Application of the Monomer Reactivity Ratios to the KineticModel Discrimination and the Solvent-Effect Determination for the Styrene/Acrylonitrile

Monomer System
ANDRZEJ KAIM Faculty of Chemistry, University of Warsaw, ul. Pasteura 1, 02-093 Warsaw, Poland

Received 10 March 1999; accepted 6 December 1999

ABSTRACT: The impact of reactivity ratios determined with the Nelder and Mead simplex method on the kinetic-model discrimination and the solvent-effect determination for the styrene/acrylonitrile monomer system was investigated. For the monomer system, the penultimate unit effect was inversely proportional to the polarity of the solvent: acetonitrile N,N-dimethylformamide methyl ethyl ketone toluene. Quantitatively, the penultimate unit effect could be correlated with an absolute value of the difference between the standard deviation of the reactivity ratios determined for the terminal and penultimate models. By application of the F test, the penultimate model was justied for copolymerization in toluene. The conclusion was less certain for polymerization in methyl ethyl ketone. With a scanning procedure based on the simplex method, it was found that an equivalent representation of the copolymer-composition data could be achieved with multiple sets of penultimate-model reactivity ratios. However, the relationship between the triad-sequence distribution and copolymer composition depended on the reactivity-ratio set chosen for the microstructure determination. The microstructure calculated with the penultimate-model reactivity ratios determined with the simplex method from the initial guess (r11 r1, r21 1/r2, r22 r2, r12 1/r1) did not obey the general bootstrap effect rule. This observation still requires some theoretical interpretation. 2000 John Wiley & Sons, Inc. J Polym Sci A: Polym Chem 38: 846 854, 2000

Keywords: styrene; acrylonitrile; radical copolymerization; reactivity ratios; solvent effects; microstructure

INTRODUCTION
The choice of the proper kinetic model for describing the composition and microstructure of copolymers produced by free-radical copolymerization has been the subject of many investigations. For most monomer systems, the terminal1 model succeeds in describing copolymer composition in terms of the monomer-feed composition (eq 1). For

example, for the styrene/methyl methacrylate (STY-MMA) monomer system, both the terminal and penultimate2 models (eq 2) present almost the same perfect t to the composition data:3,4 d M 1 M 1 M 1 r 1 M 1 M 2 d M 2 M 2 M 2 r 2 M 2 M 1

(1)

Correspondence to: A. Kaim (E-mail: akaim@alfa.chem. uw.edu.pl)

Journal of Polymer Science: Part A: Polymer Chemistry, Vol. 38, 846 854 (2000) 2000 John Wiley & Sons, Inc.

M 1 r 11 M 1 M 2 M 2 r 21 M 1 M 2 d M 1 n d M 2 M 2 r 22 M 2 M 1 1 r 12 M 1 r 12 M 2 M 1 1 r 21

(2)

846

APPLICATION OF THE MONOMER REACTIVITY RATIOS

847

where n represents the monomer ratio in the copolymer and [M1] and [M2] are the mole fractions of monomer M1 and M2, respectively. Parameters r1, r2, r11, r22, r21, and r12, dened in a conventional way, stand for the reactivity ratios of the monomers M1 and M2 in the terminal and penultimate models, respectively. Systems have been reported, however, in which the introduction of the four-parameter penultimate model (also called the explicit penultimate model2) gives an improved t to the composition data over the two-parameter (terminal) model. For example, the styrene/acrylonitrile (STY-AN) monomer system is better described57 with the penultimate model than with the terminal model. However, an examination of the rate constant of propagation and termination in the copolymerization of numerous monomer systems, including the STY-MMA monomer system, as a function of monomer composition in a feed results in the conclusion that the terminal model fails to describe the absolute rate of the propagation for the system.8 Despite this, the model still is used often for describing the composition and microstructure of most copolymers produced by free-radical copolymerization. As summarized recently,9 to justify the continued use of the terminal-model composition and triad/pentad fraction equations in such systems, the implicit penultimate model8 has been proposed. It is well-recognized that the perfect t of one or another kinetic model can be considered when it provides simultaneously a satisfactory t to the composition and the overall propagation-rate data.9 However, a question can be raised: how powerful are composition data in discriminating between the terminal and penultimate model for binary copolymerization? An interesting study on the subject was dome some time ago by Moad et al.10 Some new aspects of the present state of the art justify in my opinion a further exploration of the problem. One of them is the successful application of the nonlinear least-squares (nlls) approach based on the Nelder and Mead11 simplex method to t both the terminal and the penultimate models to the copolymer-composition data. This method enables to estimate reactivity ratios and related standard deviations with optional precision. With a scanning procedure based on the simplex method, it has been demonstrated4,12 that the equivalent representation of the copolymer composition in terms of the monomer-feed

composition can be achieved with multiple penultimate-model reactivity ratios. These multiple penultimate-model reactivity ratios belong to two different sets of parameters, Set I and Set II, resulting from two different initial guesses for r11, r22, r21, and r12 (Guess I: r11 r21 r1, r22 r12 r22; Guess II: r11 r1, r21 1/r2, r22 r2, r12 1/r1) used for the nlls approach. Moreover, the most accurate reactivity ratios were surrounded with sets of reactivity-ratio values of equal accuracy. In other words, multiple sets of penultimate-model reactivity ratios can describe the composition data with the same minimal standard deviation (). Similar results were obtained for both the STY-MMA and STY-AN monomer systems.3,4,12 Contrary to the STY-MMA system, for which the penultimate model did not yield any signicant improvement in the t to experimental data when compared to the terminal model, implications of the nonuniqueness in determining the penultimate-model reactivity ratios can be quite different for the system displaying a pronounced penultimate unit effect (e.g., the STYAN system). This is quite obvious, because for examining solvent effects formulated in terms of the relation between the copolymer composition and the monomer-feed composition (the so-called bootstrap effect by Harwood13), an adequate kinetic model and well-tted reactivity ratios are required.14 When one considers the nonuniqueness in determining the penultimate-model reactivity ratios, the choice of the penultimate-model reactivity ratios used for this purpose can be even more complicated because, by analogy with the copolymer of STY with MMA in bulk,15 the microstructure of the STY-AN copolymer may depend on the choice of the r-parameter set (Set I or Set II) used for calculation. The multiple solutions of the copolymerization equation originate from the nonlinear calculation methods applied in determination of the reactivity ratios. If this approach is assumed to be used widely today, the problem of the reactivity-ratio selection used for the description of copolymerization systems can acquire in the future an even greater importance. There is good reason for this expectation because determination methods for the terminal-model reactivity ratios based on the linearization of the MayoLewis equation1 (eq 1) have been criticized for a long time,16,17 besides being useless for the penultimate model (eq 2). In recent case studies and a comprehensive literature review18 20 on the estimation of copolymer-

848

Table I. Terminal- and Penultimate-Model Parameters for the Styrene ( M 1 )/Acrylonitrile ( M 2 ) Free-Radical Copolymerization System in Toluene (T), Methyl Ethyl Ketone (MEK), N,N-Dimethylformamide (DMF), and Acetonitrile (ACN) Estimated with the Simplex Method
Penultimate Model Terminal Model MEKe 0.334 0.078 0.117 0.240 DMFe ACNd Td Te MEKe DMFe ACNd Td Te MEKe
a

KAIM

Set I

Set IIc DMFe ACNd

Parameter

Td

Tc

0.328

r1

r2

0.369 (0.423) 0.128 (0.118)

0.103

0.488 (0.485) 0.065 (0.081)

r 11

r 22

r 21

r 12

f g ph ni j F( 99.5%) k F calcd

0.254 (0.242) 0.255 (0.133) 0.459 (0.566) 0.060 (0.109) 0.02460 0.01928 0.01241 0.01718 0.00620 0.00815 (0.049) (0.051) (0.023) 0.01645 2 2 2 2 2 4 11 13 14 15 12 11 3.68 28.39

0.427 0.206 0.220 0.249 0.200 0.448 (0.322) 0.060 0.219 0.195 0.066 0.108 0.071 (0.052) 0.603 25.635 107.572 6.018 6.178 17.648 (0.621) 0.077 1.055 1.439 1.144 2.570 1.727 (0.105) 0.00519 0.01229 0.00972 0.00772 0.01781 0.00621 (0.020) 0.00822 0.00164 0.00055 0.00001 0.01231 0.00956 0.00469 0.00063 0.00001 4 4 4 4 4 4 4 4 4 13 14 15 12 11 13 14 15 12 3.10 2.91 2.78 3.35 3.68 3.10 2.91 2.78 3.35 9.17 1.64 0.34 1.17 10.52 13.20 7.92 0.38 0.01

0.225 (0.242) 0.120 (0.108 0.532 (0.559) 0.095 (0.119) 0.01106

0.258 (0.284) 0.071 (0.063) 0.470 (0.570) 0.099 (0.132) 0.01077

0.205 (0.227) 0.109 (0.105) 0.315 (0.440) 0.142 (0.175) 0.01773

Values in parentheses come from original works. Using r 1 0.5 and r 2 0.1 as the initial guess for the (nlls) t. Using r 11 r 21 r 1 and r 22 r 12 r2 as the initial guess for the (nlls) t. c Using r 11 r 1 , r 21 1/ r 2 , r 22 r 2 , and r 12 1/ r 1 as the initial guess for the (nlls) t. d Calculated with data by Hill et al.6 e Calculated with data by Klumperman and Kraeger.7
n

f
i1

Standard deviation is dened as

models in the corresponding reactivity ratios, n is the number of experimental points, and p is the number of r parameters. g Absolute value of the difference between the standard deviation of the reactivity ratios determined for the monomer system in the solvent for the terminal and penultimate models, (terminal model) (penultimate model). h Number of r parameters in the kinetic model. i Number of copolymerization experiments. j Critical values of F for the probability level 99.5% (taken from ref. 28). k F calcd calculated according to eq. 3,27 where Models B and A stand for the terminal model and penultimate model, respectively.

F i exptl F i calcd 2 / n p where F i is the mole fraction of the monomer M i expressed for the terminal and penultimate

APPLICATION OF THE MONOMER REACTIVITY RATIOS

849

ization reactivity ratios, a further extension of the nonlinear approach is recommended and, because of advances in computing power, anticipated. Therefore, the proper choice of monomer reactivity ratios for the kinetic-model discrimination and the solvent-effect determination in the freeradical copolymerization of the monomer pairs displaying the penultimate effect can be relevant. The aim of this article is to study the impact of reactivity ratios determined with the simplex method on the kinetic-model discrimination and solvent-effect determination for a monomer system, such as the STY-AN system, displaying a penultimate effect.

EXPERIMENTAL
In this study, monomer-feed- and copolymer-composition data for the STY-AN system given by Hill et al.6 for toluene and acetonitrile and by Klumperman and Kraeger7 for toluene, methyl ethyl ketone (MEK), and N,N-dimethylformamide (DMF) were used. Calculation procedures for the reactivity-ratio determination and scanning were based on the Nelder and Mead11 simplex concept and were used as described previously.4,12,21

Figure 1. General picture of two-dimensional projections ( against rij; i,j 1,2; i j) of the ve-dimensional space (r11, r21, r22, r12, ) that resulted from the scanning experiments22 for the STY-AN free-radical copolymerization in toluene, MEK, DMF, and acetonitrile. values correspond to those for the penultimate model given in Table I.

RESULTS AND DISCUSSION

Penultimate Unit Effect in the Investigated Systems Terminal- and penultimate-model reactivity ratios for the STY (M1)/AN (M2) monomer system in toluene, MEK, DMF, and acetonitrile as estimated with the simplex method are presented in Table I. For the penultimate model, two different initial guesses for the reactivity ratios (Guess I: r11 r21 r1, r22 r12 r22; Guess II: r11 r1, r21 1/r2, r22 r2, r12 1/r1) led to two different penultimate reactivity-ratio sets, Set I and Set II. More precisely, for all investigated systems, wide ranges of penultimate-model reactivity ratios with the same minimal standard deviation () were found. Multiple penultimate-model reactivity ratios in the STY-AN free-radical copolymerization system in bulk were the subject of my previous article.12 Results obtained for a solution copolymerization in acetonitrile, DMF, MEK, and toluene were qualitatively very similar. Two-dimensional projections ( against rij; i,j 1,2; i

j) of the ve-dimensional space (r11, r21, r22, r12, ) that resulted from the scanning experiments22 over wide ranges of the value had approximately the same shape for all the solvents, as shown schematically in Figure 1. The differences for the investigated solvents consisted of the relative position of the global and local minima (see Set I and Set II in Table I). In addition, the two-dimensional projections ( against rij; i,j 1,2; i j) resulted in a similar curve with one minimum only (not shown). A comparison of the copolymerization-system descriptions accomplished via the terminal- and penultimate-model reactivity ratios (Table I) is presented in Figure 2. The composition data for all the investigated copolymerizations could be described equally well with quite different sets of the penultimate-model r parameters with very close standard deviations. An inspection of the coincidence of the composition curves for the different solvent systems revealed differences in the signicance of the penultimate unit effect in the solvent systems. The penultimate unit effect appeared to increase approximately in the following order: acetonitrile DMF MEK toluene. Therefore, it was inversely proportional to the polarity (expressed in Debye unit) of the solvent used (3.92,23 3.82,24 2.77,25 and 0.36,26 respectively). These ndings led to the conclusion that the stronger interactions between the solvent and monomers were less when the penultimate unit effect was pronounced. Generally, it seems that the scale of the observed penultimate unit effect depends on the solvent.

850

KAIM

Figure 2. Copolymerization diagrams for the STY/AN free-radical copolymerization systems at 60 C with the reactivity ratios calculated by the simplex method presented in Table I: (a) calculated for toluene with data by Hill et al.,6 (b) calculated for toluene with data by Klumperman and Kraeger,7 (c) calculated for MEK with data by Klumperman and Kraeger,7 (d) calculated for DMF with data by Klumperman and Kraeger7 [points () were generated with the penultimate-model reactivity ratios given by these authors (r11 0.227, r22 0.105, r21 0.440, r12 0.175, 0.02163)], and (e) calculated for acetonitrile with data by Hill et al.6 } experimental points; points generated according to eq 1 with the reactivity ratios for the terminal model; E points generated according to eq 2 with the reactivity ratios estimated for the penultimate model with r11 r21 r1 and r22 r12 r2 as the initial guess for the (nlls) t (Set I); points generated according to eq 2 with the reactivity ratios estimated for the penultimate model with r11 r1, r21 1/r2, r22 r2, and r12 1/r1 as the initial guess for the (nlls) t (Set II).

APPLICATION OF THE MONOMER REACTIVITY RATIOS

851

same experimental data (r11 0.205, r22 0.109, r21 0.315, r12 0.142) was 0.01773 (Table I). Because the difference between the two values was 0.0049, it is easy to distinguish on the graphic scale [ in Fig. 2(d)]. This example shows again how high the level of accuracy in the determination of reactivity ratios must be to conclude from composition data the correct kinetic-model discrimination. Hill et al.27 proposed the application of the statistical F test28 to prove if the use of the fourparameter (penultimate) model is justiable for tting the composition data. In other words, with this test a justness of the use of a higher order model can be veried. According to the proposal, the test is based on the ratio F of the residual sums of the squares (2) of Models A and B, where Model B (terminal) is a special case of Model A (penultimate) (eq 3):
Figure 2. (Continued from the previous page)
2 2 A / p A p B B 2 A / n p A

Klumperman and Kraeger7 also related the coincidence of the copolymer-composition curves for the discussed systems with the polarity of the solvent. However, they found qualitatively, contrary to our results, that the copolymer-composition curve for polymerization in DMF was not in agreement with the proposition. They tried to explain this exception with subtle differences in the interactions between the electron-donor and electron-acceptor molecules present in the polymerization system. In my opinion, a different explanation is also possible. Quantitatively, the significance of the penultimate unit effect for the monomer system in the discussed solvents can be (rigorously under certain statistical conditions) correlated with an absolute value of the difference between the standard deviation () of the reactivity ratios determined for the terminal and penultimate models ( in Table I). According to Table I, increased in the same order as the penultimate unit effect (acetonitrile DMF MEK toluene) regardless of the penultimate reactivity-ratio set (Set I or Set II) taken for comparison. Differences between the numerical values of appeared to be meaningless, but a case with the solvent effect of DMF proved the contrary. For the reactivity ratios given by Klumperman and Kraeger7 (r11 0.227, r22 0.105, r21 0.440, r12 0.175), the calculated standard deviation () was 0.02163. However, for the reactivity ratios estimated with the simplex method from the

(3)

Figure 3. STY (M1) -centered triad fractions for the STY/AN free-radical copolymerization estimated with the reactivity ratios for the terminal model (full points) and penultimate model (blank points) with r11 r21 r1 and r22 r12 r2 as the initial guess for the (nlls) t (Set I). , toluene with data by Hill et al.;6 F, E) toluene with data by Klumperman and Kraeger;7 , MEK with data by Klumperman and Kraeger;7 }, DMF with data by Klumperman and Kraeger;7 , acetonitrile with data by Hill et al.6

852

KAIM

Figure 4. STY (M1) -centered triad fractions for the STY/AN free-radical copolymerization estimated with the reactivity ratios for the penultimate model with r11 r1, r21 1/r2, r22 r2, and r12 1/r1 as the initial guess for the (nlls) t (Set II): (a) fM1M1M1 triad fraction, (b) fM2M1M2 triad fraction, and (c) fM2M1M1 triad fraction. toluene with data by Hill et al.;6 E toluene with data by Klumperman and Kraeger;7 MEK with data by Klumperman and Kraeger;7 } DMF with data by Klumperman and Kraeger;7 acetonitrile with data by Hill et al.6

where pA and pB are the numbers of the parameters for each model and n is the number of copolymerization experiments. Calculated Fcalcd ratios and critical values of F for the probability level ( 99.5%) are shown in Table I. Results showed that from a statistical point of view, the applica-

tion of the penultimate model for the description of STY-AN copolymerization in toluene (Fcalcd F(99.5%)) is justied. For copolymerization in MEK, the result of the test was not certain and depended on the penultimate reactivity-ratio set (Set I or Set II) taken for consideration. However,

APPLICATION OF THE MONOMER REACTIVITY RATIOS

853

for copolymerization in DMF and acetonitrile solutions, the penultimate model should be rejected because it does nor provide a better t to experimental data in comparison with the terminal data. Note that the F test results were in general agreement with the conclusion that the penultimate unit effect can be correlated with an absolute value of the difference between the standard deviation of the reactivity ratios determined for the terminal and penultimate models. Other consequences of treating the same experimental data by different models with a different number of parameters have been discussed previously.4,21 Microstructure of the Copolymers A comparison of the STY (M1) -centered triad fractions fM1M1M1, fM2M1M1, and fM2M1M2 for the STY-AN free-radical copolymers estimated with reactivity ratios for the terminal and penultimate models (Set I in Table I) is presented in Figure 3. In the curves for the STY-AN free-radical copolymerization in acetonitrile, DMF, MEK, and toluene, the bootstrap effect is observed; that is, the relationship between the triad-sequence distribution and copolymer composition did not depend on the solvent employed. Without a detailed study, it can be concluded, however, that points relating to the terminal model (full points) were displaced slightly (especially for the fM2M1M1 and fM2M1M2 triads) when compared to those corresponding to the penultimate model (blank points). This can be regarded as an additional proof for the presence of the penultimate unit effect in the discussed monomer systems. This observation is not new. The triad distribution versus the copolymer composition presented by Klumperman and ODriscoll14 for the copolymerization of STY with maleic anhydride in bulk, MEK, and toluene showed comparatively poor agreement with the terminal model when compared to the penultimate model. The situation changed dramatically when Set II of the penultimate-model reactivity ratios (Table I) was used for the triad-sequence calculation (Fig. 4). An examination of the discussed copolymers shows that the triad distribution depended on the solvent used for the copolymerization. The sequence distribution in the copolymers synthesized in toluene was especially different, although, as we remember from the previous section and Figure 2(a,b), the copolymer-composition curves for toluene calculated for both reactivityratio sets (Set I and Set II) overlap precisely.

Note, however, that there is a difference in tting the particular triad fractions (fM1M1M1, fM2M1M1, fM2M1M2) to the general bootstrap effect rule: triads involving both monomers M1 and M2 less accurately fulll the expectation in terms of the bootstrap effect symptoms. The results are difcult to interpret from a theoretical point of view. However, the STY-AN monomer systems in toluene indicated the most signicant penultimate unit effect among all the discussed solvents. The question arises whether the observed convergence was coincidental and, therefore, without any physical meaning, or could be interpreted in terms of the copolymerization kinetics.

CONCLUSIONS
From this study, it can be concluded that on the basis of the composition data alone, it is possible to discriminate between the terminal and penultimate models for the STY-AN monomer system under the condition of a very accurate reactivityratio determination. The nonlinear approach to the estimation of reactivity ratios based on the Nelder and Mead simplex method seems to be a very useful tool for the purpose. The results indicate that the signicance of the penultimate unit effect for a given monomer system can be correlated with an absolute value of the difference between the standard deviation of the terminaland penultimate-model reactivity ratios. The penultimate unit effect depends on the solvent and seems to be inversely proportional to its polarity. Contrary to the full equivalence in the representation of the copolymer-composition data with different but numerically equivalent penultimatemodel reactivity ratios, the relationship between the triad-sequence distribution and copolymer composition does depend on the set of penultimate-model reactivity ratios chosen for microstructure determination. This observation still needs some theoretical interpretation.
This study was supported with 120-501/68-BW-1418/ 18/98 from the Faculty of Chemistry, University of Warsaw.

REFERENCES AND NOTES

1. Mayo, F. R.; Lewis, F. M. J Am Chem Soc 1944, 66, 1594.

854

KAIM

2. Merz, E.; Alfrey, T.; Goldnger, G. J Polym Sci 1946, 1, 75. 3. Schweer, J. Macromol Theory Simul 1993, 2, 485. 4. Kaim, A.; Oracz, P. Macromol Theory Simul 1997, 6, 565. 5. Hill, D. J. T.; Lang, A. P.; ODonnell, J. H.; OSullivan, P. W. Eur Polym J 1989, 25, 911. 6. Hill, D. J. T.; Lang, A. P.; Munro, P. D.; ODonnell, J. H. Eur Polym J 1992, 28, 391. 7. Klumperman, B.; Kraeger, I. R. Macromolecules 1994, 27, 1529. 8. Fukuda, T.; Ma, Y.-D.; Inagaki, H. Macromolecules 1985, 18, 17. 9. Coote, M. L.; Davis, T. P. Macromolecules 1999, 32, 3626. 10. Moad, G.; Solomon, D. H.; Spurling, T. H. Macromolecules 1989, 22, 1145. 11. Nelder, O. A.; Mead, R. Comput J 1964, 7, 308. 12. Kaim, A. J Macromol Sci Chem 1998, 4, 577. 13. Harwood, H. J. Makromol Chem Makromol Symp 1987, 10/11, 331. 14. Klumperman, B.; ODriscoll, K. F. Polymer 1993, 34, 1032. 15. Kaim, A. Macromol Theory Simul 1997, 6, 907.

16. Tidwell, P. W.; Mortimer G. A. J Polym Sci 1965, A3, 369. 17. ODriscoll, K. F.; Reilly, P. M. Makromol Chem Makromol Symp 1987, 10/11, 355. 18. Van Herk, A. M.; Dro ge, T. Macromol Theory Simul 1997, 6, 1263. 19. Polic, A. L.; Duever, T. A.; Penlidis, A. J Polym Sci Polym Chem Ed 1998, 36, 813. 20. Giz, A. Macromol Theory Simul 1998, 4, 391. 21. Kaim, A.; Oracz, P. Polymer 1997, 38, 2221. 22. Detailed results will be published elsewhere. 23. Gosh, S. N.; Trambarulo, R.; Gordy, W. J Chem Phys 1953, 21, 310. 24. Krumgalz, B. S.; Barthel, J. G. M. Z Phys Chem 1984, 142, 167. 25. Fuchs, R.; Krenzer, L.; Gaube, J. Ber Bunsen-Ges Phys Chem 1984, 7, 642. 26. Becker, R.; Freedman, K. J Am Chem Soc 1985, 107, 1477. 27. Hill, D. J. T.; ODonnell, J. H.; OSullivan, P. W. Macromolecules 1982, 15, 960. 28. Wilson, E. B., Jr. An Introduction to Scientic Research; McGraw-Hill: New York, 1952; Chapter 8.