The Partition Function in Biquaternionic Form

Introduction

The partition function

$$ Z = 2\,\mathrm{Sc}\big(\exp(-\beta\tilde H_\Psi)\big) $$

is the central object of equilibrium statistical mechanics, written here with the scalar part in place of the matrix trace $\operatorname{tr}\exp(-\beta H)$ that the standard text uses. It normalizes the thermal state, generates the free energy, and encodes the spectrum of the Hamiltonian. The question this article addresses is narrow and concrete: does $Z$ have a biquaternion interpretation, and does the split of the biquaternion algebra into the material and informational sectors force any modification of it?

The answer has two parts, and they should be kept apart from the outset.

The first part is that, for a Hermitian Hamiltonian — an element $\tilde H_\Psi$ of the informational sector $\mathbb{B}_{\Psi}$ — the partition function is exactly the ordinary one, and the algebra contributes a reading rather than a new object. The thermal operator $\exp(-\beta H)$ is itself an element of $\mathbb{B}_{\Psi}$; the Hermitian form is $\langle\tilde{H}_{\Psi,1},\tilde{H}_{\Psi,2}\rangle_{*}=\mathrm{Sc}(\tilde{H}_{\Psi,1}\tilde{H}_{\Psi,2}^{*})$, whose value doubled is $2\langle\tilde{H}_{\Psi,1},\tilde{H}_{\Psi,2}\rangle_{*} = 2\,\mathrm{Sc}(\tilde{H}_{\Psi,1}\tilde{H}_{\Psi,2}^{*})$; and $Z$ is therefore that scalar part, doubled. What the doubling drops is the imaginary vector part of $\exp(-\beta H)$. The material sector $\mathbb{B}_{\Phi}$ does not appear in the thermal operator at all in this case; it appears in the imaginary-time direction along which $\beta$ is measured. So the honest finding is negative in the sense the subject demands: the algebra does not produce a second, rival partition function, and a "biquaternionic partition function" distinct from $Z$ is not needed.

The second part is the only place where the sector split changes the answer. If the generator is permitted an $\mathbb{B}_{\Phi}$ component — that is, if $\tilde H_\Psi$ is allowed to be non-Hermitian — then $Z$ can cease to be real and become complex, and the free energy can acquire an imaginary part. This extension is not forced by the framework, because observables are defined to be Hermitian elements of $\mathbb{B}_{\Psi}$. It is recorded here as a boundary of the interpretation, not as a result.

The article proceeds as follows: the standard partition function and thermal state are recalled; twice the scalar part is identified with the matrix trace of the $2\times2$ image; the thermal operator and the partition function are computed explicitly; the part the scalar part discards is examined; the place where the material sector genuinely enters — imaginary time — is made precise against the companion article on the KMS condition; and the non-Hermitian extension is computed.

The Partition Function and the Thermal State

For a quantum system with Hamiltonian $\tilde H_\Psi$ at inverse temperature $\beta$, the partition function and thermal state are

$$ Z(\beta)=2\,\mathrm{Sc}\big(\exp(-\beta H)\big),\qquad \tilde{Ϛ}_{\Psi}=\frac{1}{Z}\,\exp(-\beta H), $$

where the factor two is the size of the $2\times2$ image, and the state is normalised by $2\,\mathrm{Sc}(\tilde{Ϛ}_{\Psi})=1$.

The standard derived quantities are the free energy, the internal energy, and the entropy,

$$ F=-\frac{1}{\beta}\log Z,\qquad E=-\frac{\partial \log Z}{\partial \beta}=\frac{\langle\tilde{Ϛ}_{\Psi},\tilde H_\Psi\rangle_{*}}{\langle\tilde{Ϛ}_{\Psi},e_t\rangle_{*}}=2\,\mathrm{Sc}(\tilde{Ϛ}_{\Psi}\,\tilde H_\Psi^{*}),\qquad S=-2\,\mathrm{Sc}(\tilde{Ϛ}_{\Psi}\log\tilde{Ϛ}_{\Psi}), $$

and the expectation value of any observable $\tilde{H}_{\Psi,3}$ is $\langle\tilde{H}_{\Psi,3}\rangle=\langle\tilde{Ϛ}_{\Psi},\tilde{H}_{\Psi,3}\rangle_{*}/\langle\tilde{Ϛ}_{\Psi},e_t\rangle_{*}=\mathrm{Sc}(\tilde{Ϛ}_{\Psi}\tilde{H}_{\Psi,3}^{*})/\mathrm{Sc}(\tilde{Ϛ}_{\Psi})$, whose scalar part $\mathrm{Sc}(\tilde{Ϛ}_{\Psi}\tilde{H}_{\Psi,3}^{*})=\mathrm{Sc}(\tilde{Ϛ}_{\Psi}\tilde{H}_{\Psi,3})$ is real, the star being inert on $\mathbb{B}_{\Psi}$; it is non-negative when $\tilde{H}_{\Psi,3}\ge0$.

Two elementary facts about $Z$ govern everything below. First, $Z$ depends only on the spectrum of $\tilde H_\Psi$: if $\tilde H_\Psi$ has eigenvalues $E_n$, then $Z=\sum_n \exp(-\beta E_n)$. Equivalently, it is invariant under any $*$-unitary change of basis. These are not features of the biquaternion framework; they are features of the scalar part, which is the matrix trace halved, and the framework inherits them.

The framework adds two things to this standard picture. It gives the Hamiltonian and the state a definite home — the Hermitian subspace $\mathbb{B}_{\Psi}$ — and it fixes the scalar part as the carrier of what the matrix picture calls a trace, with the factor two written explicitly at normalisation. The rest of the article works out those two points and their consequences.

The Trace as a $2\times2$ Scalar Reading

The biquaternion algebra is isomorphic to the algebra of $2\times2$ complex matrices, $\mathbb{B}\cong M_2(\mathbb{C})$. In the isomorphism of the companion articles,

$$ e_t\mapsto I_2,\qquad \vec e_x\mapsto -i\sigma_1,\qquad \vec e_y\mapsto -i\sigma_2,\qquad \vec e_z\mapsto -i\sigma_3, $$

and the scalar imaginary $i$ maps to $iI_2$. The scalar part of a general biquaternion $\tilde Q=Q_te_t+Q_x\vec e_x+Q_y\vec e_y+Q_z\vec e_z$, with $Q_\mu\in\mathbb{C}$, is the coefficient of $e_t$,

$$ \mathrm{Sc}(\tilde Q)=Q_t,\qquad 2\,\mathrm{Sc}(\tilde Q)=\operatorname{tr}\mathsf M_2(\tilde Q). $$

It is a complex number for a general biquaternion and a real number when $\tilde Q\in\mathbb{B}_{\Psi}$. The scalar part is thus a scalar extraction: it retains the $e_t$ coefficient and annihilates the three vector directions $\vec e_x,\vec e_y,\vec e_z$.

On $\mathbb{B}_{\Psi}$ this extraction is the plain form used throughout the corpus, read as the plain pairing. Writing

$$ \tilde{H}_{\Psi,1}=h_{t,1}\,e_t+i\vec h_{\Psi,1},\qquad \tilde{H}_{\Psi,2}=h_{t,2}\,e_t+i\vec h_{\Psi,2},\qquad h_{t,1},h_{t,2}\in\mathbb{R},\ \ \vec h_{\Psi,1},\vec h_{\Psi,2}\in\mathbb{R}^3, $$

for two elements of $\mathbb{B}_{\Psi}$, the product has scalar part

$$ \mathrm{Sc}(\tilde{H}_{\Psi,1}\tilde{H}_{\Psi,2})=h_{t,1}h_{t,2}+\vec h_{\Psi,1}\cdot\vec h_{\Psi,2}, $$

and hence

$$ 2\,\mathrm{Sc}(\tilde{H}_{\Psi,1}\tilde{H}_{\Psi,2}^{*})=2\,(h_{t,1}h_{t,2}+\vec h_{\Psi,1}\cdot\vec h_{\Psi,2}). $$

The derivation of this formula and its role in the Born rule are the subject of the companion article The Born Rule from the Hermitian Form — Derivation and Comparison; here we only use it. Two features matter for the partition function. The formula is bilinear and symmetric, and it is real when both factors lie in $\mathbb{B}_{\Psi}$. Note also that the product $\tilde{H}_{\Psi,1}\tilde{H}_{\Psi,2}$ need not itself lie in $\mathbb{B}_{\Psi}$ — it acquires a real vector part $-\vec h_{\Psi,1}\times\vec h_{\Psi,2}$ in the quaternion product — and yet its scalar part, and therefore the doubled value, is still given by the formula above. The scalar-part formula does not require the product to remain in $\mathbb{B}_{\Psi}$; it requires only the factors to.

A remark on the alternative. Twice the scalar part is not the only natural scalar that can be extracted from an element of $\mathbb{B}_{\Psi}$; the biquaternion norm $\langle\tilde Q\rangle_{\natural} = \langle\tilde Q,\tilde Q\rangle_{\natural}=\tilde Q\tilde Q^{\natural}$ is another. On $\tilde H_\Psi=h_t\,e_t+i\vec h_\Psi$ it gives $\langle\tilde H_\Psi\rangle_{\natural}=h_t^2-|\vec h_\Psi|^2$, of inertia $(1,3)$, and it is the form whose future cone defines the state space. A later section shows that the biquaternion norm of the thermal operator is blind to the level splitting and so cannot serve as a partition function. The scalar part doubled, not the norm, is the scalar that thermodynamics uses; the companion article The Born Rule from the Hermitian Form — Derivation and Comparison makes the general case for why.

The Thermal Operator and the Partition Function

Let the Hamiltonian be a Hermitian element of the informational sector,

$$ \tilde H_\Psi=h_t\,e_t+i\vec h_\Psi,\qquad h_t\in\mathbb{R},\quad \vec h_\Psi\in\mathbb{R}^3, $$

with $|\vec h_\Psi|$ the modulus. The first computation is the square of the imaginary vector part. For a pure real quaternion, $\vec h_\Psi^2=-|\vec h_\Psi|^2e_t$, so

$$ (i\vec h_\Psi)^2=i^2\vec h_\Psi^2=(-1)\big(-|\vec h_\Psi|^2e_t\big)=|\vec h_\Psi|^2e_t. $$

The imaginary vector part therefore behaves like a real number of square $+|\vec h_\Psi|^2$ — it is a boost-type generator, not a phase — and the scalar part $h_t\,e_t$ commutes with it. Hence

$$ \boxed{\;\exp(-\beta H)=\exp(-\beta h_t)\exp(-\beta i\vec h_\Psi) =\exp(-\beta h_t)\Big(\cosh(\beta|\vec h_\Psi|)\,e_t-i\sinh(\beta|\vec h_\Psi|)\,\vec u_\Psi\Big), \qquad \vec u_\Psi=\frac{\vec h_\Psi}{|\vec h_\Psi|}\;}. $$

This is the thermal operator. It is Hermitian, since $\tilde H_\Psi^{*}=\tilde H_\Psi$ implies $(\exp(-\beta H))^\dagger=\exp(-\beta H)$, so it is an element of $\mathbb{B}_{\Psi}$: its scalar part is real and its vector part is imaginary. Equivalently, it is $\exp(-\beta h_t)$ times the unit-norm element $\tilde U_{\Psi,\mathrm b}=\cosh(\beta|\vec h_\Psi|)e_t-i\sinh(\beta|\vec h_\Psi|)\vec u_\Psi$, which is a boost biquaternion in $\mathbb{B}_{\Psi}$ — the natural home of boosts, as the companion article on $\mathbb{B}_{\Psi}$ records. The boost rapidity is $2\beta|\vec h_\Psi|$ in the parametrization $\tilde U_{\Psi,\mathrm b}=\cosh(\theta_{\mathrm{rpd}}/2)+i\sinh(\theta_{\mathrm{rpd}}/2)\vec u$, with $\vec u=-\vec u_\Psi$.

Applying the Hermitian form, read at normalisation, to $\tilde{H}_{\Psi,1}=e_t$ (the identity, which lies in $\mathbb{B}_{\Psi}$) and $\tilde{H}_{\Psi,2}=\exp(-\beta H)$, which also lies in $\mathbb{B}_{\Psi}$,

$$ \boxed{\;Z=2\,\mathrm{Sc}\big(\exp(-\beta H)\big)=2\exp(-\beta h_t)\cosh(\beta|\vec h_\Psi|)\;\;}. $$

This is the ordinary partition function of a two-level system with energies $h_t\pm|\vec h_\Psi|$, since

$$ \exp(-\beta h_t)\big(\exp(-\beta|\vec h_\Psi|)+\exp(+\beta|\vec h_\Psi|)\big)=2\exp(-\beta h_t)\cosh(\beta|\vec h_\Psi|). $$

The two expressions agree because the eigenvalues of $\tilde H_\Psi=h_t\,e_t+i\vec h_\Psi$ are $h_t\pm|\vec h_\Psi|$, in accordance with the matrix identification $\tilde H_\Psi\mapsto h_t\,I+\vec h_\Psi\cdot\boldsymbol\sigma$. The state is $\tilde{Ϛ}_{\Psi}=\exp(-\beta H)/Z$, whose scalar part is $\mathrm{Sc}(\tilde{Ϛ}_{\Psi})=\tfrac12$, so it has the Bloch form

$$ \tilde{Ϛ}_{\Psi}=\tfrac12\big(e_t+i\vec s_{\Psi}\big),\qquad \vec s_{\Psi}=-\tanh(\beta|\vec h_\Psi|)\,\vec u_\Psi,\qquad |\vec s_{\Psi}|=\tanh(\beta|\vec h_\Psi|)\le 1. $$

The thermal state is mixed for every finite $\beta$ and approaches the pure ground state $\tilde\Pi(-\vec u_\Psi)=\tfrac12(e_t-i\vec u_\Psi)$ only as $\beta\to\infty$. Its internal energy is

$$ E=\langle\tilde{Ϛ}_{\Psi},\tilde H_\Psi\rangle_{*}/\langle\tilde{Ϛ}_{\Psi},e_t\rangle_{*}=\mathrm{Sc}(\tilde{Ϛ}_{\Psi}\tilde H_\Psi^{*})/\mathrm{Sc}(\tilde{Ϛ}_{\Psi})=h_t+\vec s_{\Psi}\cdot\vec h_\Psi=h_t-|\vec h_\Psi|\tanh(\beta|\vec h_\Psi|) $$

in agreement with $-\partial/\partial \beta\log Z=h_t-|\vec h_\Psi|\tanh(\beta|\vec h_\Psi|)$, and its purity and entropy are the standard functions of $|\vec s_{\Psi}|$.

As a check on a case chosen independently of the derivation, take $h_t=0.7$, $\vec h_\Psi=(0.3,-0.5,0.8)$, $\beta=1.2$, so that $|\vec h_\Psi|^2=0.98$ and $|\vec h_\Psi|=0.98995$. The formula gives $Z=2\exp(-0.84)\cosh(1.18794)=1.547753$, while summing the exponential of the eigenvalues $h_t\pm|\vec h_\Psi|$ gives $\exp(-2.02794)+\exp(0.34794)=1.547753$; the matrix exponential of the corresponding $2\times2$ matrix agrees with the biquaternion form to machine precision. The Bloch modulus is $\tanh(1.18794)=0.829939$, and $E=h_t-|\vec h_\Psi|\tanh(\beta|\vec h_\Psi|)=-0.121597$ equals $2\,\mathrm{Sc}(\tilde{Ϛ}_{\Psi}\tilde H_\Psi^{*})$.

What the Trace Discards

The thermal operator has a scalar part and an imaginary vector part,

$$ \exp(-\beta H)=\underbrace{\exp(-\beta h_t)\cosh(\beta|\vec h_\Psi|)}_{\text{scalar}}\;e_t\;-\;i\underbrace{\exp(-\beta h_t)\sinh(\beta|\vec h_\Psi|)}_{\text{coefficient}}\;\vec u_\Psi, $$

and the scalar part keeps the first and annihilates the second. Three consequences are worth stating explicitly.

The discarded part is the coherence. The imaginary vector part of $\exp(-\beta H)$ points along $-\vec u_\Psi$, and it is fixed by the same combination $\beta|\vec h_\Psi|$ that fixes the Bloch vector $\vec s_{\Psi}=-\tanh(\beta|\vec h_\Psi|)\vec u_\Psi$ of the thermal state. It encodes the orientation of the thermal polarization. The partition function is blind to that orientation. $Z$ depends on $h_t$ and $|\vec h_\Psi|$ but not on the direction $\vec u_\Psi$; equivalently, $Z$ is invariant under the conjugation $\tilde H_\Psi\mapsto\tilde Q\tilde H_\Psi\tilde Q^{*}$ by any $*$-unitary biquaternion, which rotates $\vec u_\Psi$ while preserving the spectrum. The whole dependence of the thermal state on the axis of the Hamiltonian is invisible to $Z$ and appears only in $\tilde{Ϛ}_{\Psi}$ and in expectation values.

The discarded part is not lost. The information the scalar part removes from $Z$ is retained by the correlation functions. The expectation value $2\,\mathrm{Sc}(\tilde{Ϛ}_{\Psi}\tilde{H}_{\Psi,3})$, real because the star is inert on $\mathbb{B}_{\Psi}$, and, in the thermal field-theoretic setting, the KMS correlation functions of the companion article are sensitive to the imaginary vector part. The partition function is thus the crudest $*$-unitary invariant of the thermal operator: it is the piece invariant under the full $*$-unitary group, and everything else is delegated to the correlators.

The discarded part is not the material sector. It is tempting to say that the doubling "projects away the $\mathbb{B}_{\Phi}$ sector". Read literally, that statement is not correct for a Hermitian Hamiltonian. Since $\tilde H_\Psi\in\mathbb{B}_{\Psi}$ implies $\exp(-\beta H)\in\mathbb{B}_{\Psi}$, the thermal operator has no $\mathbb{B}_{\Phi}$ component at all; the subspace $\mathbb{B}_{\Phi}$ does not appear in it. What the scalar part discards is the imaginary vector part of $\mathbb{B}_{\Psi}$, which is $i$ times a real spatial vector — the same three real spatial directions that, with real coefficients, are the spatial part of $\mathbb{B}_{\Phi}$. The correct division is therefore between the scalar (energy-like, spectral) content of the thermal operator, which $Z$ keeps, and its coherence content, which $Z$ discards. The material sector enters elsewhere, through imaginary time.

Imaginary Time: Where the Material Sector Enters

The partition function can be written as twice the scalar part of the evolution operator at an imaginary time,

$$ Z=2\,\mathrm{Sc}\,\tilde Q(-i\beta),\qquad \tilde Q(r'_t)=\exp(-i\tilde H_\Psi r'_t), $$

the analytic continuation $r'_t\to-i\beta$ of the real-time evolution. This is not a convenience of the biquaternion framework; it is the standard relationship between the thermal weight and imaginary time, and it is the same continuation that makes the Matsubara formalism work.

It is here that the material sector $\mathbb{B}_{\Phi}$ genuinely enters. The material sector's time coordinate is $i r'_t$, imaginary by construction, so the continuation $r'_t\to-i\beta$ of the evolution is a displacement along the $\mathbb{B}_{\Phi}$ time direction: the analytic continuation to $r'_t=-i\beta$ moves the Minkowski time coordinate to $i r'_t=c\beta$, and the thermal operator is the evolution operator evaluated at that displaced point. In this sense the thermal circle of circumference $\beta$ — the compact imaginary-time direction over which the Euclidean formulation is defined — lies along the time axis of $\mathbb{B}_{\Phi}$. The KMS article states the structural point that accompanies this: the complexified time direction in which the KMS continuation takes place is the sum of the imaginary time direction of $\mathbb{B}_{\Phi}$ and the real time direction of $\mathbb{B}_{\Psi}$, so that $\mathbb{B}_{\Phi}\oplus i\mathbb{B}_{\Phi}=\mathbb{B}$.

Two further facts tie the partition function to that structure, and both are consistent with the KMS article rather than additional to it.

First, the normalization of the thermal state is what makes the modular Hamiltonian a Hermitian element. From $\tilde{Ϛ}_{\Psi}=\exp(-\beta H)/Z$,

$$ H_{\mathrm{modular}}=-\log\tilde{Ϛ}_{\Psi}=\beta\tilde H_\Psi+\big(\log Z\big)e_t, $$

which is Hermitian, hence an element of $\mathbb{B}_{\Psi}$; the log-partition function $\log Z$ appears as the real scalar part of the modular Hamiltonian. The companion article records the general fact that $H_{\mathrm{modular}}=-\log\tilde{Ϛ}_{\Psi}$ lies in $\mathbb{B}_{\Psi}$; the partition function is precisely the additive scalar that makes this true for the Gibbs state.

Second, the KMS boundary relation $F_{AB}(r'_t+i\beta)=F_{BA}(-r'_t)$ of the companion article is the imaginary-time statement whose finite-dimensional shadow is the Gibbs form used here. The partition function is the normalization of the KMS state, and the strip width in that relation is the same $\beta$ that measures the imaginary-time displacement of the thermal weight. Nothing in the present article changes the KMS boundary relation; the partition function presupposes it.

So the material sector enters the partition function not as a component of the thermal operator but as the direction and period of the thermal weight: $Z$ is computed from the $\mathbb{B}_{\Psi}$ Hamiltonian and normalized by an $\mathbb{B}_{\Psi}$ modular Hamiltonian, while the circle on which the trace is taken lies along the $\mathbb{B}_{\Phi}$ time.

Is There a Genuine Modification?

We can now separate the two cases cleanly.

Hermitian Hamiltonian: no modification. For $\tilde H_\Psi\in\mathbb{B}_{\Psi}$, the object $Z=2\,\mathrm{Sc}(\exp(-\beta H))$ is exactly the ordinary partition function. The algebra's contribution is an interpretation with three components: twice the scalar part is the matrix trace of the $2\times2$ image, the thermal operator is a scalar Boltzmann weight times a boost biquaternion in $\mathbb{B}_{\Psi}$, and $\beta$ measures a displacement along the imaginary-time axis of $\mathbb{B}_{\Phi}$. No second partition function arises, and none is required. If the framework is to have content here beyond notation, it must be content of this interpretive kind; the algebra alone does not supply a new $Z$.

It is worth confirming that the obvious rival scalar does not work. The biquaternion norm of the thermal operator is

$$ N\big(\exp(-\beta H)\big)=\exp(-2\beta h_t)\Big(\cosh^2(\beta|\vec h_\Psi|)-\sinh^2(\beta|\vec h_\Psi|)\Big)e_t=\exp(-2\beta h_t)e_t, $$

which is independent of $|\vec h_\Psi|$. The biquaternion norm sees only the scalar part $h_t$ of the Hamiltonian and is blind to the level splitting $|\vec h_\Psi|$; its logarithm is linear in $\beta$ and would give a free energy independent of $\beta$, with no thermal population of the excited level. Twice the scalar part is the scalar that thermodynamics requires, and the biquaternion norm is not a candidate partition function.

Non-Hermitian generator: a complex partition function. The one place where the sector split changes the answer is if the generator is allowed to leave $\mathbb{B}_{\Psi}$. Let

$$ \tilde H_\Psi=i a_0e_t-\vec a,\qquad a_0\in\mathbb{R},\quad \vec a\in\mathbb{R}^3, $$

which is an element of $\mathbb{B}_{\Phi}$ (imaginary scalar part $ia_0$, real vector part $-\vec a$). Because $\vec a$ is a pure real quaternion, $\exp(\beta\vec a)=\cos(\beta|\vec a|)e_t+\sin(\beta|\vec a|)a_{\mathrm{fop}}$, and the scalar part of the thermal operator is no longer real:

$$ \exp(-\beta H)=\exp(-i\theta_{\mathrm{mat}})\Big(\cos(\beta|\vec a|)e_t+\sin(\beta|\vec a|)a_{\mathrm{fop}}\Big), \qquad Z=2\exp(-i\theta_{\mathrm{mat}})\cos(\beta|\vec a|). $$

The Matsubara angle is $\theta_{\mathrm{mat}}=\beta a_0$, a real inverse temperature times a real energy component, so the angle is real.

The partition function is complex. Its modulus is $2|\cos(\beta|\vec a|)|$ and its phase is $-\beta a_0$: the imaginary scalar part of the $\mathbb{B}_{\Phi}$ generator (the imaginary-time direction, $ia_0e_t$) produces the phase, while the real vector part $-\vec a$ produces a real oscillatory factor. With $a_0=0.9$, $\vec a=(0.4,0.2,-0.3)$, $\beta=1.2$, one finds $Z=0.752585-1.408250\,i$, of modulus $1.596732$ and phase $-1.08=-\beta a_0$. Even with $a_0=0$ the result $2\cos(\beta|\vec a|)$ becomes negative for some $\beta$, so the "state" is not positive; the non-Hermitian generator does not define a thermal state at all in the ordinary sense.

This is a genuine modification, but it is not forced by the framework: observables are defined to be Hermitian elements of $\mathbb{B}_{\Psi}$, and a Hermitian $\tilde H_\Psi$ is the case treated in the body of the article. The non-Hermitian case is what one gets if one allows the generator to have an $\mathbb{B}_{\Phi}$ component, and it connects to two standard ideas without deriving either: complex temperature (continuing $\beta\to\beta+i\theta_{\mathrm{evo}}$, which turns the scalar part into $2\,\mathrm{Sc}(\exp(-\beta H)\exp(-i\theta_{\mathrm{evo}}\tilde H_\Psi))$ and exposes real-time evolution), and the imaginary chemical potential used in finite-density lattice studies, where an imaginary parameter again converts a sign problem into a phase. Whether an $\mathbb{B}_{\Phi}$-valued generator is physically meaningful in this framework, and whether the resulting complex free energy has an interpretation beyond these known techniques, is open.

The Norm's Logarithm as a Chemical Potential

The thermal weight is a scale: $\exp(-\beta H)$ is a Boltzmann factor times a boost, and in the scalar sector its logarithm is a free energy. Read on the biquaternion norm, the same logarithm has a chemical-potential form. The norm is multiplicative and definite on an invertible element, so a state of unit norm carries weight one and a general state carries a weight whose modulus is a power of $\lvert N\rvert$; the quantity $\log\lvert N\rvert$ is therefore a potential, and the total

$$ \mu_{\mathrm{tot}}=\log\lvert N\rvert+V , $$

with $V$ a central scalar external potential, is the sum of a configurational term and an external one. Read physically, $\log\lvert N\rvert$ is a chemical potential and $\mu_{\mathrm{tot}}$ the total potential of a grand-canonical description, the norm playing the role of the density and its logarithm that of the conjugate potential.

The interpretation is labelled. The article's own content is the scalar part and the thermal operator; the identification of $\log\lvert N\rvert$ with a chemical potential is an interpretation of the same object, and it joins the two other interpretations of the norm's logarithm in the corpus, the dilation potential of Conformal Invariance and the Massless Limit in Biquaternionic Form and the Weyl factor of The Local Complex Structure and the Speed of Light. The passage to the imaginary direction, where the material sector enters, is the subject of the article's Imaginary Time section and of The Matsubara Thermal Formalism in Biquaternionic Form.

Summary

The partition function $Z=2\,\mathrm{Sc}(\exp(-\beta H))$ has a biquaternion form, and its content is an interpretation of the scalar part rather than a new object.

For a Hermitian Hamiltonian $\tilde H_\Psi=h_t\,e_t+i\vec h_\Psi$ in the informational sector, the thermal operator is

$$ \exp(-\beta H)=\exp(-\beta h_t)\Big(\cosh(\beta|\vec h_\Psi|)e_t-i\sinh(\beta|\vec h_\Psi|)\vec u_\Psi\Big), $$

which is itself an element of $\mathbb{B}_{\Psi}$: a scalar Boltzmann weight times a boost biquaternion. Twice its scalar part is the matrix trace of its $2\times2$ image, so

$$ \boxed{\;Z=2\,\mathrm{Sc}\big(\exp(-\beta H)\big)=2\exp(-\beta h_t)\cosh(\beta|\vec h_\Psi|)\;}, $$

the ordinary two-level partition function with energies $h_t\pm|\vec h_\Psi|$. The thermal state is $\tilde{Ϛ}_{\Psi}=\tfrac12(e_t+i\vec s_{\Psi})$ with $\vec s_{\Psi}=-\tanh(\beta|\vec h_\Psi|)\vec u_\Psi$, and the internal energy $E=h_t-|\vec h_\Psi|\tanh(\beta|\vec h_\Psi|)$ agrees with $-\partial/\partial \beta\log Z$.

The scalar part of the thermal operator is kept by the doubling, and its imaginary vector part is discarded. That discarded part is the coherence — the orientation of the thermal polarization — and it is invisible to $Z$ but visible in expectation values and KMS correlation functions. It is not the material sector: for Hermitian $\tilde H_\Psi$ the thermal operator has no $\mathbb{B}_{\Phi}$ component. The material sector enters through imaginary time, since $Z=2\,\mathrm{Sc}\,\tilde Q(-i\beta)$ is the scalar part of the evolution operator continued along the $i r'_t$ direction, with the thermal circle of circumference $\beta$ lying along the $\mathbb{B}_{\Phi}$ time. The modular Hamiltonian $H_{\mathrm{modular}}=-\log\tilde{Ϛ}_{\Psi}=\beta\tilde H_\Psi+(\log Z)e_t$ lies in $\mathbb{B}_{\Psi}$, with the log-partition function as its real scalar part, in agreement with the KMS article.

The algebra does not force a modification. A genuine modification appears only if the generator is permitted an $\mathbb{B}_{\Phi}$ component, in which case $Z$ can become complex, with the imaginary scalar part of the generator producing a phase and its real vector part an oscillatory modulus. That extension is possible, connects to complex temperature and imaginary chemical potential, and is not forced by the framework; it is left open.

Summary of Notation

Symbol Meaning
$\mathbb{B}=\mathbb{C}\otimes_{\mathbb{R}}\mathbb{H}$ Biquaternion algebra
$e_t=1,\vec e_x,\vec e_y,\vec e_z$ Quaternion basis, $\vec e_\eta^2=-e_t$
$i$ Scalar imaginary, $i^2=-1$
$\mathbb{B}_{\Psi}$ Hermitian subspace (informational sector): real scalar, imaginary vector
$\mathbb{B}_{\Phi}$ Anti-Hermitian subspace (material sector): imaginary scalar, real vector
$\mathbb{B}_{H}$ Real-quaternion subspace (fixed points of complex conjugation)
$\tilde H_\Psi=h_t\,e_t+i\vec h_\Psi$ Hermitian Hamiltonian (observable)
$2\,\mathrm{Sc}(\tilde Q)=\operatorname{tr}\mathsf M_2(\tilde Q)$ Twice the scalar part, the matrix trace of the image
$2\,\mathrm{Sc}(\tilde{H}_{\Psi,1}\tilde{H}_{\Psi,2}^{*})=2(h_{t,1}h_{t,2}+\vec h_{\Psi,1}\cdot\vec h_{\Psi,2})$ Doubled scalar-part formula on $\mathbb{B}_{\Psi}$
$Z=2\,\mathrm{Sc}(\exp(-\beta H))$ Partition function
$\tilde{Ϛ}_{\Psi}=\exp(-\beta H)/Z$ Thermal state
$\exp(-\beta H)=\exp(-\beta h_t)\big(\cosh(\beta|\vec h_\Psi|)e_t-i\sinh(\beta|\vec h_\Psi|)\vec u_\Psi\big)$ Thermal operator
$\vec s_{\Psi}=-\tanh(\beta|\vec h_\Psi|)\vec u_\Psi$ Bloch vector of the thermal state
$F=-\beta^{-1}\log Z$, $E=-\partial/\partial \beta\log Z$ Free energy, internal energy
$H_{\mathrm{modular}}=-\log\tilde{Ϛ}_{\Psi}=\beta\tilde H_\Psi+(\log Z)e_t$ Modular Hamiltonian (in $\mathbb{B}_{\Psi}$)
$\langle\tilde Q\rangle_{\natural} = \langle\tilde Q,\tilde Q\rangle_{\natural}=\tilde Q\tilde Q^{\natural}$ Biquaternion norm
$\beta=1/(k_BT)$ Inverse temperature
$\langle\tilde{Q}_1,\tilde{Q}_2\rangle$ the general plain bilinear form, the scalar part of the general plain bilinear product, $\mathrm{Sc}(\tilde{Q}_1\tilde{Q}_2)$
$\langle\tilde{Q}_1,\tilde{Q}_2\rangle_{*}$ the general plain sesquilinear form, $\mathrm{Sc}(\tilde{Q}_1\tilde{Q}_2^{*})$; on the diagonal $\langle\tilde{Q}_2,\tilde{Q}_2\rangle_{*}=\sum_\mu\lvert Q_{2\mu}\rvert^{2}$

Further Reading

  • L. D. Landau and E. M. Lifshitz, Statistical Physics, Part 1 (Pergamon, 1980), for the partition function, the Gibbs state, and the thermodynamic potentials.
  • R. H_{\mathrm{modular}}. Pathria and P. D. Beale, Statistical Mechanics (Butterworth–Heinemann, 2011), for the standard treatment of the canonical ensemble.
  • H_{\mathrm{modular}}. Huang, Statistical Mechanics (Wiley, 1987), for the relation between the partition function and the spectrum.
  • M. Le Bellac, Thermal Field Theory (Cambridge, 1996), for the imaginary-time and Matsubara formulation in which $Z$ is a Euclidean trace.
  • R. Haag, N. M. Hugenholtz, and M. Winnink, "On the equilibrium states in quantum statistical mechanics," Communications in Mathematical Physics 5 (1967) 215–236, for the KMS characterization of thermal equilibrium.
  • M. Takesaki, Tomita's Theory of Modular Hilbert Algebras and Its Applications (Springer, 1970), for the modular Hamiltonian $H_{\mathrm{modular}}=-\log\rho$.
  • A. Roberge and N. Weiss, "Gauge theories with imaginary chemical potential and the phases of QCD," Nuclear Physics B 275 (1986) 734–745, for the imaginary-chemical-potential technique alluded to in the text.
  • C. M. Bender and S. Boettcher, "Real spectra in non-Hermitian Hamiltonians having $\mathcal{PT}$ symmetry," Physical Review Letters 80 (1998) 5243–5246, for non-Hermitian generators with controlled spectra.
  • Companion articles: The KMS Condition and the Biquaternion Framework, for the imaginary-time and modular structure; States, Observables and the Born Weight, for the doubled scalar-part formula and the action of $\mathbb{B}_{\Psi}$, for what the Hermitian form derives and what it restates; Quantum Physics in Biquaternionic Form, for the thermal state as an element of the Bloch ball.