Murua Coefficients in Magnus Expansion

• Murua coefficients are rational weights assigned to directed graphs that encode the nested-commutator structure of the Magnus expansion in quantum field theory.
• They are computed using recursive combinatorial methods, including partitioning and edge contraction rules, which apply to both tree and loop diagrams.
• In practical QFT applications, these coefficients facilitate the conversion from Dyson to Magnus representations, reducing the need for explicit cut subtractions in master integrals.

Searching arXiv for papers directly discussing Murua coefficients and their use in Magnus-expansion/QFT settings. Murua coefficients are combinatorial graph weights that arise in Magnus-expansion formulations of time evolution, where the $S$-matrix or time-evolution operator is written as an exponential of a logarithmic generator, such as $S=e^{iN}$ or $$ih\,X=\log\!\left[T\exp\!\left(-\frac{1}{ih}\int H_I(t)\,dt\right)\right]$$. In the recent quantum-field-theoretic literature, they are denoted by $\omega(\tau)$ or $w(G)$ and assign the rational prefactor multiplying a directed tree or loop graph in the expansion of the Magnusian. At tree level they weight diagrams built from retarded and advanced propagators; at loop level they extend to cut graphs and more general acyclic graph topologies. In this sense, Murua coefficients encode the nontrivial nested-commutator combinatorics of the Magnus series in a graph-theoretic language (Brandhuber et al., 4 Dec 2025).

1. Definition and placement in the Magnus expansion

In the relativistic-QFT formulation, Murua coefficients are introduced as the combinatorial factors multiplying diagrams in the perturbative expansion of the $N$-operator defined by

$S=e^{iN}.$

The corresponding matrix elements are termed Magnus amplitudes. The operative statement is that tree-level Magnus amplitudes can be written as sums over directed tree diagrams built from retarded and advanced propagators, with each tree weighted by a coefficient $\omega(\tau)$, and these $\omega(\tau)$ are identified as Murua coefficients (Brandhuber et al., 4 Dec 2025).

A closely related formulation uses the “Magnusian,” the logarithm of the time-evolution operator,

$$ih\, X=\log\!\left[T\exp\!\left(-\frac{1}{ih}\int H_I(t)\,dt\right)\right].$$

In that setting, the graph expansion is written schematically as

$S=e^{iN}$0

and more precisely in two graph bases as

$S=e^{iN}$1

and

$S=e^{iN}$2

Here $S=e^{iN}$3, $S=e^{iN}$4, and $S=e^{iN}$5 are exactly the Murua coefficients in generic, color-basis, and black-and-white-basis notation, respectively (Guo et al., 25 May 2026).

These coefficients are distinct from ordinary symmetry factors. The graph prefactor is $S=e^{iN}$6 or $S=e^{iN}$7, where $S=e^{iN}$8 is the graph automorphism factor and the Murua coefficient is the additional Magnus-specific rational weight. This distinction is central in both the relativistic-QFT and diagrammatic-Magnusian treatments (Brandhuber et al., 4 Dec 2025).

2. Tree-level structure and directed causal graphs

At tree level, Murua coefficients arise because the Magnus expansion is built from nested commutators rather than time-ordered products. The first orders are

$S=e^{iN}$9

$$ih\,X=\log\!\left[T\exp\!\left(-\frac{1}{ih}\int H_I(t)\,dt\right)\right]$$0

$$ih\,X=\log\!\left[T\exp\!\left(-\frac{1}{ih}\int H_I(t)\,dt\right)\right]$$1

and similarly at fourth order. The commutators generate Pauli–Jordan functions, which combine with $$ih\,X=\log\!\left[T\exp\!\left(-\frac{1}{ih}\int H_I(t)\,dt\right)\right]$$2-functions into retarded and advanced propagators,

$$ih\,X=\log\!\left[T\exp\!\left(-\frac{1}{ih}\int H_I(t)\,dt\right)\right]$$3

After rearrangement, each surviving contribution is a directed graph whose edge orientations encode causal flow. Murua coefficients are the rational weights of these directed graphs (Brandhuber et al., 4 Dec 2025).

Low-order values are explicit. For the one-edge tree appearing at four points,

$$ih\,X=\log\!\left[T\exp\!\left(-\frac{1}{ih}\int H_I(t)\,dt\right)\right]$$4

corresponding to

$$ih\,X=\log\!\left[T\exp\!\left(-\frac{1}{ih}\int H_I(t)\,dt\right)\right]$$5

For two-edge trees at five points, the coefficients are

$$ih\,X=\log\!\left[T\exp\!\left(-\frac{1}{ih}\int H_I(t)\,dt\right)\right]$$6

depending on the causal orientation class. Equivalently, for the line graph the assignments $$ih\,X=\log\!\left[T\exp\!\left(-\frac{1}{ih}\int H_I(t)\,dt\right)\right]$$7, $$ih\,X=\log\!\left[T\exp\!\left(-\frac{1}{ih}\int H_I(t)\,dt\right)\right]$$8, $$ih\,X=\log\!\left[T\exp\!\left(-\frac{1}{ih}\int H_I(t)\,dt\right)\right]$$9, and $\omega(\tau)$0 carry weights $\omega(\tau)$1, $\omega(\tau)$2, $\omega(\tau)$3, and $\omega(\tau)$4. For three-edge trees at six points, the coefficients include

$\omega(\tau)$5

with the all-in and all-out star topologies vanishing. The paper states that these values correspond precisely to the Murua coefficients (Brandhuber et al., 4 Dec 2025).

A fundamental structural property is that $\omega(\tau)$6 depends only on the internal directed graph obtained after deleting external legs. Another is arrow-reversal symmetry: because $\omega(\tau)$7, reversing all internal causality arrows leaves the coefficient unchanged (Brandhuber et al., 4 Dec 2025).

3. Recursive and combinatorial constructions

The modern literature treats Murua coefficients as the graph-theoretic realization of Magnus recursion. For trees, the diagrammatic-Magnusian paper recalls the classical recursive formula

$\omega(\tau)$8

The notation is part of the construction: $\omega(\tau)$9 is a set of partitions relative to a root or semi-root $w(G)$0, $w(G)$1 are Bernoulli numbers, $w(G)$2 is a linear-extension count, and $w(G)$3 records the relevant sign-generating flips (Guo et al., 25 May 2026).

The same paper extends this recursion from trees to loop graphs. In the color basis the loop-level formula is

$w(G)$4

while in the black-and-white basis it becomes

$w(G)$5

These formulas are presented as the loop-level extension of Murua’s recursive formula. The new objects $w(G)$6 are “almost-rooted” loop graphs associated with a partition $w(G)$7, and the BW version carries additional $w(G)$8 factors (Guo et al., 25 May 2026).

The same work generalizes the linear-extension function $w(G)$9 from trees to acyclic loop graphs and records the contraction identity

$N$0

This identity underlies the graph-level contraction rules for Murua coefficients and shows that the coefficients are not ad hoc weights but part of a structured recursive calculus (Guo et al., 25 May 2026).

A plausible implication is that Murua coefficients occupy a role analogous to other recursively defined coefficient systems in combinatorial algebra. Indirectly, work on $N$1-nomial and multi-$N$2-nomial coefficients develops composition-indexed inversion formulas and multinomial decompositions for lower-triangular arrays, although that paper does not mention Murua coefficients by name (0806.3626).

4. Loop extensions, cuts, and basis dependence

Beyond tree level, Murua coefficients remain graph weights but the graph class enlarges to include loops and cuts. In the relativistic-QFT treatment, the general expansion is

$N$3

and this is adopted as the definition of $N$4 for multiloop graphs as well. At one loop, Magnus amplitudes always contain exactly one Hadamard cut function, and the paper proves that $N$5-point one-loop Magnus amplitudes are determined by phase-space integrals of forward limits of $N$6-point tree-level amplitudes (Brandhuber et al., 4 Dec 2025).

The key one-loop relation is

$N$7

and explicit examples reproduce this rule: for the simplest one-loop graph the value is $N$8, matching $N$9, while triangle values match half the corresponding five-point tree coefficients. For maximally cut higher-loop pieces, the same paper states

$S=e^{iN}.$0

and conjectures the more general relation

$S=e^{iN}.$1

Thus loop-level Murua coefficients inherit tree-level values up to powers of $S=e^{iN}.$2 associated with cuts (Brandhuber et al., 4 Dec 2025).

The diagrammatic-Magnusian paper formulates the same loop problem in two complementary bases. The color basis uses directed red and blue edges and is adapted to the algebraic structure of nested commutators. The black-and-white basis uses directed black retarded edges and undirected white cut edges and makes Lorentz invariance and term-by-term Hermiticity manifest. The two coefficient functions, $S=e^{iN}.$3 and $S=e^{iN}.$4, are related by explicit basis-change formulas rather than being graph-by-graph identical (Guo et al., 25 May 2026).

That basis dependence is substantive rather than notational. In the BW basis, cut edges are combinatorially special, while in the color basis the central distinction is between the two causal colorings of directed edges. The paper’s “fuzzy propagator” formalism packages descendant graphs containing banana loops and shows that many descendant graphs inherit the same Murua coefficient as the corresponding primary graph, up to symmetry-factor accounting and, in the BW basis, the prescribed $S=e^{iN}.$5-type modifications (Guo et al., 25 May 2026).

5. Edge contraction rules and direct computation

A major development is the replacement of explicit Magnus-series manipulation by graph-only recursion. In the relativistic-QFT tree analysis, the edge-contraction rule is stated as

$S=e^{iN}.$6

together with the sum rule

$S=e^{iN}.$7

for all causal arrow assignments on a fixed undirected tree $S=e^{iN}.$8. The paper also notes that its sign convention differs from some earlier literature by a factor of $S=e^{iN}.$9, where $\omega(\tau)$0 is the number of vertices (Brandhuber et al., 4 Dec 2025).

The diagrammatic-Magnusian paper upgrades these statements to loop-level contraction identities. In the color basis,

$\omega(\tau)$1

$\omega(\tau)$2

where $\omega(\tau)$3 denotes edge contraction and $\omega(\tau)$4 denotes edge removal, રાખing only the connected remainder. In the BW basis, the rules simplify to

$\omega(\tau)$5

for a cut edge, and

$\omega(\tau)$6

for a retarded edge (Guo et al., 25 May 2026).

These identities yield an efficient recursive algorithm. In the color basis, one fixes an underlying non-colored directed topology, enumerates edge colorings, identifies at least one coloring with $\omega(\tau)$7, and then solves the family recursively by contraction. In the BW basis, one first erases cut edges whenever connectivity is preserved, then reduces the remaining retarded graph by contraction until only tree graphs or known primary representatives remain. The authors describe these contraction rules as offering the fastest way to produce Murua coefficients (Guo et al., 25 May 2026).

Further structural constraints support this recursion. The color-basis parity relation is

$\omega(\tau)$8

while the BW basis imposes

$\omega(\tau)$9

The paper also states a zero-sum rule in the color basis: for a fixed non-colored directed acyclic graph with $\omega(\tau)$0 edges, the sum of $\omega(\tau)$1 over all $\omega(\tau)$2 colorings vanishes (Guo et al., 25 May 2026).

6. Physical role in quantum field theory and gravitational scattering

Murua coefficients are not only combinatorial data; they have a direct computational role in extracting the Magnusian from perturbative QFT. In gravitational-wave scattering at fourth order in Newton’s constant, the central practical problem is converting the scattering amplitude into the Magnusian, the matrix element of the Magnus operator $\omega(\tau)$3 defined by

$\omega(\tau)$4

The paper’s main methodological claim is that a decomposition of master integrals incorporating Murua coefficients bypasses the cut subtraction previously needed for this conversion (Bautista et al., 25 Jun 2026).

The practical definition is given at the level of integrals with $\omega(\tau)$5 active propagators. The Murua value is

$\omega(\tau)$6

The superscripts indicate the causality prescription of each active propagator. A footnote extrapolates the pattern: for an integral with $\omega(\tau)$7 active propagators, a term with $\omega(\tau)$8 advanced propagators has coefficient

$\omega(\tau)$9

These weights sum to $$ih\, X=\log\!\left[T\exp\!\left(-\frac{1}{ih}\int H_I(t)\,dt\right)\right].$$0 and satisfy the edge-contraction rule (Bautista et al., 25 Jun 2026).

The computational significance is that one can construct the integrand with ordinary Feynman rules, reduce to master integrals by IBP as usual, and only then replace each master integral by its Murua value. The reason this works is that WQFT Feynman rules and IBP reduction are agnostic to causality prescriptions, while the Magnusian differs from the amplitude precisely in those prescriptions for the relevant process (Bautista et al., 25 Jun 2026).

In that application, the Magnusian maps directly to the black-hole-perturbation-theory phase shifts. The Murua prescription therefore functions as a direct Dyson-to-Magnus converter at the level of master integrals, avoiding explicit construction and subtraction of cuts. The same paper states that the Murua value of an integral is always real and often has softer poles than the original integral, or even vanishes in many cases, which helps explain the infrared-safe character of the Magnusian in that setting (Bautista et al., 25 Jun 2026).