Talk: Computational tools for biodegradable polymers and sustainable practices

Computational tools for biodegradable polymers and sustainable practices Petra Baˇ cová, Eleftherios Christofi, Vagelis Harmandaris, Sergio I. Molina Departamento de Ciencia de los Materiales e Ingeniería Metalúrgica y Química Inorgánica, Facultad de Ciencias, IMEYMAT, Campus Universitario Río San Pedro s/n., Puerto Real, Cádiz 11510, Spain PITS3D Role of computational methods in material design D. J. Audus and J. J. de Pablo, ACS Macro Letters 2017 6 (10), 1078-1082 PITS3D Material design at University of Cádiz Development of sustainable polymer-based materials oriented towards the design of products in large format additive manufacturing, PhD thesis, Pedro Burgos Pintos PITS3D [1,2] * [1] https://adicork.es/index.php/sectores/; [2] Created with BioRender.com PITS3D Poly(lactic acid) in additive manufacturing 3widely popular, easy to print 3low melting point, no heated tray 8sensitive to sunlight and high temperatures 8commercial samples of unknown composition úLvs. Dstereoisomer úadditives, polydisperse samples úempirical approach to nanocomposite design: basalt fibers increase Tgand CTE [1] [1] P. Burgos Pintos, Master thesis PITS3D (2) Recocido en el rango de 80-130ºC se puede utilizar para promover la cristalización y mejorar la temperatura de deflexión térmica de la parte impresa en 3D. Post-annealing in the range of 80-130ºC can be used to promote crystallization and improve the heat deflection temperature of the 3D printed part. 0-60 ºC SMARTFIL PLA 3D850 Physical Properties Typical Value Test Method Material Density 1,24 g/cm3ISO 1183 Chemical Name Polylactic Acid Mechanical Properties Typical Value Test Method Flexural Strenght 126 MPa ASTM D790 Flexural Modulus 4357 MPa ASTM D790 Tensile Yield Strenght 65.5 MPa ASTM D882 Notched Izod Impact 40 J/m ASTM D256 Thermal Properties Typical Value Test Method Heat Deflection Temperature 144 ºC 2ISO 75 Printing Properties Typical Value Print Temperature 210 10 ºC Hot Pad Fan Layer On (100%) PITS3D Rheology of poly(lactic acid) Ïeffect of the increasing D-content Ï100k, 423K, 10%, 25%, 50%, 100% 101 102 103 104 105 106 107 108 100101102103104105106107 %D dynamic moduli (G’,G’’)bT (Pa) angular frequency ωαΤ (rad/s) G’, D100 G’’, D100 G’, L50D50 G’’, L50D50 G’, L75D25 G’’, L75D25 G’, L90D10 G’’, L90D10 Othman, Norhayani, PhD thesis, 2012 PITS3D Rheology of poly(lactic acid) Ïeffect of tacticity Ï≈150k, 50% D, syndiotactic vs. atactic Othman, Norhayani, PhD thesis, 2012 L.-E. Chile, P. Mehrkhodavandi, S. G. Hatzikiriakos, Macromolecules (2016) 49 (3), 909-919 PITS3D Systematic bottom-up simulation approach Computational methods: (A) quantum: force field (B) atomistic: chemistry-specific interactions (C) coarse-grained: trends at mesoscale (D) continuum (A) (B) (C) (D) time length Multiscale: 3quantitative predictions of macroscopic properties 8time and resourse intensive [1] Guseva, D. V.; Glagolev, M. K.; Lazutin, A. A.; Vasilevskaya, V. V.; Polym. Rev. 2023, 0, 1–39 PITS3D Coarse-graining procedure Ïmonomer-like coarse-graining, iterative Boltzmann inversion 0 0.2 0.4 0.6 0.8 1 100101102103104105106107 C(t) t [ps] PLLA, 30mer atomistic CG, tCG=75tA 0 0.2 0.4 0.6 0.8 1 100101102103104105106107108109 C(t) t [ps] PLLA, 100mer atomistic CG, tCG=140tA Alireza F. Behbahani et al., Macromolecules (2021) 54 (6), 2740-2762 PITS3D Stress relaxation modulus G(t): PDLA Ïequilibrium simulations: G(t)=V kBT〈σαβ(t)σαβ(0)〉 Ïconnecting atomistic and mesoscale models to achieve the full relaxation spectrum PITS3D Stress relaxation modulus G(t): PLLA Ïconnecting atomistic and mesoscale models to achieve the full relaxation spectrum: speed up 100x 101 102 103 104 105 10-2 10-1 100101102103104105106107 G(t) [bar] t [ps] t-0.5 atomistic, 30mer CG, 30mer atomistic, 100mer CG, 100mer PITS3D Backmapping procedure (A) atomistic dataset, 5000 frames per system, training set: PLLA100, PDLA100 and Copo100 (B) encoding and learning úconvolutional neural network úversatile: bond vectors and lengths úlocal: no intermonomeric and intermolecular information (C) coarse-graining, monomer-like representation (D) backmapping úverify the model útesting the transferability (A) (B) (C) (D) PITS3D Neural Networks ÏTraining vs. validating set, loss function ÏWeights: importance of the corresponding feature ÏOverfitting: perfect fit of training data, bad performance against unseen data ÏUnderfitting: relevant relations between features and target outputs missing PITS3D Backmapping procedure: transferability across Mw Ïtraining set: PLLA100, PDLA100 and Copo100 Ïtransferability across Mw: PLLA30, PDLA30, Copo30 3possible extension to any molecular weight: industrially relevant Mw, polydisperse samples PLLA30 PDLA30 Copo30 PLLA100 PDLA100 Copo100 55%D 16%D PITS3D Backmapping procedure: chemical transferability Ïtraining set: PLLA100, PDLA100 and Copo100 Ïchemical transferability úcopolymer with a different copolymer sequence úcopolymer trained solely on homopolymers 3capability of creating any composition: mimicking conditions during synthesis PLLA100 PDLA100 Copo100 CopoHOMO CopoRAND PITS3D Master plan PITS3D Master Plan: Data set Model systems 4monodisperse 4control over composition: D and copo 8slow equilibration 8low Mw Experimental systems 8unknown composition 8wide range of Mw 4ready to be measured (if the method available) 4both used to train machine learning algorithms PITS3D Master Plan: Characterization Model systems 4Tgas a function of composition 4G(t) as a function of composition for low Mw 8estimated G(t) as a function of composition for low Mw 8Mw from G(t) Experimental systems 4Tgof commercial samples and sustainable nanocomposites 4G(t) as a function of composition for high Mw 8printability conditions