginary quaternion basis $\{e_1,e_2,e_3\}$ with $e_a{e}_b=-{\delta }_{ab}+{ϵ}_{abc}{e}_c$ and $e_a^{†}=-e_a$. The minimal $\mathcal{P}$ and $\mathcal{T}$ actions are

$$\mathcal{P}:\left(t,\mathbf{x};i;e_a\right)⟼\left(t,-\mathbf{x};i;-e_a\right),\mathcal{T}:\left(t,\mathbf{x};i;e_a\right)⟼\left(-t,\mathbf{x};-i;-e_a\right),$$

(A4)

where $\mathcal{T}$ is anti-linear6. The composite $\mathcal{PT}\equiv \mathcal{T}\mathcal{P}$ is therefore anti-linear and leaves $e_a$ unchanged: $\mathcal{PT}:e_a\mapsto +e_a$.

Appendix B.2. Invariance of the Quaternionic Metric

Recall the linear quaternionic deformation ${\mathbb{G}}_{\mu \nu }={\eta }_{\mu \nu }+{ϵ}_{0}cos\left(\omega t\right)e_1{T}_{\mu \nu }+{ϵ}_1{r}^{-1}{e}_2{R}_{\mu \nu }$ with $T_{\mu \nu }={\delta }_{\mu }^0{\delta }_{\nu }^0$ and $R_{\mu \nu }={\delta }_{\mu }^i{\delta }_{\nu }^i$. Using (A4):

• $cos\left(\omega t\right)$ is even under $t\mapsto -t$ ($\mathcal{T}$);
• $r^{-1}={\left({\mathbf{x}}^2\right)}^{-1/2}$ is even under $\mathbf{x}\mapsto -\mathbf{x}$ ($\mathcal{P}$);
• $e_1,e_2$ are $\mathcal{PT}$–even.

Hence

$$\left(\mathcal{PT}\right){\mathbb{G}}_{\mu \nu }\left(x\right){\left(\mathcal{PT}\right)}^{-1}={\mathbb{G}}_{\mu \nu }\left(x\right)\forall x,$$

(A5)

proving Equation (18) of the main text.

Appendix B.3. Covariance of the Extended Clifford Algebra

The generators ${\mathrm{\Gamma }}_{\mu }={\gamma }_{\mu }\otimes {\mathbf{1}}_{\mathbb{H}}$ and ${\mathrm{\Gamma }}_{\left(a\right)}={\gamma }_{*}\otimes e_a$ ($a=1,2$), with ${\gamma }_{*}=i{\gamma }^0{\gamma }^1{\gamma }^2{\gamma }^3$, obey $\{{\mathrm{\Gamma }}_A,{\mathrm{\Gamma }}_B\}=2{\mathcal{G}}_{AB}$. Because (i) $\mathcal{P}{\gamma }^{\mu }{\mathcal{P}}^{-1}={\gamma }^{\mu }$, (ii) $\mathcal{P}{\gamma }_{*}{\mathcal{P}}^{-1}={\gamma }_{*}$, and (iii) $e_a$ are $\mathcal{PT}$–even, one finds

$$\left(\mathcal{PT}\right){\mathrm{\Gamma }}_A{\left(\mathcal{PT}\right)}^{-1}={\mathrm{\Gamma }}_A,A\in \{\mu ,\left(1\right),\left(2\right)\}.$$

(A6)

Thus the full $\mathrm{Cliff}(1,5)$ algebra is $\mathcal{PT}$–covariant.

Appendix B.4. Pseudo-Hermiticity of the Dirac Operator

Let $\eta :={\mathrm{\Gamma }}^0={\gamma }^0\otimes {\mathbf{1}}_{\mathbb{H}}$ (${\eta }^{†}={\eta }^{-1}=\eta$). For the quaternionic Dirac operator

$$D=i{\mathrm{\Gamma }}^{\mu }\left({\partial }_{\mu }+\frac{1}{4}{\Omega }_{\mu }^{AB}{\mathrm{\Gamma }}_A{\mathrm{\Gamma }}_B\right),$$

metric compatibility implies ${\Omega }_{\mu }^{AB}{\mathrm{\Gamma }}_A{\mathrm{\Gamma }}_B=-{\mathrm{\Gamma }}_B{\mathrm{\Gamma }}_A{\Omega }_{\mu }^{AB}$. Using $\eta {\mathrm{\Gamma }}^{\mu }{\eta }^{-1}={\mathrm{\Gamma }}^{\mu †}$ one obtains

$$D^{†}=\eta D{\eta }^{-1},$$

(A7)

establishing pseudo-Hermiticity. Because $\eta$ commutes with $e_a$, adding the stochastic anti-Hermitian $\delta D$ of Section 8 leaves (A7) intact.

Appendix B.5. PT-Invariance of the Dirac operator

Applying (A6) and noting $\left(\mathcal{PT}\right){\partial }_{\mu }{\left(\mathcal{PT}\right)}^{-1}={{\Lambda }_{\mu }}^{\nu }{\partial }_{\nu }$ with ${{\Lambda }_{\mu }}^{\nu }=\mathrm{diag}(-1,1,1,1)$, one finds

$$\left(\mathcal{PT}\right)D{\left(\mathcal{PT}\right)}^{-1}=i{\mathrm{\Gamma }}^{\nu }{\partial }_{\nu }=D.$$

Therefore D is both $\mathcal{PT}$–invariant and pseudo-Hermitian, so its eigenvalues are real or appear in complex-conjugate pairs, as required for the heat-kernel expansion in Section 6.

Notation.

We retain explicitly only linear terms in the SU(2) activation parameters ${ϵ}_0,{ϵ}_1$ introduced in Equation (16); quadratic pieces first contribute to $a_4$. The $\mathcal{PT}$–even projector ${\mathbf{P}}_Q$ defined in Appendix E is tacitly applied whenever a “linear–quaternion slice’’ is mentioned.

Appendix C.1. Laplace form of D 2

For the $\mathcal{PT}$–invariant Dirac operator $D=i{\mathrm{\Gamma }}^{\mu }\left({\partial }_{\mu }+{\displaystyle \frac{1}{4}}{\Omega }_{\mu }^{AB}{\mathrm{\Gamma }}_A{\mathrm{\Gamma }}_B\right)$ built from the quaternionic metric ${\mathbb{G}}_{\mu \nu }$ of Equation (16), one may rewrite

$$D^2=-{\mathbb{G}}^{\mu \nu }{\nabla }_{\mu }{\nabla }_{\nu }+E,E=\frac{1}{4}\mathcal{R}\mathbf{1}+\mathrm{\Phi },$$

(A8)

where ${\nabla }_{\mu }={\partial }_{\mu }+{\displaystyle \frac{1}{4}}{\Omega }_{\mu }^{AB}{\mathrm{\Gamma }}_A{\mathrm{\Gamma }}_B,$ $\mathcal{R}$ is the Ricci scalar of ${\mathbb{G}}_{\mu \nu }$, and $\mathrm{\Phi }={\displaystyle \frac{1}{8}}{\mathrm{\Gamma }}^{\mu \nu }{\Omega }_{\mu \nu },{\mathrm{\Gamma }}^{\mu \nu }:={\displaystyle \frac{1}{2}}[{\mathrm{\Gamma }}^{\mu },{\mathrm{\Gamma }}^{\nu }].$

Appendix C.2. Seeley–DeWitt Master Formulas

For a Laplace–type operator of the form (A8) on a smooth four-manifold $\mathcal{M}$ without boundary the first two coefficients are

$$\begin{array}{cc}\hfill a_0& ={\displaystyle \frac{1}{16{\pi }^2}}{\int }_{\mathcal{M}}{d}^{4}x\sqrt{\left|\mathbb{G}\right|}{Tr}_{(S\otimes \mathbb{H})}\mathbf{1},\hfill \end{array}$$

(A9)

$$\begin{array}{cc}\hfill a_2& ={\displaystyle \frac{1}{16{\pi }^2}}{\int }_{\mathcal{M}}{d}^{4}x\sqrt{\left|\mathbb{G}\right|}{Tr}_{(S\otimes \mathbb{H})}\left(\frac{1}{6}\mathcal{R}\mathbf{1}+\mathrm{\Phi }\right).\hfill \end{array}$$

(A10)

The total trace factorises as ${Tr}_{S\otimes \mathbb{H}}={Tr}_S{Tr}_{\mathbb{H}}$; recall ${Tr}_{\mathbb{H}}\left(e_a\right)=0$ and ${Tr}_{\mathbb{H}}\left(e_a{e}_b\right)=-2{\delta }_{ab}$.

Appendix C.3. Evaluation of a 0

Since $\sqrt{\left|\mathbb{G}\right|}=1+\mathcal{O}\left({ϵ}^2\right)$ and ${Tr}_S\mathbf{1}=4$,

$$a_0={\displaystyle \frac{4{Tr}_{\mathbb{H}}\mathbf{1}}{16{\pi }^2}}\int d^{4}x+\mathcal{O}\left({ϵ}^2\right)={\displaystyle \frac{8}{16{\pi }^2}}\mathrm{Vol}\left({\mathbb{R}}^{1,3}\right)+\mathcal{O}\left({ϵ}^2\right),$$

i.e. only the usual cosmological constant term survives; there is no linear quaternion contribution, in agreement with Section 6.

Appendix C.4. Evaluation of a 2

Curvature part.

At $\mathcal{O}\left(ϵ\right)$ one finds $\mathcal{R}={ϵ}_{0}cos\left(\omega t\right)e_1{\partial }_i{\partial }_i{T^i}_i+{ϵ}_1{r}^{-1}{e}_2{\partial }_i{\partial }_i{R^i}_i=0$ because ${T^i}_i={R^i}_i=0$. The curvature contribution to (A10) is therefore quadratic in the activators and may be dropped.

Endomorphism part.

A direct contraction at linear order gives

$$E=-\frac{1}{2}{ϵ}_{0}cos\left(\omega t\right)e_1{T}_{\mu \nu }{\mathrm{\Gamma }}^{\mu }{\mathrm{\Gamma }}^{\nu }-\frac{1}{2}{ϵ}_1{r}^{-1}{e}_2{R}_{\mu \nu }{\mathrm{\Gamma }}^{\mu }{\mathrm{\Gamma }}^{\nu }+\mathcal{O}\left({ϵ}^2\right).$$

Using ${Tr}_S\left({\mathrm{\Gamma }}^{\mu }{\mathrm{\Gamma }}^{\nu }\right)=4{\mathbb{G}}^{\mu \nu }$ and keeping only the linear quaternion slice (see Appendix E) one obtains

$${\left[{Tr}_{(S\otimes \mathbb{H})}E\right]}_{1Q}=-8\left({ϵ}_{0}cos\left(\omega t\right)e_1+{ϵ}_1{r}^{-1}{e}_2\right)+\mathcal{O}\left({ϵ}^2\right).$$

Inserting into (A10) yields

$$a_2=-{\displaystyle \frac{8}{16{\pi }^2}}\int d^{4}x\left({ϵ}_{0}cos\left(\omega t\right)e_1+{ϵ}_1{r}^{-1}{e}_2\right)+\mathcal{O}\left({ϵ}^2\right).$$

(A11)

Equation (A11) is precisely the result quoted in Section 6: only the time-like spring ${ϵ}_{0}cos\omega t$ and the spatial vortex ${ϵ}_1/r$ survive after the internal trace at leading order.

Appendix C.5. PT Covariance

Both ${ϵ}_{0}cos\omega t{e}_1$ and ${ϵ}_1{r}^{-1}{e}_2$ are individually $\mathcal{PT}$–even (Appendix B); hence the integrated quantity $a_2$ is $\mathcal{PT}$–invariant. This guarantees that the effective Lagrangian derived in Section 6.3 respects the global symmetry of the model.

Appendix D.1. System–Bath Decomposition

We split the Type–IIB world–sheet fields as in Equation (45): $B=B_{\mathrm{slow}}+\tilde{B},X=X_{\mathrm{slow}}+\tilde{X}.$ The total action separates into

$$\mathcal{S}=\underset{\text{``}\mathrm{classical}\mathrm{geometry}\text{’’}}{\underbrace{{\mathcal{S}}_{\mathrm{slow}}[B_{\mathrm{slow}},X_{\mathrm{slow}}]}}+\underset{\mathrm{fast}\mathrm{bath}}{\underbrace{{\mathcal{S}}_{\mathrm{fast}}[\tilde{B},\tilde{X}]+{\mathcal{S}}_{\mathrm{int}}[\psi ,\overline{\psi };\tilde{B},\tilde{X}]}}+{\mathcal{S}}_D[\psi ,\overline{\psi };G(B_{\mathrm{slow}},X_{\mathrm{slow}})].$$

(A12)

The bath couples to the fermions through ${\mathcal{S}}_{\mathrm{int}}=\int d^{4}x{\tilde{J}}^A\left(x\right)\overline{\psi }\left(x\right){\mathrm{\Gamma }}_A\psi \left(x\right),$ where ${\tilde{J}}^A$ is linear in the fast fluctuations $\{\tilde{B},\tilde{X}\}$ and $A\in \{\mu ,(1),(2\left)\right\}$ labels the extended Clifford basis of Equation (17).

Appendix D.2. Bath Integration

Assuming the fast sector is in a Gaussian state ${\rho }_{\mathrm{fast}}$ at the coarse–graining scale ${\mu }_c$, the influence functional becomes

$$e^{-{\mathcal{S}}_{\mathrm{infl}}[\psi ,\overline{\psi }]}={〈e^{-{\mathcal{S}}_{\mathrm{int}}[\psi ,\overline{\psi };\tilde{J}]}〉}_{{\rho }_{\mathrm{fast}}}=exp\left(-\frac{1}{2}\int \int d^{4}x{d}^{4}y\overline{\psi }\left(x\right){\mathrm{\Gamma }}_A\psi \left(x\right){\Sigma }^{AB}(x-y)\overline{\psi }\left(y\right){\mathrm{\Gamma }}_B\psi \left(y\right)+\dots \right),$$

(A13)

where all odd moments vanish and ${\Sigma }^{AB}\left(z\right):={\langle {\tilde{J}}^A\left(z\right){\tilde{J}}^B\left(0\right)\rangle }_{{\rho }_{\mathrm{fast}}}.$

Rotational symmetry of the bath.

Because the fast bath is generated by small fluctuations around a flat D3–brane, the correlator depends only on $z^2$ and is diagonal in the internal quaternion indices:

$${\Sigma }^{AB}\left(z\right)={\sigma }^2{K}_{{\mu }_c}\left(z\right){\delta }_{\mathrm{Cliff}}^{AB}+\mathcal{O}\left({\epsilon }^2\right),K_{{\mu }_c}\left(z\right):={\displaystyle \frac{e^{-{\mu }_c^2{z}^2}}{{\mu }_c^4}}.$$

The dimensionless strength ${\sigma }^2\sim g_s^2{\left({\alpha }^{\prime }\right)}^2{\mu }_c^2/{\Lambda }_{\mathrm{UV}}^2$ is extracted by matching to the microscopic two–point function of $\tilde{B}$.

Appendix D.3. Hubbard–Stratonovich Representation

The quartic term in (A13) is linearised via an auxiliary, anti–Hermitian matrix field ${\Xi }_A\left(x\right)$:

$$e^{-{\mathcal{S}}_{\mathrm{infl}}}=\int \mathcal{D}\Xi exp\left[-\frac{1}{2}\int \int d^{4}x{d}^{4}y{\Xi }_A\left(x\right){\left[{\Sigma }^{-1}\right]}^{AB}(x-y){\Xi }_B\left(y\right)+i\int d^{4}x{\Xi }_A\left(x\right)\overline{\psi }\left(x\right){\mathrm{\Gamma }}^A\psi \left(x\right)\right].$$

(A14)

Because ${\left[{\Sigma }^{-1}\right]}^{AB}\propto {\delta }^{AB}$, ${\Xi }_A$ can be expanded on the same Clifford basis. Identifying $i{\Xi }_A{\mathrm{\Gamma }}^A\equiv -\delta D,$ the fermionic path integral becomes Gaussian:

$$\int \mathcal{D}\psi \mathcal{D}\overline{\psi }{e}^{-\int \overline{\psi }(D_0+{\epsilon }_0{f}_1{e}_1+{\epsilon }_1{f}_2{e}_2+\delta D)\psi }={det}^{1/2}[D_0+{\epsilon }_0{f}_1{e}_1+{\epsilon }_1{f}_2{e}_2+\delta D].$$

(A15)

Appendix D.4. Statistics of δD

Integrating out ${\Xi }_A$ with the Gaussian weight in (A14) yields9

$$\begin{array}{cc}\hfill {〈\delta D\left(x\right)〉}_{\mathrm{stoch}}& =0,\hfill \end{array}$$

(A16)

$$\begin{array}{cc}\hfill {〈\delta D\left(x\right)\delta D\left(y\right)〉}_{\mathrm{stoch}}& ={\sigma }^2{K}_{{\mu }_c}(x-y)\left(e_1\otimes e_1+e_2\otimes e_2\right),\hfill \end{array}$$

(A17)

reproducing Equation (50) of the main text. The stochastic process lives entirely in the two $\mathcal{PT}$-even quaternion directions and is white up to the cut–off ${\mu }_c$.

Appendix D.5. Cumulant Expansion and Step–Down Rule

Expanding the fermionic determinant in (A15) around $\delta D=0$:

$${det}^{1/2}\left[D+\delta D\right]=exp\left\{\frac{1}{2}TrlnD+\frac{1}{2}Tr\left[D^{-1}\delta D\right]-\frac{1}{4}Tr\left[D^{-1}\delta D{D}^{-1}\delta D\right]+\dots \right\}.$$

Because $\langle \delta D\rangle =0$, the first non–trivial contribution arises at quadratic order and shifts the heat–kernel coefficients according to

$${〈\delta a_n〉}_{\mathrm{stoch}}={\sigma }^2{\Lambda }^2{a}_{n-2}[K_{{\mu }_c}*D_0^2],n\ge 2,$$

(A18)

which is Equation (51) of Section 8. Equation (A18) justifies the hierarchy employed in Section 6: every deterministic coefficient $a_n$ feeds a stochastic correction to $a_{n-2}$, suppressed by ${\sigma }^2{\Lambda }^2$.

Appendix E.1. Internal Trace and Orthogonality

The Clifford–quaternion Hilbert space factorises as $\mathcal{H}=L^2({\mathbb{R}}^{1,3},S)\otimes \mathbb{H},$ so that any operator $\mathsf{O}$ acting on $\mathcal{H}$ decomposes as

$$\mathsf{O}=\sum _{A=0}^3{\mathsf{O}}^{\left(A\right)}\otimes e_A,{\mathsf{O}}^{\left(A\right)}:L^2({\mathbb{R}}^{1,3},S)\to L^2({\mathbb{R}}^{1,3},S).$$

(A19)

The internal trace ${Tr}_{\mathrm{int}}$ acts only on the quaternion factor: ${Tr}_{\mathrm{int}}\left(e_0\right)=2,{Tr}_{\mathrm{int}}\left(e_a\right)=0,{Tr}_{\mathrm{int}}\left(e_a{e}_b\right)=-2{\delta }_{ab}.$ Hence the orthogonality relations

$${Tr}_{\mathrm{int}}\left(e_A{e}_B\right)=2{\eta }_{AB}^{\phantom{0}},{\eta }_{AB}\equiv \mathrm{diag}(+1,-1,-1,-1),A,B=0,1,2,3,$$

(A20)

provide the metric on quaternion space.

Appendix E.2. Definition of the Projector

Given (A19) the linear-quaternion slice is

$${\left[\mathsf{O}\right]}_{1Q}:=\sum _{a=1}^2{\mathsf{O}}^{\left(a\right)}{e}_a,$$

(A21)

i.e. we retain only the components along $e_1$ and $e_2$, which are selected by the physical background (cf. Equation (4)). The projector ${\mathbf{P}}_Q$ acts as

$${\mathbf{P}}_Q\left[\mathsf{O}\right]={\displaystyle \frac{1}{2}}\sum _{a=1}^3{e}_a{Tr}_{\mathrm{int}}\left(e_a\mathsf{O}\right),{\mathbf{P}}_Q^2={\mathbf{P}}_Q,{Tr}_{\mathrm{int}}\left[e_a{\mathbf{P}}_Q\left[\mathsf{O}\right]\right]={Tr}_{\mathrm{int}}\left[e_a\mathsf{O}\right].$$

(A22)

With (A20) one checks explicitly ${\mathbf{P}}_Q\left[e_0\right]=0,{\mathbf{P}}_Q\left[e_{1,2}\right]=e_{1,2},{\mathbf{P}}_Q\left[e_3\right]=0.$

Appendix E.3. Commutation with PT

Using the transformation rules in Appendix B, both $e_1$ and $e_2$ are $\mathcal{PT}$-even, while $e_3$ is $\mathcal{PT}$-odd. Therefore, for any operator $\mathsf{O}$,

$$\left(\mathcal{PT}\right){\mathbf{P}}_Q\left[\mathsf{O}\right]{\left(\mathcal{PT}\right)}^{-1}={\mathbf{P}}_Q\left[\left(\mathcal{PT}\right)\mathsf{O}{\left(\mathcal{PT}\right)}^{-1}\right],$$

(A23)

i.e. ${\mathbf{P}}_Q$ commutes with the global symmetry and does not spoil pseudo-Hermiticity.

Appendix E.4. Quadratic Identities

When evaluating heat–kernel densities and Noether currents one often encounters products such as ${\left[{\mathsf{O}}_1\right]}_{1Q}{\left[{\mathsf{O}}_2\right]}_{1Q}$. Using (A21) and the quaternion algebra:

$$\begin{array}{cc}\hfill {\mathbf{P}}_Q\left[{\left[{\mathsf{O}}_1\right]}_{1Q}{\left[{\mathsf{O}}_2\right]}_{1Q}\right]& =-\sum _{a=1}^2\left({\mathsf{O}}_1^{\left(a\right)}{\mathsf{O}}_2^{\left(a\right)}\right)e_0+e_3\left({\mathsf{O}}_1^{\left(1\right)}{\mathsf{O}}_2^{\left(2\right)}-{\mathsf{O}}_1^{\left(2\right)}{\mathsf{O}}_2^{\left(1\right)}\right),\hfill \end{array}$$

(A24)

from which three important facts follow:

(i)

The scalar component (proportional to $e_0$) never contributes to the $1Q$ slice: it disappears after the projection and hence cannot spoil current conservation.

(ii)

The $e_3$ component is $\mathcal{PT}$-odd and is therefore eliminated whenever the integrand is constrained to be $\mathcal{PT}$-even (e.g. in the heat–kernel densities).

(iii)

As a result, products of two linear-quaternion operators do not re–enter the $1Q$ sector—an algebraic reason why a single counter-term ${\Pi }_{1Q}\left[a_4\right]$ suffices to cancel the anomaly at all loops (Section 7).

Appendix E.5. Trace Identities for Heat–Kernel Coefficients

Let $D_0$ denote the background Dirac operator and $\mathrm{\Phi }$, ${\Omega }_{\mu \nu }$ the endomorphism and curvature defined in Appendix C. Using (A22) one proves the selection rule

$${Tr}_{\mathrm{int}}\left[e_{a}F(\mathrm{\Phi },{\Omega }_{\mu \nu })\right]=0,F\mathrm{any}\mathrm{polynomial}\mathrm{in}\mathrm{\Phi }\mathrm{and}\Omega ,a=1,2.$$

(A25)

Consequently ${\Pi }_{1Q}\left[a_0\right]={\Pi }_{1Q}\left[a_2\right]=0,$ while ${\Pi }_{1Q}\left[a_4\right]\ne 0$ due to the mixed $R^2-R_{\mu \nu }^2+\dots$ structures. Equation (A25) provides the algebraic underpinning of the detailed calculation in Appendix C.

Appendix F.1. Notation and Renormalisation Scheme

We employ dimensional regularisation in $d=4-ϵ$ and adopt the MS subtraction convention. The bare (B) and renormalised (R) quantities are related by

$$g_s^{\mathrm{B}}={\mu }^{ϵ/2}{Z}_g^{-1}{g}_s,{\Lambda }^{\mathrm{B}}=Z_{\Lambda }^{-1}\Lambda ,{\left(f_n\right)}^{\mathrm{B}}=Z_{f_n}^{-1}{f}_n,n=0,2,4,$$

(A26)

where the Z–factors are expanded as $Z_x=1+\frac{z_x^{\left(1\right)}}{ϵ}+\mathcal{O}\left({ϵ}^{-2}\right)$. All loop integrals are evaluated with the $\mathcal{PT}$–even projector ${\mathbf{P}}_Q$ implicit.

Appendix F.2. Decomposition of the Divergent Action

The one–loop effective action can be written as

$${\mathrm{\Gamma }}_{1\mathrm{L}}=\sum _{n=0}^2{\displaystyle \frac{1}{ϵ}}{c}_{2n}{\mu }^{ϵ}{\Lambda }^{4-2n}{a}_{2n}\left[D^2\right]+{\mathrm{\Gamma }}_{\mathrm{fin}},c_0\equiv {\beta }_0,c_2\equiv {\beta }_2,c_4\equiv {\beta }_4,$$

(A27)

with $a_{2n}$ the Seeley–DeWitt densities of Appendix C. Splitting $a_4$ into its scalar and linear–quaternion parts, $a_4=a_4^{\left(0\right)}+a_4^{\left(1Q\right)},$ the divergent Lagrangian reads

$${\mathcal{L}}_{\mathrm{div}}={\displaystyle \frac{{\mu }^{ϵ}}{ϵ}}\left\{{\beta }_0{\Lambda }^4{a}_0+{\beta }_2{\Lambda }^2{a}_2+{\beta }_4\left[a_4^{\left(0\right)}+a_4^{\left(1Q\right)}\right]\right\}.$$

(A28)

Appendix F.3. Renormalisation Constants

Scalar sector.

Matching (A28) with the tree–level coefficients fixes

$$z_{f_0}^{\left(1\right)}={\beta }_0,z_{f_2}^{\left(1\right)}={\beta }_2,z_{\Lambda }^{\left(1\right)}=\frac{1}{2}{\beta }_2.$$

(A29)

The explicit values ${\beta }_0=\frac{1}{2{\left(4\pi \right)}^2},{\beta }_2=\frac{5}{3{\left(4\pi \right)}^2}$ follow from the standard heat–kernel trace.

Linear–quaternion sector.

Demanding the cancellation of ${\mathbf{P}}_Q\left[{\mathcal{L}}_{\mathrm{div}}\right]$ gives the unique solution

$$\begin{array}{|c|}\hline {\beta }_4=1,z_{f_4}^{\left(1\right)}=1.\\ \hline\end{array}$$

(A30)

All remaining Z–factors coincide with their scalar counterparts, i.e. there is no extra renormalisation of the quaternion axes $e_{1,2}$.

Appendix F.4. One–Loop β–Functions

The renormalised couplings run according to ${\beta }_x=\mu {\displaystyle \frac{d}{d\mu }}x\left(\mu \right)=-ϵx+x\mu {\displaystyle \frac{d}{d\mu }}ln{Z}_x^{-1}.$ From (A29)–(A30) we obtain

$$\begin{array}{cccc}\hfill {\beta }_{g_s}& =-{\displaystyle \frac{ϵ}{2}}{g}_s+\mathcal{O}\left(g_s^3\right),\hfill & \hfill {\beta }_{\Lambda }& =-\Lambda \left(1-\frac{1}{2}{\beta }_2\right)+\mathcal{O}\left(\Lambda g_s^2\right),\hfill \end{array}$$

(A31)

$$\begin{array}{cccc}\hfill {\beta }_{f_4}& =-ϵf_4-f_4,\hfill & \hfill {\beta }_Q& =0,\hfill \end{array}$$

(A32)

where ${\beta }_Q$ denotes the running of the $\mathcal{PT}$–even linear–quaternion strength $Z_Q\equiv Z_{f_4}^{-1}{f}_4$. Equation () confirms the decoupling of the quaternionic sector claimed in Section 7.

Appendix F.5. Scheme Independence of β 4

Because $a_4^{\left(1Q\right)}$ is the only divergent operator carrying a linear quaternion index, any admissible subtraction scheme satisfies

$${\left({\mathcal{L}}_{\mathrm{ct}}\right)}^{\left(1Q\right)}=-{\displaystyle \frac{1}{ϵ}}{\beta }_4{a}_4^{\left(1Q\right)}.$$

A finite change of scheme, ${ϵ}^{-1}\to {ϵ}^{-1}+c,$ shifts ${\beta }_4\to {\beta }_4-c{\beta }_4,$ but the requirement of exact Noether conservation (Section 6) forces the coefficient back to unity. Hence

$${\beta }_4=1⟹\mathrm{scheme}\mathrm{independent}\mathrm{at}\mathrm{one}\mathrm{loop}.$$

(A33)