Mechanisms in the Solvent Cleaning of Emulsion ... - ACS Publications

Solvent-based cleaning of emulsion polymerization reactors. J.Y.M. Chew , S.J. Tonneijk , W.R. Paterson , D.I. Wilson. Chemical Engineering Journal 20...
0 downloads 0 Views 424KB Size
Ind. Eng. Chem. Res. 2005, 44, 4605-4616

4605

Mechanisms in the Solvent Cleaning of Emulsion Polymerization Reactor Surfaces John Y. M. Chew,† Sjef J. Tonneijk,‡ William R. Paterson,*,† and D. Ian Wilson† Department of Chemical Engineering, University of Cambridge, New Museums Site, Pembroke Street, CB2 3RA Cambridge, U.K., and NeoResins, Sluisweg 12, P. O. Box 123, 5140 AC Waalwijk, The Netherlands

A new technique, fluid dynamic gauging (FDG), is introduced for studying the swelling and removal of polymer films from various surfaces. Films of two different polystyrene copolymers, labeled PX (30 kDA) and PA (125 kDa), were deposited in the laboratory on, inter alia, SS 316 plates. FDG was used to track, in situ and in real time, the films’ evolving thickness during swelling and removal in three agents: NaOH, MEK, and a commercial aqueous cleaning solution, TPU. The effects of initial film thickness, solvent concentration, solvent temperature, nature of the substrate, and applied stress were investigated. Swelling profiles suggest that Type II diffusion is the governing transport process. Different modes of behavior were found. In NaOH, swelling occurred after a lengthy induction period, but no surface cleaning was detected. MEK and TPU swelled the films with a negligible induction period and then cleaned the surfaces, with the former working faster. For MEK applied to PX and PA, and for TPU with PX, the film was removed by dissolution, i.e., the cohesion of the film was attacked. By contrast, TPU applied to PA removed the film by destroying the adhesion. Application of TPU to films composed of mixed layers of PX and PA showed that the deposition sequence determines removal behavior. A simple model allows prediction of the removal kinetics of composite films from those of singlecomponent films. The potential for substantially accelerated reactor cleaning by appropriate reactor scheduling is thereby identified. 1. Introduction 1.1. Fouling of Polymerization Reactors. Emulsion polymerization is widely used commercially for the production of latexes, paints, coatings, rubbers, and adhesives.1 In general, emulsion polymerization is a free radical polymerization process in a multiphase environment, yielding a dispersion of submicron polymer particles. This process uses the following ingredients: monomer(s), dispersing medium (usually water), surfactant (also referred to as emulsifier), water-soluble initiator, and possibly a chain transfer agent. Polymer fouling in reactors is caused by the formation of coagulum, i.e., polymer recovered in a form other than that of a stable product, during polymerization.1 The formation of coagulum is observed in all types of emulsion polymers, including coating latexes, synthetic rubbers, and specialty latexes. Polymer reactor fouling can be characterized into three main types: (i) fouling of reactor walls, roof, bottom, agitator, baffles, and cooling coil; (ii) suspended solids formed during polymerization and recovered from the product afterward by filtration or sedimentation; and (iii) fouling occurring after polymerization during storage or transportation. Fouling arises in all sizes of reactors and, for large-scale production, can lead to reactor downtime, unpredictable lengthening of cycle times, and reduced yield. The main causes of fouling in these systems are loss of colloidal stability of the polymer particles during or after polymerization, which may cause flocculation, and polymerization of the monomer by other mechanisms.1 The colloidal stability of polymer particles is governed * To whom correspondence may be addressed. † University of Cambridge. ‡ NeoResins.

by the combination of potentials which give rise to a barrier to aggregation: attractive van der Waals’ forces, electrostatic repulsion, and steric interactions which can be modified by ionic surfactants. Coagulation occurs when the kinetic energy of the polymer particles exceeds the potential energy barrier. Polymerization of a monomer by a different mechanism may occur, for example, within the monomer droplets, resulting in a larger particle size than desired.2 These larger particles are usually less stable and coagulate readily, especially in large reactors where agitation may not be uniform. These coagula may deposit on the walls and roof of the reactor and serve as nuclei for further polymerization and coagulation.1 Polymerization on reactor walls can be related to the smoothness of the wall surface; surface roughness promotes surface polymerization and coagulation. In this respect, highly polished stainless steel surfaces are often preferred. 1.2. Cleaning of Polymerization Reactors. Several approaches have been adopted to eliminate or reduce polymer fouling such as: addition of emulsifier/stabilizer, variation of agitation rate, alteration of recipe and reactor operating conditions, and use of polished surfaces.1 Remedies developed for one system may not be directly applicable to another system, so each case must be considered separately. Polymer foulants can be cleaned by chemical or mechanical methods or a combination of both, for instance, by soaking in a solvent to convert the deposit to a more removable form (e.g., a swollen gel) followed by a high-pressure jetting to remove any remaining residues. The related topic of polymer dissolution has received attention because of its importance in applications such as cleaning of polymer layers, microlithography, controlled drug release, and recycling of polymers.3 Dis-

10.1021/ie050105g CCC: $30.25 © 2005 American Chemical Society Published on Web 05/27/2005

4606

Ind. Eng. Chem. Res., Vol. 44, No. 13, 2005

solution of polymers differs noticeably from that of nonpolymeric materials: the latter can often be explained by simple diffusion laws and a unique diffusion rate.4 In contrast, the diffusion of solvent molecules into a polymer seldom proceeds via Fickian diffusion5 but frequently follows a Type II transport mechanism, whereas Fickian diffusion is an example of Type I.6 Pekcan and co-workers7-9 described the dissolution of glassy polymers via three steps. The first is the ingress of solvent molecules into the glassy matrix by diffusion. The solvent molecules fill the spaces present in the polymer in the glassy state as free volume. The absorbed solvent subsequently induces morphological and structural changes so that the glassy region becomes rubbery. The glassy (negligible solvent) and rubbery regions (a swollen gel where the solvent concentration is high) are separated by a sharp penetrant front. The presence of this advancing front is one of the main characteristics of Type II diffusion. Ueberreiter10 summarized the structure of the surface layers from the pure polymer to the pure solvent. In the third step, the polymer chains in the rubbery region adjacent to the polymer-solvent interface have high internal mobility and these chains will tend to disengage from the interface and migrate into the bulk solvent. Penetration of solvent molecules into an initially glassy polymer matrix can be divided into two main types:6 the penetrant may be a solvent or a nonsolvent, depending on its polarity and the solubility properties of the polymer-penetrant pair. In the case of solvents, the polymer-solvent interactions are stronger than the polymer-polymer attraction forces, so that the chain segments start to absorb solvent molecules, increasing the volume of the polymer matrix as the segments unwind from their entangled structure. For nonsolvents, Stamatialis6 reported that only penetration and swelling occur, yielding a uniform swollen layer at the final equilibrium. Fundamental studies of the kinetics of polymer swelling and/or dissolution have been reported in many applications.11 Experimental techniques employed to characterize polymer dissolution behavior include gravimetry, differential refractometry, optical microscopy, laser interferometry, ellipsometry, fluorescence, nuclear magnetic resonance, and Fourier transform infrared spectroscopy. Polymer dissolution can be affected by various parameters, including solvent temperature, polymer molecular weight, and solvent activity (thermodynamic and kinetic compatibility). In this work, fluid dynamic gauging (FDG) is employed to study the dissolution of polymer fouling layers. Specifically, surface films composed of two polystyrene (PS) copolymers are treated with three candidate solvents in situ and in real time, simulating the cleaning process of polymerization reactors. FDG is a technique for the measurement of soft solid deposit layers on solid surfaces immersed in a liquid environment.12 In this technique, a precision-machined nozzle on the end of a tube section is positioned close to the surface of the deposit layer as shown in Figure 1. The whole system is immersed in liquid and suction is applied so that liquid is drawn from the surroundings into the nozzle as shown by the dotted streamlines. For a given extent of suction, the mass flow rate of liquid gives a direct measurement of the distance between the nozzle and the gauged surface, h. The position of the

Figure 1. Construction of nozzle for fluid dynamic gauging.

nozzle relative to the solid surface is separately measured and, thus, gives direct estimates of layer thickness. A detailed description of the technique and its fluid mechanics is given in refs 12 and 13. Polymer swelling and removal can be monitored swiftly, reliably, and cheaply using FDG; it is employed here to elucidate cleaning mechanisms and solvent effects. Such knowledge can enable cleaning processes to be selected and optimized on a reliable basis. 2. Experimental Section 2.1. Apparatus. Quantitative data, i.e., thickness profiles describing the removal behavior of polymer coatings from stainless steel plates during contact with cleaning agents, were generated using FDG in the quasi-stagnant configuration (where the only fluid movement is the gauging flow). Three agents were studied: aqueous sodium hydroxide (NaOH; Fisher, U.K.), methyl ethyl ketone (MEK; Fisher, U.K.), and TPU, a commercial cleaning agent (Transol Production Unit b.v., The Netherlands). The apparatus used for MEK and TPU experiments is shown in Figure 2a. The unit is a jacketed vessel constructed from borosilicate glass and is covered with a brass lid. The heating fluid used was water. The temperature of the process fluid at two different positions within the vessel was recorded using two K-type thermocouples (one shown). The gauging nozzle is detailed in Figure 1 with the following dimensions: d ) 4.0 mm, dt ) 1.0 mm, w ) 0.5 mm, λ ) 0.1 mm, and R ) 45°. The hooks fitted onto the brass support, shown in Figure 2b, were designed so that the coated plates could be moved in/ out of the vessel easily. At the end of each experiment, the process fluid was drained through the discharge valve. The process fluid is drawn through the nozzle (A) by a siphon effect, and the discharge mass flow rate (B) is measured by a balance (C). The mass increments are then recorded using a computer (D). During an experiment, the hydrostatic head, s, of the tank is kept constant by regulating the fluid inlet (E). Experiments using NaOH were performed using identical gauge and measurement devices in another unit featuring an open holding tank constructed from Perspex, described in ref 12, as this solvent did not present major volatility or skin contact hazards. In this case, films were deposited on long steel plates (150 mm × 30 mm × 1 mm thick) rather than disks. The error in measuring the mass increment of the discharge fluid was Tg) (T < Tg) (T < Tg) (T > Tg) (T > Tg)

PA

NaOH 35 ( 3 1300 ( 100 25 ( 3 330 ( 30

21 ( 2 9(2 12 ( 1 2 ( 0.2

MEK 10 ( 3 6.0 × 102 ( 0.5 × 102 24 ( 4 2.0 × 105 ( 0.5 × 105 15 ( 3 5.0 × 103 ( 0.5 × 103 22 ( 4 1.0 × 105 ( 0.5 × 105

16 ( 3 9.0 × 103 ( 0.5 × 103 25 ( 4 3.0 × 105 ( 0.5 × 105 14 ( 3 2.0 × 103 ( 0.5 × 103 16 ( 3 6.0 × 103 ( 0.5 × 103

TPU 29 ( 4 24 ( 4 1.0 × 105 ( 0.5 × 105 2.0 × 104 ( 0.5 × 104 27 ( 4 7.0 × 104 ( 0.5 × 104

Em,swell (kJ/mol) Am,swell (µm/min) Ediss (kJ/mol) Adiss (µm/min)

ergy was extracted from a second Arrhenius model, namely

1 tNaOH,ind

( )

) Aind exp -

Eind RgT

(3)

The values in Table 5 show evident differences in the temperature dependency for the two materials, with the two activation energies (Em,swell and Eind) differing by ∼12 kJ/mol. The lower molecular weight PX also exhibits consistently shorter induction and swelling times. It is noteworthy that the activation energy values in the table lie between 2kT and 20kT (corresponding to 6.7 and 67 kJ/mol, respectively, at 300 K), reported as typical temperature dependencies for friction processes in entangled polymers by Rubenstein and Colby.15 The max data in Tables 3 and 4 indicate that both films swelled to give max > 2, with PA swelling notice-

4610

Ind. Eng. Chem. Res., Vol. 44, No. 13, 2005

Figure 8. Swelling profile of PX in MEK, showing constructions giving key parameters; conditions: TMEK ) 40 °C, δ0 ) 0.10 mm.

Figure 7. Removal profiles of (a) PX and (b) PA films in MEK; conditions: δ0 ) 0.10 mm.

ably more than PX, and a linearly increasing trend with temperature. These data sets were also analyzed using the timetemperature superposition approach for polymers, which leads to temperature dependency for viscosity (and, by analogy, other mobility processes) described by the Williams-Lander-Ferry (WLF) or Volger-Fulcher correlation (see ref 15). This approach, which relates viscosity to the free volume of the polymer and its glass transition temperature, did not give satisfactory or physically realistic agreement for several of the data sets. For instance, the regression coefficient, R2, for fitting the induction period data to eq 3 was > 0.99 whereas WLF did not give a good correlation, particularly for PA where the polymer swelled at a temperature below its Tg. Furthermore, the max data can be interpreted as the free volume of the swollen polymer gel, and the near-linear dependency on temperature described above suggested a Vogel temperature for these gels well below 0 °C, in which case the WLF description at the temperatures studied would be expected to approach the Arrhenius form. The Arrhenius description is, therefore, used as a useful quantitative description: the interpretation of these temperature dependencies is the subject of ongoing work. 3.2. MEK. 3.2.1. Swelling/Dissolution Profiles. Figure 7 shows polymer film thickness profiles for both PX and PA at different temperatures. MEK is clearly a cleaning solvent because the layer is dissolved completely. Somewhat different characteristic stages are evident, namely: Stage Isno induction period is observed; Stage IIsswelling to an approximately steady thickness; Stage IIIsa plateau stage, leading to Stage IV; Stage IVsa rapid decay/dissolution stage.

Stages II-IV are similar to those described by Pekcan and co-workers7 for PS, poly(methyl methacrylate) (PMMA), and other polymers treated with organic solvents at various temperatures. The steady film thickness in the plateau stage indicates either that the swelling and dissolution processes are occurring simultaneously at comparable rates or that the swelling has stopped and dissolution has yet to begin. In the third stage, the molecules in the rubbery layer adjacent to the polymer-solvent interface have high internal mobility and dissolve into the solvent, and dissolution is the dominant process. The figure shows that the overall cleaning times for the lower molecular weight species, PX, are smaller than those for PA. One possible explanation for this is that, for PA, where polymer chains are long and mutually entangled, more time is needed for the solvent to penetrate into the glassy polymer and the polymer molecules will also be inhibited from moving into the liquid phase due to dynamic friction between the chains. For example, in the case of PX dissolution at 60 °C in Figure 7a, the swelling (Stage II) and the plateau stage (Stage III) lasted less than 120 s, and hence, only the dissolution stage (Stage IV) was recorded during the gauging experiment; whereas for PA (Figure 7b), all three stages could be observed. Moreover, for all temperatures, the total cleaning time for PX is less than that for PA. Figure 8 indicates how quantitative parameters such as the swelling time, tMEK,swell, the plateau time, tMEK,plateau, and the dissolution time, tMEK,diss, have been generated from individual swelling profiles; tMEK,total is the overall time required for the polymer to dissolve completely. Figure 8 also shows how the swelling and dissolution rates are extracted from the data. Duplicated experiments again indicated that reproducibility was very good. 3.2.2. Effect of Stresses Imposed by Gauging Flows. Previous studies on denatured whey proteins,17 tomato paste,18 and soft calcium sulfate deposits19 showed that the shear stress exerted on the gauged surface can be sufficient to disrupt the surface, whereby the materials were sheared off from the stainless steel plates. It is useful to quantify the stresses acting on the polymer films, especially during the plateau and dissolution stages (III and IV), when the films were softer and more susceptible to deformation by the gauging flow. Also, at higher temperature the swollen films were expected to be softer due to greater chain mobility. The shear stresses exerted by the gauging flows in this study were calculated using the computational simulations

Ind. Eng. Chem. Res., Vol. 44, No. 13, 2005 4611

Figure 9. Effect of stresses imposed by the gauging flows on PX in MEK; conditions: TMEK ) 18 °C, δ0 ) 0.10 mm.

Figure 11. Effect of temperature on mean dissolution rate in MEK and TPU; conditions: δ0 ) 0.10 mm. Triangles indicate PX in MEK, circles indicate PA in MEK, and squares indicate PX in TPU.

Figure 10. Effect of initial dry film thickness on swelling time and dissolution time in MEK; conditions: TMEK ) 18 °C, δ0 ) 0.10 mm.

described by Chew et al.13 to lie between 6 and 35 Pa, which correspond to distances between the nozzle tip and the film of 0.25 mm and 0.08 mm, respectively. Figure 9 shows two cleaning profiles for PX performed under identical conditions at 40 °C. During Run 2, the gauge was moved away from the film (the distance between the nozzle tip and the film was >15 mm) at two intervals so that the shear stress exerted by the gauging flow was negligible. No readings were recorded during these two intervals. When the gauge was brought back into operation, the thickness measurements were recorded again, and there is very good agreement with Run 1. This indicates that, during normal operation at 40 °C, the shear stress exerted by the gauge did not affect polymer dissolution, so that dissolution is governed here by internal mechanisms. Similar behavior was observed for PA at 60 °C. Distortion of the films was only evident when the gauge was positioned very close to the fully swollen layer, indicating that FDG can be used to study the rheology of such layers, as outlined for other systems by Chew et al.18 3.2.3. Effect of Initial Film Thickness. Multilayered films were prepared and studied at solvent temperatures of 18 and 60 °C, allowing a comparison of behavior above and below the Tg. The effect of thickness was common to both materials: the mean rate of swelling and the rate of dissolution were insensitive to the initial thickness of the film regardless of T. Figure 10 shows that the swelling and dissolution times for both materials at 18 °C increase linearly with δ0; similar trends were observed at 60 °C. Furthermore, the lines extrapolate to pass through, or close to, the origin, implying that no extra resistance to swelling and

dissolution has been introduced by constructing the films from multiple layers. As with the results seen in the NaOH observations, max was insensitive to the initial dry film thickness and Type II diffusion clearly controls. 3.2.4. Effect of Temperature. The swelling rates for each polymer in MEK in Figure 6 are markedly higher than in the aqueous solvents, and both polymers exhibit a sharp step, more evident than that in NaOH, at temperatures near their respective Tg. RMEK,swell increases with temperature in an Arrhenius fashion, and the apparent activation energies are listed in Table 5. It is notable that the swelling rates of the two materials exhibit different temperature dependencies at T < Tg: the swelling rate of PA increases more rapidly than that of PX. For temperatures above Tg, however, the temperature sensitivities are similar. Figure 11 shows that RMEK,diss increases with temperature for both materials. There is again a sharp step near Tg; it should be noted that Tg will alter (decrease) once the polymer has swelled. The Arrhenius parameters for RMEK,diss were calculated using eq 4 and are listed in Table 5.

(

RMEK,diss ) Adiss exp -

)

Ediss RgT

(4)

The activation energies are calculated for each side of Tg in both cases in order to avoid errors introduced by trying to fit the equation across an evident discontinuity in the data. The rates of dissolution of PX and PA show similar temperature dependencies below Tg. The temperature sensitivity is, however, greater for PX above Tg. Furthermore, whereas Em,swell and Em,diss are similar for PX above Tg and for both polymers below Tg, these differ for PA above Tg. Figure 12a shows that max is consistently tTPU,swell, which means that the polymer dissolution stage occupies a larger fraction of the total cleaning time. Again, we see that the swelling time for the higher molecular weight polymer, PA, is slightly longer than that for PX, indicating greater chain entanglement in the former. This figure also shows that the tTPU,swell values are similar. 3.4. Comparison of NaOH, MEK, and TPU. The three potential solvents, paired with the two copolymers, exhibited three different types of behavior. The penetration of solvent from the bulk liquid into an initially dry polymer film can be divided into two main types: the penetrant may be a solvent or a nonsolvent, depending on the thermodynamic properties of the polymerpenetrant pair. In the case of nonsolvents, only penetration and swelling occur. With NaOH, the polymer films were not removed from the stainless steel plates following complete swelling, and this shows that NaOH is a nonsolvent. This nondissolving behavior is the first of the three types of behavior observed. With solvents, the polymer matrix forms a swollen layer, and this gel-like layer will eventually dissolve in the liquid solvent phase. The results above show that MEK is a solvent for both PX and PA. TPU behaves in the same manner, albeit more slowly, for PX. This is the second type of behavior. Application of TPU to PA results in the third behavior. A swelling stage was detected, followed by the breakage of its adhesive attachment to the substrate. The similarity of the swelling profiles of both PX and

4614

Ind. Eng. Chem. Res., Vol. 44, No. 13, 2005

PA in aqueous NaOH and sodium metasilicate give strong indications that swelling in TPU is caused by its alkalinity. The base opens the structure of the polymer matrix, thus allowing the surfactant to penetrate. For PA, the additives present then break the adhesive bonding to the substrate (demonstrated for SS 316, copper, and PA film); for PX, the surfactant breaks cohesive bonds within the polymer. Tables 3 and 4 summarize the rate of swelling, rate of dissolution, and maximum extent of swelling of PX and PA for the three solvents at selected temperatures. It should be noted that the range of shear stresses, i.e., 6-35 Pa, exerted by the gauging flows did not interfere with the swelling and dissolution processes (Figure 9). NaOH cleaning shows a rate and extent of swelling that increases with temperature. The maximum extent of swelling in NaOH is also greater than those in MEK and TPU. Note also that no swelling was observed for NaOH treatment at temperatures below 40 °C (