It is difficult to fully separate many solvents from water. An ethanol solution containing 4 wt.% water is an azeotrope: the vapour has the same composition as the liquid when boiled. Such mixtures cannot be separated by standard distillation. Pervaporation uses a membrane to separate such mixtures, writes Dr Cilian Ó Súilleabháin. 

Hot liquid feed enters a membrane module, permeate vaporises and passes through the membrane and the remaining solvent-rich liquid leaves as liquid retentate.

The first industrial pervaporation system was installed in 1983. However, there were no short-cut methods for sizing pervaporation systems. End users were reliant on the manufacturers of equipment for the sizing and design of pervaporation systems.  

Slow adoption of this technology

This is believed to be a cause of the slow adoption of this technology. The author developed equations that allow the membrane area to be calculated and systems designed for industrial applications where water is removed from solvent such as ethanol and isopropanol. This was the subject of a PhD thesis supervised by Dr Greg Foley of DCU.

Pervaporation systems usually involve a series of three to five adiabatic pervaporation modules interspersed with heat exchangers. The enthalpy required to vaporise the water causes the remaining liquid to be cooled leading to a reduction in flux. For most industrial applications, flux is proportional to water concentration and has an Arrhenius-type relationship with temperature. 

Fig 2. Adiabatic pervaporation module with multiple stages

Fig 3. Flux in a 3-stage pervaporation system with adiabatic modules and re-heaters

The dashed line is the flux at the feed temperature: this is proportional to the water concentration. Starting at the top right with the flux at the feed concentration, and moving diagonally down to the left, the flux drops due to the decreasing temperature and decreasing concentration. 

On reheating to the feed temperature after the first stage the flux rises to 'J_reheat'.  A similar pattern is repeated for each subsequent stage.

Constant permeate concentration across a range of liquid concentrations is assumed. This was been validated by analysis of data in peer-reviewed publications relating to commercial membranes.

An exact analytical expression was derived for the average flux in ideal isothermal pervaporation modules where flux is proportional to concentration.  This simple equation only requires readily available data. 

It provides a benchmark when deciding the number of stages to be used in a multi-stage system. The expression can be modified so as to predict the progress of batch isothermal pervaporation operations. This provides an easier and more accurate method for determining flux from bench-scale isothermal trials than was available heretofore.

A is the membrane area. J_f is the flux at the feed conditions, z_i and y_i are the feed and permeate mass fractions. ṁ_p and ṁ_f are the mass flow rates of the feed and permeate respectively.

There are some systems where flux is independent of concentration. Eq. (2), an analytical expression, was derived for such systems. Eq. (3) is a simple short-cut equation that is accurate to 1.1% for the full range of industrial applications and solvents for which pervaporation is used.

No analytical solution was found for the more common case of adiabatic pervaporation where flux is proportional to concentration. Short-cut equations were developed for ethanol and isopropanol, the two most common solvents for which pervaporation is used.

J_r is the flux at the retentate concentration and temperature, and J_reheat is the flux at the retentate composition and feed temperature. x_ri and z_i are the mass fractions of the water in the feed and retentate respectively. a, b, c, d, e, f, and n are constants. 

Equations were developed that are accurate to 2.2% and 2.4% for isopropanol and ethanol respectively as well as for other solvents for the full range of industrial conditions.

The equations above model ideal systems. A larger membrane area is required for actual pervaporation systems. Thus an 'area efficiency' akin to the Murphree efficiency in distillation systems is required. 

For the first time, efficiency values for an industrial-scale system were calculated using data from published literature. These corresponded with published models.

Optimisation of ideal multi-stage dehydration systems were reviewed. Minimum overall area is achieved by having successively larger areas for later stages.  

A novel short-cut method for minimising the area of ideal systems was developed. Typical systems have equally sized membrane for all stages: it was found that this does not lead to substantial increases in overall area.

A new single metric, 'The Pervaporation Membrane Index', was developed that gives appropriate weighting to flux and separation. It is defined as the average flux such that the water content reduces from 1% w/w above to 1% w/w below the azeotropic concentration, multiplied by the pervaporation separation modulus. This facilitates comparisons of membranes and should encourage the development of membranes with higher fluxes.

Conclusion

The ideal models and novel metrics developed in this work provide a major step towards the development of simple design methods for pervaporation systems. Assumptions in this work are justified by reference to peer-reviewed publications relating to commercial membranes and industrial-scale pervaporation systems.

The research will be of benefit to engineers working in industry, particularly those who are new to the field of pervaporation. The short-cut design methods will facilitate consideration of pervaporation in process feasibility analyses and front-end design, albeit that more rigorous design methods are required for the subsequent detailed design of pervaporation systems. 

The availability of such short-cut methods should increase the adoption of pervaporation in industry. The short-cut methods will also be of use for the analysis of modules and system design parameters by undergraduate students. Further research will be undertaken to obtain more performance data from industrial systems.

Author: Dr Cilian Ó Súilleabháin, Cork Institute of Technology