Zidan Zhanga,
Long Wanga,
Zilu Wanga,
Xuehao He*a,
Yu Chena,
Florian Müller-Platheb and
Michael C. Böhmb
aSchool of Science, Tianjin University and Collaborative Innovation Center of Chemical Science and Engineering (Tianjin), 300072 Tianjin, China. E-mail: xhhe@tju.edu.cn
bEduard-Zintl-Institut für Anorganische und Physikalische Chemie and Center of Smart Interfaces, Technische Universität Darmstadt, Alarich-Weiss-Str. 4, D-64287 Darmstadt, Germany
First published on 23rd October 2014
A coarse-grained molecular dynamics (CG-MD) and reactive Monte Carlo (RMC) hybrid method (CG-MD + RMC) has been developed to investigate the hyperbranched polycondensation of 3,5-bis(trimethylsiloxy)benzoyl chloride to poly(3,5-dihydroxybenzoic acid). The CG force field to describe the formation of the hyperbranched macromolecules has been extracted from all-atom molecular dynamics simulations by the mapping technology of iterative Boltzmann inversion. In the mapping process branched poly(3,5-dihydroxybenzoic acid) in an all-atom description has been employed as a target object to derive the CG force field for hyperbranched polymers. In the RMC simulations, the reactivity ratio of the functional groups has been optimized by fitting experimental data with the iterative dichotomy method (Macromolecules, 2003, 36, 97). Using such a simulation framework, detailed information including the molecular weight, the molecular weight distribution and the branching degree of a specific polymerization process has been derived. Radial distribution functions of the atomistic and coarse-grained systems are in excellent agreement. A good agreement between the present simulations and experiment has been demonstrated, too. Especially, the intramolecular cyclization fraction has been reproduced quantitatively. This work illustrates that the present reactive CG-MD + RMC model can be used for quantitative studies of specific hyperbranched polymerizations.
In the past, a variety of theoretical and simulation methods such as mean-field based kinetic models,7–13 lattice and off-lattice Monte Carlo (MC) models14–18 were developed to explore the reaction kinetics and structural properties of hyperbranched polymers. They have been studied with and without the consideration of properties such as steric effects, the relaxation of polymer chains and intramolecular cyclizations. Most previous efforts were focused on the generic properties of hyperbranched polymerizations. These approaches covered the influence of substitutions,9,12,18–21 the addition of core molecules,11,22–25 intramolecular ring units,14,17,26,27 the monomer concentration,17,21,28 and the chain rigidity as well as the reaction reversibility29 on the molecular weight distribution and degree of branching. To our knowledge, investigations of specific hyperbranched polymerizations using a molecular model with structural details have been reported neither in theoretical nor in simulation studies.30–32 Such techniques, however, are the prerequisite to understand specific hyperbranched polymerizations quantitatively.
In contrast to generic theoretical methods, molecular dynamics (MD) simulations can describe specific systems under realistic conditions. The recently developed reactive molecular dynamics (RMD) methods are a feasible way to explore complex reactive systems. In RMD approaches, chemical reactions are introduced into a classical model.33 Thus one has to restrict such simulations to reactive processes where (nuclear) quantum effects can be neglected. In recent Feynman path-integral Monte Carlo studies, it had been shown that (intramolecular) reactions of non-hydrogen atoms can be described adequately in a classical approximation for temperatures ≥298 K.34,35 In rearrangement processes of hydrogen atoms, quantum fluctuations are non-negligible in the vicinity of (and even below) room-temperature. The heavy coarse-grained beads, see below, in the present reactive approach imply that a classical approximation should be sufficient. If this is guaranteed the predictive capability of RMD and conventional MD simulations strongly depends on the accuracy of the force field. Some works using generic potentials without molecular details36–38 studied cross-linking reactions and phase separations induced by polymerization processes.39–42 The more accurate but complex atomistic force fields including RMDff,43 REBO,44 AIREBO45 and ReaxFF46 were developed with the intention to simulate degradation and decomposition reactions of polymers and energetic materials,47–54 formation processes and properties of carbon systems,46, 55–60 as well as transition metal catalyzed reactions.61,62 The merits of these reactive force fields have been summarized in a recent state of the art description.33 Review articles on earlier reactive MD methods and the force fields employed in these studies had been published before.63,64
It is well known that simple generic potentials cannot be used to study specific systems, although they are useful to derive general trends as a function of a chosen grid of parameters. Realistic potentials from, e.g., quantum chemistry calculations65 in atomistic resolution, however, require extensive computational capacities when calculating large systems. In order to enhance the efficiency in the simulation of specific systems, coarse-graining (CG) techniques66,67 have been proposed in MD and MC studies. Among other applications such CG simulations have been employed in studies of biological and soft matter systems.67–71 In a CG picture, several atoms of the atomistic system are grouped together to so-called superatoms or CG beads. As a matter of fact, CG force fields combine the advantages of generic and atomistic simulations but avoid their shortcomings. On the one hand, they reproduce certain characteristics of the systems (e.g. structural,72–75 thermodynamic68,76 or dynamic properties69,77), on the other, they improve the computational efficiency to a great extent. CG force fields have been applied recently to study polymerization reactions of specific molecules. For example, Farah et al. analyzed the living polymerization of styrene78 and cross-linking reactions near selective surfaces.79 The adopted CG force fields were developed by iterative Boltzmann inversion (IBI) with the aim to reproduce the molecular structures.75 Liu et al.80 studied the formation mechanism of a typical epoxy resin on carbon fibres by using reactive dissipative particle dynamics (DPD).81,82 Preceding atomistic MD simulations have been performed to determine the parameters required in the DPD simulations. The computational capability of RMD methods equipped with an accurate CG force field has been demonstrated recently in a study analyzing the curing of bifunctional epoxies with pentafunctional amines in the neighborhood of a surface allowing amine adsorption.83 More information on reactive MD techniques can be found in the review articles mentioned above.33,63,64
In the present work, we develop a reactive coarse-grained model by hybridizing molecular dynamics and reactive Monte Carlo methods (CG-MD + RMC) to study a specific hyperbranched polymerization quantitatively. A capture radius criterion similar to the work of Farah et al.33 has been employed in our study to initiate and control chemical reactions. In contrast to past simulation approaches,33,36–38,78 several refinements have been considered in the present reactive model. Thus the probability of a reaction is correlated with the structure of the reacting molecules. With this improved model, we simulate the polyesterification of bulk 3,5-bis(trimethyl-siloxy)benzoyl chloride (BTMSBCl) (Fig. 1a) to yield hyperbranched poly(3,5-dihydroxybenzoic acid) (PBTMSBCl) (Fig. 1b). This polyester is one of the earliest prepared hyperbranched polymers which is formed by AB2 type monomers.5,30,84 Note that A, B denote CG beads in the studied systems. The reaction kinetics is still not fully understood although it has been studied both in experiments and in the framework of a mean-field theory.30 In our simulations, the chosen reaction temperature (T) of 443 K is the same as in experiment.30 BTMSBCl has been used as the reactive monomer to retain typical characteristics of hyperbranched polyester units. The substitution effect (i.e. the influence of an electron-donating or electron-withdrawing group on a reactive step) is represented by a reactivity ratio for the coarse-grained functional groups. In the present work, a “functional group” characterizes a subunit (i.e. a bead in the CG picture) in the monomer or polymer; see Fig. 1 for a graphical representation. We have used experimental data (ratios or mole fractions of structural units, conversion rates of functional groups)30 to optimize the reactivity ratios. With our model, quantitative information on the degree of polymerization, the polydispersity index, the degree of branching and intramolecular cyclization can be obtained. Radial distribution functions calculated in the atomistic and coarse-grained systems have been compared by using a reverse mapping technique.85–87 The present simulations are also correlated with experimental results.
Monomer | Polymer, unreacted | Polymer, reacted | |||
---|---|---|---|---|---|
MA | –COCl | PA | –COCl | Pa | –CO– |
MB | –OSiMe3 | PB | –OSiMe3 | Pb | –O– |
MC | –C6H3 | PC | –C6H3 |
The total interaction potential of the coarse-grained system (Utotal) is composed of the bonded (Ubond) and nonbonded potential (Unb).
Utotal = Ubond + Unb | (1) |
The bonded potential includes three elements, i.e. the bond (Ubo), angle (Uan) and improper dihedral potential (Uim). The improper dihedral in this work allows to conserve a planar structure of the four CG particles of a monomeric unit.
Ubond = Ubo + Uan + Uim | (2) |
In our CG model, the bonded potentials are either given analytically in the harmonic approximation or provided in tabulated form. Most bonded interactions can be described as harmonic form; for the complex target angle distributions involving the –OTMS function, the tabulated form is required. The tabulated form is also used for the nonbonded (nb) potential. The initial values bo0, an0, im0 of the CG parameters for the harmonic potentials (eqn (3)) and the tabulated potentials (eqn (4)) can be calculated by simple Boltzmann inversion of distribution functions (pbo, pan, pim) obtained from atomistic reference simulations. The symbols bo, an and im denote the actual values of bondlength, angle and improper dihedral angle.
![]() | (3) |
Unb = −kBT![]() ![]() | (4) |
In the last equation g(r) is the radial distribution function (rdf). kB is the Boltzmann constant. The rdfs have been measured with respect to the bead center which has been explained at the end of Section 2.2.
In the definition of the CG force fields, the tabulated potentials have been fixed initially when optimizing the parameters of the harmonic interactions. The spring constants (kbo, kan, kim) and the structural equilibrium values (bo0, an0, im0) in eqn (3) have been tuned iteratively until the target (t) distributions of bonds, angles and improper dihedrals were fitted with sufficient accuracy. After that, the harmonic potential has been fixed and the tabulated potential has been optimized using iterative Boltzmann inversion (IBI).74
![]() | (5) |
The values Ux,i+1(r) and Ux,i(r) in eqn (5) (with x = an, nb) symbolize the potential in the (i + 1)th and ith iteration step, f and ν are two small parameters to tune the correction term in each iteration step. pan,i and gi(r) symbolize the angle and radial distribution functions in the ith iteration while pan,target and gtarget(r) stand for the target pan and g(r) from an atomistic simulation. The final distribution functions for BTMSBCl and branched PBTMSBCl that have been derived on the basis of the optimized CG force fields, are shown in the ESI, SF4–SF7.† The agreement between CG and atomistic reference distributions is very good. The parameters of the harmonic potential of BTMSBCl and PBTMSBCl are shown in S2 and S3.† The tabulated force fields are provided in the supporting files, too.
For the polymerized system with its monomeric units and polymer chains, the interaction potential between the components is required. We use the arithmetic mean of the interactions between identical components (eqn (6)).
![]() | (6) |
UM and UP represent the nonbonded interaction potentials between the monomers (BTMSBCl) and between polymer units (PBTMSBCl), respectively. The distribution functions of both resolutions at all conversions of A function, cA, have been checked to make sure that the present coarse-grained model is suitable. It can be seen in Fig. 2 that all radial distribution functions agree well in their peak positions. There are only small differences in their peak intensities which are enlarged with increasing cA (Fig. 2, lower right diagram). Thus it can be assumed that the above model assumptions and the combination rule applied in developing the CG force fields are a reasonable strategy to simulate hyperbranched polymerizations (specific radial distribution functions of A, B and C functions can be found in S8.†). In order to compare the diffusion behaviours of atomistic and CG models, the diffusion constants, DM and DP in the atomistic and CG models (here, M and P stand for monomer and branched polymer, respectively, see Fig. 1) have been calculated. The values of DM equal 539.7 and 3806.5 μm2 s−1 for the atomistic and CG model, respectively. For the polymer, DP equals 30.9 and 1001.5 μm2 s−1 for atomistic MD and CG-MD, respectively. The diffusion of CG molecules is much faster than atomistic molecules. Although polymerizations are kinetic processes the difference in the diffusion coefficient between atomistic and CG approaches causes no problem as the simulation time scale is not related to the experimental time scale.
![]() | ||
Fig. 2 Calculated nonbonded radial distribution functions of all particles, g(r), from fully atomistic and coarse-grained simulations at different conversions of A functions, cA. |
Pr = rA/arB/brBB/Bb/bb | (7) |
![]() |
![]() |
![]() |
|
---|---|---|---|
AB![]() |
rArBrBB | rArBrBb | rArBrbb |
Ab![]() |
rArbrBB | rArbrBb | rArbrbb |
aB![]() |
rarBrBB | rarBrBb | rarBrbb |
Ab![]() |
rarbrBB | rarbrBb | rarbrbb |
The hyperbranched polymerizations have been simulated in the NVT ensemble with the side length, L, of the cubic simulation box of 7.95 nm which agrees with the atomistic target system. Periodic boundary conditions are applied in all directions. In each simulation, 1000 BTMSBCl monomers were placed into the simulation box. The integration time step of all MD runs amounts to 0.005 ps. The Nose–Hoover thermostat (coupling time is 0.05 ps) was employed to keep the temperature fluctuating around 443 K. For each parallel sample, the unreacted system was equilibrated for 1.0 ps starting with a random initial state. After that, the reaction sequence was started. Whenever 1% of the A functions have reacted or when each particle has attempted to react once (one Monte Carlo step, MCS), the system is relaxed by a normal MD simulation of 0.2 ps to remove the produced reaction heat from the system and to release local strain. Reactive steps are continued, until the desired conversion rate (99%) is reached. For each reactive simulation, 30 parallel samples have been calculated and averaged to obtain the final results.
The coarse-grained nonbonded force field between monomers and polymers has been discussed in Section 2. The product mixture with its monomers and hyperbranched polymers is obtained as follows. A capture radius is 0.5 nm has been chosen. Initially all reactivity parameters (rA, ra, rB, rb, rBB, rBb, rbb) are identical (1.0). After each reaction event re-equilibration and excess heat removal is achieved by a MD run of 0.2 ps.
The polydispersity index, PDI, calculated by eqn (8) is used to characterize the weight distribution in the polymer. The quantity Pw is the weight-average degree of polymerization, Pn the number-average pendant. The branching degree is characterized by the fraction of branching points, FB, and the degree of branching, DB. The two quantities are defined in eqn (9) and (10) respectively, where ND is the number of branched units, NL the number of linear units and NA the total number of reacted units. In AB2 type hyperbranched polymerizations, intramolecular cyclizations occur; for every hyperbranched polymer only one ring can be formed. The ratio between the number of hyperbranched polymers with intramolecular cyclization (nring) and the total number of polymer molecules (ntot) defines the cyclization fraction Rfrac (eqn (11)).
PDI = Pw/Pn | (8) |
FB = ND/NA | (9) |
DB = 2ND/(2ND + NL) | (10) |
![]() | (11) |
rm+1i = rmi + δs | (12) |
In the fitting procedure, the step length δs is tuned according to the dichotomy method, which has been employed successfully in a recent work of some of the present authors.32 The reactivity parameters are optimized one by one, with the sequence from rB to rbb. In all iteration cycles, the step length δs for each reactivity parameter changes from 0.1 to 0.001.
The deviation between simulation and experimental results30 is characterized by the root mean square error, RMS, as defined in eqn (13). The mole fractions of the considered six structural units (ABB, ABb, Abb, aBB, aBb, abb) in the given equation are classified by the label o. Each one has been calculated for 17 different conversions of CG beads A labeled by the second index q. The quantities doq and d′oq are the ratios of these structural units in the simulation and in experiment. In each iteration step, 30 separate samples have been calculated and averaged to obtain the final results. When the parameter RMS reaches an almost flat region near zero, the obtained optimized ratios of the functional groups approach the experimentally determined mole fractions.
![]() | (13) |
The activating or deactivating influence on a polymerization for case (1) can be understood in the following way. Suppose that an BB unit has been chosen as the reactive fragment; the reactive CG bead is
. Then assume that we have an aB
and an ab
unit at the same separation from the reactive center. According to eqn (7), the required descriptors P1r (for reaction between
BB and aB
) and P2r (for reaction between
BB and ab
) are defined by triple products of “reactivity parameters”, i.e., rBBrarB and rBBrarb. As a result we observe P1r/P2r = rB/rb. The grading rB > rb indicates that the unreacted B function has an activating effect and the reaction prefers to happen between the ABB unit and the aBB unit. On the other hand, if rB < rb, the unreacted B function has a deactivating effect and the reaction prefers to happen between the ABB unit and the abB unit.
Fig. 3 (corresponding to cases 1, 2 and 3, respectively) describes the influence of the three reactivity ratios rB/rb, rA/ra, and rBB/rbb on the ratio of the observed structural units in the studied polymerizations. This ratio Runit can be defined with the help of eqn (14) where Nunit is the number of structural units of a given type at a certain conversion of function A and Nall the number of all structural units (Nall = 1000 in the chosen simulation sample). The unit type has been described in Section 3.2.
![]() | (14) |
The solid lines in the three diagrams of Fig. 3 are benchmark curves characterized by rB/rb = rA/ra = rBB/rbb = 1. In the three sets of simulations reactivity ratios of 0.1, 1 and 10 have been calculated. As the reaction progresses, the reactivity ratios have different effects on the ratio of structural units. Before cA reaches 0.2, all three reactivity ratios have a similar influence because in the early stage of the polymerization the reaction mainly happens between monomers forming a dimer in the first reactive step. After the initial stage of the polymerization, case (1) shows a larger influence on the curve shape and final composition. The cases (2) and (3) have only a moderate influence on the ratio of the structural unit curves. When cA approaches 1.0, rB/rb shows a remarkable influence on the ratio of aBB, aBb and abb type units. With rB/rb increasing, the aBB (terminal) and abb (dendritic) numbers decrease and the ratio of aBb (linear) units increases. The quantities rA/ra and rBB/rbb, however, exhibit only a weak influence on the ratio of the various units when cA is close to 1.0. The two latter ratios do not influence the final composition of the six structural units, but have impacts on the distribution of structural units along the conversion of the A function.
The changes in the ratio of the various units directly influence the structure of hyperbranched polymers (Fig. 4). The fraction of branching points FB increases remarkably with the decrease of rB/rb (Fig. 4d), as the unreacted B functions in a linear unit can react more easily. In comparison to rB/rb, the reactivity ratios rA/ra (Fig. 4e) and rBB/rbb (Fig. 4f) show a weaker influence on FB. Differences in the FB increase at not too high conversion rates and then decrease when cA approaches 1.0. Similar to the research results of previous works using lattice and off-lattice Monte Carlo methods,9,12,18,19,21 the present simulation results indicate that tuning the value of rB/rb is a good way to control the degree of branching of hyperbranched polymers. The simulations describing the influence of the three reactivity ratios are consistent with the mean-field predictions of Schmaljohann et al. although intramolecular cyclization and steric specification of molecules are not considered in their work.9
Fig. 4 also shows the influence of the reactivity ratios rB/rb, rA/ra and rBB/rbb on the polydispersity index, PDI (top panels). When rB/rb, rA/ra and rBB/rbb are large, the PDIs increase more slowly than PDIs under smaller rB/rb, rA/ra and rBB/rbb ratios. A similar phenomenon has been observed for FB (bottom panels). In comparison to rA/ra and rBB/rbb, the reactivity ratio rB/rb exhibits a more obvious influence on PDI and FB, Fig. 4. In the lattice Monte Carlo model,21 rB/rb shows a similar influence on the value of PDI. The PDI splitting comes from the fact that three processes contribute to the polymerization, i.e. reactions between monomers, monomer and polymer, as well as polymer and polymer. Note that we have discriminated only between different structural units; details in the chemistry are explained later. The normalized contributions of the different processes (PMM, PMP, PPP) are defined in eqn (15).
![]() | (15) |
nMM, nMP, nPP are the numbers of reactions between two monomers, a monomer and a polymer and two polymers at a certain conversion of A function.
At a low conversion of the A function, the reaction between two monomers (denoted as MM in Fig. 5) is the most probable process because the monomer concentration is high. The reaction number fraction PMM between two monomers shows an obvious decrease at low conversion of the A function cA < 0.4. Meanwhile, the reaction rate (PMP) between monomer and oligomer increases. Compared with PMP, the expected increase of the reaction probability PPP between two oligomers shows a delay. After the initial stage, all three types of reactions (PMM, PMP and PPP) have reached a plateau and the normalized elements of eqn (15) are close to 1/3 especially at smaller reactivity ratios. At a certain conversion of the A function, larger reactivity ratios lead to an increase of the reaction number fraction PMM; at the same time they reduce the reaction number fractions PMP and PPP. This can be explained by the deactivating influence of the reacted A or B functions on the unreacted functions when the reactivity ratios are large. It results in an only moderate increase of the PDIs with an enlargement of the reactivity ratios at smaller cA.
![]() | ||
Fig. 5 Dependence of the normalized contributions for the reactions between two monomers (PMM), monomer and polymer (PMP), and two polymers (PPP) on the conversion of A functions, cA, at different reactivity ratios. The definition of the normalized reaction contributions is described in eqn (15). The reactivity ratios are not changed during a simulation series; they are set equal to 1.0. |
When adopting the initial condition, the ratios of some structural units deviate noticeably from experimental results (see Fig. 6a). Within 12 successful iterations, the root mean square error, RMS, gradually decreases (ESI, S9†). When the RMS fluctuates around a constant minimum, the optimized reactivity parameters are achieved (see Table 3). We see that the reacted A (=a) function has a deactivating effect on the unreacted B function. The reacted B function (=b), however, has an activating effect on the unreacted B function and the unreacted A function. These influence factors make the degree of branching, DB, of PBTMSBCl polymers larger than the maximal value at ideal conditions (DBideal = 0.5).7 We want to emphasize that the reactivity ratios are consistent with the ratios of the Hammett constant94 of the studied groups. The Hammett constants σm of –COCl (A), –COOPh (a), –OSiMe3 (B) and –OCOPh (b) are 0.51, 0.37, 0.13 and 0.21, respectively. The higher the Hammett constant σm, the stronger is the electron-withdrawing effect. High σm values indicate a high stability of the phenolate anion structure in the rate-determining step. It results in an activating effect, i.e. in a higher ri. It should be mentioned that in our CG scheme the reacted A function (=a) represents a –CO– group while the reacted B function (=b) represents –O– group. In the estimation of the Hammett constants they refer to –COOPh and –OCOPh fragments. With the above numbers the experimental data are fitted well by the present simulations (Fig. 6b). Parallel tests have been performed by starting from different initial states and step lengths in the iterations; they all led to similar results.
For the optimized reactivity parameters, the degrees of polymerization (see legend to Fig. 7), the polydispersity index, PDI, and the degrees of branching, FB, DB, have been calculated (Fig. 7) and compared with experimental results. In experiment,30 Pn amounts to 15 at cA = 0.93, while in our simulations we derive Pn = 17.4 at the same cA. The DB ratio approaches 0.6 for a conversion of 1.0 in both experiment and simulation. The given comparisons emphasize that the simulation results are very close to experimental data. This indicates that our reactive model has the ability to study specific hyperbranched polymerizations quantitatively. Additionally our method provides information on the amount of ring polymers (Rfrac) and the size distribution of intramolecular rings. The simulation results can be found in Fig. 7 and Table 4. According to the structural properties of the monomers, the smallest intramolecular ring contains three units. With an increase of cA, the Rfrac first increase slowly at lower cA and then increase sharply when cA approaches 1.0. At cA = 0.99, Rfrac equals 0.47. The concentration of rings, Rs (defined as the number of rings of a specific size divided by the overall number of rings), with lengths parameters, s, from 3 to 8 (and an addition with s > 8) also change as a function of cA. As shown in Table 4, the Rs values of smaller rings exceed the Rs of larger rings for almost all cA numbers. With the rise of cA, the Rs of smaller rings (e.g. s = 3 and 4) gradually decrease; for larger rings (e.g. s > 8), however, they increase. Thus we can deduce that larger rings form preferably at higher conversions and a larger degree of polymerization. By using the reverse mapping technique, we furthermore compare the radial distribution functions, g(r), of all particles calculated in the coarse-grained and fully atomistic systems after 1 ns MD simulation time; they are shown in Fig. 8 (in addition to the g(r)s of all particles, g(r)s profiles of individual A, B and C functions are shown in S10†). Similar to the results in Fig. 2, the g(r)s obtained in the two spatial resolutions fit well at all conversions. It again indicates that the reactive coarse-grained hybrid model is highly effective to explore quantitatively hyperbranched polyesterification. The configuration of the smallest accessible cyclic unit in the hyperbranched polymer is shown in ESI, S11.† It reveals that the CG force fields developed in this work provide detailed information on the structure of hyperbranched polymers.
cA | Concentration of rings (Rs) | ||||||
---|---|---|---|---|---|---|---|
s = 3 | s = 4 | s = 5 | s = 6 | s = 7 | s = 8 | s > 8 | |
0.8 | 0.550 | 0.276 | 0.107 | 0.033 | 0.012 | 0.020 | 0.018 |
0.9 | 0.483 | 0.216 | 0.135 | 0.027 | 0.016 | 0.018 | 0.105 |
0.95 | 0.462 | 0.190 | 0.118 | 0.033 | 0.014 | 0.012 | 0.171 |
0.99 | 0.419 | 0.178 | 0.103 | 0.029 | 0.010 | 0.019 | 0.237 |
Footnote |
† Electronic supplementary information (ESI) available: Figures refer to the calculated system density, the calculated angle distribution functions and nonbonded radial distribution functions, the root mean square error and the all-atom configuration of the smallest accessible cyclic unit in the hyperbranched polymer. See DOI: 10.1039/c4ra10271a |
This journal is © The Royal Society of Chemistry 2014 |