Effective processability measurements of acrylonitrile butadiene rubber.

In 1990, many members of the Rubber Manufacturers Association recognized that the traditional tests used for the quality assurance of nitrile rubber were not sufficient to ensure that each production lot would process consistently in established formulations. Shipments from different lots of the same nitrile rubber grade were found to process differently despite having the same Mooney viscosity value. The RMA polymer suppliers and seal manufacturers formed the Processability Task Group in 1990 to complete the following mission:

"To identify a test (or tests) that could be utilized for monitoring the consistency of processing characteristics of synthetic elastomers, with specific emphasis on nitrile polymers."

In the initial phase of this work, several instruments were evaluated to determine which would be most useful in measuring processing characteristics. From the results of this initial phase, the task group decided that further experimental work would focus on the RPA 2000 rubber process analyzer, because of its versatility, and the MV2000 Mooney viscometer with stress relaxation, because of its relative simplicity.

The rubber process analyzer

The RPA 2000 is a special dynamic mechanical rheological tester (DMRT) designed to measure the dynamic properties of raw polymers, uncured compounds and final cured compounds. The RPA strains a sample in shear by oscillating the lower the sinusoidally in a sealed, pressurized cavity. Oscillation frequency can be set from 0.1 to 2,000 cycles per minute (cpm). The magnitude of the lower the movement is set to an angle of oscillation which generates the required strain on a sample. The lower the can oscillate from [plus or minus] 0.05 to [plus or minus] 90.00 degrees of arc. This angular oscillation corresponds to a strain of [plus or minus] 0.7% to [plus or minus] 1,256%. The very high strain can be applied to an uncured specimen with very repeatable results because of the sealed pressurized cavity design. The upper the is attached to a torque transducer. Figure 1 shows a torque response from the sinusoidal strain applied to the rubber specimen. A Fourier Transform operation is used to convert the complex torque response S* into the elastic torque S' (in-phase with the applied strain) and the viscous torque S" (which is 90 [degrees] out-of-phase with the applied strain). The RPA 2000 analyzer is the only DMRT which can test with good repeatability and reproducibility at very high strains. Other DMRTs do not have sealed pressurized sample chambers. The material at the edge of an unsealed sample cavity can distort under high strains in a non-repeatable way. This will result in poor performance. An RPA 2000 sample is confined at the edge and prevented from distorting (ref 1).

Another important practical advantage of the RPA is ease of use. It is much easier to load a sample into the RPA than a conventional DMRT, which requires more careful sample preparation. The RPA can be used in routine QC as well as research and compound development. Most DMRTs require careful setting of the gap between the dies or plates to ensure the sample makes contact with both dies. With the RPA, you simply load a sample and then walk away because excess sample is automatically squeezed out of the sample cavity.

Various tests arc possible on raw synthetic rubber samples in which frequency, strain and temperature of the test are changed. From these varied test conditions, the RPA provides elastic torque S', viscous torque S", storage modulus G', loss modulus G", real dynamic viscosity [Etha]', complex dynamic viscosity [Etha]*, tan [Delta] (calculated from G"/G") and other outputs. The RPA is routinely used in the quality control of synthetic elastomers. Figure 2 shows how the RPA can distinguish two SBR 1006 samples which have the same Mooney viscosity but quite different uncured tan [Delta] values.

Mooney viscometer with stress relaxation

The MV 2000 Mooney viscometer with stress relaxation is widely used in the rubber industry because it can measure Mooney viscosity as well as stress relaxation decay rates. Stress relaxation tests are run on the same specimen after completing the viscosity measurement. For example, an ML 1+4 test (large rotor, one minute preheat, four minutes run) can be performed on a raw rubber sample followed by a two minute stress relaxation at the end of the test. The total test time would be seven minutes. The Mooney stress relaxation can be performed automatically after the "final" Mooney viscosity measurements by very quickly stopping the rotation of the rotor and measuring the power law decay of the Mooney viscosity output with time. Mathematically, this power law decay is described in this equation:

M = k [t.sup.-Alpha] (1)

Where: M is the torque value in Mooney units; k is the torque value at one second; t is the time in seconds; [Alpha] is the rate of relaxation (the slope of the relaxation function). Also in a log log plot, this expression can be shown as:

Log M = -[Alpha] log t + log k (2)

The stress relaxation slope [Alpha] is commonly used as a measure of the stress relaxation in raw rubber and rubber compounds. Another Mooney stress relaxation parameter which can also be used is %M rt 30 or percent Mooney torque retained at 30 seconds after stopping the rotor. This expression is given as:

%[M.sub.ret 30] = (RV/MV) . 100%

Where: MV is final" Mooney viscosity at five minutes (1+4); RV is relaxation "viscosity" at 5.5 minutes (30" after the rotor is stopped).

Mooney viscosity is still the most common test property used in the rubber industry today. However, it is not unusual for two commercial synthetic elastomers to have the same Mooney viscosity but quite different Mooney stress relaxation decay rates. An example of this is shown in figure 3 where two EPDM elastomer samples have the same Mooney viscosity but quite different stress relaxation rates (ref 2).

Experimental

Phase 1

In January, 1993, Monsanto Instruments (now Alpha Technologies) was invited to participate in the RMA processability task group. The initial goal was to determine if the RPA could distinguish differences in viscoelastic properties of nine carefully selected raw nitrile rubbers which differed in the following characteristics:

* Average molecular weight;

* molecular weight distribution;

* bound acrylonitrile content;

* gel content.

Figure 4 shows an RPA comparison of three nitrile polymers with the same Mooney viscosity but different viscoelastic profiles as indicated by the differences in the uncured tan [Delta](which is simply the ratio of the viscous response divided by the elastic response). With increasing applied frequency (and increasing shear rate), this ordinal relationship changes.

Figure 5 gives the complex modulus G* response from an RPA frequency sweep which relates very well to differences in average molecular weight. On the other hand, figures 6 and 7 show how the uncured loss modulus response G" changes from differences in MWD as seen in a frequency sweep and a temperature sweep respectively.

Figure 8 shows the uncured storage modulus response from a frequency sweep for two nitriles which differ primarily in percent bound acrylonitrile content (% ACN). A similar response is also observed in figure 9 which gives the G' response from a temperature sweep. The higher bound ACN content results in a steeper slope of the G' response from either a frequency sweep or a temperature sweep. These two polymers have very similar Mooney viscosity values.

Lastly, there was concern regarding the test sensitivity of the RPA to gel. As a result, one of the submitted nitrile polymers contained a small quantity of gel (the other polymers were gel free). Figure I 0 shows the high test sensitivity of the RPA to the presence of gel. The gel present causes the uncured tan 8 to be significantly lower than the other nitrile rubbers tested. This is because the gel present causes the elastic response to be significantly higher due to the formation of a loose network.

From this work in phase 1, it was concluded that the RPA was effective in distinguishing nitrile polymers which a, in any of the four stated characteristics discussed earlier, These differences were, however, somewhat large in magnitude. The next phase of experimentation was designed to evaluate the ability of the RPA to distinguish very subtle differences among nitrile polymers.

Phase 2

In the first half of 1994, the Bayer Corporation prepared a series of nitrile polymers (Al through A6) in which all had the same Mooney viscosity of 60 MU (ML 1+4 @ 100 [degrees] C) and the same bound ACN of 28%. A description of these polymers is given in table 1. Some of these polymers were obtained through latex blending. The following polymer comparisons were made to evaluate statistical test sensitivity to subtle differences between selected pairs. The pairs are as follows:

* Polymer Al vs. A2 - A comparison established to measure the effects from latex blending to reach a given Mooney target vs. "normal" production.

* Polymer A2 vs. A3 - A comparison of effects from bimodal molecular weight distribution vs. normal distribution (unimodal distribution).

* Polymer A3 vs. A4 - A comparison of effects from blending normal molecular weight distribution vs. bimodal distribution NBR.

* Polymer A2 vs. A4 - A comparison of effects from blending bimodal distribution vs. normal distribution NBR (a more subtle difference.)

* Polymer A5 vs. A6 - A comparison of a documented "good" processing lot of NBR vs. a "poor" processing lot of NBR with the same Mooney viscosity and % ACN content.

Table 1

Identification     Bound           Mooney      Sample description
                 acrylonitrile   viscosity

A1                  28             60         Blend: (ML = 50) + (ML
                                               = 70) 28% ACN, ML
                                                    (1 +4) = 60

A2                  28             60         Normal MWD, 28% ACN
                                                   ML(1+4) = 60

A3                  28             60         Bimodal MWD, 28%
                                               ACN ML(1+4) = 60

A4                  28             60         Blend of A2 and A3

A5                  28             60         Good mixing behavior

A6                  28             60         Poor mixing behavior

Statistical test "sensitivity" was calculated by a method analogous to signal-to-noise ratios used in the electronics industry. The "signal" in this calculation is the difference between the maximum and minimum of the test measurement averages for the groups of polymer variations being compared. The "noise" is the pooled standard deviation (Sp) of the test from replicate testing (ref. 3).

"Sensitivity" = (Max X - Min X) / Sp (4)

A high "sensitivity" value indicates the subject test parameter is more sensitive to the polymer differences in the pair. A test parameter that produces a greater "signal" from the viscoelastic differences between the two polymers and/or has reduced noise due to random test variation will have increased sensitivity and discriminating power for the selected polymer pair. Normally the S/N sensitivity values should be at least six or higher for a test parameter to have good discernability.

Table 2 gives the RPA test parameters and test conditions which gave the highest test sensitivity values for the five nitrile rubber pairs studied. Figures 11 through 14 give a graphical picture of the repeatability and discriminating power for the RPA tests. As can be seen, the RPA effectively separated all five pairs even though some of their viscoelastic. differences were quite subtle. High strain conditions (applied strains at 140% or higher) overall gave the best test sensitivities for the five polymer pairs. Some other processability instruments could not discern differences with all five pairs. The MV 2000 with stress relaxation was only able to show good test sensitivity for pair number 1 (A1 vs. A2) and pair number 3 (A3 vs. A4). Table 2

Polymer        Property        RPA optimal       RPA statistical
comparisons    comparison      test condition    test sensitivity
                                                 (signal-to-noise
               and response                      ratio)

A1 vs. A2      Effects from      0.5 deg. arc        16.8
               laxtex
               blending to a     strain, 6 cpm,
               given             G' response
               Mooney target
               vs. "normal"
               production.

A2 vs. A3      Effects from      10 deg. arc         23.5
               bimodal MW        strain, 6 cpm,
               distribution vs.  tan [Alpha]
               normal
               distribution      response

A3 vs. A4     Effects from       10 deg. arc         13.6
              blending
              normal MW          strain, 6 cpm,
              distribution       tan [Delta]
              NBR                response
              with bimodal
              distribution

A2 vs. A      Effects from      90 deg. arc            5.4
              blending          strain, 6 cpm,
              bimodal           tan [Delta]
              distribution vs.  response
              normal
              distribution

A5 vs. A6     Good mixing       50 deg. arc           57.5
              vs.poor          strain, 6 cpm,
              mixing           S" response
              processing
              NBR

Pair number 5 (A5 vs. A6) was of particular practical interest for the task group because it addressed observed differences in factory internal mixing with the two nitriles even though they have the same Mooney viscosity. (Processability differences among NBR polymers is the reason the task force was formed.) Using the viscous torque S" response at 50 degrees arc angular deformation (700% strain amplitude), the RPA gave a test sensitivity (signal-to-noise ratio) of 57 which was higher than any other processability test instrument (the best out of six different test instruments).

Phase 3

In September, 1996, the task group identified two lots of commercially produced nitrile rubber with similar Mooney viscosity and bound acrylonitrile levels that were found to process differently. These nitrile polymers were identified as NBR 105 (for the "poor" processing) and NBR 106 (for the normal processing). NBR 105 is believed to have a broader, perhaps a bimodal molecular weight distribution. Also it was observed that the NBR 105 had small white moisture domains which introduced increased heterogeneity. Sufficient quantities of these two lots of rubber were set aside for a RMA interlaboratory crosscheck which included six laboratories with RPAs and seven laboratories with MV2000 Mooney viscometers with stress relaxation. The participating laboratories with the RPA all used polymer I configuration given in table 3 (all RPAs possess this configuration as a read only file). This configuration consisted of a conditioning test (timed test), a low strain sweep, a frequency sweep and a high strain sweep.

Table 3 - RPA test configuration polymer 1 for uncured properties

Subtest 1            Conditioning
            Time          5           m.m
            Temp.        100          [empty] C
            Freq.         30          cpm
            Strain       0.2          deg.
Subtest 2            Strain sweep
            Temp.        100          [empty] C
Strain      Freq.         6           cpm
deg. arc

0.1
0.2
0.5
Subtest 3             Frequency sweep

            Temp.         100           [empty] C
S'          Strain        0.5           deg.
Frequency
cpm
2
5
10
20
50
100
200
500
1,000
2,000                  Strain sweep
            Temp.          100        [Empty] C
Subtest 4   Freq.           6          cpm

Strain
deg.
0.5
2
5
10
20
50

Tables 4 and 5 show the calculated precision statement from the RPA interlaboratory cross-check results. These tables only show calculations for three pre-selected frequencies (5, 200 and 1,000 cpm) and three strains (5, 20 and 50 degrees arc). (This is similar to the new proposed ASTM Standard for the RPA which calls for frequencies of 6, 120 and 1,200 cpm.) Table 6 gives the precision statement for the MV 2000 Mooney viscometer test results (including Mooney stress relaxation).

[TABULAR DATA NOT REPRODUCIBLE IN ASCII]

The terms used in these tables are based on ASTM definitions which are defined (ref. 4):

* Sr - short term (same day) repeatability standard deviation

* r - short term repeatability, an established value, below which the absolute difference between two "within-laboratory" test results may be expected to lie, with a probability of 95%.

* (r) - repeatability estimate expressed as percentage of the mean of the property for which the estimate was obtained.

* Long Term r - one week repeatability, an established value, below which the absolute difference between two "within-laboratory" test results may be expected to lie, with a probability of 95%.

* Long term (r) - one week repeatability estimate expressed as percentage of the mean of the property for which the estimate was obtained.

* SR - reproducibility standard deviation.

* R - an established value, below which the absolute difference between two "between-laboratory" test results may be expected to lie, with a probability of 95%.

* (R) - reproducibility estimate expressed as percentage of the mean of the property for which the estimate was obtained.

* Short term % CV same day repeatability expressed as a coefficient of variation which is 100 times the short term repeatability standard deviation divided by the mean.

* Long term CV - one week repeatability expressed as a coefficient of variation which is 100 times the long term standard deviation divided by the mean.

From tables 4 and 5, coefficients of variation (CV) as low as 0.33% were calculated for the RPA. From table 6, CV values as low as 0.6% were calculated for the MV2000 Mooney viscometer with stress relaxation.

[TABULAR DATA NOT REPRODUCIBLE IN ASCII]

From equation 4, the statistical test sensitivity values were calculated for the RPA and MV2000 test parameters from the crosscheck data. Test parameters must have a calculated sensitivity of six or higher to show effective discriminating power. The RPA viscous torque S" response at 200 arc strain gave the highest test sensitivity of 36. Also, the other RPA high strain test conditions showed excellent discerning power with sensitivity values over 30. The viscous torque response at 200 cpm and 0.5 degrees arc strain gave a test sensitivity of 25 S/N.

The Mooney stress relaxation parameters, while not as sensitive as the RPA, still showed very good discerning power between the NBR 105 and NBR 106 elastomers. The percent Mooney torque retained 30 seconds after the rotor is stopped (%[M.sub.ret] 30) was the tenth highest in test sensitivity at 14 S/N. The Mooney stress relaxation slope a was the sixteenth most sensitive parameter at 6.6 S/N. The final Mooney viscosity (ML 1+4) gave an unacceptable test sensitivity of only 1.4 which means this parameter can not be used to distinguish processing characteristics between these two raw elastomers. These differences between the RPA and MV2000 are consistent with the findings reported by H. Kramer and J. Schnetger in 1996 (ref. 5). The average of cumulative results from seven participating laboratories for Mooney stress relaxation properties for the NBR 105 and NBR 106 raw elastomers is illustrated in figure 15.

NBR 105 and 106 elastomers were also tested in the ASTM recipe D 3187 which is given in table 7. Figure 16 compares the average Mooney viscosity and Mooney stress relaxation properties which are imparted to this compound by SGBR 105 and NBR 106, respectively. When the compound properties were compared in laboratory testing, it was found that the compound based on NBR 105 had a 66% longer black incorporation time (five minutes vs. three), a 10% to 20% higher the swell, higher mill shrinkage and poorer extrusion surface appearance than what was imparted by NBR 106. These laboratory results help quantify the reported poor processing of NBR 105.

Table 7- ASTM D 3187 test recipe for NBR

Ingredient                            phr
 NBR                                  100
 Zinc oxide                           3.0
 Sulfur, coated (2% MGC[O.sub.3])     1.5
 Stearic acid                         1.0
 Carbon black, ASTM SRB-B4 (N330)    40.0
 TBBS accelerator                     0.7
 Total                              146.2

Figure 17 illustrates the stress relaxation values obtained from multiple production lots of nitrile polymers (based on the same polymerization recipe) with controlled Mooney viscosity ranging from 35 to 150. As demonstrated in figure 15, polymers with similar viscosity can exhibit a fairly wide range of stress relaxation values, and can process very differently.

Figure 18 shows company members of the RMA which participated in these RPA and Mooney viscosity/stress relaxation interlaboratory cross-checks.

Conclusions

* The RPA 2000 is very effective as a processability tester in discerning differences among nitrile polymers from all three phases of these RMA studies.

* While the MV2000 is found to have less than half the test sensitivity of the RPA, it is still recognized as an effective processability tester in discerning viscoelastic differences among nitrile polymers.

* The RPA can distinguish nitrile polymers that differ in average molecular weight, molecular weight distribution, bound acrylonitrile content and gel content.

* Laboratory results on compounded stocks substantiate differences observed in raw polymer tests, and clearly indicate these elastomers should be characterized and graded with more sophisticated methods than Mooney viscosity alone.

Recommendations

The members of the RMA Processability Task Group voted unanimously to accept the following recommendations.

* The RPA 2000 instrument, using the standard preloaded Polymer 1 testing profile, should be recognized as an advanced procedure for measuring the consistency of nitrile polymers for processing characteristics.

* The MV2000 viscometer (a more prevalent instrument in the rubber industry), also provides an excellent low cost opportunity to monitor the consistency of nitrile polymers for processing characteristics, using the Mooney stress relaxation test method.

References

[1.] J.S. Dick and H.A. Pawlowski, Rubber characterization by applied strain variations using the rubber process analyzer, "Rubber World, January, 1995.

[2.] MV H. Burhin, W. Spreutels and J.A. Sezna, "2000 Mooney viscometer - Mooney relaxation measurements on raw and compounded rubber stocks, " presented at the Detroit ACS Rubber Div. Meeting, Oct. 1989.

[3.] J. Mandel, R.D, Stiehler, "Sensitivity - a criterion for the comparison of methods of test, "J. Research Natl. Bur. Standards, 53, No. 3,155-159, September, 1954.

[4.] ASTM D4483, "Standard practice for determining precision for test method standards in the rubber and carbon black industries, " 1997 Annual Book of ASTM Standards,, Vol 09.01.

[5.] H. Kramer arid J. Schnetger, "Laboratory measurement of viscoelastic parameters of relevance to the processing of rubber compounds, "Bayer AG, May, 1996 ACS Rubber Division Meeting, Montreal