Exponents (

## 1. Introduction

The modelling of separation chromatographic processes reported in the literature is, in general, related to macroscopic approaches for the phenomenological representation of the mass transfer mechanisms involved. In such models the microscopic aspects of the porous medium structure, related to the separation mechanisms, are incorporated implicitly, limiting the quality of the representation of the separation systems, which are strongly inﬂuenced by the micro porous adsorbent.

The modelling of ﬂuid ﬂow in porous media through the application of interconnected networks, which considers the global result from a system of interconnected microscopic elements, is related to the concepts of percolation theory.

The classical macroscopic models of chromatography have limitations in representing the structural parameters of the solid adsorbents, such as topology and morphology, as well as population effects of the molecules in the liquid phase, i.e. multi-molecules movement. Such important microscopic properties can be represented applying interconnected network models which can lead to a better understanding of the phenomenological aspects that contribute to the separation mechanisms in micro-porous media.

In the simulation of the molecules ﬂow through the column porous medium, a stochastic approach is utilized to represent the adsorption, diffusion and convection phenomena.

The molecules can move freely in the network structure, from one neighbor site to another, being respected the requirement that the ﬁnal position is not occupied by another molecule. Therefore, two molecules cannot occupy the same network site.

This chapter is dedicated to the modelling of continuous chromatography with a network approach combined with Monte Carlo like random walk stochastic methods. The porous structure of the solid adsorbent phase of the chromatographic column is represented by two and three dimensional networks, respectively square and cubic lattices (Oliveira et al., 2008; Biasse et al. 2010), in which population effects are taken into account, being related to the movement of multi-molecules modeled by stochastic phenomena of adsorption, desorption, diffusion and advection.

The use of network models to study chromatographic separation processes has been observed in the literature with different techniques and applications (Loh & Wang, 1995; Kier et al., 2000; Loh & Geng, 2003; Geng & Loh, 2004; Bryntensson, 2002; Oliveira et al., 2008, Biasse et al., 2010). In the work of Kier et al. (2000), a square network model was applied in the representation of the chromatographic column utilizing a cellular automata approach. The authors assumed arbitrarily pre-defined probabilities for the particles motion and interactions among them. Loh & Geng (2003) applied a cubic network model of interconnected cylindrical pores in the study of chromatographic systems of perfusion. Topological and morphological aspects, such as connectivity and pore size distribution, were analyzed, being observed a great influence of such porous adsorbent media characteristics on the mass transfer of the phenomena studied. In Geng & Loh (2004), the porous structure of the column was modeled considering three different Gaussian distributions of pore size, in order to represent the macro-pores, the micro-pores and the interstitial pores.

In this chapter, the fluid advection is assumed to be the main factor contributing to the molecules movement in the network structure. Such assumption is reasonable since the fluid movement in the chromatographic column comes from the external driving force provided by the pumping system. Two pore dimensions are assumed in the porous structure of adsorbent, the small and large cavities, which leads to steric and non-steric effects, respectively, due to the required space for the molecules movement.

The application of interconnected network models combined with stochastic phenomena of adsorption, diffusion and advection represents the main dynamical behaviors of the chromatographic processes of separation. The multi-molecules population effects allow the study of the dynamics of percolation through the chromatographic column, being therefore possible to evaluate the influence of molecules concentration on the mass transport phenomena along the chromatographic column.

## 2. Phenomenological and structural modelling

In the adsorption phenomenon, the molecule arriving at a new site can be adsorbed according to a probability of adsorption (

The adsorption and desorption phenomena are represented by

in which a molecule of solute

The simulation of the dynamic process of chromatographic separation in adsorption columns was performed combining the network modelling of the column porous structure with the stochastic modelling of molecules movement and interactions percolating the system.

In the schematic representation of the adsorption column shown in Fig. 1, the symbols C_{0} and C represent the solute concentrations at the entrance and at the exit of the column, respectively.

A square network was used to model surface phenomenon (studying equilibrium isotherms) and a cubic one to model the diffusion phenomenon. The porous medium of the chromatography column was represented by a two or three-dimensional cubic network model of interconnected sites (Vide Fig. 2). In such structure the connectivity, i.e. the number of neighbors connected to each site, is equal to four (2D) or six (3D). Each intersection node or site in the network corresponds to a potential adsorption location, in which the solute molecule may be adsorbed, being permitted only one adsorbed molecule per intersection.

In Fig. 2 one can see the graphical representation of both models used with the percolation threshold values (

### 2.1. Adsorption isotherm model

In this stage, molecules adsorption in stirred tanks was modeled, so that the adsorption surface was represented as a two-dimensional network. In Fig. 3 the process is represented, the liquid phase molecule can be adsorbed at the adsorbent surface material, being the latter represented as the network at the tank bottom.

In this model it is assumed that the liquid phase molecules can be adsorbed at the network according to an adsorption probability (

where

### 2.2. Diffusion phenomenon modelling

The diffusion phenomenon model was preformed through the "random walks" technique both for the two-dimensional and three-dimensional percolation at the square network and at the cubic network, respectively. Using this procedure one can use diffusion parameters for chromatography porous media and topological properties from network models, and obtain a fundamental universal correlation for these complex phenomena. Two different diffusion phenomena were studied at this stage: the first assumes that a certain molecule may follow any direction at the network; and the second that assumes that a molecule is only able to diffuse in the axial direction. The second case establishes diffusion relations for the axial dispersion for a chromatographic column, which is a important phenomenon that draws a lot of attention in separation chromatographic processes. In general, on the macroscopic models, only dispersion at the axial direction is taken into account, neglecting the transversal dispersion, that makes the numerical solutions to be simpler. In the present modelling, both axial and transversal dispersion are implicit on the stochastic "random walk" model, since the molecule is able to diffuse on both directions.

Equation 4 represents the diffusion mechanism, in which the molecule can diffuse in (

At first, a probability of occupation for network elements (

Simulations for different time ranges were made (10, 25, 50 and 75), being monitored the distance (

In the case of smaller molecular diffusion (

where

### 2.3. Movement rules and steric effect

In the simulation of the molecules movements, four rules (MR – movement rules) are considered to be representative of the diffusion and advection mechanisms. Such rules are schematically represented in Fig. 4.

From Fig. 4 it can be observed that the movement rules are determined by the directions considered in each situation. In the movement rules (I) and (II), the molecules can move in all directions of the structure while in the movement rules (III) and (IV) the molecules cannot move in the direction 6, which corresponds to the movement against the longitudinal main stream ﬂow direction, i.e. the upstream movement is not allowed. The MR-(II) and (IV) are similar to MR-(I) and (III), respectively, with a higher chance of movement in the direction 5 (longitudinal or axial direction).

The MR-(I) is considered to be representative of the diffusion mechanism of solute molecules through the porous structure.

This assumption is coherent as there is not a driving mechanism forcing the ﬂow in any particular direction of the network. In this case, according to the ﬁrst law of Fick (Bird et al., 2002), the solute ﬂow is determined by the concentration gradient of molecules, without signiﬁcant effects of external forces. The diffusion mechanism of MR-(I) is governed by

in which

The MR-(II), (III) and (IV) are considered to be representative of the advective mechanism as these conﬁgurations favor the ﬂow in the axial downstream direction of the column.

The advective mechanism MR-(II) is governed by

in which

The advective mechanisms MR-(III) and (IV) are determined, respectively, by

From Eq. (8) it can be observed that the solute movement in the direction 6 is prohibited, having the same probability for the moves in the other directions. In Eq. (9) as in Eq. (7), the probability of moving in the direction 5 is greater than in the other directions.

The simulation of the ﬂow of solute molecules was carried out according to the procedure described next. The molecules were introduced randomly at the column entrance (at the nodes with k = 1), maintaining the concentration constant at this section (C_{0}). The molecules concentration (C) at the column exit (k = 30) was calculated as the ratio of the number of molecules occupying the network intersections (sites) and the total number of intersections. In each step of the simulation, corresponding to a discrete time (

An important parameter to be taken into account in the stochastic modelling is the number of simulations (N), which indicates the number of times that the same procedure of the simulation is performed using the same control parameters. The increase in the number of simulations (N) leads to a decrease in the dispersion of the calculated value.

The steric effects were also investigated, that is, while the solid phase is able to adsorb one molecule at each site, one possibility is that the liquid phase site is able to contain one molecule and another possibility is that it is able to contain a unlimited number of molecules. In Fig. 5 are represented those two possibilities, being (A) for the case without the steric limitation and (B) representing the steric restriction.

## 3. Results and discussions

### 3.1. Langmuir isotherm model for surface adsorption

In Fig. 6 are represented the adsorption isotherms: in (A) are the results obtained for the simulations using a square network and in (B) are the experimental and deterministic model adapted from Silva (2004). One can observe that the results obtained using the stochastic adsorption method are representative for the studied phenomenon with the classic Langmuir isotherm.

These results show that the stochastic phenomenology is determinant to the behavior of equilibrium systems with multimolecules, and the overall result is governed by the individual actions of each component.

### 3.2. Results for the diffusion phenomenon (2D and 3D modelling)

In Figs. 7, 8 and 9 are represented the percolation evolution both in the square and cubic networks using different moving mechanisms. In Fig. 7 are represented the percolation using a square network of 50x50 nodes, for a time equal to 2000 steps, for four directions, and occupation probability

In Fig. 8 are represented results for the same conditions considered before, but allowing only 3 directions for the dispersion mechanism. One can observe a greater tendency towards the axial diffusion.

A evolution percolation method for 3D is presented in Fig. 9 for two movement mechanisms.

In Figs. 10 and 11 are presented the distance (

We can obtain the expoent

From the analysis of Figs. 10 and 11 we are able to observe that diffusivity, which presents a relation with

It is observed, in axial and radial dispersion situations, represented by Figs. 10B and 11B, a significant increase in the value of

The analysis of the power law Eq. 5 (Biasse et al., 2010), not only related to the distance (

### 3.3. Results of 3D network column with advective phenomenon

All results presented in this section were obtained for 5000 simulations. In Figs. 12 and 13 the test results varying

In Fig. 14 it is presented a comparison of the simulation results, without considering axial dispersion, with the experimental data acquired by Cruz (1997). One can observe the good results obtained by the network simulations, despite the breakthrough curve too sharp when equilibrium is reached (

Another important parameter to be considered here is the equilibrium fraction (

In Fig. 15 the same network, as the one used to obtain the results shown in Fig. 14 and Table 2, was considered assuming also the axial dispersion phenomenon.

Finally, in Fig. 16 are presented the simulation results (3D network model) for the discrete relative concentration (

The resistance to the percolation through the chromatographic column decreases as the relative concentration (

The highest medium resistance corresponds to an adsorption probability near 0.68 which is the minimum point of the curve. At this point the number of molecules percolating the system is reduced to the lowest level.

It must be noted that the medium resistance disappears at adsorption probability equal to 1 as at this point the adsorption is irreversible (no desorption is observed), and no molecule already adsorbed goes back to the liquid phase of the cavity, and therefore there is no interference in the movement of the other molecules.

## 4. Conclusions

The network stochastic surface multimolecular modelling was able to represent the behavior of Langmuir type adsorption isotherms. Such computational tool can be used to better understand the adsorption mechanisms to surfaces, resulting in the improvement of adsorbent materials.

The diffusion modelling through the 2D and 3D network "random walks" presented results obeying the power law with universal exponents well defined, being the latter related to the diffusive coefficients.

The effects of the power law were observed through the distance

A final important result is the observation of the existence of a direct relationship between the adsorption and desorption probabilities ratio with the equilibrium constant

## Acknowledgements

The authors acknowledge the financial support provided by CNPq, Conselho Nacional de Desenvolvimento Científico e Tecnológico, FAPERJ, Fundação Carlos Chagas Filho de Amparo à Pesquisa do Estado do Rio de Janeiro, and CAPES, Coordenação de Aperfeiçoamento de Pessoal de Nível Superior.

## References

- 1.
Biasse A. D. Fonseca E. X. Silva Neto. A. J. Câmara L. D. T. 2010 Power-Laws of Diffusion Phenomena In Chromatographic Columns, Advances and Applications in Mathematical Sciences2 2 313 320 - 2.
Bird R. B. Stewart W. E. Lightfoot E. N. 2002 Transport Phenomena, second ed., Wiley. - 3.
Bryntesson L. M. 2002 Pore Network Modelling of the Behaviour of a Solute in Chromatography Media: Transient and Steady-State Diffusion Properties, J. Chromatogr. A, 945,103 115 - 4.
Cruz J. M. 1997 Insulin Adsorption in Ion Exchanger Resins using Fix and Fluidzed Beds, M.Sc. Dissertation, Chemical Engineering Faculty, UNICAMP. (In Portuguese). - 5.
Geng A. Loh K. C. 2004 Effects of Adsorption Kinetics and Surface Heterogeneity on Band Spreading in Perfusion Chromatography- A Network Model Analysi s, Chem. Eng. Sci. 59,2447 2456 - 6.
Kier L. B. Cheng C. K. Karnes H. T. 2000 A Cellular Automata Model of Chromatography, Biomed. Chromatogr. 14,530 534 - 7.
Loh K. C. Geng A. 2003 Hydrodynamic Dispersion in Perfusion Chromatography- A Network Model Analysis, Chem. Eng. Sci. 58,3439 3451 - 8.
Loh K. C. Wang D. I. C. 1995 Characterization of Pore Size Distribution of Packing Materials Used in Perfusion Chromatography Using Network Model, J. Chromatogr. A, 718,239 255 - 9.
Oliveira G. P. Câmara L. D. T. Silva Neto. A. J. 2008 Modelling of the Separation Mechanisms in Chromatographic Columns through Adsorption and Advection in Two Dimensional Networks, XI Computational Modelling Meeting, Volta Redonda, Brazil. (In Portuguese) - 10.
Silva, 2000 Study of Insulin Adsorption in Ion Exclange Resin Columns: Experimental Parameters and Modelling, D.Sc. Thesis, Chemical Engineering Faculty, UNICAMP. (in Portuguese) - 11.
Stauffer D. Aharony A. 1992 Introduction to Percolation Theory Second Edition.