New Investigations of Adhesives for Tear Repair of Canvas Paintings

ABSTRACT

The thread-by-thread tear mending technique for canvas paintings represents an important field of research for the preservation of canvas paintings. This joining technology has been published and has become well established as a minimal invasive alternative for the treatment of local damage in the canvas, in contrast to the application of patches or the lining of paintings. Since then the technique has been further investigated, especially with regard to adhesive evaluation. For the first time, this paper presents the examination of five different test set-ups for the adhesive evaluation in an overall context: 1. uniaxial tensile tests on free adhesive films, 2. on mended yarns, 3. on fabric strips, 4. biaxial tensile tests, and 5. multiaxial burst pressure tests. The bonding was carried out on flax threads and canvas samples. For this purpose, single thread bonds were prepared within the specimens with a selection of different synthetic polymer dispersions and sturgeon glue modifications. Different bonding techniques were investigated, such as overlapping or butt joint mending, with and without bridging threads. The aim of the investigation was to understand the individual influencing factors in the various test series. Furthermore, the focus was on the development of a suitable test system for the evaluation of adhesives for the thread-by-thread tear mending treatment in textile painting supports. It can be stated that only the testing of several bonded threads in a fabric composite should be pursued as a long-term sensible test arrangement. This is in contrast to the previously very common set-up of testing mended single threads. Promising adhesives and adhesive modifications have been found for butt joints and overlap joints. The additional strengthening effect of bridging threads on butt joints has been proven and quantified.

KEYWORDS:

• Thread-by-thread tear mending
• canvas painting conservation
• adhesive selection
• bonding technique
• adhesive evaluation
• material testing
• uniaxial and biaxial testing

Acknowledgements

The co-authors and supervisors of the doctoral project, Prof. Dr Wulff Possart from Saarland University as well as Dipl.-Rest. Petra Demuth and Prof. Hans Portsteffen from CICS are warmly thanked this point.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Notes

1 The presented results were elaborated in the context of the Master Thesis of Hannah Flock M.A., which was published in German only (see Flock 2014, 2015). The results form the basis for a subsequent doctoral project as well as further publications and the dissertation. For this reason, an English summary is presented to the international community of conservators.

2 Heiber suggested 13% (Heiber 1996) as well as 10% wheat starch paste (Heiber 2003). 10% wheat starch paste leads to 15% solids content in 1:1-mixture with 20% sturgeon glue.

3 Recent research shows that pure cellulose ethers may also be suitable adhesives for selected fracture patterns in single thread bonding (see Gaasch 2019 and upcoming dissertation of Hannah Flock).

4 There are products existing on the market that are declared as pH-neutral, which are buffered alkaline and should have a neutral pH value: nevertheless, the qualities and stated pH values fluctuate strongly; moreover films are partly not pH-stable after drying, as first own investigations on different products already showed in 2010.

5 For this reason, a sturgeon glue solution of 40% was also included in the test series as the maximum possible solid content that could be processed.

6 In tear mending practice, dispersion adhesives are often diluted by adding water to reduce their viscosity, improve the wetting and penetration of the adhesive, and facilitate processing. The alternative dilution with cellulose ether gels is also used to delay rapid setting, to achieve dilution with less loss of adhesive strength, and to increase water swellability (see Baltscheit 2011).

7 The two EVA dispersions, Lineco and Evacon-R have neutral pH values. According to FTIR analyses by Elisabeth Jägers, the copolymeric EVA dispersion Evacon-R has a much higher proportion of ethylene, which acts as an internal plasticizer. Therefore Evacon-R is much more flexible and softer than Lineco. Lineco could be identified in the spectrum as an EVA dispersion but has a much lower proportion of ethylene. This could be the reason why Lineco is incorrectly identified as a homopolymeric PVAc dispersion on the container (see Flock 2014, 28–29, 211-215).

8 Following the recommendations of standards DIN EN ISO 527 and DIN EN ISO 139. For more detailed values, please see Flock (2014, 37, 58, 79, 99, 117, 619–630).

9 The comparatively high test speed was determined within the framework of pre-tests and with regard to the standards consulted (see DIN EN ISO 527, ASTM D 823, ASTM D 1005, ASTM D 2370). Lower test rates resulted in very long test durations, combined with undesired high strains and falsifying specimen deformations.

10 For further details regarding the two materials, please see Flock (2014, 224–240).

11 For detailed background information regarding the dosage and adhesive amount, please see Flock (2014, 308–321).

12 For more details about the mending process, please see Flock (2014, 320–342). For the preparation of bridging threads, please see Flock (2014, 249–250).

13 The average bursting time within 20 ± 5 sec is determined by the setting of the control valve, see DIN EN ISO 13938-2, 3.

14 Important investigations in the context of free film testing of adhesives for conservation for precise material identification have already been published by e.g. Down et al. (1996).

15 Comparing different animal glues of 30 (± 10) µm and 70 (± 10) µm thick free films while using the individual cross-sectional area for stress calculation (precision: 1 µm), different tendencies in mechanical properties among thinner and thicker films resulted (see Flock 2018, 55).

16 Due to the comparable film thickness, the tensile strength is not specified here as the average individual stress per sample cross-sectional area [N/mm²]. The results were comparable in their ratios and distributions regarding Fmax [N] and UTS [N/mm²] and their coefficients of variations (CV). This choice has also been made due to better comparability with the further testing results in Newton.

17 The context is as follows: Coefficient of Variation = (Standard Deviation / Expected Value) * 100. Hence, small expected values with small standard deviations can be compared to large expected values and their larger standard deviations. With this relative value, the units of measurement and associated absolute values are neglected.

18 In addition to the BEVA film melt, the ‘BEVA® D-8’ and ‘BEVA® D-8-S’ dispersions are also sometimes used by conservators in practice. However, based on previous tests, the use of this product alternative for tear mending should also be discouraged due to its low strength. For further information see e.g. Demuth et al. (2011).

19 For details on film structure, see also Flock (2014, 212–213).

20 There is a huge problem in reproducibility which is going to be discussed in further publications (e.g. Flock et al. 2020).

21 This poor measurement was due to the technical limitations of the testing machine since there was no possibility for e.g. strain measurement directly on the specimen itself or for non-contact strain measurement via digital image correlation.

22 In textile testing, the calculation of ultimate tensile strengths (UTS) of single threads or yarns according to Fmax per cross-sectional area [N/mm² = MPa] is quite uncommon. Instead, textile engineers usually work with only Fmax [N] as well as Fmax per titer [cN/tex], relating the maximum tensile force to the fineness of linear textile structures as fibres, threads, or yarns (see Flock et al. 2020, 13–15; Reumann 2000; DIN EN ISO 2062).

23 The same trend of high spreads showed up in past testing of mended single threads (e.g. Reuber 2010; Demuth et al. 2011): with a slightly higher number of samples, Reuber showed approx. 17-53% CV of Fmax in her test series (see Reuber 2010, 52).

24 Mowilith® D50 shows comparable values to Mowilith® DHS S1. The product showed especially promising results in former testings (see Reuber 2010, 93). Since the production of Mowilith® D50 has stopped, it was only tested again to find a similar substitute by reference.

25 Please compare to Note 22. Only the calculation of engineering stress, relating the force to the initial cross-sectional area before testing, would have been possible. However, it must be pointed out that here we can only speak of effective tensile stress, as there is no uniform continuum mechanical sense. Furthermore, the calculation of stresses [N/mm²] is also unusual for fabrics in textile testing (see Flock et al. 2020, 13-15; Reumann 2000; DIN EN ISO 13934-1; DIN EN ISO 13935-1). Alternatively, all Fmax values [N] can be related to the sample width [mm] for a substitute stress value [N/mm].

26 The biaxial tests were carried out on an individually built biaxial testing machine (Speicher 2010; Seibert, Scheffer, and Diebels 2014) at the Chair of Applied Mechanics, Saarland University.

27 For more details about the sample preparation, please see Flock (2014, 301–304).

28 Due to the rounded shape of the corners of the cruciform specimen, to a lesser extent the two thread systems also influence each other in these transitional areas by propagated friction effects.

29 Exemplary observations on increased gauge lengths (200 mm) due to extended cruciform arms suggest that the uniaxially loaded specimen length has rather little influence on the measurement results in the biaxial centre. In principle, however, shorter specimen arms are recommended for higher measurement accuracy when considering the biaxial specimen centre areas (Flock 2014, 114).

30 For additional information on biaxial testing and optical strain measurements, please see e.g. Losch (1971), Koenen and Saxe (1992), Chen and Egbers (1995), Young (1996a, 1996b), Young (1999), Young and Hibberd (1999), Speicher (2010), Seibert, Scheffer, and Diebels (2014), Seibert (2017).

31 However, the deviating determination of the strain values already mentioned must also be taken into account. For more detailed data, please see Flock (2014, 529–557).

32 There are two standardised test procedures: In the hydraulic design, the membrane underneath the fabric is vaulted up with a fluid. In the pneumatic process, air pressure is used (see DIN EN ISO 13938-1/2).

33 For more details about the sample preparation, please see Flock (2014, 558–582).

34 Testing referring to DIN 13938-2.

35 At this point, it should be pointed out that the ‘bulge test’ can be optimised with the possibility of field-optical 3D strain measurement. In this way, an anisotropic behaviour could also be represented and understood (see Andler 2010).

36 The name and appearance of the product as changed: ‘Lineco – Neutral pH adhesive’ to ‘Pel – Neutral pH adhesive’, now ‘Arcare J55 – Neutral pH Adhesive’, Co. Preservation Equipment Ltd. All declared as homopolymeric PVAc, but several FTIR spectra (controls from 2011-2019) show clearly the same copolymeric EVA and composition, no matter the batch or name (compare Flock 2014, 215). All FTIR analyses were performed by Prof. Dr. Elisabeth Jägers.

37 Cooperative PhD project, ‘Thread-by-thread Mending in Painting Conservation: Adhesives, Testing Systematics and Results’ of Hannah Flock, located at Cologne Institute of Conservation Sciences (CICS) at TH Köln and the Chair of Applied Mechanics (LTM) at Saarland University, Saarbrücken.

38 The publication of the Conserving Canvas Symposium Postprints is planned for 2021.