Diagonal Modular Invariants

• Diagonal Modular Invariants are invariants that match paired sectors in rational conformal field theory, algebraic invariant theory over F_p, and Hurwitz Frobenius manifolds to enforce symmetry.
• They serve as benchmarks for physical realizability in full CFT by linking identity modular matrices, Lagrangian algebras, and NIM-rep interpretations with categorical and K-theoretic methods.
• Their diverse formulations reveal key structural insights through SAGBI complete intersections, dual autoequivalences, and exponential reparametrizations that connect algebraic and physical frameworks.

Searching arXiv for the cited papers to ground the article in current records. Across the cited literature, “diagonal modular invariant” denotes several distinct constructions rather than a single universally fixed object. In rational conformal field theory and modular tensor categories, it is the identity modular invariant $Z_{\lambda\mu}=\delta_{\lambda\mu}$, pairing each simple sector with itself and serving as a candidate partition function of a full conformal field theory (Davydov, 2014). In modular invariant theory over fields of characteristic $p$, the phrase refers instead to invariants for the simultaneous action on a vector and a covector, notably in the ring $\mathbb{F}_p[V\oplus V^*]^G$ (Ren, 2023). In genus-zero Hurwitz Frobenius manifolds, diagonal invariants are invariant combinations of paired variables $(\alpha_\mu,\beta_\mu)$ under simultaneous permutation and logarithmic translation symmetries (Proserpio et al., 2024). By contrast, in arithmetic algebraic geometry “modular invariants” of families of curves are intersection-theoretic quantities $y(f)=\deg(J_f^*y)$, and the sharp-bound theory for $\lambda(f)$, $\delta(f)$, and $K(f)$ does not introduce a separate notion of diagonal modular invariant (Liu et al., 2020).

1. Diagonality in rational conformal field theory

In the RCFT sense, a modular invariant is a matrix $Z_{\lambda\mu}\in \mathbb Z_{\ge 0}$ recording multiplicities in the decomposition of the full state space into irreducible chiral and anti-chiral modules. It must commute with the modular $S$- and $p$0-matrices. The diagonal modular invariant is the special case

$p$1

so that left-moving and right-moving labels are matched exactly. It stands alongside the charge-conjugation modular invariant as one of the two most basic modular invariants. Categorically, if $p$2 is the modular category of the chiral algebra, then a full CFT corresponds to a special commutative algebra object in $p$3, called a Lagrangian algebra, and the modular invariant is the class of that algebra in the Grothendieck ring (Davydov, 2014).

This formulation makes “diagonal” precise: there is no mixing between distinct sectors. In the $p$4-theoretic treatment of full CFT, the same invariant appears as the benchmark case $p$5. For the diagonal invariant, the nimrep is the Verlinde ring itself, the boundary states are the primaries, the nimrep matrices are the fusion matrices $p$6, the $p$7-inductions are the identity, and the full system and neutral system coincide with the chiral theory. In that sense, the diagonal modular invariant functions as the undeformed reference point for more complicated full-CFT structures (Evans et al., 2010).

2. Physicality and unphysical diagonal modular invariants

A central distinction in the RCFT literature is between formal modular invariance and physical realizability. A modular invariant is physical if it actually comes from a full conformal field theory, equivalently from a Lagrangian algebra in the appropriate braided product category. It is unphysical if it satisfies the formal modular invariance constraints but does not arise from any full CFT. Charge-conjugation invariants are always physical, but diagonal invariants can fail to be physical. This directly contradicts the common expectation that the identity matrix should always define a full theory (Davydov, 2014).

The categorical criterion is expressed through braided autoequivalences. A Lagrangian algebra $p$8 has the diagonal modular invariant if and only if

$p$9

for a dualising braided tensor autoequivalence $\mathbb{F}_p[V\oplus V^*]^G$0, where

$\mathbb{F}_p[V\oplus V^*]^G$1

and $\mathbb{F}_p[V\oplus V^*]^G$2 is dualising if $\mathbb{F}_p[V\oplus V^*]^G$3 for every simple object $\mathbb{F}_p[V\oplus V^*]^G$4. Thus the diagonal modular invariant is physical precisely when the chiral category admits a braided autoequivalence implementing duality on simple objects.

For $\mathbb{F}_p[V\oplus V^*]^G$5-orbifolds of holomorphic conformal field theories, the chiral modular category is the twisted Drinfeld centre $\mathbb{F}_p[V\oplus V^*]^G$6, and for trivial cocycle one has $\mathbb{F}_p[V\oplus V^*]^G$7. In that setting, the diagonal modular invariant of $\mathbb{F}_p[V\oplus V^*]^G$8 is physical if and only if $\mathbb{F}_p[V\oplus V^*]^G$9 has a double class-inverting automorphism, meaning an automorphism $(\alpha_\mu,\beta_\mu)$0 such that for every commuting pair $(\alpha_\mu,\beta_\mu)$1, there exists $(\alpha_\mu,\beta_\mu)$2 with

$(\alpha_\mu,\beta_\mu)$3

The paper identifies abundant obstructions: $(\alpha_\mu,\beta_\mu)$4 has no class-inverting automorphism, non-abelian groups of odd order cannot admit a class-inverting automorphism, and for $(\alpha_\mu,\beta_\mu)$5, $(\alpha_\mu,\beta_\mu)$6 has none. By contrast, abelian groups admit inversion $(\alpha_\mu,\beta_\mu)$7, and symmetric groups $(\alpha_\mu,\beta_\mu)$8 are doubly ambivalent, so the diagonal invariant is physical there (Davydov, 2014).

3. Categorical, $(\alpha_\mu,\beta_\mu)$9-theoretic, and NIM-rep interpretations

The diagonal modular invariant plays a structural role in the $y(f)=\deg(J_f^*y)$0-theoretic description of full CFT. For loop-group theories, Freed–Hopkins–Teleman identify the Verlinde ring with twisted equivariant $y(f)=\deg(J_f^*y)$1-theory,

$y(f)=\deg(J_f^*y)$2

and the full-CFT data attached to a modular invariant are organized through twisted equivariant $y(f)=\deg(J_f^*y)$3-groups. In this setting the diagonal invariant $y(f)=\deg(J_f^*y)$4 is the canonical case in which the Verlinde nimrep, full system, and neutral system all collapse to the same basic object. For tori, the diagonal invariant is the case $y(f)=\deg(J_f^*y)$5, $y(f)=\deg(J_f^*y)$6; for finite groups, it corresponds to the diagonal subgroup $y(f)=\deg(J_f^*y)$7 with trivial $y(f)=\deg(J_f^*y)$8; and for $y(f)=\deg(J_f^*y)$9, the diagonal invariant is $\lambda(f)$0, with nimrep equal to the $\lambda(f)$1 graph and charge group $\lambda(f)$2 (Evans et al., 2010).

A second line of development concerns the relation between diagonal entries of a modular invariant and NIM-rep multiplicities. Given a pivotal module category over a spherical fusion category, the encircling module $\lambda(f)$3 is defined as an $\lambda(f)$4-module using the pivotal structure. The main theorem identifies it with the NIM-rep: $\lambda(f)$5 Applied to the $\lambda(f)$6 realization of modular invariants, this yields

$\lambda(f)$7

so the diagonal part of the modular invariant is exactly the encircling module. Corollary 5.2 states that for $\lambda(f)$8, the multiplicity of $\lambda(f)$9 within the NIM-rep is equal to

$\delta(f)$0

This categorifies the Böckenhauer–Evans–Kawahigashi relation between modular invariants and nimreps. The same paper also proves that for indecomposable module categories the dimension condition $\delta(f)$1 is automatic, and that $\delta(f)$2 recovers the full centre construction of Kong and Runkel (King et al., 20 Mar 2026).

A useful distinction follows. The diagonal modular invariant in the RCFT sense is the entire identity matrix $\delta(f)$3. The diagonal entries of a general modular invariant are a different object, and their relation to NIM-reps is the content of the categorical results above. The coincidence of terminology can obscure this difference.

4. Diagonal invariants in modular invariant theory over $\delta(f)$4

In modular invariant theory, “diagonal” refers not to an RCFT partition function but to the mixed vector–covector invariant problem for

$\delta(f)$5

when $\delta(f)$6 acts through a cohyperplane representation. With basis $\delta(f)$7 of $\delta(f)$8, each generator acts by

$\delta(f)$9

and on the dual basis $K(f)$0 one has the contragredient action

$K(f)$1

The invariant-theoretic diagonal pairings are

$K(f)$2

These are called diagonal modular invariants because they mix vector and covector variables through Frobenius powers (Ren, 2023).

The main structural theorem states that $K(f)$3 is a complete intersection generated by

$K(f)$4

and that $K(f)$5 is a SAGBI basis with respect to the grevlex order

$K(f)$6

The leading terms

$K(f)$7

control the defining relations. The paper determines exactly $K(f)$8 relations $K(f)$9, proves that the only nontrivial tête-à-têtes are

$Z_{\lambda\mu}\in \mathbb Z_{\ge 0}$0

and concludes that the invariant ring is a SAGBI complete intersection. The contrast with $Z_{\lambda\mu}\in \mathbb Z_{\ge 0}$1, which is complete intersection only when $Z_{\lambda\mu}\in \mathbb Z_{\ge 0}$2 and is not Cohen–Macaulay when $Z_{\lambda\mu}\in \mathbb Z_{\ge 0}$3, shows that adding dual variables can substantially improve algebraic structure (Ren, 2023).

5. Diagonal invariants in genus-zero Hurwitz Frobenius manifolds

In the Hurwitz-Frobenius setting, diagonal invariants arise from paired variables associated with simple poles. The symmetry acts by simultaneous permutation

$Z_{\lambda\mu}\in \mathbb Z_{\ge 0}$4

and by logarithmic shifts

$Z_{\lambda\mu}\in \mathbb Z_{\ge 0}$5

Because of these shifts, the relevant invariant ring is

$Z_{\lambda\mu}\in \mathbb Z_{\ge 0}$6

not merely a symmetric-polynomial ring in the $Z_{\lambda\mu}\in \mathbb Z_{\ge 0}$7. The basic generators are the exponential polarized power sums

$Z_{\lambda\mu}\in \mathbb Z_{\ge 0}$8

and especially

$Z_{\lambda\mu}\in \mathbb Z_{\ge 0}$9

These are the diagonal invariants of the theory (Proserpio et al., 2024).

Their significance is tied to the structure of the prepotential. The prepotential decomposes as

$S$0

with universal pole contribution

$S$1

and an interaction term written as a finite linear combination of the $S$2 and related polarized sums. Although individual flat coordinates can be complicated rational functions, the paper proves that the relevant combinations appearing in the prepotential are polynomial after suitable exponential reparametrization. Concretely,

$S$3

lies in the appropriate symmetric polynomial ring in the exponentials of the zero and pole coordinates. The mechanism is invariant-theoretic: the denominator becomes a Vandermonde factor, the numerator is antisymmetric, and divisibility by the Vandermonde determinant produces a symmetric polynomial (Proserpio et al., 2024).

This usage is conceptually close to the vector–covector story: “diagonal” again refers to invariance under a simultaneous action on paired variables. It is not a statement about modular $S$4- and $S$5-matrices.

6. Terminological boundary with modular invariants of families of curves

A different use of “modular invariant” occurs in arithmetic algebraic geometry. For a fibration $S$6 of genus $S$7 and a rational divisor class $S$8 on $S$9, the modular invariant is

$p$00

where $p$01 is the induced moduli map. The standard invariants are $p$02, $p$03, and

$p$04

For $p$05, $p$06 and $p$07 characterize isotrivial families, so in the non-isotrivial case these invariants are positive rational numbers (Liu et al., 2020).

The cited sharp-bound results concern lower bounds and extremal classifications, not a separate diagonal theory. In genus $p$08, a non-isotrivial fibration satisfies

$p$09

and each bound is sharp. Equality is characterized by explicit singular-fiber dual graphs, and there is a rigidity statement: if the total space is rational, then there are only finitely many genus-$p$10 fibrations with $p$11. In genus $p$12, the paper gives corresponding sharp lower bounds and classifies extremal singular-fiber graphs. A central technical input is the relation with fractional Dehn twist coefficients $p$13, including

$p$14

The paper explicitly states that it does not formulate its results in terms of a “diagonal modular invariant” in the RCFT sense. The closest analogue is the minimal or extremal configuration in which all singular fibers are smooth after stable reduction except one with one of a short list of special dual graphs. This suggests only a loose analogy with diagonal extremality, not an actual terminological identification (Liu et al., 2020).

The distinction is important. In the RCFT literature, diagonal modular invariants are matrices or algebra objects constrained by modular data. In the geometry of families of curves, modular invariants are intersection numbers on moduli space. The shared word “modular” does not imply a shared definition of “diagonal.”

7. Conceptual synthesis

Taken together, the cited works isolate three mathematically substantive meanings of diagonal modular invariants.

First, in RCFT and modular tensor categories, the diagonal modular invariant is the identity modular invariant, and the basic questions concern existence of a full CFT, realization by Lagrangian algebras, group-theoretic criteria such as double class-inverting automorphisms, and the relation between diagonal multiplicities and NIM-reps (Davydov, 2014).

Second, in algebraic invariant theory over fields of characteristic $p$15, diagonal modular invariants are Frobenius-twisted pairings between a representation and its dual. The emphasis is on explicit generators, SAGBI bases, complete-intersection presentations, and structural comparison with other modular invariant rings (Ren, 2023).

Third, in Hurwitz Frobenius manifolds, diagonal invariants are simultaneous invariants of paired residue and logarithmic pole coordinates. Their role is to organize the quotient geometry and to render the interaction part of the prepotential polynomial or polynomial-like after exponential reparametrization (Proserpio et al., 2024).

A recurring misconception is that “diagonal” always means the identity matrix and is automatically physical. The literature refutes both parts of that claim. In RCFT, the diagonal invariant may be unphysical; in invariant theory and Frobenius-manifold theory, “diagonal” describes the symmetry action on paired variables rather than a partition function; and in arithmetic geometry the term is not standard at all for modular invariants of families of curves. The common thread is exact matching under a paired action, but the ambient structures, realizability criteria, and applications are field-specific.