<i>Ab</i> <i>initio</i> molecular dynamics with noisy forces: Validating the quantum Monte Carlo approach with benchmark calculations of molecular vibrational properties

Luo, Ye; Zen, Andrea; Sorella, Sandro

doi:10.1063/1.4901430

Cited by 36 publications

(37 citation statements)

References 56 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…38, and the interested reader can refer to that. In the following we instead want to discuss in much more details the variational ansatz that we have used for the specific system here under consideration.…”

Section: Methodological Aspects Of the Qmc-based Molecular Dynamicsmentioning

confidence: 99%

“…The molecular dynamics, driven by quantum Monte Carlo forces, was introduced recently for the simulation of liquid hydrogen at high pressures 39,40 and to obtain vibrational properties of molecular systems 38 . At fixed ion coordinates R the many-body wave function depends on the N − electronic positions x = { r 1 , r 2 , · · · r N } by means of the following Jastrow Slater (JSD) ansatz:…”

Section: Methodological Aspects Of the Qmc-based Molecular Dynamicsmentioning

confidence: 99%

“…On such high performance computing machines a wave function approach based on QMC is currently becoming practical and competitive with DFT, allowing to treat geometry optimization of molecules up to 100 atoms 36 , vibrational properties 37,38 and molecular dynamics simulations 39,40 .…”

Section: Pacs Numbersmentioning

confidence: 99%

See 2 more Smart Citations

Ab initio molecular dynamics simulation of liquid water by quantum Monte Carlo

Zen

Luo

Mazzola

et al. 2015

The Journal of Chemical Physics

Self Cite

View full text Add to dashboard Cite

Although liquid water is ubiquitous in chemical reactions at roots of life and climate on the earth, the prediction of its properties by high-level ab initio molecular dynamics simulations still represents a formidable task for quantum chemistry. In this article we present a room temperature simulation of liquid water based on the potential energy surface obtained by a many-body wave function through quantum Monte Carlo (QMC) methods. The simulated properties are in good agreement with recent neutron scattering and X-ray experiments, particularly concerning the position of the oxygen-oxygen peak in the radial distribution function, at variance of previous Density Functional Theory attempts. Given the excellent performances of QMC on large scale supercomputers, this work opens new perspectives for predictive and reliable ab-initio simulations of complex chemical systems. PACS numbers:The simulation by first principles of liquid water, the key element of human life and biological processes, has been a dream for several decades after the foundation of Density Functional theory (DFT), even within the restriction of the Born-Oppenheimer approximation for the heavy nuclei. Realistic simulations are particular important because, at the experimental level, it is not possible to clarify completely what are the relationships between the so many different and rich phases of water and the physical interactions between water molecules, determined by hydrogen bonding and weak longrange van der Waals (vdW) interactions. Moreover water is involved in many biological and chemical processes, and first principle simulations are useful to investigate and rationalize such important mechanisms.The first attempted simulations date back to the pioneer works by Car and Parrinello 1-3 , within an efficient ab-initio molecular dynamics (AIMD) based on DFT. The comparison with the experiments, at that time available, provided a pretty good agreement with the oxygen-oxygen (O-O) radial distribution function (RDF), as far as the positions of the peaks were concerned, but the overall shape given by the simulation was overstructured. After these first studies, many other works reporting standard DFT-based simulations have been published, but the agreement with the experimental data is still not satisfactory on many aspects. The equilibrium density at ambient pressure (1 atm ∼ 10 −4 GPa), is far to be consistent with the expected one (1 gr/cm 3 ) though recent DFT functionals including van der Waals substantially reduce this discrepancy 4 . The simulated diffusion 5 is much lower than what is expected from experiments 6 , and, at least in some functionals (namely, PBE and BLYP), the solidification of water occurs at a temperature which is unrealistically large (∼ 410 K), so that some of the present DFT simulations of liquid water should be considered supercooled metastable phases 6,7 .The DFT results (about which we provide a brief summary in Tab. I) appear to be substantially influenced by the choice of the functional 5,8 , but also, within a ...

show abstract

Section: Methodological Aspects Of the Qmc-based Molecular Dynamicsmentioning

confidence: 99%

Section: Methodological Aspects Of the Qmc-based Molecular Dynamicsmentioning

confidence: 99%

See 1 more Smart Citation

Ab initio molecular dynamics simulation of liquid water by quantum Monte Carlo

Zen

Luo

Mazzola

et al. 2015

The Journal of Chemical Physics

Self Cite

View full text Add to dashboard Cite

show abstract

“…the error bars-of the nuclear forces. In this case it is also very simple to correct for the extra noise given by the QMC forces [22]. This is achieved in a very simple way, just by changing the temperature T used in the dynamics for the correct simulation at a given target temperature T target , simply as follows:…”

Section: Noise Correctionmentioning

confidence: 99%

“…This covariance matrix (at variance of the Hessian one) is always positive definite, and empirically it has been shown, in several system cases, to be almost proportional to the Hessian matrix [22], at least at the equilibrium structure, where the Hessian is also positive definite. For this reason it is natural to take the covariance matrix for accelerating the LD within QMC, and we have assumed S = Cov( f ) in the following.…”

mentioning

confidence: 99%

Acceleratingab initioMolecular Dynamics and Probing the Weak Dispersive Forces in Dense Liquid Hydrogen

Mazzola

Sorella

2017

Phys. Rev. Lett.

Self Cite

View full text Add to dashboard Cite

We propose an ab-initio molecular dynamics method, capable to reduce dramatically the autocorrelation time required for the simulation of classical and quantum particles at finite temperature. The method is based on an efficient implementation of a first order Langevin dynamics modified by means of a suitable, position dependent acceleration matrix S. Here we apply this technique, within a Quantum Monte Carlo (QMC) based wavefunction approach and within the Born-Oppheneimer approximation, for determining the phase diagram of high-pressure Hydrogen with simulations much longer than the autocorrelation time. With the proposed method, we are able to equilibrate in few hundreds steps even close to the liquid-liquid phase transition (LLT). Within our approach we find that the LLT transition is consistent with recent density functionals predicting a much larger transition pressures when the long range dispersive forces are taken into account.One of the most important problems in ab-initio molecular dynamics (MD) is the coexistence of much different time scales underlying physical processes, even when computing equilibrium finite temperature properties. This has been a challenge for most ab-initio simulations in systems of biophysical interest, the most striking one being protein folding [1,2], where the microscopic time scale of molecular vibration is of the order of f s, while the macroscopic time scale of the folding exceeds the µs. Also in water system simulations, the weak Van der Waals (vdW) interactions and the related hydrogen bond imply very difficult equilibration properties at ambient conditions, so that the most accurate simulations are based on force field fitting forces. We will show here that, a computationally expensive ab-initio method for dense liquid hydrogen can be combined with a sampling technique capable to reduce drastically the autocorrelation times. This method has the potential, due to its simplicity, to be applied with success also to much more complex systems and other ab-initio techniques.At variance of all previous attempts[3-10], we propose that an optimal way to get rid of different time scales is based on the use of first order Langevin dynamics:where T is the temperature, f Ri = −∂ Ri V ( R) and V R is an energy potential of the classical coordinates R, e.g. atomic positions. For S ij = δ ij this is the conventional first order Langevin dynamics (CFOLD). Preconditioned Langevin equations, formally similar to Eq. 40, have been already introduced in several fields different from molecular dynamics, such as gauge field theories[11], statistics [12] and machine learning [13], with different definitions of the preconditioning matrix S. Instead, the main idea of this work is based on the simple solution of CFOLD for the harmonic potential V (R) = −K R, where K is the elastic matrix term. In this case one is able to show that the i) the autocorrelation time is independent of temperature τ −1 corr = K min , where K min (K max ) is the lowest (largest) non-zero eigenvalue of the elastic matrix...

show abstract

Excited‐State Calculations with Quantum Monte Carlo

Feldt

Filippi

2020

Quantum Chemistry and Dynamics of Excited States

View full text Add to dashboard Cite

Ab initio molecular dynamics with noisy forces: Validating the quantum Monte Carlo approach with benchmark calculations of molecular vibrational properties

Cited by 36 publications

References 56 publications

Ab initio molecular dynamics simulation of liquid water by quantum Monte Carlo

Ab initio molecular dynamics simulation of liquid water by quantum Monte Carlo

Acceleratingab initioMolecular Dynamics and Probing the Weak Dispersive Forces in Dense Liquid Hydrogen

Excited‐State Calculations with Quantum Monte Carlo

Contact Info

Product

Resources

About