Topological origin of peak splitting in the structure factor of liquid water

Zoé Faure Beaulieu, Volker L. Deringer, Fausto Martelli · 2026

Read Topological origin of peak splitting in the structure factor of liquid water by Zoé Faure Beaulieu, Volker L. Deringer, Fausto Martelli (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2601-05891.txt.

Introduction

Zoe Faure Beaulieu and 2 co-authors trace in 2026 a split peak in liquid water's scattering pattern to the topology of its hydrogen-bond network. The splitting is usually read as competition between two distinct local settings around water molecules. Their atomistic model runs suggest a different, more structural explanation.

Decomposing the structure factor by hydrogen-bonded ring size showed that rings of five to eight members produce the observed double-peaked signal. Such rings dominate the network at low temperatures. Five-membered rings proved especially persistent, keeping a distinct structural signature even above room temperature. The result links an experimentally accessible diffraction feature to network topology, and may guide interpretation of anomalies in related tetrahedral liquids and glasses. The rings are not a side detail; in this account, they are the main cause of the split peak.

Catalog record

CREATORZoé Faure Beaulieu, Volker L. Deringer, Fausto Martelli
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE22 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2601-05891

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

Topological origin of peak splitting in the structure factor of liquid water Zoé Faure Beaulieu, Volker L. Deringer, Fausto Martelli Abstract The splitting of the principal peak in the structure factor of liquid water is commonly interpreted as evidence of a competition between two distinct local environments. Here, we show that this peak splitting arises from medium-range topological features of the hydrogen-bond network. Using atomistic simulations, we systematically decompose the structure factor into contributions from hydrogen-bonded rings of different sizes. We find that 5–8-membered rings, which dominate the network topology of liquid water at low temperatures, can directly explain the experimentally observed bimodal scattering signal. Among these, 5-membered rings are particularly persistent, maintaining distinct structural signatures even above room temperature. Our findings establish a direct link between the network topology of liquid water and experimentally accessible diffraction features, clarifying the microscopic basis of water’s behavior and suggesting a broader conceptual framework for interpreting the anomalies in tetrahedral network liquids and glasses. Introduction —The most direct experimental probe for the structure of a liquid is the X-ray or neutron structure factor, S(q) . For simple liquids modelled by Lennard-Jones potentials, the first peak in the structure factor appears at the wavenumber corresponding to the average interparticle distance. However, this trend does not hold for liquids with local tetrahedral ordering, such as water 25 , silicon 24 , and silica 32 ; 10 ; 18 : here, the main peak in the S(q) is often observed at a lower wavenumber. The latter feature is referred to as the first sharp diffraction peak (FSDP), and its origin had long remained unclear 8 ; 7 ; 16 ; 20 . Shi and Tanaka have shown that the FSDP in covalent liquids (including water) and glasses is a combination of two overlapping peaks: one arising from the density-wave characteristic of local structures lacking tetrahedral symmetry, and the other arising from the density-wave characteristic of tetrahedral local environments 24 ; 25 . This argument supports the view that water behaves according to a two-state model, i.e., as a mixture of ordered and less ordered local structures 27 ; 28 ; 29 . Upon cooling liquid water, the experimentally measured FSDP does indeed split into two distinct maxima 26 ; 9 ; 23 . Consistent with Shi and Tanaka’s interpretation, this effect reveals the separation between two populations: the low- q peak is associated with locally tetrahedrally coordinated, ordered environments; the high- q peak with more compact, disordered local domains 26 ; 9 ; 23 ; 24 ; 25 . The separation between these peaks increases upon cooling, indicating the emergence of spatial correlations and structural differentiation. These features are aligned with water’s known thermodynamic anomalies and the hypothesised liquid–liquid critical point (LLCP) 21 . Despite these insights, the topological origins of the FSDP splitting in liquid water remain incompletely understood. In particular, the role of medium-range topological motifs, such as hydrogen-bonded rings of various sizes, in shaping the structure factor has not been investigated. Recent work has emphasised the importance of topology in other disordered systems: Zhou et al. 34 and Tavanti et al. 30 demonstrated that shifts in ring distributions correlate directly with changes in diffraction features in silicate and chalcogenide glasses. Figure 1: Peak splitting in the structure factor, S(q) , of liquid water at 1 bar. (a) Simulation results obtained with MACE (solid lines) and ACE (dotted lines) potentials at 240 K (yellow) and 260 K (dark blue), alongside experimental data (dashed lines) from Ref. 26 . The corresponding experimental temperatures are indicated to the right of each curve. (b) Magnified view of the low- q (1.2 Å -1 to 3.6 Å -1 ) region, highlighting the emergence and divergence of two peaks, S_{1} and S_{2} , as shown by black dashed lines. Peak positions at each temperature are determined via Gaussian Process (GP) fits to the simulation data. At temperatures above 330 K, the GP fit does not resolve two separate maxima. (c) Temperature dependence of the peak separation S_{2}-S_{1} comparing the current MACE model (black circles) with the classical TIP4P/2005 model (squares), and experimental data from Skinner et al. 26 (triangles), Sellberg et al. 23 (squares), and Benmore et al. 4 (diamonds). Here, we employ advanced atomistic simulations to investigate precisely this question: how hydrogen-bonded rings of various sizes contribute to S(q) . By directly linking specific topological motifs to the observed bifurcation in the FSDP, our findings demonstrate that the topology of the hydrogen-bond network (HBN) plays a central role in shaping water’s structural signatures, and, by extension, its physical properties. This result establishes a clear connection between topological properties and experimental evidence, thus enhancing our understanding of water’s complex nature. Methodology —Our main results are obtained using a machine-learned interatomic potential (MLIP) model that we fitted to the dataset of Ref. 12 using the MACE architecture 3 . Molecular-dynamics (MD) simulations using this model capture the experimentally observed splitting of the FSDP into two distinct maxima (Fig. 1 a), as well as the corresponding linear increase in their separation at lower T 26 ; 9 ; 23 (Fig. 1 b). We compare our model’s predictions to experimental data from Ref. 26 , to simulations with the atomic cluster expansion (ACE) MLIP model from Ref. 12 , whose training dataset we have used here, and to the popular classical TIP4P/2005 model of water 1 , which reproduces well many properties of liquid water 2 ; 33 , including density and compressibility 13 , as well as the phase behaviour of ices 5 ; 31 ; 22 . Overall, our MACE model reproduces the experimental S(q) well, with some discrepancies at low q , consistent with differences observed in other validation metrics (Supplementary Materials): it tends to over-structure the network, possibly correlated with the predominance of ice-like configurations in the training dataset 12 . At high T , both MLIPs yield virtually indistinguishable predictions; at low T , the MACE model agrees better with experiment than the ACE one, a trend also reflected in the density isobar predictions (Fig. S2). A magnified view of the low- q region in Fig. 1 b highlights the increasing separation between these peaks, S 1 and S 2 , below 330 K. The peak positions, identified using a Gaussian Process (GP) fit, show a clear divergence as temperature decreases. This trend is evident from the distance between peaks S 2 and S 1 , which we plot in Fig. 1 c. Our MACE model reproduces the experimentally observed linear increase in separation 26 ; 23 ; 4 , while TIP4P/2005 simulations slightly underestimate it. Above 330 K, thermal fluctuations obscure the distinct local environments that cause the splitting, and the GP fit resolves only a single broad maximum, corresponding to an averaged O–O separation. Ring Topology in Liquid Water —Having established that our model reproduces the key experimental observation, we proceed to analyse the network topology of the liquid based on the MD trajectories. Figure 2: Ring statistics and neighbourhood structure in the HBN of liquid water from 260 to 350 K. (a) Probability distribution, P(n) , of observing hydrogen-bonded rings consisting of n water molecules, where n\in[3,12] , across a range of temperatures. (b)

Read the complete text