Web http://www.jcse.org/

Volume 2 Paper 27


Predictive Testing of Internally Coated Food Cans

J C P White

Corporate Technologies, Crown Cork & Seal Company Inc. Downsview Road Wantage Oxfordshire OX12 9BP England

Abstract

The use of Electrochemical Impedance Spectroscopy to make short-term predictions of the long-term performance of internally coated 2-piece food cans has been explored. In particular, the effects of thermal processing and of product on the electrochemical and corrosion performance of filled can variables have been investigated. It has been found necessary to use different parameters derived from Equivalent Circuit analysis to predict specific corrosion processes. Although the relative performance of different experimental variables can be predicted using comparatively short-term impedance tests, the power of the methodology was limited by the sensitivity of the data to the inherent variability between replicate samples.

Keywords: impedance, polymer, coating, predictive, packaging, cans

Introduction

The metal packaging industry makes extensive use of organic lacquer coatings to provide internal protection against corrosion of the metallic substrates (tinplate, tin-free steel, and aluminium) by the contents of the can. Recent developments, such as the requirement to change from solvent-based to water-based internal coatings to satisfy environmental legislation and the constant pressure to reduce raw material costs have led to renewed focus on the performance of these coatings. Traditional methods of assessing coating performance involve long-term storage tests which are expensive to set up, and are, by their very nature, time-consuming. Much effort has been expended in trying to develop reliable predictive tests to allow the more rapid introduction of new coatings into commercial use whilst maintaining confidence in their long-term performance.

Electrochemical Impedance Spectroscopy has, for a number of years, been regarded as the method of choice for investigating the behaviour of polymer-coated metals in contact with aqueous electrolyte systems. A number of papers have been published reporting its application to packaging systems [References 1-9 give a small selection of the published literature]. This paper describes some of the work carried out at Crown Cork & Seal on the use of Electrochemical Impedance Spectroscopy (EIS) as a predictive test of coating performance, relating both some of the successes so far, and some of the challenges which remain.

General Experimental Considerations

A number of experimental configurations have been employed in this work on lacquered containers, but typically, a three electrode system has been used, with the container as working electrode, a platinized titanium gauze counter electrode and either an anodised silver wire or an M/10 calomel reference electrode. Packed cans are opened in the conventional manner immediately before testing, and the electrodes inserted through the opening. This, however, has the disadvantage of allowing oxygen ingress, so "in-can" electrodes have been developed, which are mounted in the can side wall prior to filling and processing. Graphite rod counter electrodes and silver wire reference electrodes are used in this case. However, this procedure can cause damage to areas of the internal coating adjacent to the electrode mounting points and inadequate sealing of the cut edge will result in additional initiation points for corrosion.

All impedance measurements were carried out using a Solartron 1255 Frequency Response Analyser with either a Solartron 1286 or 1287 electrochemical interface. The equipment was computer controlled using in-house software, or more recently, ZPlot for Windows (Version 2.1, from Scribner Associates / Solartron). Equivalent circuit analysis was performed using EQUIVALENT CIRCUIT Version 3.96 or 3.99 (from Bernard A Boukamp, University of Twente), and more recently using ZView for Windows (Version 2.1). The equivalent circuits found most appropriate were typically of the type illustrated in Figure 1, where n=0, 1, 2 or occasionally 3. Significantly better fits were obtained from equivalent circuit analysis if all capacitances in the equivalent circuits were treated as constant phase elements. In the case of the coating capacitance, the exponential factor was always close to unity, but significant deviations were often found for the exponential factor of the double layer capacitances. These are presumably caused by diffusion effects, but have so far not been successfully modelled using Warburg impedances.

Results and Discussion

Many food products are subjected to a thermal sterilisation treatment (processing) immediately after filling, with between 15 minutes and 1.5 hours at 121 �C in a steam retort being a typical process, depending on the product. This has obvious implications for the performance of the internal coating, with electrolyte being forced into pores in the coating and initiating corrosion reactions. The effects of different processing times has been studied on 2-piece drawn and wall ironed tinplate food cans, internally spray coated with a water-based epoxy-phenolic coating. The cans were filled with a synthetic vegetable medium (10 g/L sodium chloride, 3 g/L citric acid monohydrate, 13.3 g/L tri-sodium citrate dihydrate, adjusted to pH 5.5), and processing at 121 �C in a laboratory retort for various lengths of time. Representative Bode magnitude plots are shown in Figure 2. There is an obvious decrease in impedance response on bringing the retort up to pressure, which is accompanied by a decrease in pore resistance. There is, however, no significant increase in coating capacitance, (calculated as a ratio of the values obtained after and before processing to allow for different initial values) Figure 3, which is considered indicative of filling of pores with electrolyte. At processing times beyond about 45 minutes there is a further decrease in the impedance magnitude and pore resistance (Figure 4 shows the average of the pore resistance measured after processing), together with a shift in the phase angle minimum and the onset of a rise in the coating capacitance. This is thought to be caused by water being absorbed into the coating, reaching the metal substrate, and initiating delamination, and is substantiated by the onset of visible blistering at about this time.

These marked changes on processing are perhaps to be expected but enable us to gain an indication of the susceptibility of different lacquers to barrier degradation on processing. For example, the then standard solvent-based lacquer was tested against the best available water-based lacquer and two of its developmental precursors (all lacquers were epoxy-phenolic based). Cans were filled with 1% sodium chloride, and impedance spectra run before and after processing at 121 �C for 1 hour. The pore resistance and lacquer capacitance ratio data obtained are given in Table 1.

Lacquer

Pore resistance (post-process) /W

Lacquer Capacitance ratio
(Q processed/Q unprocessed)

Solvent-based

1920

1.00 0.05

Water based

186

1.23 0.18

Precursor A

3800

91 9

Precursor B

57

3.4 2.9

Table 1. Impedance Data for Internally Coated Food Cans Filled with 1% NaCl and Processed for 1 hour at 121 �C

When cans that had been similarly filled and processed were opened after 2.5 months room temperature storage, two distinct corrosion effects were observed. Some cans showed a general underfilm discoloration (rusting), whilst others showed more localised rust spots. The lacquer variables were ranked from 1 (best) to 4 (worst) for these two effects, and the results are quoted in Table 2.

Lacquer

Ranking for localised rust spots

Ranking for general underfilm rusting

Solvent-based

1 1

Water based

3 2

Precursor A

2 4

Precursor B

4 3

Table 2. Pack Test Results for Internally Coated Food Cans Filled with 1% NaCl and Processed for 1 hour at 121 �C, opened after 2.5 months.

There is an obvious correlation between low pore resistance and severe localised rusting, and between general underfilm rusting and increases in lacquer capacitance on processing, the latter presumably due to high levels of electrolyte uptake. Thus for predictive purposes it is necessary to use specific parameters for predicting different modes of failure.

This work was extended to "real" foodstuffs that are often used as model products in long-term storage tests. Chicken soup is typical of a wide range of fatty products, mushy peas are representative of products which often cause an under-lacquer staining of tin or iron sulphide and blackcurrants are used as a model for acidic products (pH 3). Foodcans (a total of 6 replicate cans per lacquer per product), internally coated with solvent-based and water-based epoxy-phenolic lacquers, were initially filled with 1% sodium chloride, left (open) to equilibrate for 24 hours, and their EIS spectra recorded. This procedure was designed to allow an initial test of the consistency of the sample of cans with a simple, reproducible medium. The cans were then washed and dried, filled with the appropriate product, and processed under the conditions in Table 3: -

Product

Temperature

Duration

Chicken soup

121 �C

90 minutes

Mushy Peas

116 �C

110 minutes

Blackcurrants

100 �C

30 minutes

Table 3 Processing Conditions for Model Food Products

Impedance spectra were then recorded for all the can variables, using the product as electrolyte. The cans were emptied, washed and dried, and re-filled with 1% sodium chloride. After a further 24 hours equilibration time, the impedance spectra were re-recorded. The results obtained are quoted in Table 4.

In chicken soup, the results both from tests in the product and subsequently in 1% NaCl show that better performance should be obtained from the solvent-based lacquer, and this is indeed the case. When cans from a storage test run in conjunction with this exercise were opened after 2.5 months, corrosion was observed only in the headspace (the area between the product fill level and the top of the can) and the water-based lacquer showed systematically more corrosion.

At 2.5 months, the cans filled with blackcurrants show very little corrosion damage, and no systematic difference between the water and solvent-based variables. This is in line with the high values of pore resistance and comparatively small rises in lacquer capacitance observed when tests were performed "in product". In addition, a high frequency feature was observed in the Nyquist plots, which is believed to be consistent with film formation over the whole can surface. This would explain why this product is rather less aggressive than would be expected given its low pH.

The tests in mushy peas were complicated by the highly viscous, semi-solid nature of the product, and difficulty was experienced in removing the product from the cans, which may have resulted in some lacquer damage. However, the enormous increases in lacquer capacitances observed when re-tested in 1% sodium chloride indicate a high level of lacquer barrier degradation. After 2.5 months storage, blushing (an opaque appearance caused by water uptake) was apparent on both water and solvent-based lacquers, in addition to severe headspace corrosion, which was more severe in the water-based variable. A small degree of sulphide staining was also present on the can base, to a similar extent for both variables.

Test Product

Lacquer

Post-process tests in product

Post-process tests in 1% NaCl

Lacquer Capacitance ratio

Pore Resistance / W

Lacquer Capacitance ratio

Pore Resistance / W

Chicken Soup

Solvent-based

1.7

1470

1.9

4160

 

Water-based

3.3

46

5.6

49

 

(t-test)

(1.9)

 

(2.3)

 

Blackcurrants

Solvent-based

2.0

4630

1.9

220

 

Water-based

1.3

6760

1.2

650

 

(t-test)

(1.4)

 

(1.8)

 

Mushy peas

Solvent-based

1.3

210

66

1470

 

Water-based

2.4

170

700

64

 

(t-test)

(3.5)

 

(1.7)

 

Table 4. Impedance Data for Model Food Products packed in 2-piece Food Cans with Epoxy-phenolic Internal Lacquer. Based on 6 cans/Variable

One of the problems with this methodology is that opening cans prior to testing prevents the possibility of repeat measurements to monitor the longer-term behaviour. The approach of packing a large number of cans, and then opening a small sample of them for testing on each occasion (e.g. 1 day, 1 week, 1 month, 3 months) is not ideal because of the inherent replicate sample variability [10]. To try to overcome these problems, in-can electrodes (as described in the Experimental Section) were fitted to cans prior to filling with chicken soup and processing. Again, 2-piece foodcans were used, with a water-based lacquer variable and solvent-based control. Impedance spectra were run after 1 day, 1 week and 1 month after filling (but not, on this occasion, prior to filling). The cans were stored at 35 �C for the first month, in order to accelerate corrosion reactions, then subsequently at 25 �C until opening 1 year after filling. Whilst no significant difference was predicted between the performance of the two lacquers (visual examination after 1 year indicated that the water-based lacquer actually performed slightly worse), it was possible to monitor the impedance response with time for individual cans, and predict comparatively good and bad performers. Figures 5 to 8 show the time dependence of the Bode plots of two cans coated internally with a solvent-based lacquer, filled with chicken soup. Can 1 (Figures 5 & 6) was a comparatively bad performer when examined on opening, whilst Can 3 (Figures 7 & 8) was comparatively good. The major change appears to occur between 1 week and 1 month; the "bad" can shows a reduced low-frequency impedance magnitude, and the phase angle minimum at about 10Hz has disappeared. Examination of the time trends of the lacquer capacitance (Figure 9) shows that both cans display similar trends in lacquer capacitance, with Can 3 having a steeper initial fall and a steeper subsequent rise. It is not understood why the lacquer capacitance falls during the first few days of storage. Figure 10 shows the corresponding time dependence of the pore resistance, and it can be seen that Can 1 shows a greater decrease in pore resistance, particularly between 1 week and 1 month, as previously noted in the Bode plot.

A recent study involved testing a range of three developmental white internal epoxy-phenolic lacquers for food can applications. The lacquers were applied at the same film weight (417 60 mg/can) onto 2-piece food cans manufactured from tinplate with two internal tin coating weights (2.24 and 15.1 g/m�). This gave a total of 6 variables, as follows: -

Variable Code

Lacquer Applied

Initial Tin Coating (g/m�)

H1 A 2.24
H2 A 15.1
H3 B 2.24
H4 B 15.1
VC1 C 2.24
VC2 C 15.1

Table 5. Coding of Variables for White Internal Lacquer Investigation

Owing to the large number of samples in the experiment, it was not possible to use the in-can electrodes approach. Therefore the previously standard method of filling sufficient cans to enable batches of five to be tested at each time interval (initial, 1 day, 1 week, 1 month and 3 months after filling) was used. The initial tests were made on unprocessed cans, filled with 1% sodium chloride (the cans were allowed to equilibrate for two hours before testing). All other tests were made on samples taken from a batch which had been filled with chicken soup and processed for 90 minutes at 121 �C. Cans were opened and the reference and counter electrodes inserted immediately before running the impedance spectra. Figures 11 to 16 show the time dependence of the pore resistances and lacquer capacitances, grouped by lacquer type. The most striking observation is the degree of scatter within each variable, though in terms of pore resistance, the higher tin coating weight variables (H2, H4 and VC2) appear to be marginally better than their low coating weight counterparts with the same lacquer. As discussed previously, pore resistance and lacquer capacitance are indicators of the relative importance of different corrosion processes. Plotting one against the other should enable comparison of lacquers for both effects simultaneously, and, eventually, it should be possible to draw regions of "acceptability and unacceptability" on the graphs. Figures 17 and 18 show such graphs for the three variables with high tin coating weights at 1 day and 3 months. At first sight, they merely emphasise the experimental scatter, but differentiation becomes apparent in the 3 month’s data, with variable H2 shows higher pore resistance and H4 more susceptible to electrolyte uptake (higher lacquer capacitance ratios).

Another approach is to examine the parameters directly related to the actual degree of corrosion, i.e. the charge transfer resistances [2]. Because typically two resistor / constant phase element combinations placed in series with each other have been used in the equivalent circuits, the two resistance values have been added to give an overall charge transfer resistance. Figure 19 displays these data for the three variables with high tin coating weight. Some discrimination between variables is beginning to become apparent at 3 months, but not before. A similar conclusion would be obtained by comparing the impedance magnitude at low frequency, because the pore resistances are small compared to the charge transfer resistances. The charge transfer data at 1 month also reveal the superior performance of the high tin coating variables. This is shown in Figure 20 for lacquer A, the corresponding graphs for Lacquers B and C are not displayed, but are similar. The findings of these predictive tests have been borne out by the results of storage tests.

Conclusions

It has been demonstrated that: -

However, the following problems have been encountered: -

Acknowledgements

The Author would like to thank his past and current colleagues within Corporate Technologies, Crown Cork and Seal and in particular, Mr Andrew Speyer, some of whose work is presented here.

References

    1. Tait, W. S., Journal of Coatings Technology, 61, No. 768, 57 (1989)
    2. Tait, W. S., Journal of Coatings Technology, 62, No. 781, 41 (1990)
    3. Ichiba, M., Iwasa, H., Watanabe, T., Proc. 5th Int. Tinplate Conf. 1992, 288
    4. Hollaender, J., Ludwig, E., Hillebrand, S., Proc. 5th Int. Tinplate Conf. 1992, 300
    5. Junges, P., Pennera, G., Billen, L., Darre, J. M., Birck, J. C., Proc. 5th Int. Tinplate Conf. 1992, 425
    6. Willem, J-F., Lamberigts, M, Bastin, P., Proc. 6th Int. Tinplate Conf. 1996, 106
    7. Bugnard, C., Seurin, P., Trichies, N., Junges, P., Proc. 6th Int. Tinplate Conf. 1996, 297
    8. Pezzani, A., Lupi, R., Montanari, A. Cassara, A., Palmieri, A., Materials Science Forum, Vols. 289-292, 259 (1998)
    9. Catala, R, Cabanes, J. M., Bastidas, J. M., Corrosion Science Vol. 40, No. 9, 1455 (1998)
    10. Tait, W. S., Journal of Coatings Technology, 66 No. 834 pp 59 (1994)

 

Figure 1. Typical Equivalent Circuit

 

Figure 2. Bode magnitude plots for foodcans filled with synthetic vegetable medium, and processed at 121 �C for various times.

Figure 3. Lacquer Capacitance ratio for foodcans filled with synthetic vegetable medium, and processed at 121 �C for various times.

Figure 4. Pore resistance after processing for foodcans filled with synthetic vegetable medium, and processed at 121 �C for various times.

Figure 5. Bode magnitude plot for poor-performing can packed with chicken soup, measured using in-can electrodes.

Figure 6. Bode phase angle plot for poor-performing can packed with chicken soup, measured using in-can electrodes.

Figure 7. Bode magnitude plot for good-performing can packed with chicken soup, measured using in-can electrodes.

Figure 8. Bode phase angle plot for good-performing can packed with chicken soup, measured using in-can electrodes.

Figure 9. Time dependence of lacquer capacitance of solvent-based lacquered foodcans packed with chicken soup, measured using in-can electrodes.

Figure 10. Time dependence of pore resistance of solvent-based lacquered foodcans packed with chicken soup, measured using in-can electrodes.

Figure 11. White-lined cans - lacquer A. Time dependence of pore resistance.

Figure 12. White-lined cans - lacquer B. Time dependence of pore resistance.

Figure 13. White-lined cans - lacquer C. Time dependence of pore resistance.

Figure 14. White-lined cans - lacquer A. Time dependence of lacquer capacitance.

Figure 15. White-lined cans - lacquer B. Time dependence of lacquer capacitance.

Figure 16. White-lined cans - lacquer C. Time dependence of lacquer capacitance.

Figure 17. White lined cans. Plot of lacquer capacitance ratio vs. pore resistance after 1 day. High tin coating weight variables.

Figure 18. White lined cans. Plot of lacquer capacitance ratio vs. pore resistance after 3 months. High tin coating weight variables.

Figure 19. White lined cans. Time dependence of charge transfer resistance for high tin coating weight variables.

Figure 20. White lined cans. Time dependence of charge transfer resistance for lacquer A.