Chemical Degradation Pathways in Siloxane Polymers Following

Nov 27, 2018 - (b) Quartile analysis of ensemble statistics for the total work performed plotted as a function of density. The Journal of Physical Che...
0 downloads 0 Views 7MB Size
Subscriber access provided by YORK UNIV

B: Fluid Interfaces, Colloids, Polymers, Soft Matter, Surfactants, and Glassy Materials

Chemical Degradation Pathways in Siloxane Polymers Following Phenyl Excitations Matthew P. Kroonblawd, Nir Goldman, and James P. Lewicki J. Phys. Chem. B, Just Accepted Manuscript • DOI: 10.1021/acs.jpcb.8b09636 • Publication Date (Web): 27 Nov 2018 Downloaded from http://pubs.acs.org on December 2, 2018

Just Accepted “Just Accepted” manuscripts have been peer-reviewed and accepted for publication. They are posted online prior to technical editing, formatting for publication and author proofing. The American Chemical Society provides “Just Accepted” as a service to the research community to expedite the dissemination of scientific material as soon as possible after acceptance. “Just Accepted” manuscripts appear in full in PDF format accompanied by an HTML abstract. “Just Accepted” manuscripts have been fully peer reviewed, but should not be considered the official version of record. They are citable by the Digital Object Identifier (DOI®). “Just Accepted” is an optional service offered to authors. Therefore, the “Just Accepted” Web site may not include all articles that will be published in the journal. After a manuscript is technically edited and formatted, it will be removed from the “Just Accepted” Web site and published as an ASAP article. Note that technical editing may introduce minor changes to the manuscript text and/or graphics which could affect content, and all legal disclaimers and ethical guidelines that apply to the journal pertain. ACS cannot be held responsible for errors or consequences arising from the use of information contained in these “Just Accepted” manuscripts.

is published by the American Chemical Society. 1155 Sixteenth Street N.W., Washington, DC 20036 Published by American Chemical Society. Copyright © American Chemical Society. However, no copyright claim is made to original U.S. Government works, or works produced by employees of any Commonwealth realm Crown government in the course of their duties.

Page 1 of 31 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

The Journal of Physical Chemistry

Chemical Degradation Pathways in Siloxane Polymers Following Phenyl Excitations Matthew P. Kroonblawd,∗,† Nir Goldman,†,‡ and James P. Lewicki† †Physical and Life Sciences Directorate, Lawrence Livermore National Laboratory, Livermore, CA 94550, United States ‡Department of Chemical Engineering, University of California, Davis, California 95616, United States E-mail: [email protected]

1

ACS Paragon Plus Environment

The Journal of Physical Chemistry 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Abstract We use ensembles of quantum-based molecular dynamics simulations to predict the chemical reactions that follow radiation-induced excitations of phenyl groups in a model copolymer of polydimethylsiloxane and polydiphenylsiloxane. Our simulations span a wide range of highly porous and condensed phase densities, and include both wet and dry conditions. We observe that in the absence of water, excited phenyl groups tend to abstract hydrogen from other methyl or phenyl side groups to produce benzene, with the under-hydrogenated group initiating subsequent intrachain cyclization reactions. These systems also yield minor products of diphenyl moieties formed by the complete abstraction of both phenyl groups from a single polydiphenylsiloxane subunit. In contrast, we find that the presence of water promotes the formation of free benzene and silanol side groups, reduces the likelyhood for intrachain cyclization reactions, and completely suppresses the formation of diphenyl species. In addition, we predict that water plays a critical role in chain scission reactions, which indicates a possible synergistic effect between environmental moisture and radiation that could promote alterations of a larger polymer network. These results could have impact in interpreting accelerated aging experiments, where polymer decomposition reactions and network rearrangements are thought to have a significant effect on the ensuing mechanical properties.

1

Introduction

Polysiloxane-based materials, or silicones, are widely used in a range of technological applications where favorable long term mechanical response, chemical inertness, and shape filling factor over a broad temperature range are required. Applications include space-filling rubber or foam gaskets and shock absorbers to biomedical implants, lubricants, and adhesives. 1,2 Many applications of silicone rubber and foam components have specific requirements for mechanical properties such as elastic moduli (bulk, Young’s, and shear), viscoelastic response, and hardness. The mechanical properties of silicones are often tuned by controlling

2

ACS Paragon Plus Environment

Page 2 of 31

Page 3 of 31 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

The Journal of Physical Chemistry

the distribution of chain lengths, the nature of cross links and chain termini, and through addition of other media including fillers or sacrificial pore formers. 3,4 Consequently, engineering silicones are highly complicated and have multicomponent structures with features at multiple length scales that are often poorly characterized. Chemical and physical degradation processes can result in undesirable and irreversible changes that alter mechanical performance over the service lifetime and result in macroscopic deformations to components through creep, permanent set, and brittle failure. 5–9 Possible drivers for chemical reactions in silicones vary by application conditions and include thermal cycling, exposure to atmospheric or head space gases, and ionizing radiation. Accelerated aging and degradation experiments performed on polydimethylsiloxane (PDMS) and polydiphenylsiloxane (PDPS) containing materials have typically assessed macroscopiclevel mechanical properties or accessed the underlying chemistry through chromatographic or spectroscopic measurements. 8,10,11 For instance, it is known that exposure to heat, acids, or bases leads to depolymerizing chain backbiting reactions that produce off gassing cyclic siloxanes 10,11 while exposure to moisture promotes hydrolysis reactions. 8 Moisture is particularly important to consider because it is postulated to promote hydrolytic chain scissions. 8 Exposure to γ radiation is thought to further enhance chemical degradation processes and can lead to undesirable macroscopic changes including compression set. 6–8 Irradiated PDPScontaining systems also consistently off gas benzene, which indicates a localization of energy sufficient to result in dephenylation reaction events. Significant changes in material hardness and stress relaxation properties accompany spectroscopic signatures for chemical degradation, but clear ties to specific atomic-scale chemical processes remain elusive. Atomistic simulations provide fundamental mechanistic insights into reactive processes and their consequences in polymeric materials that are difficult to elucidate from experiments alone. 5,12,13 Reactive force fields used with classical all-atom molecular dynamics (MD) are capable of reaching large time and space scales, but frequently are not transferable to regions of chemical phase space beyond their fitting regime without extensive tuning. 14–16 In

3

ACS Paragon Plus Environment

The Journal of Physical Chemistry 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

contrast, quantum-based molecular dynamics (QMD) provides an accurate and transferable approach to predict condensed-phase chemistry. However, the extreme computational expense of QMD simulations with ab initio methods such as density functional theory 17 (DFT) places significant limitations on the accessible time and length scales that can be studied. Semiempirical methods such as density functional tight binding 18,19 (DFTB) can retain much of the accuracy of DFT with several orders of magnitude reduction in computational expense. The improved computational efficiency of DFTB facilitates performing many independent simulations to gather ensemble statistics 20,21 and affords relatively high throughput to ascertain the influence of various structural, thermal, and environmental factors on chemical processes within the scope of a single study. Along these lines, we investigate the initial steps of benzene off-gassing in a model PDMS:PDPS copolymer system using QMD simulations based on the DFTB method. A steered dynamics approach 22 is combined with an ensemble methodology to sample dephenylation processes in a variety of compressed and stretched polymeric chains. A wide range of system densities are investigated (0.2 g cm−3 < ρ < 1.2 g cm−3 ) to bracket the wide range of local packing environments present in real silicone rubbers and foams. Our simulations reveal complicated reaction pathways including chain scissions, intrachain cyclizations, and the formation of diphenyl moieties. The influence of moisture is characterized and found to significantly alter both the reaction pathways and product formation probabilities compared to those for a desiccated environment.

2 2.1

Methods General Simulation Details

Quantum-based molecular dynamics simulations of a model siloxane PDMS:PDPS copolymer system were performed using the self-consistent charge DFTB method. 18,19 The DFTB total energy is derived from an expansion of the Kohn-Sham energy 17 about a reference electronic 4

ACS Paragon Plus Environment

Page 4 of 31

Page 5 of 31 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

The Journal of Physical Chemistry

density to second order in charge fluctuations and is evaluated as

EDFTB = EBS + ECoul + ERep + EDisp .

(1)

Here EBS is the electronic band structure energy, ECoul captures electrostatic interactions between fluctuating partial charges, and ERep and EDisp are empirical repulsion and dispersion terms. We used the DFTB parameter set pbc-0-3 (available at http://www.dftb.org), a typical off-the-shelf parameter set for silicon-containing systems. Universal force field dispersion terms were used for EDisp . DFTB employs a tight-binding framework 23 to efficiently construct the electronic Hamiltonian and overlap integrals from tables in terms of a nonorthogonal, atomic orbital basis set. The electronic structure was evaluated at the Γpoint only using Fermi-Dirac thermal smearing 24 with the electronic temperature set equal to the instantaneous ionic kinetic temperature at each time step. Molecular dynamics simulations were performed without spin polarization, as constrained optimizations described in Section 3.1 showed that adding spin polarization had little effect on DFTB predictions for the dephenylation energetic barrier. Trajectories were integrated using Extended Lagrangian Born-Oppenheimer dynamics 25–28 driven by LAMMPS 29 with forces and stresses evaluated by the DFTB+ code. 30 Orthorhombic or cubic three-dimensionally periodic simulation cells were used in all cases. Isothermalisochoric (N V T ) simulations were performed with a Nosé-Hoover-style thermostat 31,32 at T = 300 K. The time step was set to 0.20 fs and the electronic structure was evaluated using four self-consistent charge cycles per step. Isothermal uniaxial and isotropic strains of the simulation cell lengths were performed using the N V T /SLLOD method 33,34 at rate 0.05 nm ps−1 .

5

ACS Paragon Plus Environment

The Journal of Physical Chemistry 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

(a) Precursors PDMS

Page 6 of 31

(b) Model PBC

PDPS

O-Ph I-Ph

I-Ph

O-Ph

PBC (c) Pseudo Gas Phase

Lz (nm) 3.5 2.5 1.5

0.0 2.0 nm

(d) Condensed Phase 1.57 nm (0.6 g cm-3)

1.24 nm (1.2 g cm-3)

Figure 1: (a) Precursors for PDMS:PDPS copolymers. (b) Schematic for the model 2:1 PDMS:PDPS copolymer system. Snapshots of simulation cells corresponding to (c) ELD and (d) condensed phase configurations. Atoms are colored cyan, red, and yellow for carbon, oxygen, and silicon and hydrogen atoms are not shown for clarity. The simulation cell is drawn with green lines.

2.2

Model PDPS:PDMS System

Schematics and atomistic renderings of our model PDMS:PDPS copolymer system are shown in Figure 1. All atomistic renderings were prepared using the Open Visualization Tool 35 (OVITO). Precursors for the polymerization reaction are the cyclic species octamethylcyclotetrasiloxane and octaphenylcyclotetrasiloxane, which contribute four PDMS or PDPS monomers, respectively. A 2:1 PDMS:PDPS copolymer chain was constructed that was comprised of two PDMS precursor units bonded to a single PDPS one to form a single non6

ACS Paragon Plus Environment

Page 7 of 31 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

The Journal of Physical Chemistry

terminated, quasi-infinite backbone that extends through the periodic boundary (176 atoms total). We distinguish between inner and outer phenyl groups, respectively denoted I-Ph and O-Ph in Figure 1(b), based on their covalent separation from PDMS monomer units. There are eight unique phenyl groups (four inner and four outer) as each monomer contributes two. The simulation cell dimensions were set to yield several distinct configurations as shown in panels (c) and (d). Extremely low density configurations (ρ < 0.4 g cm−3 ; labeled ELD for short) were prepared in which the backbone was uniaxially strained along z to specific cell lengths Lz while holding the transverse dimensions fixed at 2.0 nm × 2.0 nm. In these cases, the modeled polymer strand is non-interacting with any possible neighboring strands. For reference, the equilibrium Lz for a pure PDPS crystal with the same number of monomer units (i.e., 12) is ≈3.0 nm. 36 Condensed phase configurations were generated by isotropically straining the cell to cubic dimensions with side lengths 1.57, 1.42, 1.32, and 1.24 nm, which yield systems with densities 0.6, 0.8, 1.0, and 1.2 g cm−3 . The lowest condensed phase density nominally corresponds to the bulk density of a typical cellular foam 8,11 and the highest density is similar to pure PDPS crystal. 36 ELD configurations correspond to local packing environments of particularly low density, such as near pores in a silicone foam, whereas the condensed phase configurations correspond to local regions of material that are at or near bulk density. We note that while our simulation cells are small, the chemical reaction events we seek to probe are typically highly localized. Larger simulation cell sizes would be required to directly ascertain the influence of chemical reactions on bulk mechanical properties. Hydrated versions of the ELD and condensed phase configurations were generated by adding ten H2 O molecules through random insertion without overlaps while holding the cell dimensions fixed (206 atoms total). Thus, the hydrated simulation cells have slightly higher densities than their dry counterparts. We performed a final 10 ps N V T thermalization for each cell configuration following hydration or the imposition of strains prior to sampling chemically reactive processes.

7

ACS Paragon Plus Environment

The Journal of Physical Chemistry 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

2.3

Page 8 of 31

Steered MD Simulation Details

Dephenylation events were sampled using steered dynamics simulations performed with the PLUMED 1.3 plugin. 37 A moving harmonic bias potential

VBias (R; t) =

K [R(t) − Ro (t)]2 , 2

(2)

was applied to a collective coordinate R(t) at time t that was distinct to each unique phenyl group. We defined R(t) for a given phenyl group to be the separation distance between the center of mass for the phenyl carbon atoms and the bonding silicon atom. The force constant K was set to 500 kcal mol−1 Å−2 and only one phenyl group was biased at a time in a given simulation. During a steered dynamics simulation, the centroid of the harmonic bias potential, Ro (t) = Req + ut,

(3)

was moved at a constant rate u away from the nominal equilibrium position Req = 3.3 Å. In the limit of u → 0, the work done to the system corresponds to the change in free energy ∆F whereas finite u will generally lead to an overestimation of ∆F . 22 Our primary interest is to ascertain possible chemical reaction pathways following localization of energy in the phenyl moiety, so performing a very large ensemble of steered simulations in order to converge an estimate for ∆F through the Jarzynski equation 22 was not considered here.

2.4

DFT Calculations

Density functional theory (DFT) single point calculations and configuration optimizations were performed for selected structures to validate DFTB results. Calculations were performed with VASP 38 using the Perdew-Burke-Ernzerhof 39 (PBE) generalized gradient approximation functional with projector-augmented wave (PAW) potentials 40,41 and Grimme D3 dispersion corrections. 42 The electronic structure was evaluated at the Γ-point only and

8

ACS Paragon Plus Environment

Page 9 of 31 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

The Journal of Physical Chemistry

we note in text those calculations which include spin polarization. The plane wave cutoff was set to 500 eV and we applied Fermi-Dirac thermal smearing with the electron temperature set to 25.85 meV (i.e., 300 K). The self-consistent field accuracy threshold was set to 10−6 eV.

3

Results and Discussion

3.1

Model Validation

Validation testing was performed to assess the accuracy of the DFTB method with the pbc0-3 parameter set for siloxane polymer systems. Single point calculations were performed for a pure PDMS a unit cell crystal (96 atoms) using the lattice parameters and atomic configuration reported in Ref. 36. The PDPS unit cell is orthorhombic (a = 20.15 Å, b = 9.829 Å, c = 4.944 Å) and contains two chains of PDPS with two monomers per chain that are periodically infinite along the c lattice direction. Comparison of the two predicted electronic densities of state (Figure 2) shows a good matching for many occupied state features with modest shifts in energy. The energy gap between the highest-occupied and lowest-unoccupied molecular orbitals (i.e., the HOMO-LUMO gap) is also quite similar, with DFTB and DFT predicting 4.507 eV and 4.133 eV, respectively. The influence of including spin polarization on the energetic barrier to remove a phenyl group was determined by performing four series of constrained optimizations that scanned the Si–C(phenyl) bond. Geometry optimizations were performed with a force-based convergence criterion of 10−1 eV Å−1 in both DFT and DFTB. This was found to be sufficient to provide a qualitative description of the underlying reactive potential energy surface. Here, an ELD configuration of the PDMS:PDPS copolymer model (see Figure 1) was used with Lz = 3.0 nm to avoid complications from steric effects. Scans were performed by displacing the phenyl group along the the Si–C(phenyl) bond separation vector and then optimizing the configuration while holding the Si and C(phenyl) atoms fixed. Two separate scans were per9

ACS Paragon Plus Environment

The Journal of Physical Chemistry

Electronic Density of States

DFTB

DFT

-20

-15

-10

-5 0 Energy (eV)

5

10

15

Figure 2: Electronic density of states for PDPS crystal computed using DFTB and DFT. Discrete energy levels were smoothed using a Gaussian smearing function with the variance parameter set to σ 2 = 0.01 eV2 . The HOMO energy is set to zero in both cases. 140 120

Energy (kcal mol-1)

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 10 of 31

100 80 60

DFTB (non spin polar)

40

DFTB (spin polar) DFT (non spin polar)

20 0

DFT (spin polar)

1

2 3 4 5 6 Si-C(phenyl) Separation Distance (Å)

7

Figure 3: ELD optimized scans of a Si–C(phenyl) bond in at 2:1 PDMS:PDPS copolymer. Optimizations were performed using DFTB and DFT, each with and without spin polarization. formed with DFTB and DFT each in order to test the effects of electronic spin polarization. Singlet and triplet starting configurations were considered for the spin polarized calculations and the singlet state was found to be consistently lower in energy. The energetic barriers predicted by the various approaches are qualitatively similar (Figure 3). As expected, 43 the spin polarized optimizations yield lower energies than their non spin polarized counterparts at large separation distances, but converge to the same solution at small separations. Non spin polarized DFT and DFTB optimizations yield similar energies

10

ACS Paragon Plus Environment

Page 11 of 31

Energy (kcal mol-1 atom-1)

2.5 2.0

(a)

1.5 1.0 0.5 0.0 -0.5 1.0

1.5

2.0

2.5 3.0 Lz (nm)

3.5

4.0

4.5

1.5

2.0

2.5 3.0 Lz (nm)

3.5

4.0

4.5

0.3 0.0 σzz (GPa)

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

The Journal of Physical Chemistry

(b)

-0.3 -0.6 -0.9 -1.2 -1.5 1.0

Figure 4: Thermally averaged mechanical response for a strained PDMS:PDPS copolymer chain showing (a) the change in relative energy and (b) the stress along the strained direction σzz . Data points correspond to time averages over the last 10 ps of 20 ps N V T trajectories and error bars correspond to the standard deviation of the mean. at small and large separation distances, but there is a notable discrepancy for intermediate distances. Gradients on the DFTB potential energy surface are generally higher, which is consistent with our past experience 20 using another standard off-the-shelf DFTB parameter set for organic systems. It is perhaps surprising that the DFTB predictions decrease by a modest ≈5 kcal mol−1 when spin polarization is included. This is in contrast to the more substantial ≈20 kcal mol−1 reduction predicted by DFT. Consequently, we use the non spin polarized DFTB approach in all subsequent simulations owing practically to its efficiency and its reasonable accuracy compared to non spin polarized DFT. Development of a DFTB parameter set specifically tailored for siloxane polymers is the subject of future work. Mechanical response properties for the ELD copolymer chain were determined by per-

11

ACS Paragon Plus Environment

The Journal of Physical Chemistry 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

forming 20 ps N V T simulations at various Lz strains over the interval 1.5 nm ≤ Lz ≤ 4.0 nm. The relative increase in energy is shown in Figure 4(a) and the stress along the strained direction σzz is shown in Figure 4(b). We use the convention of negative stress for a tensile (stretched) state. Comparison with the atomistic renderings in Figure 1(c) shows that even significant changes in local packing of the copolymer chain between 1.5 nm and 2.5 nm results in little change in the potential energy and negligible differences in the stress state. There is a marked increase in energy and tension when the chain is strained to 3.5 nm. We note that while the chain is stable when strained to 4.0 nm, it promptly scissions upon phenyl group excitation. Based on the mechanical response, we chose three different ELD chain strains (Lz = 1.5, 2.5, and 3.5 nm) that capture distinctly different stretched and conformational regimes. These range from condensed-like chain packing (1.5 nm), to a non-tensile state with reduced steric crowding of phenyl groups (2.5 nm), to a high-energy configuration under substantial tensile load where steric effects should be minimal (3.5 nm).

3.2

Chemistry in Desiccated Systems

Ensembles of QMD simulations were performed with ELD and condensed phase configurations without added water molecules to assess the sorts of chemical events that follow phenyl group excitations in desiccated environments. Specifically, we performed eight simulations per initial configuration in which one unique phenyl group per simulation was excited using the steered MD approach described in Section 2.3. Seven different starting configurations were considered that corresponded to three different ELD chain lengths (1.5, 2.5, and 3.5 nm) and four different condensed phase densities (0.6, 0.8, 1.0, and 1.2 g cm−3 ), leading to 56 independent simulations. The steering rate for the bias potential centroid was set to u = 0.50 Å ps−1 (see Equation 3) and each simulation was 10 ps in duration, yielding a combined 560 ps of trajectory. Reaction products in each simulation were assessed through direct visual inspection of the trajectories. The variety and distributions of chemical events predicted in our simulations did not vary substantially with respect to either the chain 12

ACS Paragon Plus Environment

Page 12 of 31

Page 13 of 31 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

The Journal of Physical Chemistry

length or density, so we focus here on describing the product types. Relative occurrence rates for each product type will be discussed in connection with results for hydrated systems in Section 3.4. Three different typical end states were predicted for the excited phenyl group. A majority of the time, the excited phenyl group abstracts a hydrogen atom from a neighboring phenyl or methyl side group to form benzene. The formation of benzene was expected based on accelerated aging experiments. Due to the short time scale (10 ps) for the simulations, in some instances the excited phenyl group did not have sufficient time to abstract a hydrogen and persisted for the duration as a free phenyl ion C6 Hδ− 5 with a slight negative partial charge (approximately −0.3e). The final typical end state involved the formation of a covalent CC bond between the excited phenyl and one of the other phenyl side groups to produce a diphenyl species. We performed analogous ensembles of simulations for the highest and lowest condensedphase densities in which the excited phenyl group was steered at a slower rate (0.25 Å ps−1 ) over 20 ps. (That is, 16 additional simulations yielding another 320 ps of combined trajectory.) The only major difference between the fast and slow steering rates was that the number of persistent free phenyl ions substantially decreased for the slower rate. This is consistent with the expectation that the short timescale persistence of free ions is a kinetic effect, not a thermodynamic one. Hydrogen abstraction by the excited phenyl group leads to production of benzene. However, this abstraction also produces an undercoordinated silicon atom (with a single phenyl side group) and an undercoordinated carbon atom on a nearby methyl or phenyl side group. When hydrogen is abstracted from a methyl or phenyl group on a neighboring monomer unit, what follows is the consistent formation of cyclic intrachain “cross-links” such as those shown in Figure 5. Ring closure reactions in which the undercoordinated silicon and carbon atoms covalently bond results in a system-wide closed shell electronic configuration with silicon remaining in an sp3 hybridized state. A number of different N -cycles were produced, including

13

ACS Paragon Plus Environment

The Journal of Physical Chemistry 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

4-cycle

4-cycle 5-cycle

7-cycle

48-cycle 8-cycle

Figure 5: Example cyclic species formed following the abstraction of hydrogen by an excited phenyl group. Atoms are colored cyan, red, yellow, and white for carbon, oxygen, silicon, and hydrogen. Carbon and hydrogen atoms that are not part of the cycle have been rendered as grey. 4-cycles when hydrogen is abstracted from a neighboring methyl and cycles with 5+ atoms when hydrogen is abstracted from a neighboring phenyl. We predict that cycles with four or more atoms form for all initial backbone configurations save for the most highly stretched ELD case (Lz = 3.5 nm). For this set of simulations, the only significant difference between exciting inner verses outer phenyl groups (see Figure 1(b)) is that 4-cycles specifically only form when an outer phenyl group is excited. Our ensembles of simulations also predict that the excited phenyl group can abstract a hydrogen atom from the other phenyl group on the same monomer unit. In this case, intrachain cyclization is not a viable route to a system-wide closed shell. Instead, DFTB predicted the formation of a (meta)stable 3-member ring (or 3-cycle) in a number of ELD and condensed phase simulations (Figure 6). The first reaction step in panel (a) is the abstraction of hydrogen from the phenyl group on the same monomer unit to form benzene and a charged phenyl side group with four hydrogen atoms. This is followed in panel (b) with

14

ACS Paragon Plus Environment

Page 14 of 31

Page 15 of 31 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

The Journal of Physical Chemistry

(a)

t' = 0 ps

(b)

t' = 1.6 ps

(c) 1.79 Å 1.40 Å 122.1

o

109.7o

1.40 Å a

b c

1.41 Å

1.43 Å 1.39 Å

118.1o

1.39 Å

a = 47.1o b = 66.5o c = 66.4o

Figure 6: Reaction mechanism forming a 3-member cycle between silicon and two carbon atoms. The first step in (a) involves the abstraction of hydrogen and the second step in (b) involves a 30o rotation of the bonded phenyl group, producing the 3-cycle. The geometry from a DFT-level optimization of the 3-cycle structure is shown in (c) where the rest of the system has been removed for clarity. Color conventions follow those in Figure 5, with the excited phenyl group also rendered in color. a 30o rotation of the side group leading to the formation of a 3-cycle involving the silicon backbone atom and two adjacent phenyl carbon atoms. The time delay in our simulations between the abstraction and rotation steps is on the order of picoseconds or more. Given that 3-atom cyclic structures are highly uncommon in organic chemistry, we performed additional calculations to validate this particular DFTB result. Two DFT-level optimizations were performed (with and without spin polarization) on a condensed phase configuration with a 3-cycle generated by a DFTB steered MD simulation. Both DFT optimizations converged to the same result and confirmed that the 3-cycle is indeed a local minimum. Figure 6(c) shows the 3-cycle structure resulting from a non spin polarized DFT optimization. A high degree of strain is evident. Only the O-Si-O angle retains a neartetrahedral value; the O-Si-C angles are modestly strained and the C-Si-C angle is highly

15

ACS Paragon Plus Environment

The Journal of Physical Chemistry 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

acute. Both Si-C bond lengths in the 3-cycle are 1.79 Å, which is substantially shorter than the other regular Si-C(phenyl) single bonds in the system (≈1.87 Å). The phenyl ring also strains to accommodate the 3-cycle through an increase in the length of the shared C-C bond (1.43 Å) compared to the C-C bond lengths in the free benzene molecule (1.40 Å).

3.3

Chemistry in Hydrated Systems

The effects of humidity on chemistry resulting from phenyl excitations were assessed by performing analogous ensembles of simulations to those described in Section 3.2, but with additional water molecules present. Hydrated configurations were prepared for each ELD strain and condensed phase density by adding 10 water molecules through random insertion without atomic overlaps. These configurations were equilibrated with 10 ps of N V T simulation prior to performed steered MD simulations of dephenylation. As before, the system strain/density had little to no discernible effect on the predicted chemistry. Comparison of results for fast (0.50 Å ps−1 ) and slow (0.25 Å ps−1 ) steering rates showed the rate had practically no effect on the ensemble distribution of final products when water is present. Water significantly alters the set of predicted reactions that follow phenyl excitations. Perhaps the most drastic difference is that in ≈65% of all simulations performed here, the excited phenyl group is replaced by a silanol group. In these cases, the excited phenyl typically abstracts a hydrogen atom from a water molecule. The hydroxyl forming the silanol side group and the abstracted hydrogen may or may not come from the same water molecule. Figure 7 shows a particularly complicated set of concerted proton transfer reactions that led to the formation of benzene and silanol in one simulation. Three different water molecules align to form a hydrogen-bond wire between the undercoordinated carbon and silicon atoms. In a 20 fs time span, sequential proton transfers occur along this wire resulting in formation of a Si-OH bond. The result is a benzene molecule, two water molecules, and the formation of a silanol side group. One comparatively more rare chemical event that did not occur in any of our simula16

ACS Paragon Plus Environment

Page 16 of 31

Page 17 of 31 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

The Journal of Physical Chemistry

t' = 0 fs

t' = 10 fs

t' = 20 fs

Figure 7: Concerted proton transfers between three bridging water molecules leading to formation of benzene and a silanol side group. Green arrows highlight hydrogen motions. Color conventions follow those in Figure 5, with atoms not participating in the reaction shown as grey. tions of dry systems is the scission of the Si-O backbone (Figure 8). Similar to the reaction sequence described for silanol group formation, the first step is the abstraction of hydrogen by the excited phenyl group leading to formation of benzene and hydroxyl. However, in this situation within 100 fs the hydroxyl group associates with a nearby sp3 coordinated silicon atom on a neighboring monomer unit leading to a hypercoordinated silicon site. This hypercoordinated silicon site can be immediately adjacent to the undercoordinated silicon produced by the previous phenyl excitation. The Si-O polymer chain bond between the hydroxylated, hypercoordinated silicon and the oxygen bonded to the undercoordinated silicon quickly scissions to produce two termini, resulting in one silanol and one silanone end group, respectively. The entire scission process occurs over several hundred femtoseconds. Silanones, along with most double bonded silicon functional groups, are usually unstable and highly reactive. 2 Only recently have stable silanones been isolated experimentally. 44 Thus, while a silanone terminal group persists for the remainder of the simulation in which it is formed (