- We discovered molecular recognition mechanisms of peptides and surfactants on precious metal (Au, Pt, Pd) and oxide surfaces (silica, clay), allowing specific computational predictions of binding strength, growth directions, shape, size, and yield of nanostructures.
- Our team developed accurate and compatible force fields for oxides, fcc metals, clay and cement minerals for the simulation of inorganic-organic materials at the 1 to 100 nm scale under realistic solution conditions (INTERFACE force field).
- We characterized, predicted the reactivity of, and designed Pd nanoparticle catalysts for Stille coupling reactions and olefin hydrogenation in atomic resolution in collaboration with experimental partners at U Miami, FL.
- Our research group introduced realistic surface models for silica and apatite for common surface chemistry and pH conditions for the first time.
- We explained the action of grinding aids in cement and suggested new organic additives to achieve energy savings in ball mills in collaboration with partners at Sika AG and ETH Zurich.
- Heinz (during his doctoral and postdoctoral work) explained the assembly and phase transitions of surfactants on clay minerals, as well as relationships between surface modification and exfoliation in polymer nanocomposites.
Figure 1. Images from simulations depicting the adsorption of peptides and surfactants on clay minerals and precious metals.
We developed accurate atomistic computational models (force fields) to investigate specific recognition at metal-organic and mineral-organic interfaces in solution. Molecular dynamics simulations with these models reach far beyond the length and time scales of ab-initio and DFT methods (100 nm vs ~2 nm, simulation times of microseconds versus picoseconds) and provide understanding of nanocrystal growth and shape control in 3D atomic resolution (Adv. Funct. Mater. 2015, 25, 1374). The accuracy in surface and interface energies is typically better than that of quantum mechanical methods at a million times lower computational cost. The methods enable quantitative comparisons to imaging and binding measurements in the laboratory to understand and guide in the bottom-up assembly of materials (see Chem. Soc. Rev. 2016, 45, 412; Langmuir 2013, 29, 1754).
Figure 2. Understanding facet-specific nanocrystal stabilization and growth: a specific peptide ligand (T7) stabilizes (100) facets of cuboctahedral Pt seed crystals at intermediate concentration, leading to the formation of cubic shapes (Adv. Funct. Mater. 2015, 25, 1374).
Figure 3. Differential adsorption of water and peptide T7 depends on the location on a finite (100) nanocrystal facet in solution. a. The mobility of water molecules is higher near the edges of nanocubes compared to the center of the facets. b. T7 peptides are less mobile and show smaller local differences in adsorption energy than water, leading to peptide attraction near the edge versus repulsion near the center (this is specific for peptide T7, Adv. Funct. Mater. 2015, 25, 1374).
These approaches to computer simulations and chemical theory fill a current gap in understanding of metal nanostructures, oxides, layered and defect-containing materials. Perhaps even more important, the accurate simulation of biological-inorganic interfaces can help explain molecular signatures of disease, such as osteoporosis, coronary calcification, and silicosis in 3D atomic resolution, which bears great promise in combination with laboratory and clinical studies to explore more adequate diagnosis and treatments. At length scales up to 107 atoms and time scales up to 10-5 s, these computational methods in chemical resolution are the only ones available to explain accumulating experimental data and to transition from trial-and-error based laboratory studies to predictive models.
The approach to derive chemically accurate models for inorganic compounds and seamlessly integrate them in existing biomolecular force fields (e.g. CHARMM, AMBER, OPLS-AA, GROMACS, CVFF) is unique and generates chemical insight. We discovered, for example, a soft epitaxial match that governs the interactions of peptides, amino acids, and (non-thiol) surfactants with noble metal surfaces, which permits facet-specific molecular designs to achieve selective binding (J. Am. Soc. 2009, 131, 9704; Small 2012, 8, 1049). Using these concepts, the growth of various Pt and Rh nanostructures has been explained by choosing specific peptide templates (tetrahedra vs cubes, single crystal vs twin), including correlations of expected trends in shape, yield, and nanocrystal size. It was even feasible to correctly predict the effects of single point mutations in the peptides on the final nanoparticle morphology (Nano Lett. 2013, 13, 840; Adv. Funct. Mater. 2015, 25, 1374).
Figure 4. Computational discovery of selective peptide recognition on metal facets in solution. a. Preference of phenyl rings and guanidinium for (111) facets through epitaxial match (spacing L1). A comparable fit of these molecules for (100) facets is not possible due to incompatible symmetry (spacing L2). Instead, linear molecules can significantly adsorb to (100) facets by epitaxial match. b. Representative conformation of arginine on a gold (111) facet in aqueous solution (Soft Matter 2011, 7, 2113).
Our team also introduced pH resolved mineral surface chemistry to the computational chemistry community. We quantified the importance of surface chemistry and pH on silica and apatite surfaces, and computed binding constants of peptides in aqueous solution in quantitative agreement with adsorption isotherms across a range of surface chemistry and pH (with Carole C. Perry, Nottingham, UK, J. Am. Chem. Soc. 2012, 134, 6244; Chem. Mater. 2014, 26, 2647; Chem. Mater. 2014, 26, 5725-5734). Apatite surfaces were assumed to be deprotonated in prior simulations, along with less well justified force field parameters, corresponding to conditions of immediate cell death at pH>14. The new force fields and surface models by our team enable the computer-aided investigation of bone and tooth decay, drug design, and coronary calcification across the full range of pH values and other conditions in aqueous solution at the scale of 1 to 100 nm (J. Phys. Chem. C 2016, online).
Figure 5. Details of the variable surface chemistry of silica and dedicated models. a. Schematic illustrations of silica nanoparticle surfaces with silanol groups (SiOH) and a fraction of ionized groups such as sodium siloxide (SiONa). b. Typical Q3 glass surface at pH 3. c. Same surface at pH 7. d. Q2 quartz surface at pH 8. e. Q3/Q4 annealed silica surface at ~500 ºC (J. Am. Chem. Soc. 2012, 134, 6244; Chem. Mater. 2014, 26, 2647).
Figure 6. Prediction of specific peptide binding as a function of pH on typical silica glass surfaces with a silanol area density of 4.7 per square nanometer. (a) Data from adsorption isotherms in experiment. (b) Data from molecular dynamics simulation for the same systems.
Figure 7. Differences in surface properties of hydroxyapatite as a function of pH according to simulation with the CHARMM-INTERFACE force field (numbers in black) and experiment (numbers in blue). The cleavage energy in vacuum and the immersion energy in water are reduced towards lower pH. Simulations also show a reversal in binding mechanism of the peptide SVSVGGK to apatite surfaces. At pH 10, adsorption on the (010) prismatic plane is mediated by ionic groups and comparatively weak. At pH 5, the polar and hydrophobic motif SVSV is more attracted and binds notably stronger (J. Phys. Chem. C 2016, online).
Using reactive extensions of the CHARMM-INTERFACE force field, we explained and predicted reaction rates of Pd nanoparticle catalysts in C-C coupling reactions as a function of particle size (PCCP 2013, 15, 5488; JACS 2013, 135, 11048, ACS Nano 2015, 9, 5082). We also developed first electronically refined force fields for bcc metals and applied such models to locate the position of missing C atoms in WC systems in combination with high resolution electron tomography (Nat. Mater. 2015, 14, 1099).
Figure 8. Catalytic performance of Pd NPs in Stille coupling in experiment and according to MD simulation. a. The relative turnover frequency (TOF) in experiment can be predicted by the computed atom leaching rate for a NP. b. Illustration of the abstraction energies of atoms for a Pd NP capped with Pd4 peptide. The superficial peptide is partially shown for ease of view, and Pd atoms of lower energy are shown in lighter color (active sites, ACS Nano 2015, 9, 5082).
Earlier during his PhD and postdoc work, Heinz examined the self-assembly of surfactants on clay mineral surfaces and their dispersion in polymer matrices (with Ulrich W. Suter, Richard A. Vaia, Barry L. Farmer). This work explains the nanostructure of surfactants on clay surface as a function of CEC and chain length (Chem. Mater. 2007, 19, 59), the origin of thermal transitions observed in DSC due to partial melting of the alkyl chains and due to lateral mobility on the surface (J. Am. Chem. Soc. 2003, 125, 9500), and the role of cleavage energy of organiclaly modified layered silicates during exfoliation in polymer nanocomposites (Chem. Mater. 2010, 22, 1595). A thermodynamic model for the formation of well dispersed polymer/filler composites has been introduced and nanomechanical properties of clay minerals were analyzed, establishing the maximum in-plane modulus of mica-type silicates of 160 GPa (J. Phys. Chem. C 2010, 114, 1763) and the smallest bending radius of approximately 3 nm before failure (J. Phys. Chem. C 2011, 115, 22292).
Figure 9. Schematic of the self-assembly and thermal behavior of alkyl chains grafted to clay minerals and other substrates as a function of the packing density (Langmuir, 2008, 24, 3727).
Figure 10. Bending of clay mineral layers, showing a limit of ~3 nm for the smallest bending radius and associated bending energies (J. Phys. Chem. C 2011, 115, 22292).
A fundamental contribution to molecular simulations has been the assignment of atomic charges and the interpretation as a quantitative measure of chemical bonding (J. Phys. Chem. B 2004, 108, 18341). The method uses experimentally accessible data of electron deformation densities, an Extended Born model, and basic thermodynamic data on atomization and ionization energies that are available for compounds across the periodic table. This method enabled the systematic derivation of reliable force fields for inorganic compounds, which proved impossible before, and is the backbone for the INTERFACE approach.
Figure 11. The extended Born model employs known trends in atomization and ionization energies of the elements in the determination of atomic charges of atoms in molecules and crystals (J. Phys. Chem. B 2004, 108, 18341).
Heinz introduced a method to analyze local stress tensors down to atomic resolution in molecular dynamics and Monte Carlo simulations that is consistent with the macroscropic description of stress tensors according to the virial theorem (Phys. Rev. E, 2005, 72, 066704; Mol. Sim. 2007, 33, 747). The method finds applications in biomolecular and heterogeneous systems.
Figure 12. Calculation of local stress tensors in the presence of many-body interactions (Phys. Rev. E, 2005, 72, 066704).
The Heinz team, in collaboration with Sika AG, also introduced the first molecular models of tricalcium silicate, tricalcium aluminate, tobermorites, and other cement minerals (see Langmuir 2013, 29, 1754; J. Phys. Chem. C 2013, 117, 10417; Dalton Trans. 2014, 43, 10602). Typical deviations from experiment are <0.5% in lattice parameters, <5% in elastic moduli, and <5% in cleavage energies. Concrete is the world’s most abundant man-made material with an annual production of over 2 billion tons. The new modeling resources help understand hydration and organic interactions in high accuracy on the 1 to 1000 nm scale, aiming at lowering global CO2 emissions and improving earthquake-resistant building structures by optimization of composition and properties from the nanoscale to the macroscopic scale.