Polymer field theory

The standard continuum model of flexible polymers, introduced by Edwards (Edwards 1965), treats a solution composed of ${\displaystyle n}$ linear monodisperse homopolymers as a system of coarse-grained polymers, in which the statistical mechanics of the chains is described by the continuous Gaussian thread model (Baeurle 2007) and the solvent is taken into account implicitly. The Gaussian thread model can be viewed as the continuum limit of the discrete Gaussian chain model, in which the polymers are described as continuous, linearly elastic filaments. The canonical partition function of such a system, kept at an inverse temperature ${\displaystyle \beta =1/k_{B}T}$ and confined in a volume ${\displaystyle V}$, can be expressed as

${\displaystyle Z(n,V,\beta )={\frac {1}{n!(\lambda _{T}^{3})^{nN}}}\prod _{j=1}^{n}\int D\mathbf {r} _{j}\exp \left(-\beta \Phi _{0}\left[\mathbf {r} \right]-\beta {\bar {\Phi }}\left[\mathbf {r} \right]\right),\qquad (1)}$

where ${\displaystyle {\bar {\Phi }}\left[\mathbf {r} \right]}$ is the potential of mean force given by,

${\displaystyle {\bar {\Phi }}\left[\mathbf {r} \right]={\frac {N^{2}}{2}}\sum _{j=1}^{n}\sum _{k=1}^{n}\int _{0}^{1}ds\int _{0}^{1}ds'{\bar {\Phi }}\left(\left|\mathbf {r} _{j}(s)-\mathbf {r} _{k}(s')\right|\right)-{\frac {1}{2}}nN{\bar {\Phi }}(0),\qquad (2)}$

representing the solvent-mediated non-bonded interactions among the segments, while ${\displaystyle \Phi _{0}[\mathbf {r} ]}$ represents the harmonic binding energy of the chains. The latter energy contribution can be formulated as

${\displaystyle \Phi _{0}[\mathbf {r} ]={\frac {3k_{B}T}{2Nb^{2}}}\sum _{l=1}^{n}\int _{0}^{1}ds\left|{\frac {d\mathbf {r} _{l}(s)}{ds}}\right|^{2},}$

where ${\displaystyle b}$ is the statistical segment length and ${\displaystyle N}$ the polymerization index.

To derive the basic field-theoretic representation of the canonical partition function, one introduces in the following the segment density operator of the polymer system

${\displaystyle {\hat {\rho }}(\mathbf {r} )=N\sum _{j=1}^{n}\int _{0}^{1}ds\delta \left(\mathbf {r} -\mathbf {r} _{j}(s)\right).}$

Using this definition, one can rewrite Eq. (2) as

${\displaystyle {\bar {\Phi }}\left[\mathbf {r} \right]={\frac {1}{2}}\int d\mathbf {r} \int d\mathbf {r} '{\hat {\rho }}(\mathbf {r} ){\bar {\Phi }}(\left|\mathbf {r} -\mathbf {r} '\right|){\hat {\rho }}(\mathbf {r} ')-{\frac {1}{2}}nN{\bar {\Phi }}(0).\qquad (3)}$

Next, one converts the model into a field theory by making use of the Hubbard-Stratonovich transformation or delta-functional transformation

${\displaystyle \int D\rho \;\delta \left[\rho -{\hat {\rho }}\right]F\left[\rho \right]=F\left[{\hat {\rho }}\right],\qquad (4)}$

where ${\displaystyle F\left[{\hat {\rho }}\right]}$ is a functional and ${\displaystyle \delta \left[\rho -{\hat {\rho }}\right]}$ is the delta functional given by

${\displaystyle \delta \left[\rho -{\hat {\rho }}\right]=\int Dwe^{i\int d\mathbf {r} w(\mathbf {r} )\left[\rho (\mathbf {r} )-{\hat {\rho }}(\mathbf {r} )\right]},\qquad (5)}$

with ${\displaystyle w(\mathbf {r} )=\sum \nolimits _{\mathbf {G} }w(\mathbf {G} )\exp \left[i\mathbf {G} \mathbf {r} \right]}$ representing the auxiliary field function. Here we note that, expanding the field function in a Fourier series, implies that periodic boundary conditions are applied in all directions and that the ${\displaystyle \mathbf {G} }$-vectors designate the reciprocal lattice vectors of the supercell.

Using the Eqs. (3), (4) and (5), we can recast the canonical partition function in Eq. (1) in field-theoretic representation, which leads to

${\displaystyle Z(n,V,\beta )=Z_{0}\int Dw\exp \left[-{\frac {1}{2\beta V^{2}}}\int d\mathbf {r} d\mathbf {r} 'w(\mathbf {r} ){\bar {\Phi }}^{-1}(\mathbf {r} -\mathbf {r} ')w(\mathbf {r} ')\right]Q^{n}[iw],\qquad (6)}$

where

${\displaystyle Z_{0}={\frac {1}{n!}}\left({\frac {\exp \left(\beta /2N{\bar {\Phi }}(0)\right)Z'}{\lambda ^{3N}(T)}}\right)^{n}}$

can be interpreted as the partition function for an ideal gas of non-interacting polymers and

${\displaystyle Z'=\int D\mathbf {R} \exp \left[-\beta U_{0}(\mathbf {R} )\right]\qquad (7)}$

is the path integral of a free polymer in a zero field with elastic energy

${\displaystyle U_{0}[\mathbf {R} ]={\frac {k_{B}T}{4R_{g0}^{2}}}\int _{0}^{1}ds\left|{\frac {d\mathbf {R} (s)}{ds}}\right|^{2}.}$

In the latter equation the unperturbed radius of gyration of a chain ${\displaystyle R_{g0}={\sqrt {Nb^{2}/(6)}}}$. Moreover, in Eq. (6) the partition function of a single polymer, subjected to the field ${\displaystyle w(\mathbf {R} )}$, is given by

${\displaystyle Q[iw]={\frac {\int D\mathbf {R} \exp \left[-\beta U_{0}[\mathbf {R} ]-iN\int _{0}^{1}ds\;w(\mathbf {R} (s))\right]}{\int D\mathbf {R} \exp \left[-\beta U_{0}[\mathbf {R} ]\right]}}.\qquad (8)}$

To derive the grand canonical partition function, we use its standard thermodynamic relation to the canonical partition function, given by

${\displaystyle \Xi (\mu ,V,\beta )=\sum _{n=0}^{\infty }e^{\beta \mu n}Z(n,V,\beta ),}$

where ${\displaystyle \mu }$ is the chemical potential and ${\displaystyle Z(n,V,\beta )}$ is given by Eq. (6). Performing the sum, this provides the field-theoretic representation of the grand canonical partition function,

${\displaystyle \Xi (\xi ,V,\beta )=\gamma _{\bar {\Phi }}\int Dw\exp \left[-S[w]\right],}$

where

${\displaystyle S[w]={\frac {1}{2\beta V^{2}}}\int d\mathbf {r} d\mathbf {r} 'w(\mathbf {r} ){\bar {\Phi }}^{-1}(\mathbf {r} -\mathbf {r} ')w(\mathbf {r} ')-\xi Q[iw]}$

is the grand canonical action with ${\displaystyle Q[iw]}$ defined by Eq. (8) and the constant

${\displaystyle \gamma _{\bar {\Phi }}={\frac {1}{\sqrt {2}}}\prod _{\mathbf {G} }\left({\frac {1}{\pi \beta {\bar {\Phi }}(\mathbf {G} )}}\right)^{1/2}.}$

Moreover, the parameter related to the chemical potential is given by

${\displaystyle \xi ={\frac {\exp \left(\beta \mu +\beta /2N{\bar {\Phi }}(0)\right)Z'}{\lambda ^{3N}(T)}},}$

where ${\displaystyle Z'}$ is provided by Eq. (7).

The RG theory makes use of a series of RG transformations, each of which consists of a coarse-graining step followed by a change of scale (Wilson 1974). In case of statistical-mechanical problems the steps are implemented by successively eliminating and rescaling the degrees of freedom in the partition sum or integral that defines the model under consideration. De Gennes used this strategy to establish an analogy between the behavior of the zero-component classical vector model of ferromagnetism near the phase transition and a self-avoiding random walk of a polymer chain of infinite length on a lattice, to calculate the polymer excluded volume exponents (de Gennes 1972). Adapting this concept to field-theoretic functional integrals, implies to study in a systematic way how a field theory model changes while eliminating and rescaling a certain number of degrees of freedom from the partition function integral (Wilson 1974).

An alternative approach is known as the Hartree approximation or self-consistent one-loop approximation (Amit 1984). It takes advantage of Gaussian fluctuation corrections to the ${\displaystyle 0^{th}}$-order MF contribution, to renormalize the model parameters and extract in a self-consistent way the dominant length scale of the concentration fluctuations in critical concentration regimes.

In a more recent work Efimov and Nogovitsin showed that an alternative renormalization technique originating from QFT, based on the concept of tadpole renormalization, can be a very effective approach for computing functional integrals arising in statistical mechanics of classical many-particle systems (Efimov 1996). They demonstrated that the main contributions to classical partition function integrals are provided by low-order tadpole-type Feynman diagrams, which account for divergent contributions due to particle self-interaction. The renormalization procedure performed in this approach effects on the self-interaction contribution of a charge (like e.g. an electron or an ion), resulting from the static polarization induced in the vacuum due to the presence of that charge (Baeurle 2007). As evidenced by Efimov and Ganbold in an earlier work (Efimov 1991), the procedure of tadpole renormalization can be employed very effectively to remove the divergences from the action of the basic field-theoretic representation of the partition function and leads to an alternative functional integral representation, called the Gaussian equivalent representation (GER). They showed that the procedure provides functional integrals with significantly ameliorated convergence properties for analytical perturbation calculations. In subsequent works Baeurle et al. developed effective low-cost approximation methods based on the tadpole renormalization procedure, which have shown to deliver useful results for prototypical polymer and PE solutions (Baeurle 2006a, Baeurle 2006b, Baeurle 2007a).

To make the methodology amenable for computation, Baeurle proposed to shift the contour of integration of the partition function integral through the homogeneous MF solution using Cauchy's integral theorem, providing its so-called mean-field representation. This strategy was previously successfully employed by Baer et al. in field-theoretic electronic structure calculations (Baer 1998). Baeurle could demonstrate that this technique provides a significant acceleration of the statistical convergence of the ensemble averages in the MC sampling procedure (Baeurle 2002, Baeurle 2002a).

In subsequent works Baeurle et al. (Baeurle 2002, Baeurle 2002a, Baeurle 2003, Baeurle 2003a, Baeurle 2004) applied the concept of tadpole renormalization, leading to the Gaussian equivalent representationof the partition function integral, in conjunction with advanced MC techniques in the grand canonical ensemble. They could convincingly demonstrate that this strategy provides a further boost in the statistical convergence of the desired ensemble averages (Baeurle 2002).