The characteristic smell of old plastic: Analyzing VOCs from museum pieces

March of history

As history marches on, more and more items made from plastic are beginning to appear in museums. This raises questions of how to protect such plastic items from the ravages of time, as they will likely require quite different preservation techniques from more traditional museum pieces made from wood or metal. First, though, museum curators need to identify which plastic items actually require preservation, based on their level of deterioration. A team of British chemists has now confirmed that this can be done by analyzing the volatile organic compounds (VOCs) released by the items.

The advantage of focusing on VOCs as an indicator of deterioration is that they can be collected non-invasively, allowing the plastic items to be analyzed without damaging them. In 2016, the British team, led by Matija Strlič at University College London, developed a method for doing this, based on solid-phase microextraction (SPME) and gas chromatography-mass spectrometry (GC-MS).

Their method involves simply placing a commercially-available SPME fiber designed for collecting volatile compounds, comprising divinylbenzene/Carboxen on polydimethylsiloxane, next to a historical plastic item for a week. The collected volatiles are then released by heating the fiber in the injection port of a GC column.

Artificial aging

To determine whether the VOCs released from historial plastics can provide an accurate indication of the level of deterioration, Strlič and his team have now applied their method to a range of different plastic materials that they artificially aged. These plastic materials, which included cellulose nitrate, cellulose esters such as cellulose acetate and cellulose propionate, polyurethane foams, poly(vinyl chloride) (PVC), polystyrene and polyethylene, were chosen to reflect the range of plastic materials found in museum collections. To artificially age these materials, Strlič and his team kept samples of them at 80°C and 65% relative humidity for between two and 10 weeks.

They found that the materials released a range of VOCs that varied depending on the degree of aging. Many of these VOCs were produced via oxidation or hydrolysis, both of which cause physical deterioration in plastics. For example, cellulose nitrate released furfural, which is produced by the acid-catalyzed hydrolysis of cellulose, and these emissions increasing with greater aging. PVC released ethylhexanol, a known degradation product of a common plasticizer used in PVC, and these emissions increased for the first two weeks of aging but then held steady.

Other VOCs had less clear emission profiles over time, with many of the plastics releasing several VOCs that varied in complex ways. So Strlič and his team used an informatics technique known as linear discriminant analysis (LDA) to produce models for classifying the samples of each plastic material into one of two groups – aged for up to four weeks and aged from six to 10 weeks – based on their VOC profiles. LDA classifies samples into groups by maximizing the variance between the groups and minimizing the variance within the groups.

Tate Britain

The found that the models worked for all but two of the plastic materials, able to assign them to one of the two groups based on their VOC profiles with an accuracy of 62–83%. For PVC and polyethylene, however, the models were only able to classify them into the two groups with an accuracy of around 50%, no better than chance.

Finally, they used their method to analyze the VOCs released by three artistic objects made of plastic held at Tate Britain, an art museum in London: two were made from cellulose acetate and one was made from cellulose nitrate. Strlič and his team fed the detected VOC profiles for each of these plastic objects into the models for the respective materials, which assigned the two cellulose acetate objects to the low aging group and the cellulose nitrate object to the high aging group. This suggests that the object made from cellulose nitrate had suffered more deterioration than the other two.

Now, this is admittedly a rough-and-ready measure of deterioration, as the artificial aging process can’t capture all the effects of actual aging, but it shows that this approach can work. The problem is the accuracy of the models, but they can easily be improved by using the SPME-GC-MS method to analyze the VOCs released by actual aged plastic items.