Binding models¶
The following binding models are presented in dynamic binding mode. By replacing all occurrences of \(\mathrm{d}q / \mathrm{d}t\) with \(0\), quasistationary (rapidequlibrium) binding mode is achieved. In quasistationary binding, it is assumed that ad and desorption take place on a much faster time scale than the other transport processes such that bead liquid phase \(c_{p,i}\) (or bulk liquid phase \(c_i\) for certain unit operation models) are always in equilibrium with the solid phase \(q_i\).
Equilibrium constants¶
For the quasistationary binding mode, adsorption and desorption rate are no longer separate entities. Instead, the quotient \(k_{\text{eq}} = k_a / k_d\) of adsorption and desorption coefficient is the relevant parameter as shown for the linear binding model (see Section Linear):
The equilibrium constant \(k_{\text{eq},i}\) is used in CADET by setting \(k_{d,i} = 1\) and \(k_{a,i} = k_{\text{eq},i}\).
Note that adsorption rate \(k_{a,i}\) and desorption rate \(k_{d,i}\) are linearly correlated in both binding modes due to the form of the equilibrium constant \(k_{\text{eq}}\):
This correlation can potentially degrade performance of some optimization algorithms. While in quasistationary binding mode this is prevented by using the technique above, a dynamic binding model has to be reparameterized in order to decouple parameters:
This can be achieved by a (nonlinear) parameter transform
Reference concentrations¶
Some binding models use reference concentrations \(c_{\text{ref}}\) and \(q_{\text{ref}}\) of the mobile phase modulator (e.g., salt) in the particle liquid and solid phase, respectively. The reference values are mainly used for normalizing adsorption and desorption rates, but also for other parameters that appear with those concentrations. They amount to a simple parameter transformation that is exemplified at one equation of the steric mass action binding model
where \(c_{p,0}\) denotes the mobile phase salt concentration and
is the number of available binding sites which is related to the number of bound salt ions. Using the parameter transformation
we obtain the modified model equation
This transformation serves as a (partial) nondimensionalization of the adsorption and desorption rates and, by properly choosing the reference concentrations \(c_{\text{ref}}\) and \(q_{\text{ref}}\), may improve the optimizer performance.
Recommended choices for \(c_{\text{ref}}\) are the average or maximum inlet concentration of the mobile phase modifier \(c_0\), and for \(q_{\text{ref}}\) the ionic capacity \(\Lambda\). Note that setting the reference concentrations to \(1.0\) each results in the original binding model.
Dependence on external function¶
A binding model may depend on an external function or profile \(T\colon \left[ 0, T_{\text{end}}\right] \times [0, L] \to \mathbb{R}\), where \(L\) denotes the physical length of the unit operation, or \(T\colon \left[0, T_{\text{end}}\right] \to \mathbb{R}\) if the unit operation model has no axial length. By using an external profile, it is possible to account for effects that are not directly modeled in CADET (e.g., temperature). The dependence of each parameter is modeled by a polynomial of third degree. For example, the adsorption rate \(k_a\) is really given by
While \(k_{a,0}\) is set by the original parameter XXX_KA
of the file format (XXX
being a placeholder for the binding model), the parameters \(k_{a,3}\), \(k_{a,2}\), and \(k_{a,1}\) are given by XXX_KA_TTT
, XXX_KA_TT
, and XXX_KA_T
, respectively.
The identifier of the externally dependent binding model is constructed from the original identifier by prepending EXT_
(e.g., MULTI_COMPONENT_LANGMUIR
is changed into EXT_MULTI_COMPONENT_LANGMUIR
).
This pattern applies to all parameters and supporting binding models (see Table 3).
Note that the parameter units have to be adapted to the unit of the external profile by dividing with an appropriate power.
Each parameter of the externally dependent binding model can depend on a different external source.
The 0based indices of the external source for each parameter is given in the dataset EXTFUN
.
By assigning only one index to EXTFUN
, all parameters use the same source.
The ordering of the parameters in EXTFUN
is given by the ordering in the file format specification in Section Binding models.
Binding model feature matrix¶
A short comparison of the most prominent binding model features is given in Table 3. The implemented binding models can be divided into two main classes: Singlestate and multistate binding. While singlestate models only have one bound state per component (or less), multistate models provide multiple (possibly different) bound states for each component, which may correspond to different binding orientations or binding site types. The models also differ in whether a mobile phase modifier (e.g., salt) is supported to modulate the binding behavior.
Binding model 
Competitive 
Mobile phase modifier 
External function 
Multistate 

× 
× 
✓ 
× 

✓ 
× 
✓ 
× 

✓ 
× 
✓ 
× 

✓ 
✓ 
✓ 
× 

✓ 
✓ 
✓ 
× 

✓ 
✓ 
✓ 
× 

✓ 
✓ 
✓ 
× 

✓ 
✓ 
✓ 
× 

✓ 
✓ 
✓ 
× 

× 
× 
✓ 
× 

✓ 
× 
✓ 
✓ 

✓ 
× 
✓ 
✓ 

✓ 
✓ 
✓ 
✓ 

✓ 
✓ 
× 
✓ 

✓ 
✓ 
✓ 
✓ 