Foto del docente

Giulio Cesare Sarti

Emeritus Professor

Alma Mater Studiorum - Università di Bologna


Keywords: solubility separation of biomolecules gas separations permeability glassy polymers affinity membranes membranes


Research activity has been devoted mainly to the following areas:


1.                  membrane separations.

1.1.      membrane distillation.

1.2.      development of affinity membranes; analysis of affinity membrane apparatus.

1.3.      membrane gas separations.

1.4.      metal membranes for hydrogen separation.

1.5.      reverse osmosis.

1.6.      pervaporation.


2.                  diffusionin polymers.

2.1.      Fickian diffusion and chemical post-treatment.

2.2.      non-Fickian diffusion.

2.2.1.   localized swelling and viscoelastic diffusion.

2.2.2.   stress and deformation effects in non-Fickian diffusion.


3.                  thermodynamics and thermomechanical properties of polymeric fluids.

3.1.            solubility, diffusivity and permeability in solid polymers.

3.2.            model predictions and correlations for the solubility in glassy polymers.

3.3.            thermo-rheological constitutive equations.

3.2.      foundations of rational thermodynamics.


4                                Separation and purification of proteins and immunoglobulins

4.1.      development of affinity membranes and affinity membrane modules.

4.2.      experimental characterization of affinity membranes and affinity membrane modules.

4.3.      modelling of affinity membrane behavior.


5                                chemical processes  in microelectronics.

5.1       CVD reactions.

5.2       growth of silicon oxide and silicon nitride.


The research activity lead to over 200 scientific publications, two patents and numerous presentations to scientific international meetings.

Brief overview of research achievements


The research activity performed has been developed in a period of  over 33 years. The main focus of this description is on the more recent research interests which have been  active at least in the last decade; they involve different membrane separation processes  and the related topic concerning the diffusion in solid polymers.  The general approach is essentially based on continuum theories.


1.         Membrane separation processes.


The topic has been studied in particular with the main aim to develop new membrane processes and to inspect their potential applicability as separation techniques; more common membrane separation techniques have also been  studied with the aim to inspect their feasibility to non traditional applications.


1.1.      Membrane distillation (MD). The process is based on the use of microporous hydrophobic membranes, in contact with an aqueous stream, containing either salts or dissolved organics (VOC's) or gases: the hydrophobicity of the membrane material prevents the liquid from entering into the membrane pores, as long as the pressure prevailing over the aqueous solution is lower than the minimum entry pressure of the membrane material. Typical values of the minimum entry pressure for microporous membranes in polypropylene or in PTFE are in the order of 210 bar. When the pressure over the liquid  inlet stream is kept below the minimum entry pressure, the entrance of the membrane pores separates the liquid phase from a vapour phase trapped inside the pores; the membrane thus acts as a physical support for a meniscus through which a liquid vapour equilibrium is established giving rise to a separation process. The transport through the membrane takes place in the vapour phase supported inside the pores and may be due to molecular diffusion (as it occurs in most of the cases), to Knudsen diffusion (as it occurs at low pressures), or to convection (Poiseuille flow). Most frequently the actual driving force for mass transport inside the pores is a partial pressure gradient along the pore itself, but a total pressure gradient may also be obtained.

The research studies performed have shown that the operation may be conducted through  set-ups  which may appear rather different even if they are all ultimately interpreted based on the same physical phenomena; I have considered direct contact membrane distillation (DCMD), gas gap membrane distillation (GGMD), sweeping gas membrane distillation (SGMD), vacuum membrane distillation (VMD).

Indeed the permeate side of the membrane can be in contact with another aqueous liquid; this configuration is referred to as DCMD.  The driving force for separation may be sustained in different ways: i) by keeping a temperature difference across  the membrane, so that there are different  liquid vapor conditions at the menisci existing at the entrance and at the exit of the pores, respectively; the water flux thus will be directed from the warm side solution to the cold side, if the temperature difference is sufficiently large;   ii) by keeping a concentration difference across the membrane, e.g. by using in the permeate side an extractant which is not volatile and depresses the activity of water (e.g. salts, glycols, polyglycols).

The process has been studied both experimentally and theoretically and a careful model description has been also developed which accounts for the following transport resistances: mass transport from the bulk liquid to the membranes (responsible for the so called concentration polarization), heat transfer from the bulk liquids to the membranes surfaces (responsible for the so called temperature polarization), heat and mass transfer across the membrane. It has been shown that in several cases of interest the process is controlled by heat transfer within the liquid phases, while in other situations it may be controlled by mass transfer within the liquids and seldom by mass transfer within the membrane. A proper non trivial criterion based on dimensionless numbers has also been developed to compare the effects on the process rate of resistances different in nature and in physical dimensions, as heat and mass transfer resistances.

The energy efficiency of the process was also studied and related to the process conditions and membrane properties (thickness and porosity). The economical study  indicated that  the process is attractive to produce pure water from salt solutions when low temperature heat sources are available, and/or when low potentiality is required. The process is interesting for the low temperature concentration of fruit juices either by using temperature differences or by using extractants in the permeate side; the juices thus obtained may be very concentrated (up to 60 ° Brix) and still retain most of the original aroma compounds. When applied to fermentation broths, the process shows also interesting potential applications in the achievement of continuous fermentation through the removal of ethanol.

The results are reported in the publications N. 31, 39, 41, 43, 44, 48, 49, 50, 52, 57, 58, 60, 61, 66.


When the permeate side of the membrane is in contact with a stagnant layer  of inert gas which is then in contact with the liquid condensed on a refrigerated surface, we have the GGMD configuration. The physical resistances controlling the process are the same entering DCMC, with the addition of heat and mass transfer in the gas gap. Also for the present configuration studies have been performed similar to the ones outlined for the DCMD case. The main advantage of this configuration is associated to a more efficient use of energy since virtually only the latent heat is provided by the liquid  and the energy losses across the membrane are rather minor. This is beneficial in comparison to DCMD when no energy recovery is possible. A further advantage of the GGMD configuration is represented by the possibility to control the pressure of the gas gap to values below the atmospheric pressure, thus enhancing significantly the process rate.

The results are reported in the publications N. 48, 50, 52, 60, 66.


When the permeate side of the membrane is in contact with a stream of sweeping gas the configuration is referred to as SGMD . In this case another process variable  is introduced due to the possibility of varying the gas flow rate and thus the transfer resistances (for mass and heat transport) in the gas phase. The process has been modelled both for flat and for tubular membranes; the model simulations well represent the results obtained in the experimental study performed.  The process has been developed also in co-operation with Snamprogetti S.p.A., Milan, and a patent has been obtained (N. 53). The feasibility of the process to extract volatile contaminants from aqueous streams was also established; the results are reported in the publications N. 51, 53 (patent), 89, 91.


Finally, when the permeate side of the membrane is kept at pressures below the condensation pressure of the permeating gases the configuration is referred to as vacuum membrane distillation (VMD) and this was first studied and introduced in the literature by my group. This process has been thoroughly studied since it appears  interesting for its technical and  economical features. The main resistances are encountered by heat and mass transfer in the liquid feed and by mass transfer in the membrane. By using permeable membranes one usually operates at conditions in which the membrane does not offer any significant resistance and the process is controlled by the transport phenomena in the liquid phase. First theoretically, by extending the criterion to compare the effects of resistances of different dimensions already used for DCMD, and then experimentally it was demonstrated in particular that for streams containing VOC's, the flux of water is essentially controlled by heat transfer only, while the flux of the organic components are determined by both heat and mass transfer in the liquid. The use of capillary membranes is thus not recommended, in spite of the larger area per unit volume, due to the high resistances encountered in the laminar flow inside; the use of tubular membranes thus improves significantly (by a factor 5 at least) the process performance, as  model calculations indicate and  experiments confirm. The mathematical  model developed indicates also the particular role of the downstream pressure in the case of aqueous streams containing VOC's. At low pressures high fluxes are observed, as expected, with permeating vapours essentially containing water with little percent of VOC. With increasing the downstream pressure,  the total flux decreases essentially due to a decrease in the water flux while the flux of organic remains practically unaltered; this is so until the pressure value exceeds the water vapour pressure at the operating temperature. Therefore it is convenient to operate at sufficiently high pressures and not at a high vacuum; we can thus obtain a vapour phase which exceeds 70-80% wt in the organic, from a liquid feed containing 2% of VOC. This has been observed for a variety of aqueous mixtures containing various VOC's as MTBE, ethanol, n-propyl alcohol, acetone, methyl acetate. The economical analysis of the process indicates that in comparison to pervaporation VMD is more convenient  when operated in turbulent flow and is also more convenient than air or vapour stripping followed by adsorption on active carbon. The theoretical and experimental analysis has been performed to find out under what conditions the process is controlled either by heat transfer resistances or by mass transfer resistance, developing suitable dimensionless parameter to compare such different resistances. The feasibility of the process to concentrate fruit juices and musts has also been inspected.

The results are reported in the publications N. 56, 70, 78, 79, 83, 85, 89, 98, 99, 111, 115, 126, 130, 151.



1.2.      Affinity membrane separations.

This relatively recent separation technique has been studied in my lab through an initial co-operation with Professor G. Belfort (Rensslear Polytechnic Institute,  Troy, NY), for the selective separation of proteins from a protein solution. The process is an alternative to the affinity chromatography procedure typically followed. My studies started simply by considering the purification from cell lysates of a class of  fusion proteins  i.e. proteins which contain the protein of interest linked (through a linker as factor Xa or an intein) to another protein which can be easily recognised and adsorbed onto the membrane surface. For the latter the maltose binding protein (MBP) was considered. The desired fusion proteins can be produced as endocellular  proteins, by DNA modified E. coli strains. The same affinity membrane, selective for MBP, can be applied to all the different fusion proteins containing MBP as one domain. The study proved first the process feasibility by considering the  following steps: i) cell growth and harvesting; different strains codify for MBP fused with -galactosidase or with rubredoxin or with intein-CBD;  ii) recovery  of the protein from the cell lysate by using traditional techniques or also the affinity membranes produced; iii) surface modification of commercial membranes in order to obtain membranes which selectively bind to the MBP domain; iv) determination of the adsorption and desorption kinetics of the fusion proteins over the modified membranes; v)  determination of the sorption isotherm of the protein over the membrane; vi) obtain the protein desorption and study the efficiency; vii) model simulation for the process rate.

The study pointed out that cellulose membranes, modified by amylose chemically bound onto the surface, are selectively binding the fusion protein considered even from the direct centrifuged cell lysate. After a mild wash with water to detach proteins subject to nonspecific week bonds only, the elution step revealed the presence on the membrane of the fusion protein in rather pure conditions, with an extremely high selectivity. The same membranes apply equally well for all the three MBP fusion proteins inspected. A simple but effective model simulation has been proposed, which indicates that the sorption process  is highly non linear.

The affinity membranes obtained have been used for several fusion proteins containing the MBP domain, studying the sorption isotherms, sorption and desorption kinetics and capacity.

The work has been extended to affinity membranes containing selective ligands for lectins and immunoglobulins as IgG and IgM. Affinity membranes for lectins obtained from Momordica charantia seeds, and for peanuts agglutinin and Ricinus communis agglutinin were prepared and studied following the same lines indicated above.

Most recently this topic has been developed by considering affinity membranes suitable for the purification of monoclonal antibodies from the supernatant of industrial cell coltures. The different affinity membranes considered have been obtained from different preactivated supports made of regenerated cellulose or polyethersulfone. Also different ligands have been used, beyond Protein A, namely protein A mimetic synthetic ligands as A2P, provided by Prometics ltd., and D-PAM, provided by Xeptagen SpS., sequences of 6 peptides provided by professor. R.G. Carbonell of NCSU-Raleigh, NC. Different spacers between membrane surface and ligand have also been inspected, considering among others thyole and azido compound. The main target proteins considered are h-IgG and h-IgM. The membranes prepared have been characterized with pure protein solutions both in batch and dynamic mode, studying also the variation of adsorption capacity, sorption and elution kinetics versus feed flow rate and membrane lifetime. The membranes have then been used with complex mixtures as industrial supernatants of cell coltures as well as different sera. In parallel to the experimental analysis performed, the monoclonal antibody separation through affinity membranes has been simulated through a model implemented in ASPEN ustom Modeller and the model developed is appropriate to describe more than satisfactorily the breackthrough curves experimentally observed at various feed flow rates and concentrations, both for pure protein solutions as well as for supernatant of cell coltures. In general, the use of affinity membranes shows advantages since they are much faster than processes based on traditional beads, and do not show any problems associated to column packing or high pressure drops.

The results are partially reported in the publications N. 92, 93, 101, 112, 119, 122, 123 (patent), 124, 127, 129, 136, 145, 150, 152, 153, 157, 164, 166, 170, 172.



1.3.      Gas separations.


Gas separations can be obtained through polymeric membranes often in the glassy phase. The separation factor is determined by the combined effect of solubility and of diffusivity in the membrane. Both properties are usually measured from pure gases and very seldom directly from mixtures of gases, which are in fact used in practical applications. Most often the separation factor expected from given mixtures is simply estimated from pure gas data. It is vice versa known that in some cases the expectations based on pure gas data definitely fail even qualitatively, as it is the case of CH 4 / n-C 4H 10  in poly(1-trimethylsilyl-1-propyne)   (PTMSP). Such unexpected behaviours are largely due to the effects of the solubility of different components into the solid polymer. The problem of the determination of the solubility of gases and vapours in glassy polymers has been considered, both experimentally and theoretically; it  must be noticed that in a glassy phase the traditional  true thermodynamic equilibrium conditions  do not apply since the glass is a nonequilibrium phase.

Experimental results have been obtained for the solubility, or rather pseudo-solubility, in glassy PTMSP of different alcohols and n-alkanes, under different conditions. It has been shown that the classical dual mode model does not apply even qualitatively in order to describe the solubility of alcohols since the isotherm is s-shaped with a very small solubility coefficient at low pressures. It has been shown, however, that when the isotherms are represented in terms of  penetrant chemical potential, in place of penetrant pressure, versus penetrants mass fraction, the behaviour observed for the n-alkanes is qualitatively similar to that observed for the alcohols. Based on this hint, and on a lattice fluid theory, a model has been developed for the Gibbs free energy of a nonequilibrium glassy mixture (NELF model). That model  rests upon the use of the polymer partial density as a proper measure of the out-of-equilibrium frozen into the glass; that value is governed by the bulk rheology of the glassy membrane. The thermodynamic model developed in publications N. 96 and 102, proves to be entirely predictive; it is simply based on the volumetric properties of the pure polymers and the pure penetrants, and is in excellent good agreement  with the experimental data available for binary mixtures.

The model has been also extended to the case of mixtures of penetrants and of  polymers and is able to quantitatively explain the observed behaviour for the case of CO 2/C 2H 4 mixtures in glassy PMMA (publ. N. 104). The model is also able to account for the different isotherms observed in PTMSP by alcohols and n-alkanes. The success of this tool motivated the invitation received to deliver the  lecture  presented at the 1997 Gordon Research Conference on Membranes (Andover, NH, Aug 3-8, 1997).

The approach has also been extended to swelling penetrants for which a simple correlation method has been developed based virtually on a single solubility data point.

The model represents an important and promising tool to predict the expected separation behaviour from mixtures of gases and, on the other side, to select the best glassy polymer for given gas separations. It was also shown that the general non-equilibrium thermodynamic analysis allows to extend this approach for the solubility in glassy polymers to different models other than the lattice fluid model initially considered. Indeed application of the method also to SAFT and PHSC models, suitably modified for the non-equilibrium approach, proved equally satisfactory. A rather broad series of polymer penetrants pairs and  of polymeric blends has been used for the comparison with model calculations.

Most recently it has been shown that the model accounts also for the experimentally known, but as yet unexplained, dependence of the infinite dilution solubility coefficient in glassy polymers on penetrant critical temperature. In addition, application of the model to composite  matrices formed by a high free volume glassy polymer and solid nanoparticles has shown that the model can be used to predictin a very reliable way the solubility isotherms of several different solutes once the solubility isotherm of one single test vapor is known. The fractional free volume of the polymer matrix, which is calculated from the model also allows to calculate the proper dependence of penetrant diffusivity on the filler loading in the mixed matrix, and to describe the unexpected increase of solubility, diffusivity and permeability which is experimentally in polymers loaded with fumed silica nanoparticles. 

The permeability of gaseous penetrants in glassy polymeric membranes is presently under investigation, by developing a rigorous theoretical model able to represent well the permeability dependence on upstream pressure, including also the non-monotonous trend commonly known as "plasticization effect". The model proves also suitable for the development of a predictive procedure for the premeability of gases in glassy matrices. Particularly interesting is also the possibility to describe well the permeability of gas mixtures in glassy polymeric membranes.



1.4.      Metal membranes for hydrogen separation.


In recent years, thanks also to a relevant financial support, I have started a project focused on the study of membranes and membrane modules for the separation of hydrogen from the streams obtained from reforming reactors, at operating conditions around 400 °C to 600 °C, at pressures below 15 bars. The experimental equipment was designed anew, setup and thoroughly tested for safety. Different metal membranes have been tested based on Pd/Ag alloys, which were made available by collaborating institutions and also produced in my lab via electroless plating. The membrane properties were tested by varying the pretreatments, temperature of operation and feed gas composition. Remarkably, for some membranes which are endowed with very interesting selectivity, a relevant effect of concentration polarization in the gas phase was observed and documented in detail. The CFD simulation of the membrane module, which has been performed in parallel, helped obtaining proper modifications of the membrane module with better fluid-dynamics and overall membrane behavior. At the same time the same simulation also confirmed the presence of gradients of hydrogen concentration in the gas phase when the membrane permeability used is that of the membrane for which the experiments indicated hydrogen fluxes sensitive to the feed flow rate. The activity so far lasted less than three years and publication of data was possible only for a part of the results obtained (see ref. 173)


1.5.      Reverse osmosis.


Reverse osmosis is a rather well established operation for desalination purposes. However its potentials in the treatment of process streams deriving from chemical processes have not yet been investigated in sufficient detail. The increasing concern for water reuse and the stringent requirements for water discharge make this application interesting. The problem has been studied at the pilot plant level, by considering two different process waters, one deriving from a Montedipe plant in Mantua (IT), the other deriving from a process stream from a Ciba-Geigy plant in Pontecchio (Bologna, IT). In both cases the aqueous streams contained also a complex mixture of organic components with a broad molecular weight spectrum. The feasibility of the process has been established and the design variables optimised. The work has lead to the plant design and economical evaluations. The results have been partially published in ref. N. 67 and partially are held as proprietary results.


1.6.      Pervaporation.


The pervaporation (PV) process has been studied by considering water/ethanol mixtures and also by examining the separation of organic/organic mixtures. In the first case the applicability of PTMSP membranes was considered, in view of the high permeability of that membrane versus the alcohol. The main reference application is the  set-up of a continuous fermentation with a continuous removal of ethanol from the broth. The effects of ethanol concentration were inspected, as well as the ageing effects observes in the membrane when contacted with liquid feeds. The experimental result indicated that there is an appreciable influence of the membrane thickness, which cannot be scaled  according to the rules governing Fickian diffusion (ref. 69, 80).

The potential application of PV to the separation of organic/organic mixtures  has been examined by using a modified PPO membrane containing hydrohyl groups, (Eniricerche), in order to render it suitable for the separation of MTBE from methanol. The process gave rise to interesting separation factors ranging from about 7 to over 20 depending upon the concentration range of the feed mixture. It has been observed that the flux of MBTE is non monotonous with increasing the MTBE content of the mixture but undergoes a minimum. The flux of methanol, vice-versa, was found  to be monotonous and linearly decreasing with increasing the MTBE mass fraction in the feed. The permeability of either component was calculated by extending to the case of PV the permeability concept introduced for gas separations, based on the thermodynamic properties of MTBE/methanol mixtures. The values and the trends observed were consistent with the fact that methanol is a swelling agent for the modified PPO membrane (ref. 88).



2.         Diffusion in polymers.


This subject is partially related to the previous one, for the cases in which the determination of the diffusion coefficient is important to calculate the permeability and/or the separation factors. In several other  cases the main motivation is different and resides in specific applications as packaging, coating, desolventisation of polymers, stress cracking of polymer products, drug delivery systems.


2.1.      Fickian diffusion and chemical post-treatment.


The experimental determination of the diffusion coefficient of various penetrants in PTMSP films was obtained through a specific transient gravimetric technique especially designed to be compatible also with the fast sorption rates observed for some penetrants in that matrix. A series of alkyl alcohols and of alkanes was considered which show a very interesting unusual behaviour. The n-alkanes presented  a concentration dependance for the diffusion coefficient endowed with a rather flat maximum; vice versa, with increasing the penetrant concentration, the diffusivity of alcohols showed initially a rather pronounced minimum followed by a subsequent maximum. It has been shown that such different behaviours in the same matrix are associated to the thermodynamic factor for diffusion (i.e. the partial derivative of the penetrant chemical potential with respect to the log of the penetrant mass fraction). The latter quantity was experimentally determined through the solubility isotherm which has a rather different trend for alkanes and for alcohols. The mobility coefficient, i.e. the proportionality constant between  the mass flux and the chemical potential  gradient, can thus be calculated as  the ratio between diffusivity and the thermodynamic factor. Remarkably, the mobility coefficient shows indeed exactly the same monotonous and parallel trend in all the cases inspected, thus indicating that the major differences observed for the diffusion coefficients are due to the thermodynamic factor alone. The above behaviour was also shown to be consistent with the high free volume present in the PTMSP matrix and, in addition, indicated a simple way to correlate if not to predict the diffusion coefficient of different penetrants in PTMSP. The results have been partially published in refs. N. 95, 97,103, 105.

The effects of diffusion processes in the thermal post-treatment of poly(butylen terephthalate) (post polycondensation) were also studied both experimentally and theoretically. The major chemical reactions in thermal post-polycondensation were considered and the effect of the diffusion towards the external surfaces of the low molecular species involved (1,4-butanediol, terephthalic acid and water) were studied. In comparison with the experimental data obtained, the model was thus able to account for the increase in the polymer molecular weight versus post-treatment time at any position inside the polymer matrix. Thus it gave a correct guidance for the post-polycondensation treatment of PBT products. The results have been  published in ref. N. 32, 35, 46.

2.2.      Non-Fickian diffusion.


The different non-Fickian mass transport behaviours observed in solid polymers have been extensively studied in the past by G.C. Sarti, both theoretically and experimentally. From an experimental point of view the mass transport kinetics of liquid n-alkanes in glassy polystyrene has been studied at different temperatures; the mechanical properties of the samples were well characterised with regard to the critical stress for craze growth and propagation. In all the cases inspected the presence of a discontinuity front was always observed, marking the separation between penetrated and unpenetrated glassy matrix. The front position changes in time according to a power law (t n) with an exponent n varying between 0.5 (Fickian like )  to unity (Case II) behaviour. In the high temperature range, still below the glass transition temperature, an anomalous behaviour was observed very close to Fickian like behaviour; as the temperature decreases the exponent n increases towards the Case II behaviour. The mass transport process results from the combined effects of the diffusion resistance encountered across the penetrated region, and the swelling resistance offered by the matrix around the swelling front. The latter factor would give rise to a constant penetration rate per se (i.e. n=1), while the diffusive resistance per se would give rise to an exponent equal to 0.5. The kinetics of the advancing front was shown to be nicely correlated to the craze propagation kinetics. Analogous studies have been performed in PMMA by using methanol, ethanol and propanol as the penetrants. By considering planar geometries, both weight uptake and swelling measurements were performed, monitoring the position of the internal swelling front (separating the penetrated from the unpenetrated region) as well as of the external sample surfaces. The fast swelling agent, methanol, gives rise to an anomalous diffusion behaviour very close to a Fickian like response; in this case the process is controlled almost completely by the diffusion resistance. For  ethanol and propanol, viceversa, the swelling resistance is controlling, for the usual sample thickness considered (around 1mm), thus giving rise to a CaseII kinetics i.e. to a constant sorption rate.

In all cases the existance of an internal moving front indicates the presence of a shock concentration wave travelling towards the sample midplane, at a velocity decreasing in time for the case of methanol and at constant velocity for the higher alcohols.

The effect of the pre-elongation treatments was also studied by prestretching the samples to different elongation values and then freezing in the elongation prior to the sorption process. Higher pre-elongations result in higher sorption rates, as otherwise expected; moreover, with increasing the elongation frozen in the glassy matrix, a higher importance of the swelling resistance was also observed giving rise to behaviors closer to CaseII the higher the elongation ratio considered.

These results have been discussed in publications N. 18, 22, 26, 34, 40, 87.


The mentioned different non-Fickian behaviours have been modelled according to different model approximations which are hereafter schematically listed and discussed.


Latest news

At the moment no news are available.