---
title: 'Multi-Invariants: Concepts & Applications'
url: https://www.emergentmind.com/topics/multi-invariants
type: topic
---

# Multi-Invariants: Concepts & Applications

Searching arXiv for recent papers on “multi-invariants” across the main mathematical and physical usages.
I’ll look for recent arXiv entries using “multi-invariants” and closely related phrases such as “multiple invariants,” “local unitary invariants,” and “Galois extensions by multi-invariants.”
“Multi-invariants” is not a single universally standardized object. In current mathematical and physical literature, the term denotes several families of constructions in which a single invariant is replaced by a structured collection of invariants, or in which an invariant is built from multiple replicas, components, channels, or ramification types. In arithmetic statistics, multi-invariants separate tame ramification by local type and lead to multivariable counting problems for Galois extensions [2211.16698, 2507.12342]. In multipartite quantum theory, they are local-unitary-invariant polynomials obtained by contracting several copies of a state and its conjugate according to permutation data, with Coxeter and dihedral structures playing a central role [2411.00935, 2509.06348, 2509.00593, 2601.16258]. In topology and knot theory, the same general impulse appears as component-indexed, type-indexed, or endpoint-indexed refinements of classical invariants [1909.04159, 2504.09368, 2204.11234]. In applied settings, the term also appears in the construction of moment invariants for multi-channel data, purity-like invariants of multi-photon states, and approximate invariants of nonlinear beam dynamics [2201.00877, 2506.12857, 2103.06351].

## 1. Scope of the term and recurring formal patterns

Across these literatures, “multi-invariants” typically signals one of two formal moves. The first is a **multivariate refinement**: instead of one scalar invariant, one keeps a vector \((\operatorname{inv}_1,\dots,\operatorname{inv}_m)\) that resolves several local or componentwise behaviors separately. The second is a **multi-copy contraction**: one constructs a polynomial from several copies of an object and its conjugate, with the pattern of contraction itself carrying the invariant data. These two patterns are conceptually distinct but structurally related.

| Domain | Basic object | Indexing principle |
|---|---|---|
| Arithmetic statistics | \(G\)-extensions \(L/K\) | Ramification types \([I_i,\gamma_i]\) |
| Multipartite quantum theory | \(\psi^{\otimes n_r}\bar\psi^{\otimes n_r}\) | Permutations, replica symmetry, \(\psi\)-graphs |
| Link theory | Link or multi-virtual diagram | Components, endpoint order, crossing types |
| Multi-channel analysis | \(F(X):\Omega\subset\mathbb R^M\to\mathbb R^N\) | Coordinate/value transform behavior |

The arithmetic formulation in “Malle’s conjecture with multiple invariants” defines one invariant for each nontrivial ramification type and predicts a box-counting asymptotic of the form \(C\,X_1\cdots X_m\) when all \(\operatorname{inv}_i\le X_i\) [2211.16698]. The quantum formulation in “Multi-invariants and Bulk Replica Symmetry” takes \(n_r\) copies of \(\psi\) and \(n_r\) copies of \(\bar\psi\), contracts indices via permutations \((g_1,\dots,g_q)\in S_{n_r}^q\), and then studies normalized quantities \(\hat E=E^{1/n_r}\) together with the replica symmetry group of the contraction pattern [2411.00935]. In link theory, the same refinement principle appears when a polynomial assigns distinct variables \(t_i\) to different ordered components, or when a diagrammatic invariant depends on several commuting virtual types rather than a single undifferentiated virtual crossing [1909.04159, 2504.09368]. In multi-channel signal analysis, the invariants are designed to be unchanged under coupled transformations of both coordinates and values, rather than only one of the two [2201.00877].

A persistent conceptual theme is that the underlying geometry or combinatorics becomes visible only after the single coarse invariant is replaced by a structured family. This suggests that “multi-invariants” function less as a single definition than as a methodological class.

## 2. Arithmetic multi-invariants and counting Galois extensions

The arithmetic version begins with a finite group \(G\), a number field \(K\), and the tame local behavior of a \(G\)-extension \(L/K\). “Malle’s conjecture with multiple invariants” associates to each nontrivial ramification type \([I_i,\gamma_i]\) an invariant
\[
\operatorname{inv}_i(L)=\prod_{\mathfrak p\text{ of ramification type }[I_i,\gamma_i]} \operatorname{Nm}(\mathfrak p),
\]
so that each coordinate records the product of norms of primes having a fixed tame inertia type [2211.16698]. The paper proposes the multivariate heuristic
\[
\sum_{\substack{\textnormal{$G$-ext. } L/K:\ \operatorname{inv}_i(L)\le X_i\ \forall i}} \frac1{\#\operatorname{Aut}(L)} \sim C\,X_1\cdots X_m,
\]
together with boundary and annulus versions, and proves the conjecture for all finite abelian groups. A major structural result is that refined Artin conductors carry essentially the same information as the vector \((\operatorname{inv}_1,\dots,\operatorname{inv}_m)\) [2211.16698].

“Counting \(D_4\)-field extensions by multi-invariants” gives a concrete nonabelian realization of this philosophy for
\[
D_4=\langle r,s\mid s^2=r^4=1,\ rs=sr^{-1}\rangle.
\]
There are four nontrivial conjugacy classes of cyclic subgroups,
\[
\langle s\rangle,\qquad \langle rs\rangle,\qquad \langle r\rangle,\qquad \langle r^2\rangle,
\]
and the paper therefore studies four invariants attached to the inertia types of tamely ramified odd primes. Because \(D_4\) has a nontrivial outer automorphism interchanging \(\langle s\rangle\) and \(\langle rs\rangle\), the objects counted are pairs \((M,\sigma)\) with a specified isomorphism
\[
\sigma:\operatorname{Gal}(M/\mathbb Q)\xrightarrow{\simeq} D_4
\]
rather than bare fields \(M\) [2507.12342].

For a pair \((M,\sigma)\), if \(I_p\) denotes the image under \(\sigma\) of the inertia group at a rational prime \(p\), the four invariants are
\[
\begin{aligned}
\operatorname{inv}_1(M)&= \prod_{p : I_{p}\text{ conjugate to } \langle s \rangle} p, \\
\operatorname{inv}_2(M)&= \prod_{p : I_{p}\text{ conjugate to } \langle rs  \rangle} p, \\
\operatorname{inv}_3(M)&= \prod_{p : I_{p} \text{ conjugate to } \langle r \rangle} p, \\
\operatorname{inv}_4(M)&= \prod_{p : I_{p}\text{ conjugate to } \langle r^2 \rangle} p .
\end{aligned}
\]
These products run only over tamely ramified odd primes; \(2\) is wild for \(D_4\)-extensions and is excluded from the invariant itself [2507.12342].

The paper identifies the subfield tower
\[
\mathbb Q\subset K_1,K_2,K\subset L\subset M
\]
with
\[
K_1=\mathbb Q(\sqrt{m_1m_2}),\qquad K_2=\mathbb Q(\sqrt{m_1m_3}),\qquad K=\mathbb Q(\sqrt{m_2m_3}),
\]
where \(m_1,m_2,m_3\) are square-free and pairwise coprime, with \(m_1>0\). For odd \(p\), the inertia type is determined by ramification in \(K_1\) and \(K_2\), and the first three invariants become
\[
\operatorname{inv}_1(M)=m_2',\qquad \operatorname{inv}_2(M)=m_3',\qquad \operatorname{inv}_3(M)=m_1',
\]
while \(\operatorname{inv}_4(M)\) is the odd part of the extra ramification in \(M/L\) beyond the biquadratic subfield [2507.12342].

The ordering is box ordering:
\[
\operatorname{inv}_i(M)\le X_i\qquad (i=1,2,3,4).
\]
Under the mild lower bound
\[
\min_{1\le i\le 4}\{X_i\}\gg (\log \operatorname{X})^A\qquad\text{for some }A>111,
\qquad \operatorname{X}=\max_{1\le i\le 3}X_i,
\]
the main theorem proves
\[
\# \{(M, \sigma): \operatorname{inv}_i(M)\le X_i\}
=
{\frac{27}{8}\prod_{p>2} \left(1-\frac{1}{p}\right)^4 \left(1+\frac{4}{p}  \right) X_1 X_2 X_3X_4
\cdot \left(1 + O\left((\log \operatorname{X})^{\frac{111-A}{19} \right)\right)\right)}.
\]
Thus the asymptotic has exponent \(1\) in each variable and a completely explicit Euler-product constant [2507.12342]. The paper states that this verifies Gundlach’s version of Malle’s conjecture in the \(D_4/\mathbb Q\) setting and that the leading constant matches the Loughran–Santens prediction after accounting for the \(8\) choices of \(\sigma\) [2507.12342].

## 3. Replica, Coxeter, dihedral, and stabilizer-state multi-invariants

In multipartite quantum theory, multi-invariants are local-unitary-invariant polynomials built from several copies of a pure state and its complex conjugate. For a \(q\)-partite state
\[
|\Psi\rangle=\sum_{i_1,\dots,i_q}\psi_{i_1\cdots i_q}\,|e_1^{i_1}\rangle\otimes\cdots\otimes |e_q^{i_q}\rangle,
\]
“Multi-invariants and Bulk Replica Symmetry” defines a multi-invariant \(E\) by taking \(n_r\) copies of \(\psi\) and \(n_r\) copies of \(\bar\psi\), with all indices contracted pairwise according to permutations \((g_1,\dots,g_q)\), \(g_a\in S_{n_r}\), and then normalizing by
\[
\hat E=E^{1/n_r}.
\]
There is a gauge redundancy
\[
E(g_1,\dots,g_q)=E(gg_1,\dots,gg_q),\qquad g\in S_{n_r},
\]
and the replica symmetry is the subgroup that preserves the contraction pattern up to this common left multiplication [2411.00935].

The most symmetric case is when the replica symmetry acts freely and transitively on the replicas. Then the replicas may be identified with the elements of a finite group \(G\), the contractions are given by right-regular actions of chosen generators, and the left-regular action yields the replica symmetry. The paper also adjoins an orientation-reversing reflection \(p\) to form an extended replica symmetry group \(\hat R\), with generators \(r_a=p\cdot g_a\) satisfying \(r_a^2=e\); in this form the ordinary replica symmetry becomes an index-\(2\) subgroup of a Coxeter-group quotient [2411.00935].

The holographic question studied there is which multi-invariants admit a dominant bulk saddle preserving replica symmetry for arbitrary region configurations. The paper states that the most general solution under that requirement is given by the **Coxeter invariants**, and that orbifolding by the preserved symmetry yields a bulk cone-manifold whose fixed loci form codimension-\(2\) networks of conical singularities. A major consequence is that there are **infinitely many infinite families** of multi-invariants that evaluate identically on the holographic vacuum state because the same orbifold \(H^3/K\) can arise from different free normal subgroups \(S\triangleleft K\), hence from handlebodies of different genus [2411.00935].

The Coxeter theme reappears in “Monotones from multi-invariants: a classification,” where multi-invariants are encoded by bipartite edge-labeled \(\psi\)-graphs. For \(n_r\) white and \(n_r\) black vertices, the normalized invariant is
\[
\hat{\mathcal I}=\mathcal I^{1/n_r},\qquad \hat\nu=1-\hat{\mathcal I}.
\]
The key monotonicity criterion recalled in the paper is that if the \(\psi\)-graph is connected and edge-convex, then \(\hat\nu\) is a pure-state entanglement monotone under LOCC. The main conjecture is that edge-convex multi-invariants are exactly those labeled by finite Coxeter groups; the paper proves this for \(A_n\), \(B_n(=C_n)\), \(D_n\), and notes that the dihedral family \(I_n\) had already been proved, leaving six exceptional connected cases written as \(E_{6,7,8}\), \(F_4\), and \(H_{3,5}\) [2509.06348]. This gives a precise graph-theoretic bridge between monotonicity and Coxeter symmetry.

Two especially explicit quantum multi-invariants are studied in “Genuine multi-entropy, dihedral invariants and Lifshitz theory.” For a tripartite pure state, the Rényi multi-entropy is defined from \(n^2\) replicas by
\[
S^{(3)}_n(A:B:C)=\frac{1}{1-n}\frac1n \log\frac{\mathcal Z^{(3)}_n}{(\mathcal Z^{(3)}_1)^{n^2}},
\]
with a specified grid-like permutation pattern on the three parties, and the **genuine multi-entropy** is
\[
G_n^{(3)}(A:B:C)=S_n^{(3)}(A:B:C)-\frac12\big(S_n(A)+S_n(B)+S_n(C)\big).
\]
For \(z=2\) Lifshitz ground states, the paper finds
\[
G_n^{(3)}(A:B:C) = \frac{2-n}{2n}\Big(I_{1/2}(A:B)-2\mathcal E(A:B)\Big),
\]
where \(I_{1/2}\) is the Rényi mutual information at index \(1/2\) and \(\mathcal E\) is the logarithmic negativity. It also proves that the **dihedral invariant** satisfies
\[
\mathcal D_{2n}(A:B)=S^R_{2,n}(A:C),
\]
so that the dihedral replica construction is exactly the \((m=2,n)\) Rényi reflected entropy of the complementary bipartition [2509.00593].

“Multi-invariants in stabilizer states” adapts this entire framework to stabilizer states. For a pure \(q\)-partite state, it defines
\[
\mathcal Z(\sigma_1,\dots,\sigma_q)=\langle\psi|^{\otimes n}(\sigma_1\otimes\cdots\otimes \sigma_q)|\psi\rangle^{\otimes n},
\]
with the usual left-right relabeling redundancy. For stabilizer states, the identity
\[
|\psi\rangle\langle\psi|=\frac{1}{2^n}\sum_{g\in G}g
\]
makes the invariants tractable, and the paper gives an efficient numerical algorithm based on graph-state measurement rules, with complexity stated as roughly \(\mathcal O(n^3)\). For tripartite stabilizer states, the GHZ-extraction theorem yields the closed formula
\[
\mathcal E_n^{\mathrm{ME}}
=
\log_2\!\Bigg(
\frac{|G|}
{(|G_A||G_B||G_C|)^{n/2}\,|G_{AB}\cdot G_{BC}\cdot G_{AC}|^{(n-2)/2}}
\Bigg),
\]
and the paper further develops a counting argument for Coxeter multi-invariants of \(q\)-partite stabilizer states [2601.16258].

A plausible implication is that, within quantum information, “multi-invariants” has become a nexus joining replica combinatorics, Coxeter symmetry, entanglement monotones, holographic geometry, and stabilizer-state computability rather than a merely formal generalization of Rényi-type quantities.

## 4. Multi-component and multi-type invariants in low-dimensional topology

In knot theory and related low-dimensional topology, multi-invariant constructions refine classical invariants by recording component labels, endpoint order, or multiple virtual crossing types. The common principle is that a single scalar link invariant is replaced by an object sensitive to the extra combinatorics of several components or several crossing species.

“A multivariable Casson-Lin type invariant” constructs a signed count of irreducible \(\mathrm{SU}(2)\) representations of the link group with prescribed meridional traces. For a colored braid \(\beta\) whose closure is \(L\), the invariant is defined as the algebraic intersection number
\[
h_\beta^c(\alpha)=\bigl\langle \widehat{\Lambda}_n^{\alpha,c},\widehat{\Gamma}_\beta^\alpha\bigr\rangle_{\widehat{H}_n^{\alpha,c}},
\]
after removing the abelian locus and quotienting by the free \(SO(3)\)-action. For a \(2\)-component ordered link \(L=K_1\cup K_2\) with linking number \(1\), the main formula is
\[
h_L(\alpha_1,\alpha_2)
=
-\frac12\Bigl(\sigma_L(\omega_1,\omega_2)+\sigma_L(\omega_1,\omega_2^{-1})\Bigr),
\qquad (\omega_1,\omega_2)=(e^{2i\alpha_1},e^{2i\alpha_2}),
\]
under the stated nonvanishing conditions on the multivariable Alexander polynomial. The paper also gives deformation criteria relating limits of irreducible \(\mathrm{SU}(2)\) representations to zeros of \(\Delta_L\) [1805.03050].

“The Multi-variable Affine Index Polynomial” assigns a variable \(t_i\) to each ordered component \(L_i\) and defines
\[
p_L(t_1,\ldots,t_n)
=
\sum_{i=1}^n\ \sum_{c\in T_i\text{ over }T_j}
\operatorname{sgn}(c)\bigl(t_i^{W_{\pm}(c)}-1\bigr),
\]
where the variable is determined by the component of the overstrand. If all variables are set equal, one recovers Kauffman’s compatible-link invariant; if the link is a virtual knot, the construction reduces to the original Affine Index Polynomial. The paper proves that the multi-variable Affine Index Polynomial is a Vassiliev invariant of order one [1909.04159].

For multi-virtual links, the algebraic refinement is different. “Algebraic invariants of multi-virtual links” introduces **operator quandles**
\[
(Q,*,A_T),\qquad A_T=\{\alpha,\beta,\gamma,\dots\},
\]
where the automorphisms indexed by the virtual crossing types pairwise commute. This leads to an operator quandle coloring invariant and an operator quandle \(2\)-cocycle invariant for multi-virtual links. The paper also constructs an explicit infinite family
\[
Q_m(e)=\mathbb Z_3\times \mathbb Z_2^m
\]
of connected operator quandles in which the right translations \(R_{(0,a)}\), \(a\in\mathbb Z_2^m\), are pairwise distinct and pairwise commuting, and uses these invariants to classify small multi-virtual knots and to exhibit an infinite family of pairwise nonequivalent multi-virtual knots with a single classical crossing [2504.09368].

“Invariants of multi-linkoids” extends knotoids to diagrams with both open and closed components on a closed orientable surface. Its Kauffman bracket polynomial is
\[
<L>(A^{\pm1},\lambda)
=
\sum_{\sigma_i}\langle L\mid \sigma_i\rangle
(-A^2-A^{-2})^{\|c\|}\lambda^{\|l\|},
\]
where \(\|c\|\) and \(\|l\|\) count closed and open components in the state. For ordered multi-linkoids, the ordered bracket replaces the single \(\lambda\) by endpoint-pair variables \(\lambda_{ij}\), giving a strictly finer invariant in explicit examples. The paper also defines a Kauffman bracket skein module and relates multi-linkoids in \(S^2\) to simple generalized \(\Theta\)-graphs, which then support colored \(T\)-invariants [2204.11234].

A parallel state-sum refinement appears in “Multi-Skein Invariants for Welded and Extended Welded Knots and Links.” There, each classical crossing has three states rather than two, and after solving the coefficient constraints imposed by \(R2\), \(F1\), and the extended welded moves, the normalized invariants take the form
\[
Y(L)=r^{v(L)}(ar-\nu b)^{-w(L)}[L]
\]
for welded links and
\[
Y(L)=r^{v(L)}(ar-b)^{-w(L)}[L]
\]
for extended welded links, with the explicit skein relations listed in the paper [1809.05874]. A further general mechanism is given in “Multi-switches and virtual knot invariants,” which attaches to a virtual link diagram \(D\) and a virtual biquandle multi-switch \((S,V)\) an algebraic system \(X_{S,V}(D)\) invariant under generalized Reidemeister moves [2001.06971].

These constructions show that topology uses “multi-” primarily to encode extra discrete structure: multiple components, multiple endpoint types, multiple virtual types, or multiple state resolutions.

## 5. Multi-channel, optical, and dynamical invariants

In applied mathematics and physics, multi-invariants often quantify which parts of a complex system remain unchanged under restricted transformation classes. Here the emphasis shifts from classification by isomorphism to robust feature extraction, decomposition of state space, or near-integrable dynamics.

“Gaussian-Hermite Moment Invariants of General Multi-Channel Functions” studies general multi-channel functions
\[
F(X):\Omega\subset\mathbb R^M\to\mathbb R^N
\]
under two transform models: the **rotation-affine transform**
\[
Y=R_{in}X,\qquad G(Y)=A_{out}F(X)+T_{out},
\]
and the **total rotation transform**, where \(A_{out}=R_{out}\). The unified framework uses the differential operators
\[
\phi_{12}=\nabla_1^T\nabla_2,\qquad \psi_{12\cdots M}=\left|\left(\nabla_1,\nabla_2,\cdots,\nabla_M\right)\right|,
\]
together with the primitives
\[
\Gamma_{12}=F(X_1)^TF(X_2),\qquad
\Lambda_{12\cdots N}=\left|\left(F(X_1),F(X_2),\cdots,F(X_N)\right)\right|,
\]
to generate Gaussian-Hermite moment invariants (MGHMIs). A key theorem states RA-relative invariance when \(\Gamma\)-type terms are excluded and TR absolute invariance in the rotational case. The paper derives independent invariant sets with \(13\) MGHMIs for RGB images, \(7\) TR-invariants and \(7\) RA-invariants for \(2\)D vector fields, and \(14\) MGHMIs for color volume data. On the synthetic RGB test set, the reported MGHMIs achieve \(100\%\) classification accuracy across tested \(\sigma\) values, with mean relative errors typically below \(1\%\) and maximum around \(1.47\%\) [2201.00877].

“Experimental Observation of Purity-Like Invariants of Multi-photon States in Linear Optics” considers a fixed \(n\)-photon, \(m\)-mode Fock space under realizable linear-optical unitaries. Using a Hermitian transfer matrix decomposition
\[
\mathrm H(M)=\mathcal T\oplus \mathcal P,
\]
the paper identifies three conserved quantities,
\[
I_n=\operatorname{tr}(H_0\rho)^2=\frac1M,\qquad
I_{t'}=\sum_{i=1}^{m^2-1}\operatorname{tr}(H_i\rho)^2,\qquad
I_p=\sum_{i=m^2}^{M^2-1}\operatorname{tr}(H_i\rho)^2,
\]
satisfying
\[
I_n+I_{t'}+I_p=\operatorname{tr}(\rho^2).
\]
Here \(I_{t'}\) is the traceless tangent contribution associated with single-photon dynamical degrees of freedom, whereas \(I_p\) is the perpendicular contribution tied to genuine multi-photon interference. The experiment on two-photon, two-mode polarization states confirms conservation of these quantities under sampled linear-optical unitaries [2506.12857].

In accelerator physics, “Design of double- and multi-bend achromat lattices with large dynamic aperture and approximate invariants” uses the term in a deliberately weaker sense. The target quantities are the Courant–Snyder actions
\[
J_x=\frac12(\bar x^2+\bar p_x^2),\qquad J_y=\frac12(\bar y^2+\bar p_y^2),
\]
which are exact invariants in a linear uncoupled lattice but only **quasi-invariants** in the nonlinear achromat setting. The optimization tunes sextupoles and octupoles to minimize turn-by-turn fluctuations of \(J_x\) and \(J_y\) and to minimize tune diffusion via NAFF, using NSGA-II. The paper states that five virtual particles are launched, with four objectives per launch, for a total of \(20\) objectives. The resulting DBA and MBA lattices exhibit large dynamic aperture, trajectories confined to invariant-like tori, robustness to resonances and errors, and a large amplitude-dependent tune spread; for the NSLS-II application, the reported single-bunch instability threshold increases from about \(2\!-\!3\) mA without BBFB and \(5\!-\!6\) mA with BBFB to greater than \(8\) mA without BBFB and \(12\) mA with BBFB [2103.06351].

Taken together, these examples show that in applied settings a “multi-invariant” may be exact, relative, or only approximate, provided it resolves several structurally distinct conserved sectors.

## 6. Related invariant programs and conceptual distinctions

A related but not identical literature studies invariant families in contexts where the adjective “multi-” modifies the underlying object rather than the invariant formalism itself. These works are closely adjacent to the multi-invariant program and clarify its boundaries.

“Permutation Centralizer Algebras and Multi-Matrix Invariants” develops the representation-theoretic framework underlying gauge-invariant observables built from several matrix species. Its central algebra is
\[
A(m,n)\equiv \text{the subalgebra of } \mathbb C[S_{m+n}] \text{ invariant under conjugation by } S_m\times S_n,
\]
and its Wedderburn–Artin decomposition explains the counting of restricted Schur operators. The dimension formula
\[
|A(m,n)|=\sum_{R_1\vdash m,\;R_2\vdash n,\;R\vdash m+n} g(R_1,R_2;R)^2
\]
expresses the block structure in terms of Littlewood–Richardson coefficients, and the paper further analyzes the center, maximally commuting subalgebra, and star product on fixed-degree invariants [1601.06086]. This is not a replica-based multi-invariant theory, but it exhibits the same shift from single invariants to structured invariant algebras.

A second adjacent program is the basis-independent study of CP in multi-Higgs models. “CP-odd invariants for multi-Higgs models and applications with discrete symmetry” writes the scalar potential as
\[
V={\phi^\ast}^aY_a^b\phi_b+{\phi^\ast}^a{\phi^\ast}^cZ_{ac}^{bd}\phi_b\phi_d
\]
and constructs basis invariants by complete index contraction. Under CP, such invariants go to their complex conjugates, so a CP-odd invariant has the form
\[
\mathcal I=I-I^\ast.
\]
The paper systematizes these constructions up to six \(Z\) tensors and uses them to diagnose explicit and spontaneous CP violation in potentials with triplet fields under \(A_4\), \(S_4\), \(\Delta(27)\), \(\Delta(54)\), \(\Delta(3n^2)\), and \(\Delta(6n^2)\) [1703.05750]. Again, the connection to multi-invariants is structural rather than terminological: the formalism studies invariant families attached to a multi-field system, but the indexing principle is tensor contraction rather than ramification type or replica symmetry.

A further nearby antecedent in quantum information is “Local Unitary Invariants of Generic Multi-qubit States,” which gives complete sets of LU polynomial invariants from generalized Bloch tensors. The paper proves that generic two-qubit states require at most \(12\) polynomial invariants and generic three-qubit states at most \(90\) [1507.02832]. This is not presented as a multi-invariant theory in the later replica sense, but it supplies a precursor notion of invariant completeness for multipartite states.

These adjacent literatures indicate a useful distinction. In the strictest contemporary sense, especially in arithmetic statistics and multipartite entanglement, a multi-invariant is a deliberately structured refinement indexed by several local types or by replica-contraction data. In a broader and older sense, it may denote any invariant formalism that becomes meaningful only after one simultaneously tracks several channels, components, or tensor sectors. The modern term therefore names a family resemblance rather than a single theorem schema: multivariate refinement in arithmetic, replica symmetry in quantum theory, type-resolved skein or quandle data in topology, and sector decompositions in applied invariant theory.

Source: https://www.emergentmind.com/topics/multi-invariants