Jorg

Physical Aging Kinetics of Syndiotactic Polystyrene as Determined From Creep Behavior JORG BECKMANN and GREGORY B. MCKENNA* Polymers Division National Institute of Standards and Technology Gaithersburg, Maryland 20899 and BRIAN G. LANDES, DAVID H. BANK, and ROBERT A. BUBECK Dow Chemical Company Midland, Michigan 48674 Creep experiments in uniaxial extension have been performed to explore the kinetics of the physical aging process in semicrystalline syndiotactic polystyrene (sPS) having two processing histories. Classical time-aging time superposition behavior was found for both materials at temperatures from 70 to 95°C, with the shift rate m decreasing as temperature was increased. Virtually no aging was seen at 95°C, the DSC determined glass transition, Tg. This behavior was atypical for a semicrystalline polymer and reminiscent of the behavior of glassy amorphous thermoplastics. Some evidence for a separate crystalline aging mechanism . Tg, which manifests itself as only vertical shifts without timescale shifts, is seen in experiments at T . 100°C. Finally, the two different materials age differently, suggesting that some control of aging can be obtained by altering processing conditions or morphology.

INTRODUCTION he cooling of an amorphous polymer through the glass transition Tg results in a non-equilibrium glassy state that then proceeds to spontaneously evolve towards a temporally distant equilibrium (1, 2). Associated with the evolution of the state variables of, e.g., volume or enthalpy, are observable changes of macroscopic properties such as viscoelastic response, yield strength, impact resistance, etc. (3), which has come to be referred to as physical aging. One anticipates similar behavior for the glassy response of semicrystalline polymers. However, it has been observed that for semicrystalline polymers physical aging persists even above the glass transition temperature (3– 8). A partial explanation for this behavior was offered by Struik (3– 8). His argument is based on the observation that in constant frequency experiments, in which the temperature is ramped from below to above the Tg, the loss factor (tan d) of the semicrystalline polymers is broadened and appears to extend well above the Tg range of the equivalent amorphous polymer. Struik interpreted such behavior in terms of part

T

* Corresponding author.

of the amorphous phase being constrained by the presence of the crystalline regions, hence having a decreased mobility and correspondingly increased Tg. This model is also consistent with concepts of a “rigid” amorphous phase in semicrystalline polymers, but does not agree with the small changes in glass transition observed upon large deformation of, e.g., rubber. Another complicating feature of the problem is the fact that recrystallization phenomena are possible .Tg. Here we report on studies of the physical aging process in syndiotactic polystyrene, for which the amorphous and crystalline phases have similar densities (9), and one might consequently expect that constraints on the amorphous phase would be less than those found in other semicrystalline polymers where amorphous/crystalline densities are significantly different. As shown below, we find qualitatively different aging behavior for the sPS from that exhibited by other semicrystalline polymers. TIME-AGING TIME SUPERPOSITION Typically, after quenching glass forming materials from .Tg to the aging temperature, the evolution of the viscoelastic response with increasing aging time can be represented by a time-aging time superposition

POLYMER ENGINEERING AND SCIENCE, SEPTEMBER 1997, Vol. 37, No. 9

1459

Jorg Beckmann et al. Table 1. Description of Injection Molded Samples of Syndiotactic Polystyrene. Sample Designation SPS140 SPS300F SPSA or SPS140A SPSB

Mold Temperature

Sample Thickness, inch (mm)

140°F to 150°F (60°C to 66°C) 285°F to 295°F (141°C to 146°C) — —

⁄ (3.2) ⁄ (3.2) ⁄ (1.6) 1⁄16 (1.6) 18

18 1 16

principle similar to time-temperature superposition. We can examine this by considering the following: assume that the creep behavior at each aging time can be represented using a Kohlrausch-Williams-Watts (KWW) function (10, 11):

D ( t ) 5 D 0 exp@ t/ t 0 # b

clear surface with a turbid core turbid clear surface with a turbid core turbid

where t0(te) is the value of t0 in Eq 2 at aging time te and t0(tref ) its value at the reference aging time. Having obtained the values of ate, these are then analyzed in the conventional manner of making double logarithmic plots of log(ate) vs. log(te), the gradient of which has been defined by Struik (3) as:

(1)

where D(t), the creep compliance, 5 e/s where e is the measured strain and s is the applied stress, t0 is a characteristic retardation time and b a shape parameter for each creep curve. D0 is a fitting parameter. For a given temperature and strain, and assuming that b is independent of aging time, it is possible to perform time-aging time superposition of the data by reducing the curves to a reference aging time via a horizontal shift along time axis. The aging time shift factor ate is then defined from the KWW function as:

log@ a te # 5 log @ t 0 ( t e ) / t 0 ( t ref )#

Visual Appearance

(2)

m 5 d log( a te ) /d log( t e ) ,

(3)

where m is often referred to as the shift rate. In general, if the shape of the creep curve is unaltered as aging time progresses, we consider that timeaging time superposition (TAS) can apply. Hence, the TAS is not limited to the KWW type of creep function, but any function that describes the data can be used. Finally, it is normally found that slight shifts in the magnitude of the creep compliance are often needed to superimpose the data. These vertical shifts represent a change in the strength of the retardation (creep)

Fig. 1. Double logarithmic representation of creep compliance vs. time for samples of syndiotactic polystyrene (SPS300F) “normalized” at 160°C and quenched to 70°C where “sequential” aging tests were performed. Aging times were: (h) 1800 s; (E) 3600 s; (‚) 7200 s; () 14,400 s; ({) 28,800 s; (1) 57,600 s; (3) 115,600 s; (✳) 230,400 s. Reduced curve produced by combination of vertical and horizontal shifts is shown displaced arbitrarily for clarity. Applied stress 5 5 MPa. 1460

POLYMER ENGINEERING AND SCIENCE, SEPTEMBER 1997, Vol. 37, No. 9

Physical Aging Kinetics of Syndiotactic Polystyrene

Fig. 2. Double logarithmic representation of creep compliance vs. time for samples of syndiotactic polystyrene (SPS300F) “normalized” at 160°C and quenched to 90°C where “sequential” aging tests were performed. Aging times were: (h) 1800 s; (E) 3600 s; (‚) 7200 s; (ƒ) 14,400 s; ({) 28,800 s; (1) 57,600 s; (3) 115,600 s; (✳) 230,400 s. Reduced curve produced by combination of vertical and horizontal shifts is shown displaced arbitrarily for clarity. Applied stress 5 5 MPa.

process, and, in the context of the KWW function, can be written as:

b te 5 D 0 ( t e ) /D 0 ( t e,ref )

(4)

These will be found to be particularly important .Tg. EXPERIMENTS Materials Four samples of syndiotactic polystyrene were provided to NIST by Dow in the form of dumbbell- shaped specimens produced by injection molding (12). Their designations and some comments describing their visual appearances are presented in Table 1. Two of these samples were used in the physical aging experiments: The SPS300F, which was injected into a hot mold (300°F; 149°C) and contained a nucleating agent, and the SPSA, which was injected into a mold held at 140°F (60°C) and contained no nucleating agent. All samples were “normalized” for 1 h at 160°C before quenching to the aging (testing) temperature because of possible effects cold crystallization could have on the physical aging measurements that would lead to difficulty in experimental reproducibility. Creep Tests The mechanical tests were carried out in uniaxial extension in creep conditions. The mechanical tests were performed in load control using a computer-con-

trolled servo-hydraulic testing machine (Instron model 132.25), equipped with an oven specially designed for a rigorous temperature control. The temperature was controlled using a Cole Parmer DigiSense Model 2186-20 temperature controller. Oven stability was 6 0.2°C during each experiment. The temperature gradient over the gauge length of the sample was 6 0.2°C as determined from thermocouple measurements. Sample strain was measured with an extensometer Instron (12) Model 2620-259 attached to the sample. Aging Tests The aging experiments were carried out by first annealing the specimens at 160°C for 1 h. They were then placed in the test machine at the aging test temperature, and the sequence of aging “probe” stresses in creep was applied. Experiments were performed at aging test temperatures of 60, 70, 80, 82.5, 85, 90, 95, 100, 110, and 120°C. The experimental temperature range encompasses region 1, region 2, and the very lower end of region 3 of the four characteristic temperature regions defined by Struik (4 – 8). The loading history applied to the sample after the quench, followed the protocol developed by Struik: after quenching the samples to the aging test temperature, the creep tests were started sequentially at increasing aging times, te. The duration of the load application, t l, was also increased to keep the ratio tl/te

POLYMER ENGINEERING AND SCIENCE, SEPTEMBER 1997, Vol. 37, No. 9

1461

Jorg Beckmann et al.

Fig. 3. Double logarithmic representation of creep compliance vs. time for samples of syndiotactic polystyrene (SPS300F) “normalized” at 160°C and quenched to 95°C where “sequential” aging tests were performed. Aging times were: (h) 1800 s; (E) 3600 s; (‚) 7200 s; (ƒ) 14,400 s; ({) 28,800 s; (1) 57,600 s; (3) 115,600 s; (✳) 230,400 s. Time-aging time superposition was not possible for these data. Applied stress 5 5 MPa.

Fig. 4. Double logarithmic representation of creep compliance vs. time for samples of syndiotactic polystyrene (SPS300F) “normalized” at 160°C and quenched to 110°C where “sequential” aging tests were performed. Aging times were: (h) 1800 s; (E) 3600 s; (‚) 7200 s; (ƒ) 14,400 s; ({) 28,800 s; (1) 57,600 s; (3) 115,600 s; (✳) 230,400 s. Reduced curve produced by (primarily) vertical shifts is shown displaced arbitrarily for clarity. Applied stress 5 5 MPa. 1462

POLYMER ENGINEERING AND SCIENCE, SEPTEMBER 1997, Vol. 37, No. 9

Physical Aging Kinetics of Syndiotactic Polystyrene

Fig. 5. Double logarithmic representation of creep compliance vs. time for samples of syndiotactic polystyrene (SPS300F) “normalized” at 160°C and quenched to 120°C where “sequential” aging tests were performed. Aging times were: (h) 1800 s; (E) 3600 s; (‚) 7200 s; (ƒ) 14,400 s; ({) 28,800 s; (1) 57,600 s; (3) 115,600 s; (✳) 230,400 s. Reduced curve produced by (primarily) vertical shifts is shown displaced arbitrarily for clarity. Applied stress 5 5 MPa. Table 2. Vertical Shift Factors bte Required to Superimpose the Creep Curves at Different Aging Times and at 110 and 120°C for Samples SPS300F and SPS140A.

te/S

SPS 140A T 5 120°C bte

SPS300F T 5 120°C bte

SPS 140A T 5 110°C bte

SPS 300F T 5 110°C bte

1800 3600 7200 14,400 28,800 57,600 115,200 230,400

0.99 0.95 0.95 0.96 0.96 0.98 0.99 1.00

0.90 0.89 0.91 0.92 0.93 0.95 0.98 1.00

0.87 0.90 0.92 0.94 0.96 0.97 0.98 1.00

0.94 0.95 0.95 0.95 0.96 0.98 1.00 1.00

constant at 0.1. An engineering stress of 5 MPa was applied to the samples. The zero aging time was chosen to be the time at which the samples attained the testing (aging) temperature. RESULTS 300F Polymer: Time-Aging Time Superposition In this section we present the results from aging experiments in which the material was “probed” using sequential creep tests (13). Figure 1 shows the aging of the SPS300F at 70°C. Similar behavior was observed at both 60 and 80°C. The typical aging as seen in Fig. 1 as a shift of the creep behavior towards longer times

was observed for these temperatures. Furthermore the curve shape can be described with the KWW Eq 1. Many amorphous polymers have been found to follow Eq 1 with a value of b 5 1⁄3 and, for the sPS at 60, 70 and 80°C, we also find b 5 1⁄3. Figure 1 also shows the “offset” “master curve,” which is the result of the superimposed creep curves measured at different aging times shifted horizontally on the logarithmic timescale. In experiments at 82.5, 85 and 90°C, we found that the KWW function did not fit the full range of the data. At 82.5°C, the KWW function describes the creep response only up to ;4000 sec and the exponent b decreases slightly to 0.28 in the short time interval. The deviation of the creep response from the KWW behavior is more pronounced at 85 and 90°C. At 85°C, the KWW function can only fit the creep behavior up to 500 sec with an exponent of b 5 0.1. Above this time a constant double logarithmic creep rate m 5 0.2 was observed (power law response). Despite these factors, the individual curves at each of these temperatures, while not represented by the KWW function, are still superimposable to create a master curve, as depicted in Fig. 2 for the 90°C data. At 95°C (Fig. 3) we see that the short time creep responses are virtually independent of aging time and the double logarithmic creep rate at long creep times decreases with increasing aging time. The opposite behavior was observed for the responses at 100 and 105°C. The creep curves are different in the short

POLYMER ENGINEERING AND SCIENCE, SEPTEMBER 1997, Vol. 37, No. 9

1463

Jorg Beckmann et al.

Fig. 6. Double logarithmic representation of the aging time shift factor ate vs. aging time te for syndiotactic polystyrene (SPS300F) at the temperatures indicated. Data not shown at 95°C and 100°C because time-aging time superposition could not be performed on data from these temperatures.

timescale but tend towards a common plateau value as the aging time is increased. The change of shape of the creep responses at different aging times does not allow a master curve construction. Hence, time-aging time superposition in the range 95 to 105°C is not possible. Figures 4 and 5 show the creep behavior of sPS for different aging times at 110 and 120°C, well above the Tg of the sPS. Interestingly, master curves could be constructed for the creep responses depicted in Figs. 4 and 5 using only vertical shift factors with no horizontal shift. This implies a stiffening of the material without any change in the viscoelastic response itself, Table 3. Initial Value of Double Logarithmic Shift Rate m for sPS 300F and 140A Materials at Different Temperatures. Temperature m 5 d log(ate)/d log(te) m 5 d log(ate)/d log(te) (°C) for 300F Material for 140A Material 60 70 80 82.5 85 90 95 100 110 120

1464

0.89 0.87 0.68 0.37 0.35 0.32 no superposition no superposition 0 0

not measured 0.95 0.82 0.79 0.73 no superposition no superposition no superposition 0 0

which may be due to increasing crystallinity. The vertical shifts for 110 and 120°C are presented in Table 2. To characterize the aging in detail for the different regions the double logarithmic aging rate m (Eq. 3) can be determined from aging time shift factors ate obtained in producing the master curves in the temperature regions where time-aging time superposition is applicable. Figure 6 shows how the aging of the SPS300F depends on temperature for the creep data described in the preceding paragraphs. We readily see that the slopes of the curves decrease as the aging temperature increases and that at 110 and 120°C the aging (of the viscoelastic response) ceases. The reader is cautioned to recall that between 95 and 105°C timeaging time superposition could not be applied, although the responses did not change dramatically with aging time in that region. In Table 3 are gathered the shift rates m at different temperatures, calculated from the initial slope of the curves shown in Fig. 6. Importantly, Fig. 6 and Table 3 indicate a pronounced temperature dependence of m at temperatures between Tg – 25°C and Tg 1 25°C. 140A Polymer: Time-Aging Time Superposition This section describes the results of the physical aging characterization of the 140A material. The overall behavior is similar to that of the SPS300F material

POLYMER ENGINEERING AND SCIENCE, SEPTEMBER 1997, Vol. 37, No. 9

Physical Aging Kinetics of Syndiotactic Polystyrene

Fig. 7. Double logarithmic representation of creep compliance vs. time for samples of syndiotactic polystyrene (SPS140A) “normalized” at 160°C and quenched to 70°C where “sequential” aging tests were performed. Aging times were: (h) 1800 s; (E) 3600 s; (‚) 7200 s; (ƒ) 14,400 s; ({) 28,800 s; (1) 57,600 s; (3) 115,600 s; (✳) 230,400 s. Reduced curve produced by combination of vertical and horizontal shifts is shown displaced arbitrarily for clarity. Applied stress 5 5 MPa.

Fig. 8. Double logarithmic representation of creep compliance vs. time for samples of syndiotactic polystyrene (SPS140A) “normalized” at 160°C and quenched to 85°C where “sequential” aging tests were performed. Aging times were: (h) 1800 s; (E) 3600 s; (‚) 7200 s; (ƒ) 14,400 s; ({) 28,800 s; (1) 57,600 s; (3) 115,600 s; (✳) 230,400 s. Reduced curve produced by combination of vertical and horizontal shifts is shown displaced arbitrarily for clarity. Applied stress 5 5 MPa. POLYMER ENGINEERING AND SCIENCE, SEPTEMBER 1997, Vol. 37, No. 9

1465

Jorg Beckmann et al.

Fig. 9. Double logarithmic representation of creep compliance vs. time for samples of syndiotactic polystyrene (SPS300F) “normalized” at 160°C and quenched to 90°C where “sequential” aging tests were performed. Aging times were: (h) 1800 s; (E) 3600 s; (‚) 7200 s; (ƒ) 14,400 s; ({) 28,800 s; (1) 57,600 s; (3) 115,600 s; (✳) 230,400 s. Time-aging time superposition was not possible for these data. Applied stress 5 5 MPa.

described previously. Even though a 160°C normalization was performed prior to the aging experiments, there are subtle differences in detail that indicate an influence of the processing conditions on the aging process. Figure 7 depicts the aging response at 70°C for the 140A material. As in the case of the 300F material, time aging time superposition holds for this material and at the temperatures investigated up to 80°C, a KWW function with constant b ' 1⁄3 fits the data. In measurements at 82.5 through 85°C the behavior is no longer systematically KWW in nature, but the curves can still be superimposed by a combination of vertical and horizontal shifts (see Fig. 8). In Fig. 9 we show the creep responses at different aging times for 90°C. Non-superposability is observed for the different aging times in the 90°C experiments as well as those carried out at 95 and 100°C, a result that is somewhat different from what was observed for the 300F material where superposability was possible up to 95°C. Figure 10 shows the aging response at 120°C. In this data set, as well as that at 110°C, the superposability of the creep responses at different aging times is primarily performed by vertical shifting, indicating no significant change in the retardation times due to aging at these temperatures. Rather, vertical shifts indicate a change in the intensity of the retardation processes. As discussed later, this behavior is atypical of semicrystalline polymers. The vertical shift factors for the SPS140A material are presented in Table 2. 1466

Figure 11 summarizes the aging data for the 140A material in a plot of log (ate) vs. log (te). We note that the aging rates [slopes of the log (ate) vs. log (te) curves] differ subtly from the data of Fig. 6 for the 300F material. In effect, at 85°C the aging rate for the 140A material is somewhat higher than for the 300F material at the same temperature. Because the curves at 90 to 100°C and 95 to 100°C are non-superposable, the comparison of the aging rates is not so straightforward. Close examination of the data, however, suggests that the 300F material is more stable at these temperatures than is the 140A material. The aging shift rates m for both the 140A and 300F materials are compared in Table 3. DISCUSSION AND IMPLICATIONS The aging of sPS, as measured by sequential creep experiments subsequent to a quench from a “normalization” temperature of 160°C is significantly different from that observed in most semicrystalline polymers. The experimental observations suggest that the sPS behaves much like an amorphous polymer below Tg, while above Tg the response is unlike either amorphous materials or other semicrystalline polymers. These are discussed point by point below. Behavior Below Tg: The creep compliance behavior of the sPS at different aging times is similar to those observed for amorphous polymers in general, viz., the viscoelastic response can be represented by a time-

POLYMER ENGINEERING AND SCIENCE, SEPTEMBER 1997, Vol. 37, No. 9

Physical Aging Kinetics of Syndiotactic Polystyrene

Fig. 10. Double logarithmic representation of creep compliance vs. time for samples of syndiotactic polystyrene (SPS140A) “normalized” at 160°C and quenched to 120°C where “sequential” aging tests were performed. Aging times were: (h) 1800 s; (E) 3600 s; (‚) 7200 s; (ƒ) 14,400 s; ({) 28,800 s; (1) 57,600 s; (3) 115,600 s; (✳) 230,400 s. Reduced curve produced by (primarily) vertical shifts is shown displaced arbitrarily for clarity. Applied stress 5 5 MPa.

aging time superposition and the double logarithmic shift rate m is near to unity well below Tg and decreases to zero as temperature increases to and above Tg. Behavior Near Tg: Near the Tg (95 to 105°C) the creep behavior does change as aging time increases. While the changes in creep are small, time-aging time superposition can still not be applied to the results. In this temperature range, there is little volume recovery response. The observed behavior in this range is different from that in other semicrystalline polymers (cf., Refs. 4 – 8). In other semicrystalline materials the aging of the viscoelastic response continues as strongly as (sometimes more strongly than) well below Tg. Behavior Above Tg: For creep compliance response changes upon aging, however, there is no apparent change in the characteristic viscoelastic retardation time. Rather, all of the changes can be expressed as a simple increase in the material stiffness, which is presumably due to changes in the crystalline phase, e.g., degree of crystallinity, crystallite size distribution or secondary recrystallization. Such a response is different from either that of the amorphous material or other semicrystalline materials. For the former, physical aging of any sort ceases, while for the latter, aging effects seem to be stronger above Tg (see, e.g., Ref. 8.) than below it. Implications: The above picture presents an interesting set of possibilities for the sPS. While the changes in viscoelastic properties below Tg are similar

to those associated with aging in both amorphous and semicrystalline polymers, the behavior above Tg is different. Upon aging, only the stiffness increases, while the viscoelastic component of the response remains the same. Consequently, the viscoelastic dissipation in the material remains constant and one might anticipate less embrittlement upon aging at high temperatures than might be otherwise anticipated, although fracture toughness investigations have not been a part of this study. Effects of Molding History: Finally, we note that the two different materials investigated here exhibited somewhat different aging behaviors. In particular, the 140A material showed a more rapid change in the viscoelastic response with increasing aging time in the vicinity of Tg. The reasons for this are unclear at this time. The differences in the morphologies have not been investigated systematically. However, it should be noted that the 140A material did not contain a nucleating agent, while the 300F material did. Furthermore, the as-received materials were visually different. The 140A material had a clear surface with a turbid core, while the 300F material was completely turbid. The differences in aging are subtle, but suggest that either a) the material with the nucleating agent resists aging in the vicinity of Tg better than does the sample with no nucleating agent or b) molding at a temperature . Tg (300F material) results in a more stable material than molding , Tg (140A material).

POLYMER ENGINEERING AND SCIENCE, SEPTEMBER 1997, Vol. 37, No. 9

1467

Jorg Beckmann et al.

Fig. 11. Double logarithmic representation of the aging time shift factor ate vs. aging time te for syndiotactic polystyrene (SPS300F) at the temperatures indicated. Data not shown at 90°C, 95°C and 100°C because time-aging time superposition could not be performed on data from these temperatures.

ACKNOWLEDGMENTS This work was sponsored, in part, through a Cooperative Research and Development Agreement between the Dow Chemical Company and the National Institute of Standards and Technology. Dr. Beckmann thanks the University of Potsdam, Potsdam, Germany, which also provided partial support for his stay at NIST. The authors would also like to thank Dr. Craig Carriere of the Dow Chemical Company for many stimulating discussions during the course of this work. REFERENCES 1. A. J. Kovacs, Fortsch. Hochpolymer Forschung, 3, 394 (1963). 2. G. B. McKenna, in Comprehensive Polymer Science. Vol. 2, Polymer Properties, p. 311, C. Booth and C. Price, eds., Pergamon, Oxford, England (1989). 3. L. C. E. Struik, Physical Aging in Amorphous Polymers and Other Materials Elsevier, Amsterdam (1978). 4. L. C. E. Struik, Polymer, 28, 1521 (1987). 5. L. C. E. Struik, Polymer, 28, 1534 (1987). 6. L. C. E. Struik, Polymer, 30, 799 (1989).

1468

7. L. C. E. Struik, Polymer, 30, 815 (1989). 8. I. Spinu and G. B. McKenna, Polym. Eng. Sci., 34, 1808 (1994). 9. Dow Chemical Company internal data. 10. F. Kohlrausch, Pogg. Ann. Phys., 12, 393 (1847). 11. G. Williams and D. C. Watts, Trans. Faraday Soc., 66, 80 (1970). 12. Certain commercial materials and equipment are identified in this paper to specify adequately the experimental procedure. In no case does such identification imply recommendation or endorsement by the National Institute of Standards and Technology, nor does it imply necessarily that the product is the best available for the purpose. 13. The creep and aging data reported here are typical of those obtained by us (8, 14) and others (4 –7) on similar equipment and with similar methods. Typically the creep compliance measurements and the ability to determine superposability are limited by the machine capabilities and temperature stability of the measurement and the resulting aging time shift factors are reliable to within about 0.1 log10 units in isothermal measurements. Statistical estimates of the fitting parameters to the KWW function are less meaningful because the parameters are interdependent (not orthogonal in parameter space) and we refrain from putting uncertainty estimates on them. 14. A. Lee and G. B. McKenna, Polymer, 29, 1812 (1988).

POLYMER ENGINEERING AND SCIENCE, SEPTEMBER 1997, Vol. 37, No. 9