betaFeedback

WorksPhysics and Astronomy

Optimal reaction coordinates and kinetic rates from the projected dynamics of transition paths

Abstract

Finding optimal reaction coordinates and predicting accurate kinetic rates for activated processes are two of the foremost challenges of molecular simulations. We introduce an algorithm that tackles the two problems at once: starting from a limited number of reactive molecular dynamics trajectories (transition paths), we automatically generate with a Monte Carlo approach a sequence of different reaction coordinates that progressively reduce the kinetic rate of their projected effective dynamics. Based on a variational principle, the minimal rate accurately approximates the exact one, and it corresponds to the optimal reaction coordinate. After benchmarking the method on an analytic double-well system, we apply it to complex atomistic systems: the interaction of carbon nanoparticles of different sizes in water.

Introduction

Physico-chemical transformations such as phase transitions, chemical reactions, and biomolecular conformational changes are characterized by metastable states separated by free-energy barriers, so that transitions between states are rare events. Atomistic computer simulations of rare events – especially molecular dynamics (MD) – can play an important role, complementary to experiments, by predicting mechanisms, free-energy landscapes, and kinetic rates.

However, MD simulations face two prominent challenges. First, a gap – often of many orders of magnitude – between the short time scale of atomic motion (femtoseconds) and the long time scale of rare events hampers direct, brute-force simulations. Second, the high-dimensional nature of configuration space makes the analysis of the atomistic trajectories and the extraction of relevant information intrinsically difficult. [1, 2]

To overcome or at least alleviate these challenges and gain insights into the transformation processes, including quantitative thermodynamic and kinetic properties, it is necessary to reduce the dimensionality of the problem by projecting on an appropriate low-dimensional space. To this aim, collective variables (CVs) – generic functions of atomic coordinates able to track interesting structural changes – are often introduced, heuristically or by machine learning. [3]

Among all possible CVs describing a transition, the optimal CV, or “reaction coordinate" (RC), is widely considered the so-called committor function: in MD simulations, this function is defined as the probability pB(x)p_B(x) that a system will reach state BB before state AA starting from atomic positions xx in 3N3N-dimensional configuration space, with initial momenta randomly drawn from the equilibrium distribution [4, 5, 6, 7, 8, 9, 10].

The committor is considered optimal since i) it allows predicting the fate of atomic configurations towards reactants or products, and ii) it preserves the kinetics when employed to build a reduced model of the dynamics (vide infra). The concept dates back to the 1930’s, when Onsager studied the recombination of a pair of ions in the presence of a uniform electric field [11].

The committor can be used to design tests for the “quality" of a RC: for instance, a good RC qq is expected to display a distribution of values pB(x)p_B(x) sharply peaked at 0.5 when considering a set of different configurations xx corresponding to a same location q(x)=q∗q(x) = q^* in RC space, i.e., precisely identifying the transition state configurations. [4, 12, 13]

Some remarks are in order. Given a generic CV qq, the corresponding free energy landscape is always well-defined mathematically, via the equilibrium density F(q)=−kTlog⁡peq(q)F(q) = -kT \log p_{\mathrm{eq}}(q): if the CV is able to resolve (at least partially) reactants and products, such F(q)F(q) will display a barrier. However, different CVs can contain different amounts of information about the transition mechanism. Moreover, the height of the barrier will depend on the particular choice of the CV [14, 15] (see Figure 1).

Analytic double-well free-energy landscapes, Langevin trajectories, and inferred free-energy and diffusion profiles

Figure 1. (a) Analytic double-well free-energy landscape, with arrows indicating one-dimensional projections along the xx axis, the diagonal, and the yy axis, respectively. (b) 100 Langevin bi-dimensional trajectories, projected on a single CV, relax from the barrier-top into A (blue) or B (orange), and are used to optimize effective one-dimensional Langevin models via likelihood maximization. (c) For each projection, the free-energy and diffusion profiles inferred from the latter models are shown (continuous lines with standard error over 10 reconstructions) alongside the exact F(q)F(q) profiles and an alternative estimate of D(q)D(q) (dashed lines), from the Kramers-Moyal formula [42] D(q)≈(⟨Δq2⟩q−⟨Δq⟩q2)/2τD(q) \approx\left(\langle\Delta q^{2}\rangle_q-\langle\Delta q\rangle_q^{2}\right)/2\tau where Δq\Delta q is the CV displacement in a time interval τ\tau, with binned averages along qq.

The barrier along the optimal RC is supposed to contain more useful information from the viewpoint of the calculation of the rate, keeping in mind that any invertible transformation of the optimal RC results again in an optimal RC, since such transformation cannot produce or destroy microscopic information: in the case of a non-linear transformation, the free-energy barrier and the diffusion coefficient are different for the two RCs [16, 14], the differences compensating each other in such a way that the projected dynamics keeps the same kinetic properties.

It is therefore important to distinguish the problem of converging the calculation of the free energy barrier along any given CV from a statistical viewpoint (a sampling problem) from the problem of identifying among all CVs the optimal RC and computing the corresponding barrier, for the purpose of estimating the rate.

Several methods, based on different theoretical principles, have been put forward to estimate optimal RCs from MD data. Given the customary identification of the optimal RC with the committor function, a possibility consists in optimizing a parametrization of the committor function based on pB(x)p_B(x) values estimated from transition path sampling (TPS), exploiting, e.g., the histogram test [13], maximum-likelihood approaches [17, 9], or artificial intelligence [18, 19].

We remark, however, that precisely estimating committor values very close to zero or to one by shooting requires an unfeasible large amount of trajectories: this hampers the accurate reconstruction of the optimal RC in presence of large free-energy barriers elsewhere than in the vicinity of the barrier top. This problem is addressed in the reweighted path ensemble approach [20].

Several approaches have been developed to generate flexible RC representations able to capture slow degrees of freedom connected to rare events [21, 22, 23, 24, 25, 26, 27]. Starting from the seminal work of Ma and Dinner [18], where a neural network was trained with committor data, recent years saw a flourishing of machine-learning approaches to RC optimization. [28, 29, 30, 31, 32]

In this work we address with a single methodology several prominent tasks of molecular simulations: identifying an optimal RC, estimating the free energy landscape and the diffusion along such RC, and estimating the rate of the process.

Results in the literature indicate that i) a model of the projected dynamics along the optimal RC gives more direct access to accurate kinetic rates [13, 8, 9], and ii) the accurate rate is the minimal one – attained for the optimal RC – with respect to the rates computed from the projected dynamics along each of all possible CVs [10]. The method we present exploits these principles: putative CVs are generated in a Monte Carlo scheme, and the kinetic rates of associated effective low-dimensional models are estimated and systematically minimized, leading to optimal RCs and accurate kinetic predictions. An important ingredient of our approach is the automatic construction of stochastic (Langevin) models of CV evolution: for this purpose, we adopt the maximum likelihood algorithm proposed in Ref. 33, based on short TPS-like MD trajectories. We remark that such task is non-trivial from a numerical viewpoint [34], and alternative approaches exist [35, 36, 37, 38].

Theoretical methods

Variational principle: the optimal coordinate provides the minimal rate

We start by considering ergodic diffusive dynamics (as described by overdamped Langevin equations) in a high-dimensional space, for a system with two metastable states AA and BB. The transition rate from state AA to BB is denoted by kA→Bk_{A \to B}. Next, we consider the effective dynamics resulting from the projection of the full high-dimensional dynamics onto a low-dimensional manifold of collective variables (CVs): such effective dynamics is an approximation of the original one, formulated again in terms of diffusive equations [39, 10]. the reaction rate of the effective dynamics between states A~\widetilde{A} and B~\widetilde{B}, defined in the low-dimensional space, is k~A~→B~\widetilde{k}_{\widetilde{A} \to\widetilde{B}}. Zhang, et al.[10] proved that the transition rate of the full dynamics is always smaller or equal to the one computed using effective dynamics. In other words, the optimal reaction coordinate yields a minimal rate,

kA→B≤k~A~→B~(1)k_{A \to B} \leq\widetilde{k}_{\widetilde{A} \to\widetilde{B}} \tag*{(1)}

The optimal projection, preserving the value of the rate constant kA→B=k~A~→B~k_{A \to B} = \widetilde{k}_{\widetilde{A} \to\widetilde{B}}, is achieved when the CV corresponds to the committor function or any invertible function of the latter.

As a complementary result, Ref. 40 formally proved that it is possible to obtain accurate rates from an overdamped Langevin model when projecting high-dimensional underdamped Langevin dynamics on a single CV, the committor, irrespective of the existence of timescale separation and metastability in the system. Considering that typical MD simulations, where Hamilton’s equations are coupled with a thermostat, are akin to high-dimensional underdamped Langevin equations, the previous result should hold also in conventional MD.

Our approach consists in optimizing a RC defined as a function of several trial CVs by minimizing the kinetic rate of the projected dynamics with respect to all possible definitions of the RC. Here we consider the projection on a one-dimensional CV, assuming that the variational principle connecting the optimal CV to the minimal rate holds also when the input data is represented by MD simulations with a thermostat. The projected dynamics is approximated by an overdamped Langevin equation, consistently with Refs. [10, 40]:

q˙=−βD(q)∂F(q)∂q+∂D(q)∂q+2D(q) η(t) ,(2)\dot{q} = -\beta D(q) \frac{\partial F(q)}{\partial q} + \frac{\partial D(q)}{\partial q} + \sqrt{2D(q)}\,\eta(t) \, , \tag*{(2)}

with D(q)D(q) the diffusion coefficient, F(q)F(q) the free energy profile and η(t)\eta(t) a Gaussian white noise of zero mean and unit variance. The use of this stochastic model has several appealing features: the only parameters of the equation consist in the functions F(q)F(q) and D(q)D(q), the mass and velocities are not explicitly needed, the Markovian character allows for simple likelihood expressions, and the mean first passage time (MFPT) of the A→BA \to B transition, inverse of the kinetic rate kk, can be directly obtained by numerical integration [41, 14]:

k−1=∫q0bdxeβF(x)D(x)∫axdy e−βF(y)(3)k^{-1} = \int_{q_0}^{b} dx \frac{e^{\beta F(x)}}{D(x)} \int_{a}^{x} dy\, e^{-\beta F(y)} \tag*{(3)}

where q0q_0 is the starting point in AA for the MFPT computation, while aa and bb are respectively the reflecting and absorbing boundaries.

Algorithm for reaction coordinate optimization

The phase-space information used for RC optimization is represented by MD trajectories spanning transitions between the reactants and products regions of configuration space. We adopt TPS as a source of input data for several reasons: [12, 13, 9] i) ergodic trajectories (spontaneously spanning the transitions) are unfeasible in presence of free-energy barriers ≫kBT\gg k_{\mathrm{B}}T, while TPS has an affordable cost for many systems; ii) TPS trajectories are free from biasing forces and, at statistical convergence, faithfully reproduce ergodic trajectories; iii) recent numerical evidence indicate that ∼100\sim100 TPS-like MD trajectories projected on a suitable CV are sufficient to reconstruct accurate free-energies and rates by means of overdamped Langevin models [33].

The proposed RC optimization algorithm starts from the projection of TPS trajectories on a basis set of NN potentially-relevant CVs Q(t)=(Q1(t),Q2(t)...QN(t))\mathbf{Q}(t) = (Q_1(t), Q_2(t)...Q_N(t)). All these coordinates are put on the same footing by shifting and scaling so that for each of them the range of variation on the actual MD trajectories is between 0 and 1 (oriented so that A→BA \rightarrow B for growing values). Note that such a linear transformation of the coordinate does not deform the corresponding free energy landscape besides an irrelevant additive constant: passing from xx to y=ax+by=ax+b the probability densities transform according to e−β[F(x)−F~(y)]=ρ(x)/ρ~(y)=dy/dx=ae^{-\beta[F(x)-\widetilde{F}(y)]} = \rho(x)/\widetilde{\rho}(y) = dy/dx = a. Moreover, the new diffusion coefficient is scaled, D~y/Dx=(dy/dx)2=a2\widetilde{D}_y/D_x = (dy/dx)^2 = a^2 [42], as can be easily seen in the simple case of a driftless constant-DD diffusion process: ⟨[x(t)−x(0)]2⟩=2Dxt\langle[x(t)-x(0)]^2\rangle= 2D_xt implies that ⟨[y(t)−y(0)]2⟩=2(a2Dx)t=2D~yt\langle[y(t)-y(0)]^2\rangle= 2(a^2D_x)t = 2\widetilde{D}_yt.

A trial RC q=∑1NwiQi≡w⋅Qq = \sum_{1}^{N} w_i Q_i \equiv\mathbf{w}\cdot\mathbf{Q} as a linear combination of the basis-set CVs is generated from random weights normalized as w2=1\mathbf{w}^2 = 1. A Monte Carlo loop is then started, at each step proposing a new RC qnewq_{\mathrm{new}} obtained by modifying the weights of the last accepted step through small random variations w→w+δw\mathbf{w} \rightarrow\mathbf{w}+\delta\mathbf{w} (each δwi\delta w_i being drawn from a uniform distribution between [−0.05,0.05][-0.05, 0.05]).

A newly proposed RC is accepted or rejected based on a Metropolis criterion aimed at minimizing the kinetic rate estimated from a maximum-likelihood Langevin model. The latter is optimized following the method in Ref. [33]: for a sufficiently small time interval τ\tau, the transition probability density pp (propagator) between points qq and q′q' in CV-space can be approximated as [43]

p(q′,t+τ∣q,t)≈12πμe−(q′−q−ϕ)2/2μ(4)p(q',t+\tau\mid q,t) \approx\frac{1}{\sqrt{2\pi\mu}}e^{-(q'-q-\phi)^2/2\mu} \tag*{(4)}
ϕ=aτ+12(aa′+Da′′)τ2,μ=2Dτ+(aD′+2a′D+DD′′)τ2(5)\phi=a\tau+\frac{1}{2}(aa'+Da'')\tau^2,\quad\mu=2D\tau+(aD'+2a'D+DD'')\tau^2 \tag*{(5)}

where the prime indicates d/dqd/dq, a=−βDF′+D′a=-\beta DF'+D' is the drift in Eq. (2), and the approximation includes terms up to order τ2\tau^2. We verified that resorting to the less accurate first-order propagator ϕ=aτ=(−βDF′+D′)τ\phi=a\tau=(-\beta DF'+D')\tau, μ=2Dτ\mu=2D\tau gives unsatisfactory results for the systems considered in this work.

Based on Eq. (5), the likelihood of the MD trajectory, sampled with a time resolution τ\tau and projected on a CV qq is given by

−log⁡L(θ)=∑k=1M−1{12log⁡[2πμk(τ)]+[qk+1−qk−ϕk(τ)]22μk(τ)}(6)-\log\mathcal{L}(\theta)=\sum_{k=1}^{M-1}\left\{\frac{1}{2}\log[2\pi\mu_k(\tau)]+\frac{[q_{k+1}-q_k-\phi_k(\tau)]^2}{2\mu_k(\tau)}\right\} \tag*{(6)}

where θ\theta represents the set of all parameters of the Langevin equation (i.e., the profiles F(q)F(q) and D(q)D(q)), MM is the number of trajectory configurations, and ϕk(τ)≡ϕ(qk,τ)\phi_k(\tau)\equiv\phi(q_k,\tau), μk(τ)≡μ(qk,τ)\mu_k(\tau)\equiv\mu(q_k,\tau). The optimal Langevin model for the given qq and τ\tau, obtained by minimizing −log⁡L(θ)-\log\mathcal{L}(\theta) as a function of the parameters with an iterative stochastic algorithm (see Ref. [33] for details), directly yields the free energy and diffusion profiles.

The kinetic rate is computed from the Langevin model using Eq. (3), with integral boundaries defined in the following way: aa is the smallest observed value of qq (for each CV the transition A→BA\rightarrow B is in the direction of increasing qq values), while q0q_0 and bb are the average final values of qq for shooting trajectories ending in AA and BB, respectively, i.e., the bottom of the corresponding free-energy minima.

Having obtained the rate corresponding to a given proposed CV, a Metropolis-like test is employed to accept or reject the CV based on the following expression for the probability:

P=min⁡[1,(koldknew)α×τoldnoiseτnewnoise](7)P=\min\left[1,\left(\frac{k_{\mathrm{old}}}{k_{\mathrm{new}}}\right)^\alpha\times\frac{\tau_{\mathrm{old}}^{\mathrm{noise}}}{\tau_{\mathrm{new}}^{\mathrm{noise}}}\right] \tag*{(7)}

this expression tends to minimize the rate (with α\alpha playing the role of an adjustable inverse Monte Carlo temperature, as discussed in the Results) and simultaneously – coherently with the hypothesis of a Markovian Langevin-like behavior – the memory of the stochastic process. The latter is quantified by τnoise\tau_{\mathrm{noise}}, the auto-correlation time of the “observed" noise trajectory GkG_k estimated from the qkq_k trajectory via the optimal Langevin model:

Gk=qk+1−qk+[βD(qk)F′(qk)−D′(qk)]τ2D(qk)τ(8)G_k = \frac{q_{k+1}-q_k + [\beta D(q_k)F'(q_k)-D'(q_k)]\tau}{\sqrt{2D(q_k)\tau}} \tag*{(8)}

(see Ref. [33] for details).

To summarize, the algorithm performs the following steps:

  1. Project TPS trajectories relaxing from a barrier top onto a basis set of CVs {Qi}i=1,…,N\{Q_i\}_{i=1,\ldots,N}

  2. Construct a first trial RC qq as a sum of the basis CVs with random weights

  3. Generate a new CV qnewq_{\mathrm{new}} by randomly modifying the weights

  4. Estimate the free energy and diffusion profiles as well as the A→BA \rightarrow B rate from an optimal Langevin model of the observed trajectory qnew(t)q_{\mathrm{new}}(t) using likelihood maximization

  5. Accept or reject the new CV with the probability in Eq. (7), aimed at minimizing the rate and enforcing Markovianity

  6. Go to step 3 (iterate until convergence of the rate)

At convergence, the algorithm provides an optimal RC, as well as the corresponding free-energy and diffusion profiles and kinetic rate, all starting solely from a set of 100 TPS-like short MD trajectories. If the basis set contains all the relevant degrees of freedom for the transition process, the optimal rate should approach the exact rate.

Analytic double-well potential

A two-dimensional double-well free-energy surface is defined, for benchmark purposes, as the sum of two Gaussian-shaped wells:

F(x,y)=−C[e−(x−x0)22σx2e−(y−y0)22σy2+e−(x−x1)22σx2e−(y−y1)22σy2](9)F(x,y)=-C\left[e^{-\frac{(x-x_0)^2}{2\sigma_x^2}}e^{-\frac{(y-y_0)^2}{2\sigma_y^2}}+e^{-\frac{(x-x_1)^2}{2\sigma_x^2}}e^{-\frac{(y-y_1)^2}{2\sigma_y^2}}\right] \tag*{(9)}

with C=20 kBTC=20\,k_{\mathrm{B}}T, Gaussian centers (x0,y0)=(0.1723,0.5058)(x_0,y_0)=(0.1723,0.5058), (x1,y1)=(0.8060,0.5058)(x_1,y_1)=(0.8060,0.5058), and widths σx2=0.03921\sigma_x^2=0.03921, σy2=0.2519\sigma_y^2=0.2519.

We consider the region x∈[0,1]x\in[0,1] and y∈[0,1]y\in[0,1], see Figure 1a. The diffusion matrix was set to D=0.015⋅(1001) ps−1\mathbf{D}=0.015\cdot\begin{pmatrix}1&0\\0&1\end{pmatrix}\,\mathrm{ps}^{-1}.

The reference MFPT was estimated directly from brute-force trajectories. For this purpose, 10 long overdamped Langevin simulations with a time step of 0.1 fs were performed. For the aggregate simulation time (10 μ\mus), 16 039 jumps from state AA to BB were observed, yielding a MFPT of 5600 ±\pm 40 ps and a transition rate kA→B=1.79±0.01⋅10−4 ps−1k_{A\rightarrow B}=1.79\pm0.01\cdot10^{-4}\,\mathrm{ps}^{-1}.

Input trajectories for RC optimization were obtained by shooting 100 overdamped Langevin trajectories from the barrier top. From these trajectories, 51 relaxed to state AA (left well) and 49 relaxed to state BB (right well). The basis set CVs, in this case, are simply xx and yy, yielding two one-dimensional projections x(t)x(t) and y(t)y(t) of the 100 relaxing trajectories.

Fullerene dimers in water

The association and dissociation of C₆₀ and C₂₄₀ fullerene dimers (OPLS-AA force-field [44]) in water solution (SPC force-field [45]) have been simulated by MD using GROMACS v2019.4 [46, 47] patched with PLUMED 2.5.3 [48]. The simulations are similar to those in Ref. 33: for more computational details than what is summarized here, we refer to the latter article.

In the first system, two C₆₀ molecules were solvated by 2398 water molecules in a simulation box of 3.607³ nm³ with periodic boundary conditions. In the second system, two C₂₄₀ molecules were solvated by 5375 water molecules in a simulation box of 5.223 nm35.22^{3}\ \mathrm{nm}^{3}. MD simulations were performed with a time step of 1 fs in the NPTNPT ensemble at 298 K [49] (thermostat time constant = 1 ps) and 1 atm [50] (barostat time constant = 4 ps).

The reference free-energy profiles of the association/dissociation of the C60_{60} fullerenes in water as a function of the 8 basis CVs were computed from unbiased simulations of 1 μ\mus total aggregated time, from the population histogram: F(Qi)=−kBTlog⁡ρeq(Qi)F(Q_i)=-k_BT\log\rho_{\mathrm{eq}}(Q_i). Error bars were estimated as the standard deviation of the mean over 5 independent replicas.

We used as input data for RC optimization a set of 100 aimless shooting [51, 17] trajectories of 20 ps, generated using the script publicly available at https://github.com/physix-repo/aimless-shooting, employing a separation of 0.1 ps between successive shooting points.

For the C60_{60} dimer we define the dissociated state based on the distance between centers of mass as d≥1.34d \geq1.34 nm and the associated state as d≤1.17d \leq1.17 nm; for the C240_{240} dimer we define the dissociated and associated states as d≥2.01d \geq2.01 nm and d≤1.9d \leq1.9 nm, respectively. The reference rate constants for the dissociation of the C60_{60} and C240_{240} dimers from the associated state were estimated using the reactive flux formalism over 1000 aimless shooting trajectories (see Ref. 33 for details), obtaining for C60_{60} a MFPT of 6.1±1.26.1 \pm1.2 ns and for C240_{240} a MFPT of 9.4±1.1 μ9.4 \pm1.1\ \mus.

The basis set CVs employed for projecting the fullerene dimers trajectories are the following:

  1. dd: the distance between the centers of mass of each fullerene molecule.

  1. cccc: the number of carbon-carbon contacts, estimated summing continuous coordination functions between any atom of the first fullerene (set S1S_1) and any atom of the second fullerene (set S2S_2)

cc=∑i∈S1∑j∈S2Cij,Cij=1−(rijr0)n1−(rijr0)m(10)cc=\sum_{i\in S_1}\sum_{j\in S_2}C_{ij},\qquad C_{ij}=\frac{1-\left(\frac{r_{ij}}{r_0}\right)^n}{1-\left(\frac{r_{ij}}{r_0}\right)^m} \tag*{(10)}

where rijr_{ij} is the distance between atoms ii and jj, with parameters r0=0.35r_0=0.35 nm, n=6n=6 and m=10m=10.

  1. c2wc2w: the number of carbon-water contacts, defined similarly to cccc in Eq. (10), in this case with r0=0.6r_0=0.6 nm, set S1S_1 including all carbon atoms of the two fullerene molecules, and set S2S_2 including all oxygen atoms of the water molecules.

  1. c1wc1w: the water-carbon contacts for a single fullerene molecule, defined as c2wc2w except for the inclusion of a single fullerene in set S1S_1.

  1. scsc: the approximate carbon pair entropy, estimated using the expression [52]

sc=−2πρkB∫0rmax⁡[g(r)ln⁡g(r)−g(r)+1]r2 dr(11)sc=-2\pi\rho k_B\int_{0}^{r_{\max}}[g(r)\ln g(r)-g(r)+1]r^2\,dr \tag*{(11)}

where ρ\rho is the density, rmax⁡r_{\max} is a cutoff set to 0.65 nm, and g(r)g(r) is the pair distribution function of carbon atoms, estimated via Gaussian kernels as

g(r)=14πNρr2∑i≠j12πσ2e−(r−rij)2/(2σ2)(12)g(r)=\frac{1}{4\pi N\rho r^2}\sum_{i\ne j}\frac{1}{\sqrt{2\pi\sigma^2}}e^{-(r-r_{ij})^2/(2\sigma^2)} \tag*{(12)}

where NN is the number of carbon atoms and σ=0.01\sigma=0.01 nm.

  1. swsw: the approximate water pair entropy, estimated with the same equations and parameters as scsc applied to oxygen atoms.

  1. uccucc: the Van der Waals carbon-carbon potential energy, computed over all carbon pairs.

  1. ucwucw: the Van der Waals carbon-water potential energy, computed between all the carbon atoms and all the solvent atoms.

Results and discussion

In the Results section, for each putative RC in the optimization algorithm, the likelihood of a Langevin model was maximized employing 3⋅1063\cdot10^{6} and 2⋅1062\cdot10^{6} iteration steps for the C60_{60} and C240_{240} fullerenes, respectively, adopting a time resolution τ=1\tau=1 ps for the projected MD trajectory. The latter was tested as sufficient to yield time-decorrelated noise when using the dimer center-of-mass distance dd as CV.

Results and discussion

Two-dimensional double well

To test the validity of the algorithm developed, we first benchmark it on an analytic model: the double-well system detailed in the Methods section. By construction the dynamics is of overdamped-Langevin nature, the trajectories being obtained by integration of such stochastic differential equation over the two-dimensional F(x,y)F(x,y) landscape in Figure 1 with a constant isotropic diffusion matrix.

After projecting the trajectories on a one-dimensional CV, several non-trivial effects can be anticipated from theory. Figure 1 illustrates for instance three simple choices of the CV, corresponding to the xx axis, the yy axis, or the 45∘45^\circ diagonal.

Langevin trajectories shown in Figure 1(b) clearly resolve two distinct end-states when adopting q≡xq \equiv x, while the separation becomes less clear upon deviation of qq from the xx axis until vanishing for q≡yq \equiv y.

The exact one-dimensional F(q)F(q) profile obtained by projecting on a generic CV q(x,y)q(x,y) can be computed from probability marginalization as

−βF(q)=∫dx∫dy e−βF(x,y)δ[q−q(x,y)](13)-\beta F(q) = \int dx \int dy\, e^{-\beta F(x,y)} \delta[q-q(x,y)] \tag*{(13)}

Figure 1(c) shows that CV changes result in significant differences in the one-dimensional free-energy profiles. In particular the barrier is maximal for q≡xq \equiv x, it is strongly reduced for the diagonal, and it vanishes for q≡yq \equiv y; analogously, the distance between the two minima is largest for q≡xq \equiv x and it vanishes for q≡yq \equiv y. These observations support the intuitive idea that xx is the optimal RC for this problem, fully capturing the progress of the transition mechanism. Moreover, the reduced barriers for alternative CVs – simply the effect of overlapping contributions of the two wells – suggest faster rates for the corresponding effective dynamics. The effective one-dimensional diffusion coefficient D(q)D(q) (Figure 1(c)) is also affected by projection, increasing when qq deviates from xx, however its effect on the rate is linear, less important than barrier variations, which have an exponential effect.

Starting from xx and yy as basis CVs, 10 independent runs of the RC optimization algorithm are performed, aimed at minimizing the rate of the effective projected model (at time resolution τ=0.01\tau= 0.01 ps), each starting with a different random linear combination as initial CV (see Figure 2).

RC optimization by rate minimization applied to the double-well benchmark

Figure 2. RC optimization by rate minimization applied to the double-well benchmark: (a) the rate evolution is shown for 5 independent optimizations with α=2,3,4\alpha= 2, 3, 4, respectively; (b) the histogram of the rate values in the stationary regime, i.e., discarding the initial relaxation; (c) number of correlated time steps (at 0.01 ps resolution) of the effective noise estimated from the original trajectories using the optimal Langevin model for each CV.

The parameter α\alpha (effective Monte Carlo inverse temperature in Eq. (7)) is varied from 2 to 4: in all tested cases the initial rate decreases almost monotonously, often by several orders of magnitude, until reaching an equilibrium behavior with small fluctuations in a small interval. Low α\alpha values increase the acceptance of putative CVs in presence of an increase of the rate, whereas high α\alpha values decrease such acceptance causing the stochastic optimization to converge more rapidly towards a minimal rate, with smaller equilibrium fluctuations. α\alpha plays therefore the role of an adjustable parameter providing some control over the convergence of the rate minimization algorithm.

Irrespective of the initial putative CV the algorithm successfully and consistently retrieves an optimal combination including at least 95% of xx. Given that after the initial relaxation a stationary distribution of equilibrated kk values is reached, with a finite width that depends on the parameter α\alpha, and introducing a degree of fuzziness, different criteria could be envisaged to identify the optimal rate values and the corresponding optimal RCs. In the case of the double well system, identifying the predicted rate with the average value of the stationary distribution and its uncertainty with the standard deviation yields 2.6±1.0⋅10−4 ps−12.6 \pm1.0 \cdot10^{-4}\ \mathrm{ps}^{-1} for α=3\alpha= 3, in good agreement with the reference brute-force rate of 1.8⋅10−4 ps−11.8 \cdot10^{-4}\ \mathrm{ps}^{-1}.

However, the results for the solvated fullerene dimers (see next section), a realistic complex system, indicate the possibility of a few spurious outliers at the lowest kk values, resulting from imperfect optimization of the Langevin model for a few putative CVs. This suggests an alternative criterion to identify the optimal rate predicted by the algorithm, as the 5th percentile in the stationary distribution, i.e., the rate below which 5% of the rates are found in the distribution. This criterion predicts (for α=3\alpha= 3) a 5th percentile value of 1.6⋅10−4 ps−11.6 \cdot10^{-4}\ \mathrm{ps}^{-1}: taking as optimal prediction the average over the 5 rates closest to such value for each of the 10 independent optimization runs, we get a predicted rate of 1.60±0.04⋅10−4 ps−11.60 \pm0.04 \cdot10^{-4}\ \mathrm{ps}^{-1} again in good agreement with the reference.

We remark that, even though the original dynamics in (x,y)(x,y)-space is Markovian, non-Markovian effects could appear on the effective dynamics when projecting on CVs different from the committor, requiring in principle an underdamped (or generalized) Langevin description [40]. Therefore, the term penalizing deviations from

Markovianity in Eq. (7) plays, in general, an important role, helping to reject proposed CVs that would introduce significant memory effects rendering inappropriate the overdamped model. Moreover, as shown in Ref. 33, from the viewpoint of the inference of Langevin models non-Markovian behavior appears to be correlated with barrier overestimation and thus rate underestimation, a spurious effect that must be avoided to exploit correctly the variational principle connecting the minimal rate to the optimal RC.

Our simulations show that for all values of α\alpha the vast majority of the CVs accepted during the optimization process have an effective noise free from time-correlation (Figure 2), with occasional exceptions, confirming the ability of the algorithm to enforce Markovianity.

Interaction between fullerene nanoparticles in water

We applied the RC optimization algorithm to the case of the dissociation process of fullerene dimers in water solution, respectively C60\mathrm{C}_{60} and C240\mathrm{C}_{240}. These systems are rather complex, including thousands of atoms, and feature an associated state, corresponding to a free-energy minimum, characterized by the carbon nanoparticles in close contact without the mediation of water molecules, and a dissociated state where each fullerene is fully-solvated. The transition state region features a water molecule bridging between the fullerenes (Figure 3), corresponding to the top of a sizable free-energy barrier when a CV able to resolve the metastable states, like the distance dd between centers of mass, is employed [33].

Characteristic atomic configurations for the C60 and C240 dimers in bulk water solution

Figure 3. Characteristic atomic configurations for the C60\mathrm{C}_{60} and C240\mathrm{C}_{240} dimers in bulk water solution. Water molecules distant <4< 4 Å from carbon atoms are shown. The progress from the associated state (left), passing through the transition state (center), until the dissociated state (right), is followed by CVs normalized to span between zero and one. A water molecule in the transition state, bridging between fullerenes and breaking the direct hydrophobic interactions between them, is highlighted.

We start by proposing a pool of eight basis CVs to be used as building blocks for RC optimization: the set ranges from geometry-inspired CVs like the distance dd and the total number of carbon-carbon or carbon-water contacts (cccc, c2wc2w, c1wc1w) to physics-inspired CVs like the approximate two-body carbon or water entropy (scsc, swsw) and the carbon-carbon or carbon-water interaction energy (uccucc, ucwucw). It is not obvious a priori which CVs or combination thereof could be the best approximations of an optimal RC.

The TPS trajectories relaxing from the barrier top, together with the free-energy and diffusion profiles inferred from likelihood maximization of Langevin models for each of these CVs are reported for the C60\mathrm{C}_{60} dimer in Figure 4, and display, as expected, strong variations depending on the CV choice (see the Supporting Information for C240\mathrm{C}_{240}). Clearly, CVs able to resolve well the associated and dissociated states feature a barrier, contrary to CVs lacking such resolving power. Overall, the F(q)F(q) landscapes compare well with the reference brute-force ones. A time resolution τ=1\tau= 1 ps is adopted for Langevin modelling, as it yielded Markovian behavior in combination with the CV dd for both fullerene sizes in Ref. 33.

MD trajectories and free-energy, diffusion, and kinetic-rate profiles for eight collective variables

Figure 4. MD trajectories from TPS of the C60_{60} fullerene dimer in water relaxing to the associated (blue) or dissociated (orange) state (left), with the corresponding F(q)F(q), D(q)D(q) profiles and kinetic rates (right), for each CV of the basis set ((a) to (h), definitions in the Methods section). The profiles F(q)F(q) and D(q)D(q) with continuous lines are inferred from likelihood maximization (average and standard error over 10 independent runs). The kinetic rate kk is calculated using Eq. (3). The dashed F(q)F(q) are obtained from equilibrium MD histograms.

The basis CVs that results in the lowest kinetic rates, of the order of 10−310^{-3} ps−1^{-1}, are dd, cccc and uccucc; the CVs providing the highest (hence, less accurate) rates are instead swsw and c1wc1w, of the order of 10−110^{-1} ps−1^{-1}. This allows a first ranking of the quality of the basis CVs. The position-dependent diffusion coefficient D(q)D(q) is roughly between 0.01 and 0.03 ps−1^{-1} for all CVs except swsw, that reaches almost 0.08 ps−1^{-1} (Figure 4).

We applied the RC optimization algorithm (500 accepted steps, with 10610^{6} Langevin optimization iterations for each putative CV) starting from different initial random combinations of the basis CVs and testing the parameter α=2,3,4\alpha= 2, 3, 4. As shown in Figure 5, for both fullerene sizes the rate relaxes to a stationary distribution after about 200 steps: rate fluctuations are wider in these systems than for the double well, however the lower part of the distribution has the same order of magnitude of the reference rates.

RC optimization results for the C60 and C240 fullerene dimers in water

Figure 5. RC optimization results for the C60_{60} and C240_{240} fullerene dimers in water. (a,e) 5 independent rate minimization runs are displayed, the grey bar corresponding to the reference rate (thickness = error bar) estimated via reactive flux. (b,f) Histogram of the rate values from 10 optimizations in the stationary regime (i.e., discarding the first 300 iterations) for different α\alpha values. (c,g) Number of correlated time steps (at 1 ps resolution) of the effective noise. (d,h) Weights distribution for the optimal RCs from 10 independent runs: for each run, the 5 RCs closest to the 5th percentile rate are included.

Employing the 5th percentile of the stationary kk distribution as criterion (see previous section) at α=3\alpha= 3, we obtain as optimal rates the values 1.26±0.15⋅10−41.26 \pm0.15 \cdot10^{-4} ps−1^{-1} and 1.8±1.1⋅10−81.8 \pm1.1 \cdot10^{-8} ps−1^{-1} for the dissociation C60\mathrm{C}_{60} and C240\mathrm{C}_{240} dimers, respectively, that compare well with the references reactive-flux rates 1.6±0.3⋅10−41.6 \pm0.3 \cdot10^{-4} ps−1^{-1} and 1.1±0.1⋅10−71.1 \pm0.1 \cdot10^{-7} ps−1^{-1}, respectively (see Methods section).

The Markovianity appears enforced in an effective way, with the vast majority of the accepted CVs having uncorrelated effective noise and a few ones displaying <10< 10 correlated steps (see Figure 5(c, Figure 5g)).

Finally, the composition of the optimal RCs for the solvated C60_{60} and C240_{240} dimers is reported in Figure 5(d, Figure 5h): since the basis CVs are not uncorrelated between them, the optimal RC is not uniquely defined and different weights in the combination q=∑i=1NwiQiq = \sum_{i=1}^{N} w_i Q_i can lead to RCs with similar quality and similar rates.

In fact, performing a principal component analysis on the members of the set of optimal RCs (i.e., the

50 RCs with rate closest to the 5th percentile in Figure 5(b,Figure 5f)} provides explained variance ratios for the first four principal components [0.40, 0.24, 0.14, 0.11] for C60_{60} and [0.56, 0.17, 0.16, 0.04] for C240_{240} (see Supporting Information). This shows that it is not possible to easily reduce the dimensionality in the optimal space of RCs by means of one or two dominating components.

Nevertheless, some basis CVs contribute more than others, on average, to the optimal RCs: in particular, Figure 5(d,Figure 5h) suggest that dd, cccc, scsc, and uccucc play a prominent role for both fullerene sizes. The latter CVs do not contain water degrees of freedom, pointing, somehow, counter-intuitively, to a limited role of the solvent in the optimal RC for these systems.

Concluding remarks

We presented an algorithm for the automatic optimization of the RC for activated processes starting from ≈100\approx100 short TPS MD trajectories and a basis set of putative CVs. The approach is based on the variational principle in Ref. 10: the rate predicted by a Langevin model of the effective projected dynamics is minimized by the optimal RC, coinciding in this limit with the true MD rate.

As an essential ingredient, we adopt the method in Ref. 33 to infer an optimal overdamped Langevin model for each putative RC, thus reliably recovering the corresponding free-energy and diffusion profiles as well as the kinetic rate.

Since we make the assumption that an overdamped stochastic equation be able to faithfully model the projected dynamics, we explicitly disfavoured the appearance of memory effects in the acceptance probability expression applied to newly proposed CVs, thus keeping the Markovianity under control.

Numerical tests on an analytic double-well system as well as on MD simulations of carbon nanoparticles in explicit water solution indicate that optimal RCs are systematically identified in a robust and efficient way: this allowed us to predict kinetic rates in the microsecond time scale from ∼10\sim10 ps-long trajectories.

A foremost advantage of the present algorithm is that RC optimization is done as post-processing of a MD data set: a large number of putative CVs can be screened with limited computer resources without the need to re-run expensive MD calculations. Moreover, compared to alternative approaches based on machine learning the committor, there is no need to estimate committor values over a large sample of atomic configurations, a computationally-expensive endeavour that, by construction, is limited to the region close to the barrier top. On the contrary, the proposed approach only requires a small number of reactive trajectories and, by construction, it learns the optimal RC across the full span of the transition paths joining metastable states, even where the committor assume values very close to zero or to one.

Here the basis set of CVs is built from heuristic considerations, while in future works it could be advantageous to devise an unsupervised building procedure. Likewise, from the simplistic formulation of optimal RCs as linear combinations of basis CVs, it would be beneficial for complex processes to adopt more flexible, nonlinear formulations, possibly including neural networks. As a final word of caution, the effective description adopted here in terms of overdamped Langevin equations is appropriate for a subset of all possible rare-events systems and processes: underdamped or generalized (non-Markovian) stochastic models can be necessary in other cases.

We expect that the approach introduced here could help discovering optimal RCs and predicting accurate rates for challenging open problems like crystal nucleation and protein-protein or protein-ligand dissociation, whose complexity, so far, eluded systematic and predictive computational studies.

Acknowledgement

We gratefully acknowledge very insightful discussions with Christoph Dellago, Gerhard Hummer, Hadrien Vroylandt, Arthur France-Lanord, Alessandro Barducci, Bettina Keller, Tony Lelièvre and Edina Rosta. This work was performed with the support of the Institut des Sciences du Calcul et des Données (ISCD) of Sorbonne University (IDEX SUPER 11-IDEX-0004). Calculations were performed on the GENCI-IDRIS French national supercomputing facility, under grant numbers A0110811069, A0120901387, A0130811069.

Supporting information

Optimal RC composition for the benchmark double-well system with $\alpha = 3$

Figure 6. Optimal RC composition for the benchmark double-well system with α=3\alpha= 3

MD trajectories and profiles for the C240 fullerene dimer

Figure 7. MD trajectories from TPS of the C240\mathrm{C}_{240} fullerene dimer in water relaxing to the associated (blue) or dissociated (orange) state (left), with the corresponding F(q)F(q), D(q)D(q) profiles and kinetic rates (right), for each CV of the basis set ((a) to (h), definitions in the Methods section). The profiles F(q)F(q) and D(q)D(q) with continuous lines are inferred from likelihood maximization (average and standard error over 10 independent runs). The kinetic rate is calculated using Eq. (3).

Time auto-correlation of the effective noise of the CV basis set for C$_{60}$

Figure 8. Time auto-correlation of the effective noise of the CV basis set for C60_{60}, evaluated on 100 TPS MD trajectories. The solid lines are the average correlation of 10 independent optimal models (see main text) while the shaded line represents the standard deviation.

Time auto-correlation of the effective noise of the CV basis set for $C_{240}$

Figure 9. Time auto-correlation of the effective noise of the CV basis set for C240C_{240}, evaluated on 100 TPS MD trajectories. The solid lines are the average correlation of 10 independent optimal models (see main text) while the shaded line represents the standard deviation.

Principal component analysis on the 50 members of the optimal set of RCs for $C_{60}$ and $C_{240}$

Figure 10. Principal component analysis on the 50 members of the optimal set of RCs (a) for C60C_{60} and (b) for C240C_{240}.

References

  1. [1]Fabio Pietrucci. Strategies for the exploration of free energy landscapes: Unity in diversity and challenges ahead. Rev. Phys., 2:32–45, 2017.DOI
  2. [2]Aldo Glelmo, Brooke E Husie, Alex Rodriguez, Cecilia Clementi, Frank Noé, and Alessandro Laio. Unsupervised learning methods for molecular simulation data. Chem. Rev., 121(16):9722–9758, 2021.DOI
  3. [3]Paraskevi Gkeka, Gabriel Stoltz, Amir Barati Farimani, Zineb Belkacemi, Michele Ceriotti, John D Chodera, Aaron R Dinner, Andrew L Ferguson, Jean-Bernard Maillet, Hervé Minoux, et al. Machine learning force fields and coarse-grained variables in molecular dynamics: application to materials and biological systems. J. Chem. Theory Comput., 16(8):4757–4775, 2020.arxiv.org/abs/2004.06950
  4. [4]Peter G Bolhuis, David Chandler, Christoph Dellago, and Phillip L Geissler. Transition path sampling. Annu. Rev. Phys. Chem, 53:291–318, 2002.pubmed.ncbi.nlm.nih.gov/11972010
  5. [5]Young Min Rhee and Vijay S Pande. One-dimensional reaction coordinate and the corresponding potential of mean force from commitment probability distribution. J. Phys. Chem. B, 109(14):6780–6786, 2005.pubmed.ncbi.nlm.nih.gov/16851763
  6. [6]Philipp Metzner, Christof Schütte, and Eric Vanden-Eijnden. Illustration of transition path theory on a collection of simple examples. J. Chem. Phys, 125(8):084110, 2006. doi: 10.1063/1.2335447.DOI
  7. [7]Weinan E and Eric Vanden-Eijnden. Transition-path theory and path-finding algorithms for the study of rare events. Annu. Rev. Phys. Chem., 61:391–420, 2010.
  8. [8]Polina V Banushkina and Sergei V Krivov. Optimal reaction coordinates. Wiley Interdiscip. Rev. Comput. Mol. Sci., 6(6):748–763, 2016.
  9. [9]Baron Peters. Reaction coordinates and mechanistic hypothesis tests. Annu. Rev. Phys. Chem., 67:669–690, 2016.DOI
  10. [10]Wei Zhang, Carsten Hartmann, and Christof Schütte. Effective dynamics along given reaction coordinates, and reaction rate theory. Faraday Discuss., 195:365–394, 2016.DOI
  11. [11]L Onsager. Initial recombination of ions. Phys. Rev., 54:554–557, 1938.DOI
  12. [12]Gerhard Hummer. From transition paths to transition states and rate coefficients. J. Chem. Phys, 120(2):516–523, 2004.pubmed.ncbi.nlm.nih.gov/15267886
  13. [13]Robert B Best and Gerhard Hummer. Reaction coordinates and rates from transition paths. Proc. Natl. Acad. Sci, 102(19):6732–6737, 2005.pubmed.ncbi.nlm.nih.gov/15814618
  14. [14]Swetlana Jungblut and Christoph Dellago. Pathways to self-organization: Crystallization via nucleation and growth. Eur. Phys. J. E, 39(8):1–38, 2016.DOI
  15. [15]Johannes CB Dietschreit, Dennis J Diestler, Andreas Hulm, Christian Ochsenfeld, and Rafael Gómez-Bombarelli. From free-energy profiles to activation free energies. J. Chem. Phys, 157(8):084113, 2022.DOI
  16. [16]Sergei V Krivov and Martin Karplus. Diffusive reaction dynamics on invariant free energy profiles. Proc. Natl. Acad. Sci, 105(37):13841–13846, 2008.DOI
  17. [17]Baron Peters, Gregg T Beckham, and Bernhardt L Trout. Extensions to the likelihood maximization approach for finding reaction coordinates. J. Chem. Phys, 127(3):034109, 2007.pubmed.ncbi.nlm.nih.gov/17655433
  18. [18]Ao Ma and Aaron R Dinner. Automatic method for identifying reaction coordinates in complex systems. J. Phys. Chem. B, 109(14):6769–6779, 2005.pubmed.ncbi.nlm.nih.gov/16851762
  19. [19]Hendrik Jung, Roberto Covino, and Gerhard Hummer. Artificial intelligence assists discovery of reaction coordinates and mechanisms from molecular dynamics simulations. arXiv preprint arXiv:1901.04595, 2019.arxiv.org/abs/1901.04595
  20. [20]Jutta Rogal, Wolfgang Lechner, Jarek Juraszek, Bernd Ensing, and Peter G Bolhuis. The reweighted path ensemble. J. Chem. Phys, 133(17):174109, 2010.DOI
  21. [21]Guillermo Pérez-Hernández, Fabian Paul, Toni Giorgino, Gianni De Fabritiis, and Frank Noé. Identification of slow molecular order parameters for markov model construction. J. Chem. Phys, 139(1):07B604_1, 2013.
  22. [22]Pratyush Tiwary and BJ Berne. Spectral gap optimization of order parameters for sampling complex molecular systems. Proc. Natl. Acad. Sci, 113(11):2839–2844, 2016.DOI
  23. [23]Frank Noé and Cecilia Clementi. Collective variables for the study of long-time kinetics from molecular trajectories: theory and methods. Curr. Opin. Struct. Biol, 43:141–147, 2017.DOI
  24. [24]Mohammad M. Sultan and Vijay S Pande. tica-metadynamics: accelerating metadynamics by using kinetically selected collective variables. J. Chem. Theory Comput., 13(6):2440–2447, 2017.DOI
  25. [25]A. Pérez de Alba Ortiz, A Tiwari, RC Puthenkalathil, and B Ensing. Advances in enhanced sampling along adaptive paths of collective variables. J. Chem. Phys, 149(7):072320, 2018.DOI
  26. [26]Hao Wu and Frank Noé. Variational approach for learning markov processes from time series data. J Nonlinear Sci, 30(1):23–66, 2020.arxiv.org/abs/1707.04659
  27. [27]Ziwei He, Christophe Chipot, and Benoît Roux. Committor-consistent variational string method. J. Phys. Chem. Lett., 13(40):9263–9271, 2022.DOI
  28. [28]João Marcelo Lamim Ribeiro, Pablo Bravo, Yihang Wang, and Pratyush Tiwary. Reweighted autoencoded variational bayes for enhanced sampling (rave). J. Chem. Phys, 149(7):072301, 2018.arxiv.org/abs/1802.03420
  29. [29]Wei Chen and Andrew L Ferguson. Molecular enhanced sampling with autoencoders: On-the-fly collective variable discovery and accelerated free energy landscape exploration. J. Comput. Chem., 39(25):2079–2102, 2018.DOI
  30. [30]Hythem Sidky, Wei Chen, and Andrew L Ferguson. Machine learning for collective variable discovery and enhanced sampling in biomolecular simulation. Mol. Phys., 118(5):e1737742, 2020.DOI
  31. [31]Ferry Hooft, Alberto Pérez de Alba Ortíz, and Bernd Ensing. Discovering collective variables of molecular transitions via genetic algorithms and neural networks. J. Chem. Theory Comput., 17(4):2294–2306, 2021.DOI
  32. [32]Lixin Sun, Jonathan Vandermaus, Simon Batzner, Yu Xie, David Clark, Wei Chen, and Boris Kozinsky. Multitask machine learning of collective variables for enhanced sampling of rare events. J. Chem. Theory Comput., 18(4):2341–2353, 2022.arxiv.org/abs/2012.03909
  33. [33]Karen Palacio-Rodriguez and Fabio Pietrucci. Free energy landscapes, diffusion coefficients, and kinetic rates from transition paths. J. Chem. Theory Comput., 18(8):4639–4648, 2022.DOI
  34. [34]Federica Ferretti, Victor Chardes, Thierry Mora, Aleksandra M Walczak, and Irene Giardina. Building general langevin models from discrete datasets. Phys. Rev. X, 10(3):031018, 2020.DOI
  35. [35]Gerhard Hummer. Position-dependent diffusion coefficients and free energies from bayesian analysis of equilibrium and replica molecular dynamics simulations. New J. Phys., 7(1):34, 2005.DOI
  36. [36]Qi Zhang, Jasna Brujić, and Eric Vanden-Eijnden. Reconstructing free energy profiles from nonequilibrium relaxation trajectories. J. Stat. Phys., 144(2):344–366, 2011.DOI
  37. [37]Lorenzo Boninsegna, Feliks Nüske, and Cecilia Clementi. Sparse learning of stochastic dynamical equations. J. Chem. Phys., 148(24):241723, 2018.DOI
  38. [38]Francois Sicard, Vladimir Koskin, Alessia Annibale, and Edina Rosta. Position-dependent diffusion from biased simulations and markov state model analysis. J. Chem. Theory Comput., 17(4):2022–2033, 2021.
  39. [39]Frédéric Legoll and Tony Lelièvre. Effective dynamics using conditional expectations. Nonlinearity, 23(9):2131, 2010.DOI
  40. [40]Jianfeng Lu and Eric Vanden-Eijnden. Exact dynamical coarse-graining without time-scale separation. J. Chem. Phys., 141(4):07B6191, 2014.DOI
  41. [41]Robert Zwanzig. Nonequilibrium statistical mechanics. Oxford university press, 2001.
  42. [42]Hannes Risken. Fokker-planck equation. Springer, 1996.
  43. [43]Alexander N Drozdov. High-accuracy discrete path integral solutions for stochastic processes with noninvertible diffusion matrices. Phys. Rev. E, 55(3):2496, 1997.DOI
  44. [44]William L Jorgensen, David S Maxwell, and Julian Tirado-Rives. Development and testing of the opls all-atom force field on conformational energetics and properties of organic liquids. J. Am. Chem. Soc., 118(45):11225–11236, 1996.DOI
  45. [45]Herman JC Berendsen, James PM Postma, Wilfred F van Gunsteren, and Jan Hermans. Interaction models for water in relation to protein hydration. In Intermolecular forces: proceedings of the fourteenth Jerusalem symposium on quantum chemistry and biochemistry held in Jerusalem, israel, april 13–16, 1981, pages 331–342. Springer, 1981.DOI
  46. [46]Herman JC Berendsen, David van der Spoel, and Rudi van Drunen. Gromacs: a message-passing parallel molecular dynamics implementation. Comput. Phys. Commun., 91(1-3):43–56, 1995.DOI
  47. [47]Mark James Abraham, Teemu Murtola, Roland Schulz, Szilárd Páll, Jeremy C Smith, Berk Hess, and Erik Lindahl. Gromacs: High performance molecular simulations through multi-level parallelism from laptops to supercomputers. SoftwareX, 1:19–25, 2015.DOI
  48. [48]Gareth A Tribello, Massimiliano Bonomi, Davide Branduardi, Carlo Camilloni, and Giovanni Bussi. Plumed 2: New feathers for an old bird. Comput. Phys. Commun., 185(2):604–613, 2014.DOI
  49. [49]Giovanni Bussi, Davide Donadio, and Michele Parrinello. Canonical sampling through velocity rescaling. J. Chem. Phys., 126(1):014101, 2007. doi: 10.1063/1.2408420. URL https://doi.org/10.1063/1.2408420.
  50. [50]Michele Parrinello and Aneesur Rahman. Polymorphic transitions in single crystals: A new molecular dynamics method. J. Appl. Phys., 52(12):7182–7190, 1981.DOI
  51. [51]Baron Peters and Bernhardt L. Trout. Obtaining reaction coordinates by likelihood maximization. J. Chem. Phys., 125(5):054108, 2006. doi: http://dx.doi.org/10.1063/1.2234477. URL http://scitation.aip.org/content/aip/journal/jcp/125/5/10.1063/1.2234477.
  52. [52]Pablo M Piaggi, Omar Valsson, and Michele Parrinello. Enhancing entropy and enthalpy fluctuations to drive crystallization in atomistic simulations. Phys. Rev. Lett., 119(1):015701, 2017.DOI

Paper details

vlnc:W_naqr3zak
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 0

downloads · 0

A view counts once per browser per day, a download each time the PDF is requested.

  • Physics and Astronomy

Not recorded for this paper

View statement
Looking for related papers…
Contents