TIP3P-ST: Optimized 3-Site Water Model

• TIP3P-ST is a rigid three-site water model with re-optimized geometry and charges, yielding improved bulk properties like dielectric constant and surface tension.
• Molecular dynamics simulations validate its enhanced performance in bulk and at charged silica interfaces, showing closer agreement with experimental dielectric values.
• A 65 K structural temperature shift aligns TIP3P-ST’s hydrogen-bond topology with TIP4P/2005, highlighting its distinct yet systemically tuned behavior compared to standard TIP3P.

TIP3P-ST is a rigid three-site water model treated as a re-optimized TIP3P variant for improved bulk properties, especially the dielectric constant and surface tension, and evaluated in recent molecular dynamics studies of aqueous solid–liquid interfaces (Tavakol et al., 4 Aug 2025). A related but distinct structural context comes from the temperature-shift analysis of standard TIP3P, in which TIP3P’s hydrogen-bond network can be mapped onto that of TIP4P/2005 by a structural temperature shift of $65~\text{K}$; that earlier work does not define a separate reparameterized TIP3P-ST model, but it provides an important conceptual background for the “ST” designation as a temperature-shifted structural interpretation of TIP3P behavior (Shevchuk et al., 2012).

1. Nomenclature and model identity

In the 2025 interfacial study, the model appears as both TIP3-ST and TIP3P-ST; the abstract and discussion sometimes use the shorter form, whereas the Methods and tables use TIP3P-ST consistently (Tavakol et al., 4 Aug 2025). Within that study, TIP3P-ST is a 3-site, rigid TIP3P variant whose parameters were re-optimized specifically to improve bulk properties, including the dielectric constant and surface tension. It is therefore not identical to standard TIP3P, even though both are three-site, oxygen-centered Lennard-Jones water models.

This distinction matters because the same literature corpus also contains an earlier, structurally oriented use of the “ST” idea. The 2012 analysis compares standard TIP3P with six other classical water models and shows that their structural populations collapse onto master curves after model-dependent temperature shifts are applied. For standard TIP3P, the shift is particularly large, which motivates a structural-temperature interpretation of TIP3P behavior, but not a new force field under that name (Shevchuk et al., 2012).

2. Molecular specification and simulation protocol

The parameterization used for TIP3P-ST in the 2025 study is summarized in Table 1 of that work (Tavakol et al., 4 Aug 2025).

Quantity	TIP3P-ST value
$\sigma_O$	$3.19257~\text{\AA}$
$\varepsilon_O$	$0.143~\text{kJ/mol}$
$q_H$	$+0.425\,e$
$q_O$	$-0.85\,e$
$l_{\text{OH}}$	$\sigma_O$0
$\sigma_O$1	$\sigma_O$2
Flexibility	No
Electrostatics	LJ/Coul + PPPM

As in standard TIP3P, Lennard-Jones interactions are placed on oxygen rather than hydrogen. The model is rigid, with bonds and angles constrained, and long-range electrostatics are handled by PPPM. The short-range non-bonded treatment uses Lennard-Jones plus short-range Coulomb interactions with a CHARMM potential cutoff function in which the force and energy smoothly reach zero in the region between inner and outer radius values (Tavakol et al., 4 Aug 2025).

For bulk dielectric calculations, the simulation protocol uses 3000 water molecules in a 5 × 5 × 5 nm$\sigma_O$3 cubic box, NPT conditions at 300 K and 1 atm, and a Nosé–Hoover thermostat and barostat. NaCl concentrations extend up to 2.5 M in bulk calculations. Production runs last 50 ns, with the final 30 ns used for dielectric analysis. For silica–solution solid–liquid interfaces, production runs last 30 ns, with NaCl concentrations up to 0.84 M (Tavakol et al., 4 Aug 2025).

A specific limitation of the interfacial implementation is that Na$\sigma_O$4 and Cl$\sigma_O$5 Lennard-Jones parameters were not optimized specifically for TIP3P-ST in that work. Instead, the simulations used the SPC/E-based parametrization of Yagasaki et al. because of the unavailability of the relevant parameter. Silica interactions were taken from the Emami et al. Interface FF, originally calibrated for TIPS3P-PPPM (Tavakol et al., 4 Aug 2025).

3. Structural temperature shift and the relation to standard TIP3P

The 2012 analysis of water structure introduces a microscopic classification based on hydrogen-bond topology out to two solvation shells. Each water molecule is assigned to one of four states: $\sigma_O$6, corresponding to a fully coordinated local network extending through the second shell; $\sigma_O$7, corresponding to a fully coordinated first shell with defective or looped second-shell connectivity; $\sigma_O$8, corresponding to a three-coordinated central water; and $\sigma_O$9, comprising all remaining local environments. By construction,

$3.19257~\text{\AA}$0

Hydrogen bonds are defined geometrically by $3.19257~\text{\AA}$1 and $3.19257~\text{\AA}$2, with at most four hydrogen bonds per water (Shevchuk et al., 2012).

Within that framework, standard TIP3P exhibits the same qualitative temperature evolution as the other models: $3.19257~\text{\AA}$3 increases as temperature decreases, $3.19257~\text{\AA}$4 rises and then falls after a maximum, and $3.19257~\text{\AA}$5 and $3.19257~\text{\AA}$6 decrease upon cooling. The key result is not a different structural sequence but a shifted temperature window. For TIP3P, the maximum of $3.19257~\text{\AA}$7 occurs at

$3.19257~\text{\AA}$8

and the temperature of maximum density is

$3.19257~\text{\AA}$9

well below the experimental $\varepsilon_O$0 (Shevchuk et al., 2012).

Using TIP4P/2005 as reference, the study assigns standard TIP3P a structural temperature shift

$\varepsilon_O$1

Operationally, this means that standard TIP3P at nominal temperature $\varepsilon_O$2 has structural populations $\varepsilon_O$3 and $\varepsilon_O$4 comparable to those of TIP4P/2005 at approximately $\varepsilon_O$5. The same study finds that TIP3P has the largest shift among the compared models and the least stabilization of the fully coordinated $\varepsilon_O$6 state. For configuration $\varepsilon_O$7, the free energy is written as

$\varepsilon_O$8

and for $\varepsilon_O$9 the correlation between $0.143~\text{kJ/mol}$0 and $0.143~\text{kJ/mol}$1 is reported as very strong, with Pearson correlation $0.143~\text{kJ/mol}$2 across models (Shevchuk et al., 2012).

The tetrahedral order parameter is also evaluated through the standard Errington–Debenedetti form,

$0.143~\text{kJ/mol}$3

with ensemble average $0.143~\text{kJ/mol}$4. After temperature shifting, model families with the same geometry show near-identical radial distribution functions and matched $0.143~\text{kJ/mol}$5 values under matched structural conditions, whereas three-site and four-site families remain distinct. This establishes an important boundary condition for interpreting “ST”: temperature shifts align hydrogen-bond topology, but they do not erase geometry-dependent differences in RDF shape or tetrahedral fine structure (Shevchuk et al., 2012).

4. Bulk dielectric behavior

The 2025 study evaluates bulk dielectric response using the standard dipole-fluctuation relation

$0.143~\text{kJ/mol}$6

Among the water models considered, SPC/Fw, H2O/DC, TIP3P-ST, OPC3 and FBA/e yield values closest to the known water dielectric constant at $0.143~\text{kJ/mol}$7, with experiment taken as $0.143~\text{kJ/mol}$8 (Tavakol et al., 4 Aug 2025).

For pure water, TIP3P-ST gives

$0.143~\text{kJ/mol}$9

This places it slightly above experiment, by approximately $q_H$0–$q_H$1, but still in the range classified as satisfactory in that comparison. The same study contrasts this with markedly larger deviations for several other models, including flexible TIP3P-Fw and TIPS3P-PPPM (Tavakol et al., 4 Aug 2025).

The salt-dependent behavior was followed explicitly for SPC/Fw, H2O/DC, TIP3P-ST, and TIPS3P-PPPM over 0, 0.5, 1.0, 1.5, 2.0, and 2.5 M NaCl. All four models show a monotonic decrease of $q_H$2 with increasing salt concentration. For TIP3P-ST, the reported trend is that it overestimates $q_H$3 for concentrations below 0.5 M and predicts values similar to SPC/Fw at higher salt concentrations. On that basis, TIP3P-ST is counted among the four “most suitable” models for dielectric behavior over the low-to-moderate concentration range emphasized in the study (Tavakol et al., 4 Aug 2025).

5. Charged silica interfaces and Poisson–Boltzmann correspondence

TIP3P-ST is one of four models selected for explicit simulation of a charged silica–NaCl(aq) interface. The interfacial system uses a Q$q_H$4 silica surface, pH 7, and 0.67 SiO$q_H$5–Na$q_H$6 groups per nm$q_H$7. The simulation cell has cross-section 3.7 × 3.7 nm$q_H$8, length 16.5 nm normal to the interface, and a silica slab thickness of 2.3 nm; each system contains two solid–liquid interfaces and is studied at 0.042, 0.10, 0.21, 0.42, and 0.84 M NaCl (Tavakol et al., 4 Aug 2025).

The ion profiles obtained with TIP3P-ST follow the expected electrostatic pattern for a negatively charged silica surface. Cl$q_H$9 is depleted near the surface, Na$+0.425\,e$0 accumulates to form the electric double layer, and increasing salt concentration compresses the diffuse layer. The free-energy minimum for moving Na$+0.425\,e$1 from bulk to the interface is reported at $+0.425\,e$2–$+0.425\,e$3 from the silica surface, independent of model and salt concentration. The Stern layer is defined as the region from the silica surface to that minimum, and the integrated Stern layer charge increases with salt concentration. At 0.84 M, all models except TIPS3P-PPPM converge to a Stern-layer charge around $+0.425\,e$4, slightly above half the surface charge magnitude of silica $+0.425\,e$5 for each side) (Tavakol et al., 4 Aug 2025).

The study then compares molecular dynamics with Gouy–Chapman/Poisson–Boltzmann theory. The ion free energy is written as

$+0.425\,e$6

and PB profiles are generated using either the experimental dielectric constant or the MD-derived dielectric constant. For TIP3P-ST, the MD free-energy minima agree well with PB predictions, particularly when the effective surface charge includes the MD-determined Stern-layer charge. Agreement improves at higher salt concentration. Ion density profiles from MD and PB also agree well for distances greater than approximately $+0.425\,e$7 from the surface, whereas close to the surface PB fails to capture explicit Stern-layer physics, including complete co-ion exclusion and the discrete accumulation of counter-ions (Tavakol et al., 4 Aug 2025).

At low salt, specifically around 0.042 M, TIP3P-ST tends to overestimate Na$+0.425\,e$8 density close to the surface relative to PB. As salt concentration increases, this tendency is gradually reversed, although continuum theory still does not reproduce finite-size and excluded-volume effects in the near-surface region (Tavakol et al., 4 Aug 2025).

6. Limitations, misconceptions, and usage context

A common misconception is to treat TIP3P-ST as merely a relabeled standard TIP3P. The available evidence does not support that simplification. The 2025 model is a distinct rigid three-site parameterization with modified geometry, charges, and oxygen Lennard-Jones terms, whereas the 2012 study concerns standard TIP3P and introduces a structural temperature-shift interpretation rather than a new force field (Tavakol et al., 4 Aug 2025, Shevchuk et al., 2012).

A second misconception is that agreement with Poisson–Boltzmann theory follows directly from having a near-experimental dielectric constant. The interfacial results are more specific. For TIP3P-ST, good agreement in adsorption free energies requires inclusion of the Stern layer charge in the effective surface charge used by PB. Moreover, continuum theory remains unable to represent the explicit near-surface ion layering seen in MD, even when free-energy minima are matched (Tavakol et al., 4 Aug 2025).

The principal limitations emerge at elevated salt concentration. For salt concentrations higher than $+0.425\,e$9, the 2025 study reports formation of random ion–ion pairs in bulk and interfacial regions, leading to increased noise, lower reproducibility, and reduced applicability of Gouy–Chapman/PB theory. For TIP3P-ST specifically, the authors state that for the counter-ion condensation parameter the MD–PB agreement stops beyond $q_O$0. This places TIP3P-ST as slightly less robust than SPC/Fw and H2O/DC for PB-like condensation behavior at intermediate salt, while still performing substantially better than TIPS3P-PPPM in overall electric-double-layer fidelity (Tavakol et al., 4 Aug 2025).

The structural literature adds a complementary limitation. Temperature shifts can align hydrogen-bond populations across models, but they do not eliminate geometry-family effects. In the 2012 analysis, three-site models such as TIP3P cluster together after shifting, yet remain distinct from four-site models in RDF shape and tetrahedral order at matched structural conditions. This suggests that any interpretation of TIP3P-ST as “TIP3P corrected by temperature” is incomplete: topology can be shifted onto a master curve, but geometric details remain model-family dependent (Shevchuk et al., 2012).

Within the scope of the cited work, TIP3P-ST is therefore best understood as an improved rigid three-site water model with a bulk dielectric constant close to experiment, a reasonable dielectric decrement with NaCl up to approximately 0.5 M, and good consistency with PB-based interfacial free-energy analysis when Stern-layer charge is incorporated. Relative to standard TIP3P and TIP3P-FB, it is reported to outperform both in matching the experimental dielectric constant of water and in the rational behavior at charged silica interfaces, while still inheriting the general high-salt limitations associated with ion pairing, borrowed ion parameters, and the finite resolution of continuum electrostatics (Tavakol et al., 4 Aug 2025).