Testing and analysis of rubber-to-metal bonded parts.

The robust nature of rubber-to-metal bonded parts qualifies them for applications in critical systems with hostile environments. If a part fails, it's imperative to determine the root cause of the failure. The constituents of rubber compounds make analysis difficult using traditional methods such as reflectants Fourier Transform Infrared Spectrometer (FTIR) or energy dispersive x-ray (EDS). Analysis of the system is often more difficult when failure occurs at the adhesive bond interface. The author, using x-ray photon scattering spectroscophy (XPS) and an ISI SX-40 scanning electron microscope (SEM) in his analysis, was able to deduce the most likely cause of failure to be premature crosslinking of the adhesive. By eliminating other possible causes of unbond, such as contamination, a more robust manufacturing process can be developed.

Carbon black filled rubber parts may challenge many analytical methods because the compounded rubber may include up to twenty individual ingredients. Often, after mixing, it is difficult to extract ingredients from the polymer base due to their interactions with other ingredients such as carbon black.

Analysis of extracted ingredients may give erroneous results due to contamination by other ingredients from the compound. The carbon black components in the blend often interfere with spectural analysis methods because carbon black tends to absorb spectural frequencies, such as infrared.

Testing of bonded rubber-to-metal parts amplifies the problems two-fold. The bonding adhesive is also a multi-component polymeric blend, much like the rubber. Also involved are contaminants associated with the substrate. For example, a stamped steel insert may be contaminated with hydrocarbons. If the metal is porous, oil may bleed out of the metal grain over time and undermine a bond system. Or, if the oil is present initially, it may prevent adhesive wet-out, thus inhibiting adhesion to the substrate.

Traditional analysis lacking?

Some techniques such as reflectants FTIR and EDS would not be adequate for testing rubber bonded to metal. EDS, using a windowless system, does not have sufficient resolution to detect lower weight elements. Typically, with a windowless system, resolution begins to soften around sodium.

Techniques such as FTIR are subject to error due to occlusion of peaks by stronger signals. Both of these methods tend to penetrate the samples to a degree that an accurate compositional analysis of the surface cannot be obtained. This information and data are most critical in bond failure analysis.

XPS, SEM most definitive

To gain the most definitive picture of failed bonded surfaces, SEM and x-ray photon scattering methods have proven the most valuable.

Using SEM techniques, an excellent picture of the failed surface may be obtained. With bonded rubber compounds, often both the adhesive and the rubber are black, making it difficult to determine where the rubber begins and the adhesive ends. Since SEM images are produced by bouncing electrons off a surface, at a given angle to a detector, artifacts as small as one micron are detectable.

This is sufficient to detect differences between adhesives and rubber, or detect other materials on the surface of a part. This information is crucial in determining how a part failed. A good visual representation of the surface can help determine whether the bond failed adhesively or cohesively. It can also be used to determine to what degree a part may have been bonded before it failed.

XPS, or ESCA, is an excellent method for determining chemical composition of a surface. This method penetrates the surface of the sample to a depth of only 50 microns. XPS measures the binding energies of the atoms present on the surface, yielding a very distinctive signature of the atomic species present. XPS is sensitive enough to generate signals that provide both a quantitative and qualitative picture of the surface composition over a wide elemental range. Computer analysis techniques convert these signals to atomic percentages to allow reconstruction of the various materials present.

Case study example

The following case study analyzes black rubber bonded to a steel part. The part was exposed to a hostile environment (hydrocarbon fluids at 150 [degrees] C) for a relatively short period of time before failure occurred. Parts of this type are expected to last approximately eight years; this part failed in under six months.

Initial visual analysis indicated failure primarily in an adhesive mode. However, no apparent cause of the failure was easily identifiable. Gaining a clear understanding of the cause of the adhesive failure at this point was difficult because the part was saturated with hydrocarbon oil.

An ISI SX-40 scanning electron microscope was used to analyze the mating failed surfaces and a cross section of the bonded surface. The photographs in figures 1 and 2 show the failed surfaces magnified up to 150 times. These pictures show surfaces were relatively smooth and interacted very little. What appeared to be pulled-out material was present on one surface with fishers on the opposite face. This would indicate the part was partially bonded before it failed.

The cross section (figure 3) of the remaining bonded area indicated that under normal conditions, the adhesive could not be distinguished from the rubber at a magnification of 45 times. This would indicate a high degree of commingling is necessary between rubber and adhesive to achieve a good bond.

At this point the unbond was believed to be caused by an overbake of the adhesive. In this situation, the poor bond is due to premature crosslinking of the adhesive prior to the application of rubber. The crosslinking of the adhesive causes mechanisms for adhesion to be compromised and inhibits physical interaction.

To confirm this, we subjected both a control part, with bond failure due to adhesive over bake, and the failed part to analysis by XPS. Only failed mating surfaces were examined by this method. Typical areas scanned by this method were 1 mm x 3 mm. It is not necessary to neutralize with the electron gun due to only minimal charging effects. Only a single sample. was tested for each condition.

Samples were scanned in a region between 0 and 1,100 eV. Multiplex scanning of all significant peaks seen in the survey was accomplished. Binding energies associated with all observed photopeaks were referenced to the values for C-C/C-H hydrocarbon species at 284.6 eV. Values for atomic concentrations were determined using PHI software version 4.0. Atomic concentrations were determined using the formulation: Atomic % for element X = Ix

Sx Tx

n Ii

[epsilon] Si Ti

i = I

I = photopeak area S = sensitivity factor T = total acquisition time n = number of photopeaks considered

The XPS analysis of the mating surfaces of the defective part (tables 1 and 2) showed the surfaces to be very similar in elemental composition. Values from the failed part, when compared to the control pan (tables 3 and 4), indicated the hydrocarbon fluid did not affect the analysis. This is most likely because the high vacuum, necessary in this testing technique, removed the hydrocarbon fluid from the surface.

     Table 1 - XPS analysis of failed part,
          rubber side of failed area

Element        Area   Sensitivity   Concentration

         (cts-eV/s)        factor             (%)

C1s           45306         0.296           73.36
O1s           18461         0.711           13.46
N1s             313         0.477            0.34
Si2p           3843         0.339            5.88
Ca2p           1316         1.833            0.37
P2p             342         0.486            0.36
S2p             293         0.666            0.23
Al2p              0         0.234            0.00

     Table 2 - XPS analysis of failed part,
       adhesive/metal side of failed area

Element        Area   Sensitivity   Concentration

         (cts-eV/s)        factor             (%)

C1s           43017         0.296           73.96
O1s           25671         0.711           18.37
N1s             702         0.477            0.75
Si2p           3436         0.339            5.16
Ca2p           2482         1.833            0.69
P2p             559         0.486            0.59
S2p             570         0.666            0.44
Al2p             23         0.234            0.05

     Table 3 - simulated failure unbond area,
                   rubber side

Element        Area   Sensitivity   Concentration

         (cts-eV/s)        factor             (%)

C1s           40822         0.296           73.29
O1s           24327         0.711           18.18
N1s             594         0.477            0.66
Si2p           4297         0.339            6.74
Ca2p           1320         1.833            0.38
P2p             238         0.486            0.26
S2p             161         0.666            0.13
Al2p             92         0.234            0.21
F1s             287         1.000            0.15

     Table 4 - simulated failure unbond area,
                adhesive/metal side

Element        Area   Sensitivity   Concentration

         (cts-eV/s)        factor             (%)

C1s           44106         0.296           72.63
O1s           29689         0.711           20.35
N1s             718         0.477            0.73
Si2p           4155         0.339            5.97
Ca2p            673         1.833            0.18
P2p              77         0.486            0.08
S2p              71         0.666            0.05
Al2p              3         0.234            0.01

The trace elements present in analysis of both surfaces of both parts support the conclusion that the parts failed by the same mechanism. Test results support the idea that both samples failed in a boundary layer between the rubber and the adhesive. Trace elements, known not to be native to the rubber compound, indicate the source to be the adhesive.

This was further supported by the SEM photographs that showed pull-out in the adhesive layer.

Conclusions

From the presence of adhesive components on all mating surfaces tested, we can effectively eliminate contamination as the cause of the unbond. The presence of the adhesive and the pulled-out material present on the failed part would indicate the part may have been weakly bonded. Finally, the evident lack of adhesive and rubber interspersion would indicate one or both components had achieved sufficient viscosity to resist flow. The combined evidence tends to support our initial observation that all adhesive had prematurely crosslinked. This data will allow us to act correctly on the cause and help produce a more robust part and process.

James Jacks has been with Acadia Polymers for the past four years. His job responsibilities have included materials development and analysis. He has been associated with the Virginia Polytechnic Univ. through a grant program he coordinates. The primary focus of the grant has been the study of materials and bond integrity in application environments.