Methods for estimating k(T,P) values¶
The objective of each of the methods described in this section is to reduce the master equation into a small number of phenomenological rate coefficients \(k(T,P)\). All of the methods share a common formalism in that they seek to express the population distribution vector \(\mathbf{p}_i\) for each unimolecular isomer \(i\) as a linear combination of the total populations of all unimolecular isomers and bimolecular reactant channels.
The modified strong collision method¶

rmgpy.pdep.msc.
applyModifiedStrongCollisionMethod
(network, str efficiencyModel='default')¶ The modified strong collision method utilizes a greatly simplified collision model that allows for a decoupling of the energy grains. In the simplified collision model, collisional stabilization of a reactive isomer is treated as a singlestep process, ignoring the effects of collisional energy redistribution within the reactive energy space. An attempt to correct for the effect of collisional energy redistribution is made by modifying the collision frequency \(\omega_i(T,P)\) with a collision efficiency \(\beta_i(T)\) estimated from the lowpressure limit falloff of a single isomer.
By approximating the reactive populations as existing in pseudosteady state, the master equation is converted to a matrix equation is at each energy. Solving these small matrix equations gives the pseudosteady state populations of each isomer as a function of the total population of each isomer and reactant channel, which are then applied to determine the \(k(T,P)\) values.
In practice, the modified strong collision method is the fastest and most robust of the methods, and is reasonably accurate over a wide range of temperatures and pressures.
The reservoir state method¶

rmgpy.pdep.rs.
applyReservoirStateMethod
(network)¶ In the reservoir state method, the population distribution of each isomer is partitioned into the lowenergy grains (called the reservoir) and the highenergy grains (called the active space). The partition generally occurs at or near the lowest transition state energy for each isomer. The reservoir population is assumed to be thermalized, while the activespace population is assumed to be in pseudosteady state. Applying these approximations converts the master equation into a single large matrix equation. Solving this matrix equation gives the pseudosteady state populations of each isomer as a function of the total population of each isomer and reactant channel, which are then applied to determine the \(k(T,P)\) values.
The reservoir state method is only slightly more expensive than the modified strong collision method. At low temperatures the approximations used are very good, and the resulting \(k(T,P)\) values are more accurate than the modified strong collision values. However, at high temperatures the thermalized reservoir approximation breaks down, resulting in very inaccurate \(k(T,P)\) values. Thus, the reservoir state method is not robustly applicable over a wide range of temperatures and pressures.
The chemicallysignificant eigenvalues method¶

rmgpy.pdep.cse.
applyChemicallySignificantEigenvaluesMethod
(network, list lumpingOrder=None)¶ In the chemicallysignificant eigenvalues method, the master equation matrix is diagonized to determine its eigenmodes. Only the slowest of these modes are relevant to the chemistry; the rest involve internal energy relaxation due to collisions. Keeping only these “chemicallysignificant” eigenmodes allows for reduction to \(k(T,P)\) values.
The chemicallysignificant eigenvalues method is the most accurate method, and is considered to be exact as long as the chemicallysignificant eigenmodes are separable and distinct from the internal energy relaxation eigenmodes. However, this is often only the case near the highpressure limit, even for networks of only modest size. The chemicallysignificant eigenvalues method is also substantially more expensive to apply than the other methods.