Nanoscale Structure–Property Relationships of Cyanate Ester as a Function of Extent of Cure

Abstract

Cyanate esters are key thermosetting resins for composite materials that require structural integrity and resistance to elevated temperatures. Because cyanate ester composites require relatively high processing temperatures, they are susceptible to the formation of process-induced residual stresses, which compromise their overall strength and durability. Process modeling is a key strategy for optimizing processing parameters to minimize such residual stresses. A necessary component of effective and efficient process modeling of composites is computationally established resin property evolution relationships for a range of processing parameters. In this study, the physical, mechanical, and thermal properties of a cyanate ester resin are established as a function of processing time and temperature using experimentally validated molecular dynamics modeling. The results show that the properties are strongly dependent on the processing temperature. At processing temperatures above 160 °C, the properties quickly approach their fully cured values, whereas at processing temperatures below 140 °C, the chemical cross-linking is significantly inhibited, and processing times to complete cure are relatively long. The evolution of the physical, mechanical, and thermal properties as a function of processing time is established, which is critical data needed as input into multiscale process modeling and optimization of cyanate ester composites for computationally driven composite design.

1. Introduction

Polymer matrix composites (PMCs) are of significant interest in the aerospace, automotive, and electronics communities because of their excellent specific properties, fatigue resistance, corrosion resistance, low dielectric constant, and low coefficient of thermal expansion. , Cyanate ester thermoset resins contain cyanate (−O–CN) functional groups, which form strong networks of aromatic triazine rings that provide excellent thermal stability and a relatively high glass transition temperature (T _g) between 250 and 300 °C. , Thus, cyanate ester resins are ideal for high-temperature applications.

During the processing of cyanate ester PMCs, the curing matrix undergoes shrinkage due to chemical and thermal effects, leading to the formation of residual stresses within the heterogeneous microstructure of the composite. , Although experimental-based process modeling has been performed for cyanate ester resins, − the accumulation of empirical data for comprehensive process optimization is relatively cumbersome and expensive. Fortunately, it is possible to efficiently optimize the processing parameters (hold temperatures, hold times, ramp rates) such that these residual stresses are minimized via computational process modeling. This optimization requires complete knowledge of the evolution of physical, mechanical, and thermal properties of the matrix during cure, which is not currently available for cyanate ester resins. Expressing the property evolution in terms of simple processing-property relationships is a critical component of computational process modeling.

Molecular dynamics (MD) simulation is a powerful technique to simulate thermoset resin behavior at the nanoscale and to predict the evolution of properties during curing. − Patil et al. , used MD to study the evolution of physical and mechanical properties of epoxy systems as a function of extent of cure. Shenogina et al. predicted the physical, mechanical, and thermal properties of a bisphenol-A-based epoxy as a function of extent of cure. Gaikwad et al. used MD to predict the physical and mechanical properties of benzoxazine resin during cure. Moore et al. performed MD simulation to investigate the cross-linking mechanism and water uptake of polycyanurate, including LECy (bisphenol-E-based di­[cyanate ester]), BADCy bisphenol-A-based di­[cyanate ester], and SiMCy (silicon-containing BADCy). Kemppainen et al. investigated the effect of different MD packing methods on the potential energy evolution, cured density, volumetric shrinkage, and the fractional free volume (FFV) of a bisphenol-E-based cyanate ester. Although these studies presented valuable MD simulation protocols, they did not provide the required property evolution data for the cyanate ester process modeling.

This study implements MD simulation and cure kinetics to predict the physical, mechanical, and thermal properties of a low-viscosity bisphenol-E-based cyanate ester as a function of extent of cure and processing time. All of the predicted properties are validated with experimental measurements from the literature. The outcome of this study provides the necessary information for process modeling of cyanate ester composite materials for future integrated computational materials engineering (ICME) and Materials Genome Initiative (MGI) applications. ,−

2. Material and Methods

This section describes the procedures for creating the initial MD models, simulated cross-linking, property prediction, and process modeling. All of the MD models were simulated using the Large-Scale Atomic/Molecular Massively Parallel Simulator (LAMMPS) software package, version April 2022. The potential energy of the MD models was described using the reactive interface force field based on the polymer consistent force field (PCFF-IFF-R). , This force field has been successfully used to simulate bulk polymer systems in multiple studies. − ,,− The LUNAR (LAMMPS Utility for Network Analysis and Reactivity) software package was used for many of the MD model setup and analysis procedures described herein. LUNAR is freely available on GitHub. The MD models were atom-typed using LUNAR’s “atom_typing” module and then parametrized using LUNAR’s “all2lmp” module. The models were then initialized, polymerized, and equilibrated using LAMMPS. Finally, for thermomechanical property predictions, the equilibrated models were then converted to PCFF-IFF-R using LUNAR’s “morse_bond_update” module. The Nose–Hoover thermostat and barostat were employed throughout all MD simulations. −

2.1. Molecular Model Building

The molecular structure of the cyanate ester monomer is shown in Figure . A total of 441 monomers were randomly distributed within a large periodic simulation box, resulting in a total of 14,112 atoms in a low-density gas state. Four independent replicates were constructed for statistical sampling. Each replicate was created using the same atomic coordinates from the initial system but with varied initial velocity seeds to ensure a statistically independent atomic configuration. The Conjugate-Gradient method was used to minimize each replicate in the low-density gas state, with energy and force tolerances set at 10⁴ (unitless) and10^–6 kcal/mol-Å, respectively. With this number of atoms in each replicate, the model sizes were large enough to provide both sufficient precision and efficiency based on the findings of Kashmari et al. They showed that decreasing (below 15,000 atoms) MD model sizes exhibited increasing standard deviations of predicted mechanical properties between model replicates. Above the 15,000 atom threshold, the standard deviations did not change significantly, but the simulation times increased substantially.

1.

(a) Chemical structure of 1,1-bis­(4-cyanatophenyl)­ethane monomer. This bisphenol-E cyanate ester system (Bis-E CN) is the base component of the commercially marketed AroCy L-10 (Huntsman) monomer and (b) the fully cured MD model. This bis-E CN system is a low-cure-temperature-cure cyanate ester system, ideal for current applications of interest, such as additive manufacturing.

To establish the densified form of each replicate of the resin, the ″fix/deform″ command in LAMMPS was employed to gradually reduce the simulation box dimensions at room temperature over 15 steps. The rate of reduction of simulation box dimensions started at 200 Å/ns and was gradually reduced to 0.1 Å/ns, eventually achieving the target mass density of 1.17 g/cm³. This target density was chosen based on the expected mass density of most liquid monomer systems. This densification procedure used a 1.0 fs time step for each replicate. Following the densification, the corresponding spatial density profiles were generated and demonstrated a uniform distribution of monomers in the simulation boxes. The densified systems were equilibrated in the constant pressure and temperature (NPT) ensemble at room temperature (27 °C) at a pressure of 1 atm using 1.0 fs time steps for a duration of 2 ns. After equilibration, the average mass density over the four replicates was 1.1660 ± 0.0003 g/cm³.

2.2. Polymerization

Following the relaxation, the cross-linking reaction was simulated at five processing temperatures: 120, 140, 160, 180, and 240 °C. Each replicate was heated at a rate of 50 °C/ns from room temperature to the processing temperature. Cyanate esters undergo cyclotrimerization (Figure ) during the cross-linking process. The REACTER protocol within LAMMPS was utilized to directly simulate the cross-linking. This protocol involves the use of pre- and postreaction templates to describe the molecular structure before and after each chemical reaction, along with mapping files to monitor changes in molecular structure at reaction sites. The pre- and postreaction templates were generated using LUNAR. Simulated reactions are controlled based on specified minimum and maximum cutoff distances between reactive atoms as well as a bond formation probability. The bond formation cutoff distances were 3, 4, and 7 Å, with reaction probabilities of 0.001, 0.1, and 1.0 for the first, second, and third reactions, respectively. The low probabilities were chosen for the 3 and 4 Å cutoffs to prevent overly aggressive and rapid cross-linking that could potentially lead to unphysical networks. The relatively high probability was chosen for the 7 Å cutoff to accelerate the reaction given the short time frames that MD simulations can cover. These simulations were conducted in the constant volume and temperature (NVT) ensemble with 0.5 fs time steps over 4.0 ns.

2.

Largest molecular cluster evolves as the extent of cure increases with the cyclotrimerization reaction.

The extent of cure (α) was determined as the ratio of the number of covalent bonds formed to the maximum potential number of bonds in the system. A sequential cross-linking approach was selected for this study, where each extent of cure increment model was built upon the previous extent of cure increment model. The simulated cross-linking progressed in 0.1 increment steps until it reached an extent of cure of 0.5. Above that point, the increments were decreased to 0.05 until reaching the final maximum extent of cure value. The evolution of a representative polymer MD model during cyclotrimerization is illustrated in Figure . The figure shows the evolution of molecular clusters, each cluster representing an independent chain of bonded monomers, which gradually merge during the cross-linking. The visualization software OVITO, with the “Cluster analysis” modifier, was utilized to generate these images.

2.3. Gel-Point Prediction

Flory defined the gel point as the state in the polymer in which a cross-linked network exists and prevents steady-state flow. Three established approaches of predicting the gel point from MD simulations are to track (1) the molecular weight of the largest cluster, (2) the molecular weight of the second-largest cluster, and (3) the weight-averaged reduced molecular weight (RMW) of the system as a function of extent of cure. RMW refers to the molecular weight of the entire network, except the largest cluster. The gel point is identified either by observing inflection points in the molecular weight of the largest cluster or by identifying peak values of the other two metrics. In this study, all three methods were employed to predict the gel point. Figure presents the results averaged over all replicates, qualitatively indicating an average gel point of 0.6 for all three metrics. Table shows the specific predicted gel points for each replicate by using the three metrics. A value of α = 0.6 was chosen as the gel point for postgelation volumetric shrinkage calculations.

3.

Molecular mass predictions as a function of extent of cure for gel-point prediction. The error bars indicate the standard deviations across four replicates.

1. Gel Points Identified by Each Criterion and Replicate.

replicate	largest cluster	second-largest cluster	RMW
1	0.60	0.60	0.55
2	0.60	0.55	0.55
3	0.60	0.60	0.60
4	0.60	0.60	0.60
average	0.60	0.58	0.58
standard deviation	0.0	0.03	0.03

To eliminate any residual stresses arising during the simulated cross-linking, the replicates were gradually cooled from their curing temperatures to room temperature (27 °C) at a rate of 50 °C/ns. All replicates were then relaxed at a temperature of 27 °C and pressure of 1 atm using the NPT ensemble for a duration of 2 ns. The resulting average mass density was 1.247 ± 0.009 g/cm³ for the fully cured systems at room temperature, which is in agreement with the experimentally reported values by Guenthner et al. and Davis et al. (1.23–1.24 g/cm³). ,

2.4. Mechanical Property Prediction

Mechanical deformation simulations were conducted to predict the elastic and strength properties. The bulk modulus, shear modulus, Young’s modulus, Poisson’s ratio, and yield strength for each extent of cure across all replicates were predicted by adapting the same simulation and analysis protocols outlined by Odegard et al. Specifically, the bulk modulus values were determined by applying 1 and 5000 atm NPT simulations at 27 °C, and the shear modulus values were calculated from shear deformations in the three principal planes with a strain rate of 2 × 10⁸ s^–1. A bilinear break point was determined in the shearing simulations by observing the strain at which the stress–strain slope changed significantly, usually around 7% shear strain. The bulk and shear modulus values were subsequently used to predict the Young’s modulus and Poisson’s ratio using the standard relations for linear elastic isotropic materials. The yield strength was assessed using von Mises stress computed from the shear deformations. The corresponding yield strength was the von Mises stress at the same breakpoints determined for the shear modulus calculations. The Young’s modulus of some partially cured polymer systems, as predicted by MD simulations, is often overestimated due to the viscoelastic response of the polymer above and (especially) below the gelation point. This issue becomes less prominent as the polymer network expands and becomes stiffer. The procedure developed by Patil et al. was used to map the MD predictions for predicted shear modulus to the laboratory-scale values as a function of extent of cure.

2.5. Thermal Property Prediction

To predict the glass transition temperature (T _g) and the linear coefficient of thermal expansion (CTE), each simulation replicate was gradually heated from −23 to 427 °C at a heating rate of 50 °C/ns in the NPT ensemble under 1 atm pressure. It is well-known that experimental measurements of T _g in polymer materials are affected by heating and cooling rates because of the time-dependent molecular relaxation response, as discussed in detail elsewhere. It was observed by Odegard et al. that the statistical error associated with T _g prediction with MD among replicates is often larger than the effect of the heating/cooling rate. Therefore, no heating–cooling rate corrections were applied to the T _g predictions herein. It is also important to note that recently Patil et al. demonstrated little effect of heating rate on predicted T _g values for an epoxy system with heating rates of 25, 50, and 75 °C/ns. The T _g was identified by fitting a bilinear regression to the simulation box volume vs temperature response, with the point of transition corresponding to the T _g. CTE above and below T _g was calculated using data obtained from the heating simulations using

$$\mathrm{CTE}=\frac{1}{3}\beta =\frac{1}{V}\frac{\partial V}{\partial T}$$ 1

where β and V are the volumetric coefficient of thermal expansion and the initial volume, respectively. The slopes of the bilinear regression lines used for the T _g determination were used as the final term in eq . The bilinear regression lines were fit over the entire range of temperatures spanning −23 to 427 °C.

2.6. Process Modeling

Once the mechanical and thermal properties were predicted as a function of the extent of cure, the evolution of the properties was determined as a function of processing time and temperature. To analyze the cure kinetics of the materials, the data from Robles et al. was utilized. Robles et al. carried out differential scanning calorimetry (DSC) scans at four distinct temperatures (120, 140, 160, and 180 °C). The DSC results were then fitted to a hybrid cure kinetic model, as represented in eqs and

$$\frac{\mathrm{d}\alpha }{\mathrm{d}t}=\frac{K{\alpha }^m{(1-\alpha )}^n}{1+\exp [C(\alpha -({\alpha }_{\mathrm{C}0}+{\alpha }_{\mathrm{CT}}T))]}$$ 2

$$\frac{\mathrm{d}\alpha }{\mathrm{d}t}=K_1{\alpha }^{m_1}{(1-\alpha )}^{n_1}+\frac{K_2{\alpha }^{m_2}{(1-\alpha )}^{n_2}}{1+\exp [C(\alpha -({\alpha }_{\mathrm{C}0}+{\alpha }_{\mathrm{CT}}T))]}$$ 3

where α is the extent of cure; m and n are the first and second exponential constants, respectively; the diffusion constant is represented by C; α_C0 and α_CT account for the critical extent of cure at T = 0 K (including the effects of catalysts , ) and with temperature, respectively; and K is defined by the Arrhenius equation

$$K=A\exp \frac{-E_{\mathrm{a}}}{\mathit{RT}}$$ 4

where A, R, T, and E _a are the pre-exponential factor, gas constant, temperature, and iso-conversional curve slope, respectively. The iso-conversional curve slope acts as an activation energy for cure.

Using the cure kinetics data from Robles et al. and eqs , , and , the correlation between the extent of cure and processing time for each of the four temperatures was determined and is depicted in Figure . The data in Figure were used to relate the predicted properties as a function of the extent of cure to the same properties as a function of processing time and temperature. The solid curves in Figure were directly determined from the equations.

4.

Extent of cure as a function of processing time during isothermal cure at different curing temperatures.

3. Results and Discussion

Figure a shows the predicted volumetric shrinkage as a function of the extent of cure at room temperature and the four processing temperatures with second-order polynomial fits included. Table presents the polynomial equations at the corresponding temperatures for direct use in multiscale process modeling. As curing temperatures increase, volumetric shrinkage values increase, which is likely because of the larger thermal expansion of liquid monomers relative to the polymerized network and, thus, a larger change in volume at higher temperatures as monomers convert into the network. Figure b shows the predicted mass density of the four replicates as a function of the extent of cure. Accurate mass density prediction is a crucial first step in accurately predicting thermomechanical properties of polymer materials. , The predicted MD results are compared in Figure b to the experimentally obtained mass density of the fully cured system from the literature. A linear trend is observed below the gel point for both the mass density and volumetric shrinkage. Above the gel point, the mass density and volumetric shrinkage show a smaller dependence on the extent of cure because of the formation of a three-dimensional network that effectively transmits mechanical loads and hinders the overall deformation of the material system.

5.

(a) Volumetric shrinkage and (b) mass density prediction with comparison to experiment at different temperatures. The experimental data point is at room temperature. The polynomial fits to the data are included. The error bars refer to the standard deviations of the four replicates.

2. Fitted Polynomial Equations for Volumetric Shrinkage and Mass Density as a Function of Extent of Cure at Each Processing Temperature .

temperature (°C)	volumetric shrinkage (%)	mass density (g/cm3)
27	–8.636α2 + 15.075α	–0.118α2 + 0.202α + 1.161
120	–9.023α2 + 21.354α	–0.111α2 + 0.269α + 1.073
140	–7.366α2 + 20.931α	–0.082α2 + 0.254α + 1.061
160	–8.709α2 + 21.174α	–0.099α2 + 0.283α + 1.041
180	–7.604α2 + 23.110α	–0.076α2 + 0.269α + 1.026
240	–7.735α2 + 25.020α	–0.059α2 + 0.266α + 0.973

^aBinomial fits were used for simplicity and accuracy.

The Young’s modulus evolution during cure is presented in Figure a and compared with experimental values reported in the literature. − The curing temperature for these data, as well as the data in subsequent figures, is 240 °C, which corresponds to the upper processing temperature for typical cyanate esters. The property data in Figure was evaluated at room temperature. As expected, Young’s modulus varies nonlinearly with respect to the extent of cure, especially near the gel point. The predicted Young’s modulus of the polymer system at the maximum extent of cure (α = 0.98) is 3390 ± 98 MPa, which is within 15% of the reported values in the literature. Similar trends for the predicted shear modulus are observed above and below the gel point, as shown in Figure b. For the fully cured polymer (α = 0.98), the predicted shear modulus is 1210 ± 42 MPa.

6.

(a) Fitted logistic sigmoidal regression of (a) Young’s modulus, (b) shear modulus, (c) yield strength, and (d) Poisson’s ratio at varying extents of cure. The error bars refer to the standard deviation of the four replicates.

The predicted yield strength at varying extents of cure is presented in Figure c. The predicted yield strength for the fully polymerized system at room temperature is 102 ± 20 MPa. The experimental values reported in the literature for this type of cyanate ester range from 70 to 130 MPa, which agrees with the value obtained in this study. − The yield strength increases as covalent bonds form and tighten the polymer network. Before reaching the gel point (α = 0.6), the viscous liquid cannot bear a significant load, resulting in a relatively low yield strength. In the postgelation regime, the yield strength increases due to the increasing connectivity of the polymer network.

The Poisson’s ratio values are plotted in Figure d as a function of the extent of cure. The Poisson’s ratio decreases nonlinearly with increasing extent of cure, and similar to Young’s and shear modulus, the change in Poisson’s ratio predictions reduces above the gel point. The calculated Poisson’s ratio for α = 0.98 is 0.382 ± 0.011 at 27 °C. The predicted Poisson’s ratio for the monomer liquid, α = 0.0, is 0.475 ± 0.018. In comparison, the experimentally measured value for many monomer liquids is near 0.5, thus validating the predictions.

Figure also includes fitted curves for Young’s modulus, shear modulus, yield strength, and Poisson’s ratio as a function of extent of cure. The fits are logistic sigmoidal regressions with the corresponding R ² values of 0.988, 0.987, 0.976, and 0.988, respectively (herein, reported R ² values are with respect to the average data points). Table provides the fitting coefficients for each mechanical property at room temperature for use in multiscale process modeling. The logistic regression equation is

$$y=\frac{(A_1-A_2)}{1-{\left(\frac{\alpha }{x_0}\right)}^{\mathrm{p}}}+A_2$$ 5

Predictions of T _g at varying extents of cure, above the gel point, are presented in Figure a, along with experimentally measured T _g values reported in various studies. ,, A linear regression is applied, yielding an R ² value of 0.935. Table includes the relation between T _g and the extent of cure for use in multiscale process modeling. As the polymer network becomes stiffer, the mobility of the polymer chains decreases, resulting in higher T _g values. For the fully polymerized system (α = 0.98), the predicted T _g is 267 ± 36 °C, which agrees with the experimental values.

3. Logistic Coefficients for Each Mechanical Property at Room Temperature.

logistic coefficients	Young’s modulus (GPa)	shear modulus (GPa)	yield strength (MPa)	Poisson’s ratio
A 1	0.040	0.040	0.000	0.495
A 2	2.992	1.067	89.770	0.396
x 0	0.632	0.634	0.616	0.577
p	13.803	13.279	19.216	4.384

7.

(a) Predicted and experimentally measured ,, T _g values and (b) CTE below and above T _g. The error bars refer to the standard deviation of the four replicates.

4. Fitted Linear Regression Equations for T _g and CTE Values below and above T _g .

T g	253.760α + 24.233
CTEbelow	–11.653α + 19.703
CTEabove	–29.961α + 46.021

Figure b shows the CTE values below and above T _g as a function of the extent of cure with fitted linear regressions with corresponding R ² values of 0.972 and 0.926, respectively. The CTE values above and below T _g decrease steadily as the polymerization progresses. As expected, the CTE values below T _g are lower than those above T _g because of the reduced free volume in the glassy state. For the fully polymerized system, the CTE values above and below T _g are 15 ± 3 and 8.2 ± 0.9 ppm/°C, respectively. Table includes the relations of CTEs below and above T _g with respect to the extent of the cure.

Figures – present the evolution of the physical, mechanical, and thermal properties of the cyanate ester as a function of the extent of cure. Although this information is important for understanding the relationship between chemical cross-linking and property evolution, it is not very helpful from a processing viewpoint. Because the extent of cure is difficult to track during the processing of composite materials, it is more useful to understand how the physical, mechanical, and thermal properties evolve as a function of processing parameters. Thanks to the information provided in Figure , we can directly relate the extent of cure to processing time at various temperatures.

Figure a,b illustrates the predicted volumetric shrinkage and mass density evolution during processing at four different processing temperatures, utilizing the MD predictions from Figure and the cure kinetics from Figure . The results suggest that higher processing temperatures lead to both faster and higher volumetric shrinkage, which can be attributed to the faster curing process and the polymer’s higher CTE at elevated temperatures (see Figure ). A similar trend is observed in the evolution of the mass density at various curing temperatures. As the temperature increases, the density changes more rapidly and approaches a higher value.

8.

Computational prediction of the evolution of (a) volumetric shrinkage and (b) density as a function of the processing time for varying processing temperatures. These figures were obtained by combining the MD predictions from Figure and the cure kinetics from Figure .

Figure a–d shows the variation in mechanical properties during processing, utilizing the MD predictions from Figure and the cure kinetics from Figure . Similar to the trends observed in volumetric shrinkage and density predictions, the evolution of the mechanical properties is significantly influenced by the processing temperature. With an increase in processing temperature, the curing process accelerates, which explains the reduced processing times required to reach an equilibrium value at elevated temperatures. At 120 °C, the fully cured values of the mechanical properties are not reached because of the incomplete cure at that temperature observed with the data in Figure .

9.

Computational prediction of the evolution of (a) Young’s modulus, (b) shear modulus, (c) Poisson’s ratio, and (d) yield strength as a function of processing time for varying processing temperatures. These figures were obtained by combining the MD predictions from Figure and the cure kinetics from Figure .

Figure a illustrates the change in T _g during processing beyond the gel point, utilizing the MD predictions from Figure and the cure kinetics from Figure . Similar to the trends seen in other physical and mechanical properties, the development of T _g is strongly affected by the processing temperature. As the temperature increases, both the rate of change and the final T _g values are affected due to the accelerated polymerization and, consequently, a higher extent of cure. At 120 °C, the polymer does not exceed its gel point (Figure ), which accounts for the relatively low T _g observed.

10.

Computational prediction of the evolution of (a) T _g, (b) CTE below T _g, and (c) CTE above T _g as a function of the processing time for varying processing temperatures. These figures were obtained by combining the MD predictions from Figure and the cure kinetics from Figure .

Figure b,c shows the progression of the CTE values below and above T _g, respectively. In both instances, CTE evolves more rapidly at higher processing temperatures. Similar to the mechanical properties, the faster curing rate at elevated temperatures promotes the quicker evolution of CTE toward its fully cured value.

4. Conclusions

This study provides a model-driven approach to determine the physical, mechanical, and thermal properties of a cyanate ester resin as a function of processing parameters (time, temperature). Using molecular simulation and experimentally determined cure kinetics, predicted evolution curves of volumetric shrinkage, mass density, elastic properties, yield strength, T _g, and CTE are established for four different processing temperatures up to 1000 min. The results indicate two major things. First, the molecular modeling method produces predicted mass density, Young’s modulus, and T _g that are within 15% of the literature values and are thus experimentally validated for physical, mechanical, and thermal properties. Second, there is a strong dependence of the properties on the processing temperature. For processing temperatures of 160 and 180 °C, the properties quickly approach their fully cured values. At 140 °C, the physical and mechanical properties approach their fully cured values at a significantly slower rate. At 120 °C, all of the properties fall short of the fully cured values, as the cyanate ester does not fully gel at this temperature. This work is important for providing valuable process-modeling insight for this cyanate ester resin and for demonstrating the power of molecular modeling in a modeling-driven framework for process parameter optimization. The latter point is particularly important in ICME and MGI-based approaches for efficient development and optimization of next-generation composite materials.

CRediT: Khatereh Kashmari conceptualization, data curation, formal analysis, investigation, methodology, visualization, writing - original draft, writing - review & editing; Josh Kemppainen formal analysis, methodology, software; Sagar U Patil conceptualization, formal analysis, methodology; Julieta Barroeta Robles resources; Pascal Hubert resources; Gregory M Odegard conceptualization, formal analysis, funding acquisition, project administration, resources, supervision, writing - review & editing.

The authors declare no competing financial interest.