We have carried out combined classical and ab initio molecular dynamics (MD)simulations in order to investigate the structural and vibrational properties of severalborosilicate glasses. We have considered rather simple ternary compositions withvarying SiO 2 , B 2 O 3 or Na 2 O concentrations, or more complex compositionscontaining equally CaO, Al 2 O 3 or MgO. The ab initio calculations have been carriedout within the density functional theory framework as implemented in the VASPcode. The classical MD simulations were carried out using different effective pairpotentials.We have studied the local structure of the various structural units, and inparticular we have focused on the structures around the boron atoms and how theseare embedded into the network. We have investigated how the Na atoms aredistributed around the [3] B triangles and [4] B tetrahedra. Furthermore, we have foundthat the Na distribution associated to a BO 4 tetrahedron is different from thatcorresponding to a SiO 4 tetrahedron in that the former gives rise to a distribution thatis significantly more structured.The vibrational properties have been equally studied within the ab initio approach,and we have identified the contributions of the various species as well as those ofthe local structural units. We have also calculated the dielectric function ε(ω) as wellas the absorption spectra. The latter are in good quantitative agreement withexperimental data. The results obtained in this work confirm that the atomisticsimulations, in particular the ab initio ones, give access to a better understanding ofcomplex borosilicate glasses since their structural and vibrational properties can beextracted with a good accuracy and compare very well to experimental data.

We have parameterized empirical potentials for molecular dynamics (MD) simulations ofmulti-component silicate glasses. The main motivation has been to improve predictions ofstatic properties like elastic moduli and dynamic properties like vibrational density of states(VDOS) that MD simulations have generally not been able to estimate correctly, while stillusing a simple functional form for computational efficiency.Our approach has been to fit the potentials to data extracted from accurate first principlescalculations to predict both the static and dynamic properties correctly, by explicitly incor-porating the radial distribution function (RDF) and the VDOS into the cost function of thefitting scheme. The current optimization scheme is an extension of a recent work using asinput only the structural data from the ab initio simulations, and which has been applied inthe past to obtain a reliable potential for amorphous silica [1,2].The newly developed potentials will be used to study the elastic response of multi-componentoxide glasses to external stimuli such as high temperatures, high pressures and high strains,and their deformation modes under different loading conditions.A. Carré, J. Horbach, S. Ispas, W. Kob, Europhys. Lett. 82 (2008) 17001.A. Carré, S. Ispas, J. Horbach, W. Kob, Comput. Mater. Sci. 124 (2016) 323

Density functional theory is used to calculate the vibrational properties of pure silica and sodo-silicateglasses with 20, 25, and 33 mol% of Na2O. The infrared and Raman spectra are calculated and the full responses aredecomposed into principal structural components (PSC). Those are for example the SiO4n--tetrahedra with n nonbrigdingoxygens defining the Qn-species at the origin of the structured feature at high frequency, and the Si-O-Sibridges leading the broad Raman R-band at intermediate frequencies. Our results confirm that Si-O-Si bending inbridges with large angle vibrate preferentially at lower frequencies than those with low angle. In addition, the spectralresponse of the Q2-species is bimodal and overlaps with that of the Q3, while the Q4 response covers almost all ofthe spectral range of the Qn-band. The ab-initio individual spectral responses of the PSC are used to reconstruct theexperimental Raman responses. Contrary to the commonly used multi-Gaussian decomposition, this approachprovides unambiguous band-assignments and hence a more accurate way to probe the structural and chemicalproperties of glasses from their spectroscopic signature.

We use computer simulations to investigate how a catalytic reaction in a polymer sol can induce the formation of a polymer gel. To this aim we consider a solution of homopolymers in which freely-diffusing catalysts convert the originally repulsive A monomers into attractive B ones. We find that at low temperatures this reaction transforms the polymer solution into a physical gel that has a remarkably regular mesostructure in the form of a cluster phase, absent in the usual homopolymer gels obtained by a quench in temperature. We investigate how this microstructuring depends on catalyst concentration, temperature, and polymer density and show that the dynamics for its formation can be understood in a semi-quantitative manner using the interaction potentials between the particles as input. The structuring of the copolymers and the AB sequences resulting from the reactions can be discussed in the context of the phase behaviour of correlated random copolymers. The location of the spinodal line as found in our simulations is consistent with analytical predictions. Finally, we show that the observed structuring depends not only on the chemical distribution of the A and B monomers but also on the mode of formation of this distribution.