The application of Mooney relaxation to QC.

For many years, elastomer manufacturers have sought process control tools for the processability of their products. Customers subject these elastomers to processing of many different types, employing a large range of deformation modes and stress or strain loadings. Although the term processability is often used in imprecise fashion, it generally refers in some fashion to the time dependent viscoelastic response of the elastomer to these loadings. Beyond this, little can be said to further restrict the definition, since commercially important processes cover a wide range of temperatures and stresses (or strain rates), and the time scale at which the elastomer response is important can vary from a fraction of a second to days or months. In most cases, where specific interactions are absent, time/temperature superposition may be invoked to simplify the situation by combining time and temperature effects into a single time/temperature domain. One method of depicting the elastomer's response over a broad range of this domain is a graphical representation of the viscous and elastic components of the dynamic modulus versus strain rate. For most elastomers, an attempt to represent all of this information numerically, even to moderate resolution, would require definition of a large number of parameters.

Quality control considerations

Elastomers manufacturers must determine the effects of molecular structure on the rheology of their products and identify the process controls that govern the molecular structure. Typically, this is done during product research and development. Analytical tools used during R&D include nuclear magnetic resonance (NMR) spectroscopy, gel permeation chromatography (GPC) and other powerful analytical tools to characterize polymer structure, while dynamic stress analyzers, capable of a variety of deformation modes, are used to fully characterize rheology over as many as eight decades of frequency or time scale. During these studies, it is not unusual to discover that relatively few degrees of freedom exist to alter the molecular structure in ways that noticeably affect the product rheology. While a moderate number of process controls usually exist that can affect molecular structure, some will have no detectable effect on rheology, while others will have identical effects on rheology and can be considered equivalent for process control.

Once the R&D is over, the elastomer manufacturer evaluates analytical requirements for process control. Considerations that were unimportant during R&D often become significant. One of these is analysis turnaround time. Another is test repeatability. Since quality control tests will often be performed over the entire commercial life of the product, analysis cost also becomes important. By contrast, the need for full rheological characterization becomes much less important. How much less important depends on the number of degrees of freedom for varying rheology that were uncovered during R&D. Often, this number of degrees of freedom lies between two and four.

Ideally, the number of independent quality control test parameters should be matched to the degrees of freedom in the process. During shifts in the process, quality control tests would then unambiguously define the nature of the shift and suggest a corresponding process control that could be used to correct the situation. Since most processes have more than one degree of freedom, Mooney viscosity alone is usually insufficient to guarantee that other rheological properties are well controlled.

Manufacturers have often responded to this situation by developing proprietary rheological tests to enhance control over their processes. Not surprisingly, many of these tests involve stress relaxation or strain recovery, since such tests provide information on the elastic response of the elastomer, and thus are ideal complements to Mooney viscosity. Instrument manufacturers have also developed equipment for this purpose. Notable examples in this latter category include capillary flow instruments with the swell capability (ref. 1), compression relaxation tests like the Monsanto SRPT, strain recovery instruments like the Williams Plastometer (and its improved version by Haake and Bayer -- the Defo Elastometer (refs. 2 and 3)), and the step-strain relaxation method by BFGoodrich (DSR) (ref. 4). All this effort has no doubt led to good process control in many situations, but most proposed solutions have met with problems like high cost, long analysis time, or poor repeatability or reproducibility that have prevented them from becoming industry standards.

The case for Mooney relaxation

Mooney relaxation (MLR) is not a new rheological test. Historically, it has suffered from poor sensitivity, repeatability, and reproducibility, but recent improvement in test equipment, including motor braking, reduced rotor friction, and computer control for frictional correction and data analysis have prompted reconsideration of MLR as a process control tool for elastomer processability (refs. 5 and 6).

Of the many sample deformation modes employed in viscoelastic testing, step loadings (either stress or strain) provide the maximum amount of rheological information. This is because in step loadings, no relaxation of the elastomer occurs during the loading phase, and thus no potential information is lost. However, the equipment required to perform step loading measurements is more complex than some of the alternative loading modes. and sample preparation is more demanding for step loadings, since residual stresses from the sample preparation itself must be carefully removed before performing the measurement.

Mooney relaxation, which suffers from a sub-optimal loading mode, provides advantages in analysis time, ease of sample preparation and overall cost, and may ultimately offer the best compromise for quality control. While Mooney machines are not inexpensive, they are required in virtually every elastomers laboratory. From an operating cost perspective, MLR is nearly free for those who need to measure Mooney viscosity, assuming the Mooney viscosity test temperature and times are compatible with the relaxation measurement. The major concerns with the MLR test are whether stress relaxation following cessation of flow provides adequate sensitivity, and whether a single point elasticity measurement provides enough information for process control. These latter questions were addressed in two separate studies, the results of which are presented here.

Experimental

This section provides experimental details of two quality control studies in which Mooney relaxation was evaluated and found to be a good alternative. The first study involves ethylene-propylene-diene rubber (EPDM) and the second involves star branched butyl rubber (SB-IIR).

EPDM study

For the EPDM study, the quality control objectives were well defined at the outset. For EPDM manufacture, it was desired to evaluate MLR as a replacement for a proprietary stress relaxation test that had been in commercial use as a quality control test for over 20 years. Over this time, a wide variety of experimental programs and direct experience with customers have validated that this proprietary test, in conjunction with Mooney viscosity, is sufficient to identify changes in rheological behavior of a given EPDM grade that would have practical impact on customers. This test involves measuring the time required for a fixed percentage of stress decay following a complex compressive strain loading, and will be identified here as stress relaxation time ([t.sub.sr]). It is known that more than two degrees of freedom exist for varying the molecular structure of many EPDM grades, and thus process shifts in Mooney viscosity and/or [t.sub.sr] do not uniquely define the changes in molecular structure responsible for process shifts. However, good systems exist to monitor process variables, and it is rare that manufacturing personnel are unable to identify the cause of a process shift, or at least to identify appropriate control tools to adjust the process to target. In this situation. MLR was seen to hold promise for providing information equivalent to [t.sub.sr] more rapidly. more precisely and at lower cost. Additionally, it was hoped that publication of an ASTM standard for MLR testing would make the test accessible to customers, and that it would prove to have good interlab reproducibility as well as intralab repeatability. By contrast, details of the proprietary method for [t.sub.sr] have not been divulged, since experience has shown that interlab reproducibility is poor unless extraordinary measures are taken to standardize both testing hardware and minute procedural details. Most of this sensitivity is known to come from sample preparation.

The EPDM experimental program was designed to select one or more MLR parameters that best correlated with the proprietary [t.sub.sr], and which gave repeatability and reproducibility as good as or better than [t.sub.sr]. During the first phase of evaluation, 100 EPDM samples, varying widely in molecular structure, were chosen for study. Both R&D and commercial samples were included in this set. All the testing was carried out in a single laboratory. Samples were tested for Mooney viscosity, MLR and [t.sub.sr], and measurements of some samples were replicated. Parameters measured from the MLR curves included decay times, [t.sub.90] and [t.sub.20]. slope (a), intercept (k), and MLR area (A). These parameters are defined as illustrated on figures 1 and 2. The decay times were dropped from consideration in the early stages when a preliminary check of repeatability showed them to be considerably poorer than MLR area, and to have demonstrably less ability to distinguish one sample from another. The data for slope, intercept and MLR area are summarized in table 1. These data were used to evaluate alternative MLR parameters for correlation to [t.sub.sr] and for repeatability. Following this phase, a second phase of testing involving four laboratories was conducted to assess reproducibility. A separate set of 47 commercial samples was used for this second phase. Each lab tested each sample one time, and the results are summarized in table 2.

[TABULAR DATA 1 & 2 OMITTED]

During both phases of the experimental program, Mooney viscosity and MLR analyses were conducted as described in ASTM D 1646-93a, utilizing Monsanto MV-2000E Mooney viscometers. The test temperature was 125[degrees]C, the warm up time was one minute, and the analysis time for Mooney viscosity was four minutes. The MLR phase commenced immediately following the Mooney viscosity, and was continued for 100 seconds.

Empirically, it has been noted that the MLR curves of many elastomers follow a power law relationship over a moderate time range (eg. 1 to 100 seconds). Such a relationship is not expected to hold over the entire time/temperature domain, but is observed to hold in many rheological tests for at least part of the flow region. The time/temperature range of the MLR test falls in the flow region for most commercial elastomers. The MV-2000E software makes provisions for such cases, and performs a logarithmic transformation of the data, which is then fitted by a least squares regression line:

(1) log[M] = a log[t] + log[k] where, M = Mooney viscometer torque in Mooney units (MU); a = slope of least squares regression line; log [k] = intercept of least squares regression line; t = time in seconds.

This construction is illustrated in figure 2. Note that equation 1 can also be expressed as:

(2) M = k [t.sup.a]

In this article, k is referred to as the intercept, and a as the slope.

Star branched butyl-IIR study

The SB-IIR study was motivated by the need to develop a process control tool for the star content of the product. During the R&D phases of SB-IIR, the weight fraction of star polymer in the product was measured directly by GPC. Because the star polymer has a molecular weight much higher than the base IIR elastomer, it is straightforward to characterize by GPC. However, GPC analyses are notoriously difficult in terms of repeatability and reproducibility, and analysis turnaround times are often long. Thus, GPC testing, which is ideal during the R&D phase, is not well suited for quality control. With the advent of commercial production of SB-IIR, there was an incentive to investigate an alternate to GPC for measuring star content.

Rheological tests were considered as logical candidates for indirect measurement of star content. Because of their molecular size and shape, the star molecules profoundly affect the rheology of SB-IIR, even at low weight fraction in the polymer. They contribute to both the viscous and elastic responses of the product; but because of their molecular weight, they contribute more significantly to the elastic response. In this respect, the star molecules differ markedly from the linear IIR molecules. During the R&D phase of SB-IIR it was learned that process constraints fixed some structural parameters, such as star arm length and number of arms per star. It was expected, therefore, that measurement of two independent rheological parameters would be adequate for process control of star content. Since Mooney viscosity had already been identified as a specification parameter and a quality control test for SB-IIR, MLR promised to be an ideal complement to it.

For the SB-IIR study, pilot plant samples were obtained with varying star content and average molecular weight. All samples were carefully characterized by GPC to determine star content. These samples were then analyzed for Mooney viscosity and MLR. The equipment and procedure were the same as that described in the EPDM study, with the following exceptions: 1) All samples were milled for two minutes at 120[degrees]C immediately before testing, 2) the relaxation was started after 10 minutes of run time instead of four, and 3) relaxation continued for 60 seconds instead of 100. Parameters chosen from the MLR curves were slope and intercept. This data were finally reduced to a model for prediction of star content, utilizing slope and Mooney viscosity (ML -- 1+10 (125[degrees]C)) as predictor variables. The resulting contour map is shown in figure 3.

Discussion

EPDM study

In the EPDM study, the MLR curves followed a power law relationship closely for all the samples in both phases 1 and

2. This finding indicated that the stress relaxation behavior of the EPDM samples could be completely characterized by a maximum of two parameters, i.e., slope and intercept.

Since the relaxation spectrum for the EPDM samples could be fully characterized by the slope and intercept parameters, and since there was a desire for simplicity in this quality control application, the investigators postulated that perhaps a simple combination of these parameters would produce a single parameter that would retain most of the sensitivity to rheological variation without giving up too much power to discriminate between sources of variation. While this is an oversimplification, it cannot be ruled out on principle for manufacturing systems with limited degrees of freedom. A number of empirical constructions were attempted, and one that showed promise was identified. Geometrically, this parameter represented the area (in MU-seconds) under the relaxation curve from 1 to [t.sub.f] seconds, and was thus named MLR area. Mathematically, MLR area is a combination of the slope (a) and intercept (k) parameters, and is obtained by integration of equation 2:

(3) [Mathematical Expression Omitted] where, after integration,

(4) [Mathematical Expression Omitted]

For all the data reported in the EPDM study, a value of 100 seconds was used for [t.sub.f]. There is no theoretical justification for the MLR area parameter. Heuristically, it gives weighting to both the slope and intercept parameters, but its primary justification is simply good performance against the criteria of repeatability and correlation to [t.sub.sr]. Other empirical combinations of the slope and intercept parameters were investigated, but none was found with a superior combination of correlation to [t.sub.sr] and good repeatability.

A statistical analysis of the data from the EPDM phase 1 data set is presented in tables 1 and 2. Table 1 shows that MLR area has better repeatability than either [t.sub.sr] or slope, and repeatability equivalent to intercept, expressed as coefficient of variation. Table 2 shows that MLR area correlates better with [t.sub.sr] than either slope or intercept. The superior precision of MLR area is not fully understood, but apparently the effects of stochastic instrumental and procedural errors have offsetting effects on the slope and intercept. Mathematical models of some hypothetical error sources, such as stick-slip friction of the Mooney machine rotor, did predict this type of error offsetting, although no physical evidence for the actual presence of these error sources has been obtained.

On the basis of the phase 1 results, MLR area was selected as a candidate parameter to replace [t.sub.sr]. It is important to note that the significance of MLR area in predicting [t.sub.sr] is not compromised when Mooney viscosity is added into the model. This is illustrated by the F-ratio values in table 2. Since [t.sub.sr] has been validated as an independent measure of processability over the years, this finding illustrates that MLR area also supplies useful processability information supplemental to Mooney viscosity.

After the selection of MLR area as the best candidate test to replace [t.sub.sr] for quality control, phase 2 of the study was conducted. From table 3 it can be seen that this round robin among four laboratories found a reproducibility standard deviation (SR) of about 4% for MLR area, which is reasonably good for rubber tests in general, and more importantly, compares favorably to the [S.sub.R] for [t.sub.sr].

Table 3 - MLR interlab crosscheck-test method
  reproducibility
Parameter          Units             [S.sub.R]   C.V. (%)
Monney viscosity   MU                   0.64       1.3
MLR area           MU-s                18.67       4.5
Slope              log{MU]/log [s]      .037       6.1
Intercept          MU                   0.66       3.1
[t.sub.sr]         S                    13.4      27.1

Following completion of the Phase 2 study, a process was initiated to systematically replace [t.sub.sr] testing with MLR area. This conversion process should be complete by the time of this publication.

SB-IIR study

In the SB-IIR study, the MLR curves for high star content polymers were found to follow a power law relationship at relaxation times of 4-60 seconds. Specifically, it was reasoned that parameters taken from the early part of the curve (prior to four seconds) would represent small molecule effects, while parameters taken from the later part of the curve (4-60 seconds) would reflect larger molecules. Since the goal of this study was to predict the SB-IIR star content, the later part of the curve was considered to be the more important. Based on this principle. the power law model slope from 4-60 seconds was chosen to represent the star molecules.

The analysis of the data from figure 3 clearly shows that the slope (a), taken in conjunction with Mooney viscosity, is a good predictor of weight fraction stars. Some of the other MLR parameters that were evaluated also correlated with star content, but to lesser degrees. The relationship between star content, slope and Mooney viscosity is illustrated on figure 3 as a contour map.

Once slope was identified as the preferred parameter for prediction of star content, its precision was analyzed. Based on > 100 repetitions of a standard control polymer with multiple technicians and three Mooney viscometers, the standard deviation was determined to be 0.013 for slope, which can also be expressed as ~ 1 wt% star content for a typical SB-IIR.

It was of prime interest in this study to determine the overall quality of the correlation as well as the repeatability of the method. This was accomplished by constructing a quadratic model to predict weight fraction stars from Mooney viscosity and slope. Using linear least squares regression, the parameters of the model were estimated from the data, and then a standard analysis of variance (ANOVA) was employed to determine the standard error of the estimate. For this data set, the standard error of the estimate was estimated to be 0.6 wt.% star content. This unexpectedly good result may have been the result of generating this data with a single technician on a single Mooney viscometer.

The results were sufficiently good to permit MLR slope to be selected as the quality control parameter of choice. Subsequently, as SB-IIR entered commercial production, this measurement was introduced into the quality control laboratory of the manufacturing plant and put into use as a primary process control tool.

While the SB-IIR study was underway, the MLR area parameter was being concurrently evaluated in the EPDM study. Following successful development of MLR area as a quality control tool, interest arose in evaluating it as an alternative to slope for SB-IIR. This evaluation is planned, but has not yet been conducted.

After successful completion of the EPDM study, it was anticipated that MLR area might find good use as a quality control tool for other elastomers. This, in combination with a desire to capitalize on the potential for industry standardization, led to proposing the MLR area parameter to ASTM D 11.12 as a candidate for inclusion into the D 1646 standard. After review in 1993, the full society voted to recommend inclusion of the MLR area, along with power law slope and intercept, in the power law decay model.

Summary

Elastomers manufacturers have long sought suitable quality control tools for the processability of their products. While much is known about the relationship between the molecular structure of elastomers and their processability, direct characterization of molecular structure is usually not suitable for process control. Rheological tests, can, and have been used extensively in lieu of direct molecular characterization for process control, but many of them also suffer from high cost, long analysis times, or poor repeatability or reproducibility. Mooney relaxation testing has historically been constrained by low sensitivity and poor repeatability and reproducibility, but recent improvements in test equipment have dramatically improved this situation, prompting a reevaluation of Mooney relaxation for process control.

The results of two studies are presented that address very specific quality control objectives. The first is a study aimed at replacing a proprietary stress relaxation test long used for EPDM manufacture with MLR. The second involves a newly commercialized product, star-branched butyl (SB-IIR), and examines the potential for MLR as a control tool for the weight fraction of star molecules in the SB-IIR.

The EPDM study concluded that MLR offered adequate sensitivity to detect process variation beyond that detected by Mooney viscosity, and that MLR offered good repeatability, fast analysis time and low cost. A new MLR parameter, MLR area, was identified that gave good repeatability and a high degree of correlation to a proprietary stress relaxation test used for process control. Because of the favorable findings, the replacement of the proprietary stress relaxation test with MLR is in progress. Details of the MLR area parameter are included in this article. MLR area was submitted to, and subsequently adopted by ASTM D 11.12 for inclusion in the Mooney test standard ASTM D 1646-93a.

The SB-IIR study found that the MLR slope parameter, combined with Mooney viscosity, provided the information required to predict polymer star content. This parameter was found to have adequate repeatability, and when combined in a model with Mooney viscosity, was able to predict star content with a standard error of ~1 wt. Consequently, this parameter has been adopted for commercial quality control of SB-IIR.

References

[1.] P.J. DiMauro and J.A. Sezna, "New instruments for the rubber industry," paper presented at Rubbercon 1993. Feb. 8-10,1993, New Delhi, India. [2.] G. Schramm, "Rheological testing of raw rubber to indicate links between their molecular structure and their processability," paper presented at ASTM D-11.12 meeting, San Antonio, TX, Dec. 6, 1990. [3.] R. Koopman and G. Marwede, "Rubber testing and its role in quality assurance," paper presented at IRC '91, Essen, Germany, Nov. 11, 1991. [4.] D. Hasman, "Theory, and merit of stress relaxation testing and a review of the dynamic stress relaxometer," paper presented at ASTM D-11.12 meeting, San Antonio, 7-X, Dec. 6, 1990. [5.] P.J. DiMauro, J. deRudder and J.P. Etienne, "New rheometer and Mooney technology," Rubber World, January, 990. [6.] H. Burhin, W. Spreutels and J. Sezna, "MV2000 Mooney viscometer: Mooney relaxation measurements on raw and compounded rubber stocks," ACS Rubber Division, Detroit, October, 1989.