--- title: 'TriForces: Triadic Structures in Graphs, Magnetism & MLIPs' url: https://www.emergentmind.com/topics/triforces type: topic --- # TriForces: Triadic Structures in Graphs, Magnetism & MLIPs TriForces is a nonstandard label applied to several distinct three-part constructions in recent research. In graph theory it naturally designates forcing triples—3-element forcing families in the Chung–Graham–Wilson quasirandom-graph framework, including explicit forcing triples with no forcing subpairs [2312.05969]. In condensed-matter physics it names a specific triple-\(\mathcal Q\) partial magnetic order in \({\rm UNi_4B}\) stabilized by quadrupolar interactions [2209.00984]. In machine learning for materials it denotes a model-agnostic three-stream augmentation for atomistic GNNs that separates composition, structure, and interaction information [2605.20581]. The available literature also uses the same triadic motif, sometimes explicitly and sometimes interpretively, in triangle-based quasirandomness, three-nucleon forces, tripartite entanglement, triangle-center theory, flavour gauge extensions, triality-based generation models, and triangle-free triple systems. ## 1. Terminological scope The literature associates the label with several unrelated technical objects rather than a single unified definition. | Context | Meaning | Representative paper | |---|---|---| | Quasirandom graphs | forcing triples with no forcing pairs | [2312.05969] | | UNi\(_4\)B magnetism | triple-\(\mathcal Q\) partial magnetic order called “triforce” order | [2209.00984] | | Atomistic MLIPs | three-stream augmentation for transferable representations | [2605.20581] | A plausible implication is that “TriForces” functions primarily as a descriptor of triadic organization. In the graph-theoretic case, the triad is a minimal forcing family whose every 2-element subfamily fails to force quasirandomness. In the magnetic case, the triad is a particular triple-\(\mathcal Q\) real-space order. In the machine-learning case, it is an explicit factorization of latent information into three streams. Other occurrences preserve the same three-part logic but refer to different mathematical or physical objects. ## 2. Forcing triples in quasirandom graph theory Within the Chung–Graham–Wilson framework, a family \(\mathcal H\) of graphs is forcing if the conditions \[ \operatorname{hom}(H,G_n)=\bigl(p^{|E(H)|}+o(1)\bigr)n^{|V(H)|}\qquad \forall H\in\mathcal H \] already imply quasirandomness of the graph sequence \(\{G_n\}\), equivalently \(t(H)=p^{|E(H)|}\) for all \(H\in\mathcal H\) forces property \((P0)\) [2312.05969]. The classical example is \(\{e,C_4\}\), and the paper “A note on forcing triples with no forcing pairs” establishes the existence of genuinely 3-element forcing families such that none of their 2-element subfamilies is forcing. The main result is Theorem 1.1: \[ \text{There are forcing triples with no forcing pairs.} \] More explicitly, for any fixed connected non-bipartite graph \(T\) with distinguished vertex \(0\) and \(N=|V(T)|\), the paper defines \[ \mathcal{H}_1(T)=\{T',\,db_{0,1}(T'),\,C_{2N}\}, \] \[ \mathcal{H}_2(T)=\{e,\,T',\,db_{1,2}(db_0^2(T)'')\}, \] \[ \mathcal{H}_3(T)=\{T,\,T',\,db_{1,2}(db_0(T^{(N)})'')\}, \] and proves that each \(\mathcal H_i(T)\) is forcing while no two graphs in \(\mathcal H_i(T)\) form a forcing pair [2312.05969]. The constructions use labeled-vertex operations \(F^{(k)}\) and doubling \(db_I(F)\). The forcing direction is obtained through Jensen- and Cauchy–Schwarz-type inequalities that convert homomorphism counts for doubled graphs into constraints on intermediate conditional counts, ultimately forcing the correct edge density \(\#e=(p+o(1))n^2\). Once the edge density is pinned down, classical forcing pairs such as \(\{e,C_{2N}\}\) or \(\{e,C_4\}\) imply quasirandomness. The non-forcing direction uses the Lovász–Szegedy finite graphon model and two nonconstant 2-vertex weight patterns, together with the criterion \(f(F)=t(F)^{1/|E(F)|}\), to construct non-quasirandom sequences on which any chosen 2-element subfamily has matching density parameter \(p\) [2312.05969]. Among the three families, \(\mathcal H_3(T)\) has two structural features emphasized in the paper: all three graphs have the same chromatic number as \(T\), and the number of vertices in each graph is linear in \(|T|\). This contrasts with earlier forcing-pair constructions in which the auxiliary graph \(H\) can be exponential in \(|T|\). In this precise graph-theoretic sense, TriForces are minimal 3-element forcing families whose forcing power is irreducibly ternary. ## 3. “Triforce order” in \({\rm UNi_4B}\) In \({\rm UNi_4B}\), “triforce order” denotes a specific triple-\(\mathcal Q\) partial magnetic order on the U triangular lattice, obtained in a localized pseudo-triplet crystalline-electric-field model with quadrupole degrees of freedom [2209.00984]. The ordered state is proposed as an alternative to an earlier toroidal triple-\(\mathcal Q\) order. Within a \(3\times 3\) magnetic unit cell containing 9 U sites, 6 of 9 sites carry finite in-plane magnetic dipole moments, while the remaining 3 of 9 sites are magnetically disordered but host ordered quadrupole moments. The six magnetic sites form two interpenetrating triangles, one small inverted triangle and one larger triangle, and the in-plane spins on each triangle form a \(120^\circ\) non-collinear configuration. The authors name this pattern “triforce” because the arrangement resembles three mutually touching triangles [2209.00984]. The magnetic order is described by \[ \mathbf{M}(\mathbf{r}) = \sum_{n=1}^3 \mathbf{M}_n \cos(\mathbf{k}_n\cdot\mathbf{r}+\delta_n), \] with \(\mathbf{M}_n=M\,\mathbf{v}_{n\perp}\). For the triforce domain discussed in the paper, \[ \boldsymbol{\delta}=\{0,0,2\pi/3\}. \] The quadrupole field \[ \mathbf{Q}(\mathbf{r}) = \sum_{n=1}^3 \mathbf{Q}_n \cos(\mathbf{k}_n \cdot \mathbf{r}+\delta'_n) +\mathbf{Q}'_{\mathrm K}\cos(\mathbf{k}_{\mathrm K}\cdot\mathbf{r}) +\mathbf{Q}''_{\mathrm K}\sin(\mathbf{k}_{\mathrm K}\cdot\mathbf{r}) \] uses \[ \boldsymbol{\delta}'=\{-\pi/2,-\pi/2,\pi/6\}, \] together with \(K\)-point components that generate quadrupole order both on the six magnetic sites and on the three magnetically disordered sites [2209.00984]. The paper stresses that triforce and toroidal triple-\(\mathcal Q\) states have identical spin structure factors at the primary wave vectors \(\mathbf{k}_n\). The amplitudes \(\mathbf M_n\) are the same, while the phase factors differ. Accordingly, neutron diffraction at the primary Bragg vectors does not distinguish the two states. What changes is the real-space phase structure and the quadrupolar pattern. In the toroidal state, the three core sites are disordered both magnetically and quadrupolarly at mean field; in the triforce state they are magnetically disordered but quadrupolarly ordered [2209.00984]. The stabilization mechanism is analyzed by a Landau expansion in dipole and quadrupole fields, with a local cubic coupling \[ F^{\text{loc}}_3 = -\frac{c}{3N}\sum_{\mathbf r} m^2(\mathbf r)\, q(\mathbf r)\cos[2\theta(\mathbf r)+\phi(\mathbf r)], \] which acts as a dipole–quadrupole lock-in term. The relative stability of single-\(\mathcal Q\), toroidal triple-\(\mathcal Q\), and triforce triple-\(\mathcal Q\) states is governed by negative quartic corrections obtained by integrating out quadrupolar modes at \(\Gamma\), \(\mathbf{k}_n\), and \(K\). For \(a^{\rm Q}>a^{\rm Q}_{\rm K}\), the phase-shifted triple-\(\mathcal Q\)(2) family, which includes the triforce pattern, is favored [2209.00984]. The triforce state also has a distinct cluster-multipole decomposition. The magnetic sector contains \(A_{2u}^-\) and \(B_{1g}^-\) components, while the electric sector contains \(A_{1g}^+\) and \(B_{2u}^+\). When realistic orthorhombic distortion and site-dependent crystalline-electric-field effects are included, the resulting canted triforce order is consistent with the observed current-induced magnetization in \({\rm UNi_4B}\), including the anisotropy problem that motivated re-examination of the pure toroidal interpretation [2209.00984]. ## 4. TriForces as a three-stream framework for atomistic GNNs In atomistic machine learning, TriForces is a model-agnostic augmentation for machine-learning interatomic potentials that separates composition and structure information and combines this separation with self-supervised learning to preserve transferable representations [2605.20581]. The central node representation is \[ \mathbf{h}_i = [\mathbf{h}_i^{\text{comp}} \;\|\; \mathbf{h}_i^{\text{struct}} \;\|\; \mathbf{h}_i^{\text{int}}], \] where the three streams correspond to composition, structure, and interaction. The composition stream is coordinate-blind. A structure is compressed into unique element tokens \(\{(z_t,c_t)\}_{t=1}^T\), with learned element embeddings and a count-weighted Transformer. For head \(h\), \[ a_{ts}^{(h)} = \frac{\mathbf{q}_t^{(h)}\cdot \mathbf{k}_s^{(h)}}{\sqrt{d_h}} + \log c_s, \qquad \alpha_{ts}^{(h)} = \text{softmax}_s\left(a_{ts}^{(h)}\right). \] The \(\log c_s\) bias makes attention over unique tokens exactly equivalent to attention over a multiset in which token \(s\) is repeated \(c_s\) times [2605.20581]. The structural stream is type-agnostic and rotation-invariant. It builds a SOAP-like local density from displacements \(\mathbf r_{ij}\), radial basis functions, real spherical harmonics, and multi-scale cutoffs, forming coefficients \[ C_{a\ell m}(i) = \sum_{j \in \mathcal{N}(i)} D_a(\mathbf{r}_{ij}) \, Y_{\ell m}(\hat{\mathbf{r}}_{ij}), \] and the power spectrum \[ P_{aa'\ell}(i) = \sum_m C_{a\ell m}(i) C_{a'\ell m}(i). \] A small invariant message-passing stack then produces \(\mathbf h_i^{\text{struct}}\). The interaction stream is the original MLIP backbone, such as Orb-v3, eSEN, or MACE, operating on both species and positions [2605.20581]. Self-supervised pretraining combines three losses: \[ \mathcal{L} = \mathcal{L}_\text{denoise} + \lambda_\text{mask}\,\mathcal{L}_\text{mask} + \lambda_\text{LeJEPA}\,\mathcal{L}_\text{LeJEPA}. \] The denoising term predicts Gaussian position noise; the masking term reconstructs masked atom types; and the LeJEPA-style latent objective aligns graph- and node-level embeddings across augmented views while using SIGReg regularization to avoid collapse. The paper argues that the architectural separation is the main source of gains when abundant supervised data is available, whereas self-supervision becomes especially important in low-data transfer and for retrieval quality [2605.20581]. The reported empirical gains are substantial. On OMat24 in the limited-data regime, TriForces reduces energy MAE by 57% at 20K samples only, from 81.3 to 34.6 meV/atom in the cited eSEN direct setting, and improves force MAE from 151.4 to 121.8 meV/\(\AA\) [2605.20581]. On full OMat24, Orb-v3 conservative improves from 107 meV/atom and 150 meV/\(\AA\) to 19.4 meV/atom and 95.5 meV/\(\AA\), while eSEN conservative improves from 80.3 meV/atom and 84.2 meV/\(\AA\) to 18.8 meV/atom and 78.0 meV/\(\AA\). On MatBench, TriForces variants achieve best or near-best results on 6 out of 8 tasks, and on QM9 the molecularly pretrained variants improve standard targets such as dipole moment, polarizability, and HOMO energy [2605.20581]. The latent-space analysis is equally central. Frozen TriForces embeddings support crystal-system accuracy of 96–100%, majority-element prediction near 100%, and mean nearest-neighbor distance MAE of 0.05–0.08 \(\AA\), whereas the cited baseline MLIPs obtain roughly 55–73%, 61–62%, and 0.75–1.10 \(\AA\), respectively. Separate retrieval in composition and structure space also behaves as intended: the composition stream yields high element-set recall, while the structure stream yields the best space-group recall [2605.20581]. In this usage, TriForces denotes an explicitly factorized latent representation rather than a physical force. ## 5. Other research uses of the triadic motif In quasirandom-graph theory, triangle-based forcing provides a related but distinct use of the motif. Reiher and Schacht proved that \((K_3,C_4')\) is a forcing pair and that if \((K_2,F)\) is forcing, then \((K_3,F^\triangle)\) is also forcing, where \(F^\triangle\) is obtained by replacing every edge of \(F\) by a triangle with a fresh vertex. This establishes triangles as effective forcing objects and systematically transfers edge-based forcing pairs into triangle-based ones [1711.04754]. In nuclear theory, “TriForces” has been used interpretively for three-nucleon forces. The microscopic Hamiltonian \[ H=\sum_i T_i+\sum_{i