---
title: 'DRalgo: Dimensional-Reduction EFT Toolkit'
url: https://www.emergentmind.com/topics/dralgo
type: topic
---

# DRalgo: Dimensional-Reduction EFT Toolkit

DRalgo is an algorithmic implementation and Mathematica package for constructing an effective, dimensionally reduced, high-temperature field theory for generic models. In the thermal phase-transition literature, it denotes a workflow that integrates out nonzero Matsubara modes and, when required, temporal gauge scalars to obtain three-dimensional effective field theories at the hard, soft, and ultrasoft scales; performs matching to next-to-leading order; computes beta functions, anomalous dimensions, Debye screening masses, and perturbative thermal potentials; and exports these results for numerical phase-transition, bubble-nucleation, and gravitational-wave calculations [2205.08815][2501.01286].

## 1. Definition, scope, and development

The original DRalgo release presented the package as “a package for effective field theory approach for thermal phase transitions” and described it as an algorithmic implementation that constructs an effective, dimensionally reduced, high-temperature field theory for generic models. Its stated core capabilities are automatic matching to next-to-leading order, including two-loop thermal corrections to scalar and Debye masses as well as one-loop thermal corrections to couplings; the possibility of integrating out additional heavy scalars; leading-order beta functions for general gauge-charges and fermion-families in both the fundamental and effective theory; and computation of the finite-temperature effective potential within the effective theory [2205.08815].

Subsequent work expanded the package from super-renormalizable three-dimensional operators to the one-loop matching of generic three-dimensional dimension-five and dimension-six operators for arbitrary models containing scalars, fermions, and gauge fields. In that extension, DRalgo is described as a Mathematica-based toolkit that provides the complete three-dimensional operator bases up to mass dimension six, generates the tensor structure of each Wilson coefficient in terms of the original four-dimensional couplings, computes the one-loop thermal sum-integrals in a general $R_\xi$ background-field gauge, and solves the matching relations for any subset of operators in a fully symbolic $d=3-2\varepsilon$ framework [2605.15176].

A possible source of confusion is purely terminological. In the supplied material, unrelated descriptions also attach the string “DRalgo” to a distributed augmented Lagrangian method and to a distributed retrieval-augmented generation framework. In the arXiv phase-transition literature considered here, however, DRalgo denotes the dimensional-reduction package and its extensions.

## 2. Dimensional-reduction framework

The package is built around high-temperature dimensional reduction. At temperatures $T \gg$ any zero-temperature mass, a four-dimensional gauge theory can be reduced to a three-dimensional Euclidean effective field theory by integrating out nonzero Matsubara modes and, optionally, additional heavy fields whose masses satisfy $M \gtrsim \pi T$. The result is a super-renormalizable three-dimensional theory for the static modes [2205.08815].

The scale hierarchy is organized explicitly:

| Scale | Parametric size | DR step |
|---|---:|---|
| hard | $\mu_{4d}\approx \pi T$ | full Matsubara tower |
| soft | $\mu_{3d}^{s}\approx gT$ | after integrating out nonzero modes |
| ultrasoft | $\mu_{3d}^{Us}\approx g^2T$ | after integrating out temporal $A_0$ scalars |

This separation is central to the package’s treatment of infrared sensitivity. In the standard four-dimensional high-$T$ expansion, bosonic zero modes produce infrared divergences and motivate daisy resummation. The dimensional-reduction formulation replaces that ad hoc reorganization with a systematic EFT construction in which ultraviolet and infrared physics are separated cleanly. In the high-$T$ regime, the dimensional-reduction procedure is stated to be equivalent—order by order in a combined coupling/high-$T$ expansion—to thermal-mass daisy resummation, but with the advantage of a controlled EFT framework [2505.04744].

The reduced theory can be written schematically as
$$
\mathcal L_3=\tfrac12(\nabla\phi_3)^2+\sum_{i,j}\tfrac14 F_{ij}^2+V_{3,0}(\phi_3),
$$
with tree-level potential
$$
V_{3,0}(\phi_3)=\tfrac12 m_3^2\phi_3^2+\tfrac14\lambda_3\phi_3^4.
$$
The one-loop correction is UV-finite in three dimensions,
$$
V_{3,1}(\phi_3)=-\frac{1}{12\pi}\sum_a n_a\,[m_a^2(\phi_3)]^{3/2},
$$
where the sum runs over the light three-dimensional bosonic degrees of freedom, including scalars and longitudinal gauge fields [2412.04881].

## 3. Matching, running, and effective potentials

DRalgo implements hard-to-soft and soft-to-ultrasoft dimensional reduction for a general renormalizable four-dimensional Lagrangian. In the original package description, the matching order is next-to-leading order: all two-loop thermal corrections to scalar masses and to Debye masses, and all one-loop thermal corrections to scalar quartics, gauge couplings, Yukawas, and $A_0$ self-couplings [2205.08815].

The package also computes all required four-dimensional and three-dimensional beta functions, thermal masses, anomalous dimensions, and two-loop thermal corrections to scalar masses. It automatically generates analytical expressions for the three-dimensional parameters, commonly denoted by hats, such as $\hat\mu_H^2(T)$, $\hat\lambda_H(T)$, and $\hat g_{HS}(T)$, together with the NLO contribution to the three-dimensional potential,
$$
V_{3d}(v;T)=V_{3d}^{(0)}(v;T)+V_{3d}^{(1)}(v;T),
$$
and supports export of the matching formulae and three-dimensional potentials in Mathematica or Python form [2501.01286].

A central mapping relates the three-dimensional EFT back to the thermal four-dimensional effective potential:
$$
V_4(\phi_4,T)=T\,V_3(\phi_3),\qquad \phi_4=\sqrt{T}\,\phi_3.
$$
In the leptoquark application, the same relation is written as
$$
v_i^{3d}\to v_i^{4d}/\sqrt T,\qquad V_{\rm eff}^{4d}(v^{4d},T)=T\cdot V_{\rm eff}^{3d}(v^{3d},T).
$$
This mapping is what allows DRalgo-generated EFTs to be passed directly into bounce-action and gravitational-wave solvers [2501.01286].

The supercooled-phase-transition analysis of the Abelian Higgs model makes explicit how DRalgo combines matching with renormalisation-group improvement. There, consistent RGE evolution of the couplings is described as essential for a meaningful interpretation of the results, and the four-dimensional parameters are matched onto three-dimensional EFT parameters order by order in a loop expansion. Representative one-loop matchings are
$$
g_3^2=g^2T-\frac{g^4T}{48\pi^2}L_b,
$$
$$
\lambda_3=\lambda T+\frac{T}{16\pi^2}\bigl[g^4(2-3L_b)+6g^2\lambda L_b-10\lambda^2L_b\bigr],
$$
while the scalar mass and Debye mass are matched at two loops [2511.02910].

The same study distinguishes several three-dimensional loop-ordering schemes: “3D 1-L (LO),” “3D 1-L (NLO),” “3D 2-L (NLO),” and “3D 2-L (Mixed).” This suggests that DRalgo is not merely a code generator for one fixed perturbative truncation, but a framework for controlled comparisons between different EFT expansion schemes when residual scale dependence and convergence are under scrutiny.

## 4. Software architecture and user workflow

The package is implemented in Mathematica and exposes a model-building and matching workflow in terms of named routines. A typical usage pattern begins by loading or defining the four-dimensional model, for example through `ImportModelDRalgo[...]` after tensor allocation and construction of invariant structures. The hard-to-soft step is then executed with `PerformDRhard[]`, after which the package can print soft EFT couplings and masses through calls such as `PrintCouplings[]`, `PrintScalarMass["LO"]`, `PrintScalarMass["NLO"]`, `PrintDebyeMass["LO"]`, `PrintDebyeMass["NLO"]`, and `BetaFunctions4D[]`. If desired, `PerformDRsoft[{heavyIndices}]` integrates out soft heavy scalars or temporal gauge fields to produce an ultrasoft EFT, with corresponding accessors such as `PrintCouplingsUS[]`, `PrintScalarMassUS[...]`, and `BetaFunctions3DUS[]`. The perturbative three-dimensional effective potential is then assembled via `CalculatePotentialUS[]` and retrieved with `PrintEffectivePotential["LO"/"NLO"/"NNLO"]` [2205.08815].

The internal logic of later pipelines makes the same structure explicit in algorithmic terms. In the leptoquark study, the DRalgo-based thermal potential is constructed by first RG-evolving four-dimensional parameters to $\mu_{4d}=C_{4d}T$, typically with $C_{4d}=\pi$; then performing hard-to-soft matching; then choosing an ultrasoft matching scale $\mu_{3d}^{Us}=C_{3d}m_D^{\rm min}$, with $C_{3d}=1$ by default; then integrating out temporal fields; and finally building the three-dimensional effective potential before վերադարձing the four-dimensional form $T\,V_{3d}$ [2501.01286].

DRalgo’s output format is intentionally designed for downstream use. The package can export matching relations and potentials in Mathematica or Python form, and the higher-dimensional-operator extension continues this design by returning Mathematica rule lists for Wilson coefficients. In that extension, `Dimension5Matching[...]` and `Dimension6Matching[...]` solve the hard-mode matching problem symbolically, while `HardThermal1LoopInt["B",s,\alpha,d]` and `HardThermal1LoopInt["F",s,\alpha,d]` expose the underlying bosonic and fermionic thermal master integrals [2605.15176].

## 5. Integration with phase-transition and gravitational-wave toolchains

DRalgo has become part of several end-to-end software stacks for cosmological phase transitions. The most explicit interfaces described in the literature are Dratopi, PhaseTracer2, and PT2GWFinder.

| Interface | How DRalgo is used | Notable detail |
|---|---|---|
| Dratopi | reads DRalgo-generated Python code and builds ultrasoft EFT parameters | wraps a modified `CosmoTransitions` |
| PhaseTracer2 | hand-translates DRalgo Mathematica output into C++ model methods | uses `solveBetas()` and `get3d_parameters()` |
| PT2GWFinder / DRTools | imports DRalgo output and reconstructs $V(\phi,T)$ | `ComputeDRPotential[...]` returns $T\,V_{3D}(\phi/\sqrt T,T)$ |

Dratopi is described as “a thin Python layer” that reads in the DRalgo-generated Python code, numerically solves the four-dimensional RGEs from an input scale $\mu_0$ to $\mu_{4d}=\pi T$ while enforcing user-specified constraints such as perturbativity and bounded-from-below conditions, uses the DRalgo matching routines to build the ultrasoft three-dimensional EFT parameters $p_{3d}^{Us}(T)$, provides an auto-diagonalization of field-dependent mass matrices for the NLO thermal potential, and wraps `CosmoTransitions` with minor patches so that the potential
$$
V_{4d}(v,T)\equiv T\cdot V_{3d}(v/\sqrt T;p_{3d}^{Us}(T))
$$
is called at nonzero $T$ and the condition $m_{\rm Debye}^2>0$ defines the temperature range. It also offers “sanity checks,” notably the ratio $\max m_{3d}^{Us}(v,\mu)/\mu$, to monitor high-$T$ perturbativity [2501.01286].

PhaseTracer2 incorporates dimensionally reduced effective potentials for models obtained from DRalgo. Rather than file I/O, it ingests the output through two user-written member functions in a model class deriving from `EffectivePotential::Potential`: `solveBetas(x,t)`, which solves the one-loop $\overline{\rm MS}$ beta functions from the input scale $Q_0$ to $\mu_{4D}=\pi T$, and `get3d_parameters(T)`, which evaluates the four-dimensional couplings at $\mu_{4D}=\pi T` and applies the DRalgo matching formulae to return $\{m_3^2,\lambda_3,g_3^2,\dots\}$. Once those methods are supplied, phase finding, bounce-action, thermodynamic-parameter, and gravitational-wave routines remain unchanged [2412.04881].

PT2GWFinder exposes DRalgo through a Mathematica sub-package called `DRTools`. The high-level command
```mathematica
ComputeDRPotential[benchmarkRules, {Tref, Tmin, Tmax},
  "SubRules" -> ...,
  "OrderDR" -> "NLO",
  "OrderVeff" -> "NNLO",
  "RescaleTo4D" -> True]
```
imports raw DRalgo expressions, solves the four-dimensional RG equations numerically over the chosen range, applies the matching formulae at each temperature via `DRStep`, reconstructs the analytic three-dimensional potential via `DRPotentialN`, and optionally rescales it to the four-dimensional potential. The primary returned object is the closed-form function
$$
V(\phi,T)=T\,V_{3D}\!\bigl(\phi/\sqrt T,T\bigr),
$$
which can then be passed directly to the package’s transition finder [2505.04744].

## 6. Applications, validation, and later extensions

The package has been applied to models with qualitatively different thermal histories. In a minimal leptoquark extension of the Standard Model that explains active neutrino oscillation data while satisfying current flavor physics constraints, the DRalgo–Dratopi chain automated the construction of a three-dimensional high-$T$ EFT from the original four-dimensional model, computed the resulting three-dimensional thermal potential at NLO, and fed that potential into bubble-nucleation and gravitational-wave routines. The study reports diverse phase-transition patterns, including color symmetry-breaking scenarios in the early Universe, and states that a detectable signal in the $\mathrm{mHz}$–$0.1~\mathrm{Hz}$ frequency range features color-restoration and leptoquark masses near $1.5~\mathrm{TeV}$ [2501.01286].

That application also provides explicit performance and robustness observations. The DRalgo+Dratopi chain was able to scan $O(10^5)$ model points, producing $\sim 2\times10^4$ viable phase transitions with $\beta/H<10^5$ in a matter of hours on a modern multicore workstation. Varying the hard-matching prefactor $\mu_{4d}/(\pi T)\in[0.5,2]$ changed the predicted GW peak amplitude by $\Delta h^2\Omega_{\rm GW}\lesssim 0.5$, which is presented as evidence that the explicit removal of large logarithms and inclusion of NLO terms shrinks theoretical uncertainties compared to standard four-dimensional high-$T$ potentials. The high-$T$ perturbativity diagnostic $m_{US}/\mu$ clustered around $O(1$–$2)$ for all transitions, and imposing $m_{US}/\mu<1$ eliminated only very few points on the edge [2501.01286].

Independent validation comes from software integrations. In PhaseTracer2, DR-EFT potentials are described as simpler than full four-dimensional one-loop potentials because they avoid the special functions $J_B$ and $J_F$ and instead enter through algebraic $\phi^{3/2}$ terms. In a typical two-field example, the paper reports a $\sim 2$–$3\times$ speed-up in the phase-structure scan relative to the full four-dimensional Coleman–Weinberg plus daisy potential, negligible overhead in RG running at $\sim 0.1\,\mathrm{ms}$ per temperature point, agreement of critical temperatures and latent heats at $\mathcal O(10\%)$ with the full four-dimensional $\overline{\rm MS}$ plus Parwani result for benchmark singlet-model points, and improved stability in strong supercooling because infrared modes are integrated out analytically [2412.04881]. In PT2GWFinder, the same $80\times80$ grid scan in $(\lambda_S,g_D)$ that took about $6\,\mathrm m\,40\,\mathrm s$ with the Coleman–Weinberg plus daisy implementation ran in roughly $5\,\mathrm m\,47\,\mathrm s$ with the DR approach, while yielding qualitatively similar and in many regions quantitatively similar phase-transition observables [2505.04744].

The supercooled dark-sector study sharpened the package’s role in precision comparisons between approximation schemes. There, the three-dimensional EFT with consistent expansion in the four-dimensional parameters is reported to have significantly reduced scale dependence. For the benchmark “dark photon” model, varying scales in the four-dimensional high-temperature one-loop scheme produces a fractional variation $\Delta(S_3/T)/(S_3/T)\sim 20\%$, while the three-dimensional two-loop scheme yields $\sim 5\%$. The paper further states that the four-dimensional high-temperature one-loop and three-dimensional two-loop schemes give nearly identical $S_3/T$ curves and agree on the nucleation temperature to within a few percent over the range $g\in[0.5,0.8]$, while the propagated nanohertz gravitational-wave peak amplitude and frequency differ by less than $10\%$; by contrast, the simple analytic parametrisation can deviate by tens of percent in the limit of large supercooling [2511.02910].

The 2026 higher-dimensional-operator extension broadens DRalgo’s remit from super-renormalizable three-dimensional EFTs to complete operator bases through dimension six. This extension is motivated by the statement that strong first-order cosmological transitions often require subleading hard-mode effects to be under analytic control, and that these higher-dimensional operators are crucial for quantifying the convergence of the high-temperature expansion and stabilizing nonperturbative lattice studies of three-dimensional EFTs. The package now automates matching for arbitrary models of scalars, fermions, and gauge fields, but still reports coefficients in a redundant off-shell basis with explicit gauge-parameter dependence; elimination of redundancies and associated field redefinitions are described as being left to the user or to future DRalgo updates [2605.15176].

Taken together, these developments place DRalgo at the intersection of thermal EFT construction, perturbative control of infrared physics, renormalisation-group improvement, and end-to-end phenomenology of cosmological first-order phase transitions. A plausible implication is that its main scientific role is not only to produce reduced theories, but to standardize the passage from an arbitrary renormalizable four-dimensional Lagrangian to numerically tractable three-dimensional actions whose uncertainties can be diagnosed through matching-scale variation, loop-order comparisons, and higher-dimensional-operator estimates.

Source: https://www.emergentmind.com/topics/dralgo