National Gallery Technical Bulletin Volume 23, 2002 National Gallery Company London Distributed by Yale University Press This volume of the Technical Bulletinis published with the generous support of the Samuel H. Kress Foundation. Series editor Ashok Roy © National Gallery Company Limited 2002 All rights reserved. No part of this publication may be transmitted in any form or by any means, electronic or mechanical, including photocopy, recording, or any infor- mation storage and retrieval system, without the prior permission in writing of the publisher. First published in Great Britain in 2002by National Gallery Company Limited St Vincent House, 30Orange Street London wc2h 7hh www.nationalgallery.co.uk British Library Cataloguing in Publication Data A catalogue record for this journal is available from the British Library isbn1857099419 issn01407430 525039 Edited by Diana Davies Project manager Jan Green Design by Tim Harvey Printed in Italy by Conti Tipocolor FRONTCOVER Garofalo, Saint Augustine with the Holy Family and Saint Catherine of Alexandria(NG 81), (detail of plate4, p. 23). TITLEPAGE Garofalo, The Virgin and Child with Saints William of Aquitaine, Clare (?), Anthony of Padua and Francis (NG 671), (detail of plate3, p. 22). The Mechanical Behaviour of Adhesives and Gap Fillers for re-joining Panel Paintings christina young, paul ackroyd, roger hibberd and stephen gritt The primary aimof this research is to study the than 1mm) it is necessary to re-join the panels with behaviour of adhesive/filler mixtures used to an adhesive bulked with filler. In many cases, the re-join thin panels that may subsequently move as a adhesive/filler mix serves to fill the void (possibly result of changes in the external environment. including worm holes) and hold the panels together Moisture uptake and loss in panel paintings lead to as a single structural entity. Usually, after treatment, expansion and contraction of the wood, and further the panel will be mounted in such a manner as to change in the curvature of each section of panel is support its weight and allow a certain degree of free often associated with this dimensional change. If movement of the wood with the aim of preventing the panel is constrained by a cradle or framing further damage. However, most systems introduce system, then splits in the wood and separation or some restraint, even if minimal.1 Constraint prima- failure of the panel joints can occur. Joints and old rily induces a bending force on the panel joints, repairs in the wood are often the weakest points in although tensile and shear forces may occur because the support and are the most likely sites for failure. of uneven distortion within the panels (see The installation of air conditioning, resulting in Appendix). Similar forces can also occur during the more stable environmental conditions, has made handling of unsupported panels. this kind of damage to the panels less frequent for In some cases a damaged panel painting will paintings within gallery collections. However, the have partial cleavage of a joint with original paint National Gallery contains a considerable number of covering the undamaged section. It is preferable not paintings on wood and a good deal of time is to cause separation of the joint with consequential devoted to reducing the vulnerability of these damage to the original paint. Thus, realignment and objects to possible damage caused by climatic repair of the cleaved section may be constrained by changes. Furthermore, many paintings treated by the original joint and surrounding wood. In brief, if conservators – for example those treated at the treated panels change dimensionally or geometri- Courtauld Institute of Art by the authors of this cally, tensile, compressive, shear and bending forces study – return to uncontrolled environments, such can be exerted on the joints and hence on the filled as churches. It is therefore important to re-examine gap. the properties of adhesives and gap fillers that are commonly used in the structural treatment of panel The ideal joint paintings. The most appropriate structural conservation In general, manufacturers’ data for adhesives show treatment for a panel painting will depend on the that shear strength is higher than the peel strength.2 nature of the damage, the type of wood and panel Hence, the strongest type of joint is a lap joint that construction, the surrounding ground and paint puts the adhesive in shear when forces act on it and layers, and the environment to which the painting is minimises peel at the edges of the bond (fig.1). The likely to return. In general, for situations where length of the glue line in a lap joint is increased so there are hairline cracks or where two parts of a that most of the adhesive will be elastically rather panel can be brought back into immediate contact, than plastically deformed to ensure minimal perma- a thin line of adhesive is introduced. If wood has nent deformation.3 Adhesion is dependent on both been lost, or removed, or if large deformations have the chemistry and the surface texture of the surfaces occurred, it may not be possible to fit the individual to be adhered. Surface preparation to eliminate weak planks back into their original configuration and boundary layers is important, and for good adhesion the hence gaps may be present. In cases where the split surfaces of the materials to be joined must be free of loose or gap between adjacent panels is large (greater layers and, ideally, provide a ‘mechanical key’. NATIONAL GALLERY TECHNICAL BULLETIN VOLUME 23 | 83 Christina Young, Paul Ackroyd, Roger Hibberd and Stephen Gritt will depend on the movement of the wood surrounding the joint. The wood around the join can be severely degraded at the edges by worm damage, fungal attack, splintering, fatigue and remnants of old adhesive from previous treatments. Some of these factors reduce the strength and elas- fig. 1 Diagram of shear and peel forces. ticity of the wood and will affect re-adhesion. Thus, this type of join is far from ideal. As a minimally invasive approach is a desirable criterion for treat- ment, modification or removal of the original wood to provide a better bond design is not an option. The ideal adhesive The mechanical properties of the adhesives required for panel painting conservation will vary from case to case. However, in general the following criteria apply: • The peel and shear strength should be commen- surate with, but not stronger than, the wood surrounding the joint, so that there is a low risk of causing failure within the original panel. This is counter to the normal criteria for adhesive selection for structural applications. • The elastic modulus should be high enough so that the adhesive is loaded below its yield stress. At the same time, the adhesive should be suffi- ciently flexible to allow the panels to respond to the environment without a build-up of stress around the joint. In practice, the balance between elastic modulus and flexibility is hard to achieve. • Under continuous load, creep should be minimised. • The adhesive should fail in a ductile manner and be resistant to rapid crack growth. fig. 2 Diagram of tensile and compressive forces on joint. • It should have a good ability to wet the surface of the wood. • It should possess good handling and curing char- In contrast to this ideal practice, most European acteristics, in particular, an adequate working panels of the Renaissance and Baroque periods were time, without an unduly long curing time. originally constructed of several planks that were (There may be a conflict between handling char- simply glued and butt jointed and sometimes acteristics and good wetting properties.) reinforced with dowels, fabric strips or battens.4 The bonding edges have small surface areas in rela- The ideal filler tion to the size and weight of the individual planks and hence the joint will be highly stressed. In the Ideally the adhesive/filler mix will be a homoge- case of a simple butt joint on a restrained panel neous mixture. The filler serves a number of painting, bending will create peel forces at the edges functions when mixed with the adhesive. In of the joint, tensile on one side and compressive on handling, it bulks out the adhesive (increasing the the other (fig. 2). There may also be shear forces viscosity) to give a stiff paste, which helps to prevent acting if one side of the join is restrained relative to the materials from flowing out of the gap during the other. The actual contribution from each force application or curing. It reduces the percentage 84 | NATIONAL GALLERY TECHNICAL BULLETIN VOLUME 23 The Mechanical Behaviour of Adhesives and Gap Fillers for re-joining Panel Paintings shrinkage when added to adhesives with a volatile dard (ASTM) tests that satisfy the criteria for engi- component. Additionally, it allows the cured neering applications. However, these are not directly mixture to be sanded or carved and provides a applicable for panel painting joints. Furthermore, surface that will accept either a simple surface fill or the literature on adhesives with fillers is minimal. the direct application of retouching media. The Although there are publications in the conserva- filler will also affect the mechanical properties of tion literature outlining the practical problems of the cured adhesive, altering its cohesive strength and re-joining panel paintings and listing the equipment modulus. The actual values will of course be devised to assist in the processes, there is little infor- dependent on the ratio of filler to adhesive and it mation on the technical performance of the should be possible to alter the ratio to suit a given adhesives employed for this purpose.6 A great deal application. Hygroscopic fillers may allow the joint of published information on wood adhesives is to contract or expand with varying relative humid- available from the timber industry but this tends to ity. If the sensitivity differs greatly from that of the concentrate on the fabrication of composite boards, adjoining panel then this will set up internal stresses such as blockboard and plywood, or the gluing of that can lead to failure. It is perhaps preferable that joints in modern furniture. Some of this informa- the filler is inert and that the adhesive accommo- tion is useful in understanding the general dates hygroscopic movement in the original wood. behaviour of certain adhesives but it has limited The structural, and hence the mechanical, prop- relevance to the conservation of easel paintings and erties of the filler mixture are essentially the same as other wooden artefacts.7 those for an ideal adhesive. The final mixture should The adhesives that have been employed in also meet basic criteria for conservation materials, conservation range from traditional natural materi- that is, it should be: als, such as animal or fish glues and casein • inert to humidity and temperature change, in adhesives, to commercially manufactured synthetic terms of stiffness, strength, and resistance to resins: polyvinyl acetate emulsions (PVA), urea fracture formaldehyde (UF), and epoxy resins. • reversible Animal skin and bone glues have been used for • physically and chemically stable in both the centuries in the construction of many wooden short and long term objects, including easel paintings, and continue to • resistant to fungal and bacterial attack be employed in their repair.8 In the National • non-toxic Gallery’s collection there are panel paintings with The choice of adhesives and fillers to be studied was original joints, made with animal glues, which based on issues arising from treatments of panel remain intact even after five hundred years or more. paintings carried out in the conservation studios of These are a testament to the durability of the mate- the National Gallery and the Courtauld Institute, rials within certain limits and their long-established and also upon conversations with conservators who record of performance is one of the prime reasons regularly repair panel paintings.5 The rationale why they continue to find favour. Another reason is behind a given conservator’s preferred system comes that they remain reasonably reversible, unlike the from many years of practical experience and from synthetic products. Their drawbacks, however, are observations of the treatment of failed repairs. In also well established: they are prone to attack by practice, panels may require treatments for a combi- fungi, bacteria and insects; their mechanical and nation of different problems, such as checks or adhesive properties alter according to fluctuations in splits along the grain, failed joints with varying gaps humidity; traditional preparations have uncomfort- and worm damage. The same adhesive may be used ably short working times because they are normally for each type of failure but with varying amounts of applied hot and the increased viscosity on cooling filler. It became clear during the discussions that makes them difficult to handle. conservators used and modified their chosen filler Other natural adhesives, such as fish glues and mixtures but could not objectively assess how the casein glues, were not tested in this study. Fish glues, type or quantity of filler affected the mechanical although they share similar disadvantages to animal properties of the adhesive. The handling properties glues, are thought to impart greater flexibility and were key criteria when choosing their systems; also since commercially prepared formulations remain important was any technical information that could liquid at room temperatures they have better be obtained from the literature. The data available handling properties. However, because of their from manufacturers of adhesives is based on stan- limited use as gap-filling adhesives they were not NATIONAL GALLERY TECHNICAL BULLETIN VOLUME 23 | 85 Christina Young, Paul Ackroyd, Roger Hibberd and Stephen Gritt included in these trials. Casein glues have a long been applied to a large range of museum objects. history of use as woodworking adhesives, and are Studies on the effects of ageing on epoxies have still employed in panel painting conservation in expressed concern over the pronounced yellowing of Italy and Germany.9 They are less reversible than these resins on exposure to heat and light.17 This is animal glues and remain, to some degree, sensitive of considerable concern when they are employed as to environmental change.10 consolidants for ethnographic and archaeological PVA resins were first proposed for the structural objects, where they are likely to remain visible on conservation of museum objects in the early 1950s the surface, but is a lesser concern when re-joining and a variety of these commercially prepared adhe- panel paintings as the resin is not exposed at the sives have since been used for the structural repair of picture surface. Some cured epoxy resins can be swelled panel paintings as well as other wooden artefacts.11 by solvents, but such solvents are likely to damage paint The National Gallery uses Evostick Resin W. layers and, as is the case with the other synthetic wood Generally, these materials have good working prop- adhesives, they are difficult to remove.18 erties and give strong adhesive bonds that are Barclay and Grattan’s work on fillers for wooden thought to impart some flexibility to the join. Dried artefacts gives valuable information on some of the PVA emulsions may be swelled with water, particu- properties of adhesives and gap fillers.19 The results larly with the addition of ethanol, but in the case of are pertinent to the materials used for treating easel easel paintings, these solvents are likely to damage paintings, but the aims are not directly applicable to the surrounding paint and ground layers and swell re-gluing panel joints with gaps. Normally, gap adjacent wood.12 Also, because they are difficult to fillers for ethnographic objects and wooden sculp- remove mechanically from a tight joint, they are in tures do not function as an integral part of the effect irreversible. Ageing studies have been carried structural repair, as they do in re-joining panels, but out on several types of PVA woodworking adhesives are simply intended to fill losses in the surface. A which have shown a tendency towards increasing variety of filling materials both natural and embrittlement, yellowing and reduced solubility synthetic have been used in panel paintings conser- under heat and light ageing.13 vation. Natural materials, such as chalk, wood dust, UF resins have been used as adhesives for wood cellulose powder, rye flour and coconut fibres, have conservation since the late 1940s and as consoli- been favoured, although glass beads and glass or dants from the early 1950s.14A UF resin, Extramite, phenolic microballoons may be used with epoxy has been used by the National Gallery Conservation resins to improve their gap-filling properties.20 Department for panel repair. This is a commercial The focus of the study presented here is to wood-bonding adhesive, which is designed to understand how thin wood panel joins repaired with provide a strong waterproof bond. It is applied in gap fillers behave where some constraint to movement an aqueous form and, once cured, it is insoluble in is present. In addition, the behaviour of hairline solvents and is difficult to remove even by mechanical joints under the same circumstances is assessed. scraping. Epoxy resins are the only other class of resin Experimental procedure adhesives to have been commonly used for panel painting conservation. Since they provide strong A four-point bend test was chosen as it provides a joints with a variety of substrates, and have been controlled method of exerting a constant bending shown to have some good gap-filling properties moment on a panel joint.21Both the joint itself and when used with an appropriate filler, they have been the wood either side of the joint experience the widely adopted in the conservation field, not only as same bending moment (see Appendix) and hence it adhesives but also as consolidants.15 First mentioned is possible to ascertain whether the filler or the in the conservation literature in 1952, they were wood fails first for the same bending force. The test subsequently investigated for repairing panel paint- also allows one visually to inspect the peel action of ings in 1954 by Arthur Lucas and Norman the bond at the edges and the nature of the failure. Brommelle at the National Gallery.16That investiga- The load at which the joint fails demonstrates the tion concluded that traditional animal glues were ‘practical’ properties that might be expected for better suited for the purpose on the grounds that each adhesive and type of filler in a real situation. they possessed better handling properties and were The strength and flexibility of the joint, the nature more reversible. Since the mid-1950s an improved of the failure modes (shear, peel, tensile, compres- range of epoxies have become available and have sive), and whether the failure is cohesive or adhesive, 86 | NATIONAL GALLERY TECHNICAL BULLETIN VOLUME 23 The Mechanical Behaviour of Adhesives and Gap Fillers for re-joining Panel Paintings TABLE 1Adhesives TABLE 2Fillers Hide glue: proteinaceous aqueous glue Rye flour: Neil’s Yard Healthfoods organic Standard concentration for this application: Wood flour: Sieved National Gallery wood dust 20g hide glue in 100ml of warm water Cellulose powder PVA: non-waterproof Evostick Resin W Polyvinyl Acetate Emulsion 23 Coconut flour: Imported broken coconut shell. Standard concentration: direct from the bottle Particle size 150–3microns UF: Humbrol Extramite.24Contains approximately Microballoons. SP Systems hollow phenolic resin 5% China clay and wood flour. Designed to give spheres. Particle size 50 microns. (Not normally gap-filling properties up to 1mm without any used with polyester or vinylester resins because additional filler. they can be subject to styrene attack which may Standard concentration: cause the spheres to collapse.) 1part Extramite, 2parts water by weight Epoxy: Araldite 2014. A grey epoxy paste 25 Manufacturer’s data: lap shear strength on aluminium at 18°C 25N/mm2. Peel strength on aluminium at 23°C 3N/mm Sample preparation Standard concentration: 2parts resin, 1part hardener by volume The test samples were chosen to represent panel joints with both hairline cracks and large voids or ill-matching joins where a filler is required to bulk can all be assessed using this test. out the gap. The panel samples were constructed Standard ASTM testing procedures have not from 5 × 50 × 50 mm naturally aged oak blocks, all been adopted as they do not represent the loading approximately radially cut from a single piece of conditions of a wooden panel with a simple butt oak, and are representative of some Netherlandish join. For a complete understanding, separate tests and Dutch panel paintings. All joints were made on the adhesives and fillers are also required. with the growth rings parallel to the join. The adhe- sive and fillers used are listed in Tables 1 and 2 and plate1 Filler mixtures after curing. the adhesive/ filler combinations tested are listed in Table 4. An empirical guide to their handling and carving properties is also included in this last table. Adhesives were used in standard concentrations, as indicated in Table 4, and each filler mixture was added to 5ml of adhesive dispensed from a syringe. Sufficient filler was added to produce handling properties ranging from a stiff and workable paste to a less viscous mixture which had some degree of flow. plate 1 shows the filler mixtures after curing. The amount of filler added was calculated by Araldite 2014 Extramite: Cellulose Rye flour: Wood floor: w20agt:e r10ml pRoewsind eWr: R0.e7s1ign: W5ml R0.e9s1ign: W5ml weight. Low-tack Scotch Tape was attached to the 2.09g: 5ml edges of both surfaces of each wooden block (fig. 3). This prevented penetration of the adhesive into the faces of the wood. An additional strip of tape was attached to each pair of blocks on the under- neath faces to align them 2 mm apart. This also prevented loss of adhesive and any excess seeping out of the join when using low-viscosity mixtures. In all samples the adhesive was first brushed along the edges to be adhered to improve wetting of the surfaces. This priming adhesive was applied in the Coconut flour: Microballoons: Microballoons: Coconut flour: Coconut flour: R1.e2s8ign: W5ml R0.e6s3ign: W5ml R1.e2s7ign: W5ml MReiscirno Wballoons: MReiscirno Wballoons: normal concentration except in the case of the 15:m 1l(0.78g): 15:m 1l(1.37g): Resin W where a dilution of 5 ml water to 2 ml NATIONAL GALLERY TECHNICAL BULLETIN VOLUME 23 | 87 Christina Young, Paul Ackroyd, Roger Hibberd and Stephen Gritt Samples of each pure adhesive in the standard concentrations were also prepared to produce a thin glue line in a hairline joint (see Table 1). The wooden blocks were prepared as described above and the adhesive applied to both surfaces of the fig. 3 Diagram of sample preparation. join. The samples were laterally clamped together in the Teflon jig and further clamps were placed directly on top of each block to ensure good align- ment across the join. After 24 hours the samples were removed from the jig, the tape peeled away and excess adhesive pared away with a scalpel. fig. 4 Photograph of Teflon jig in gluing jig. Resin W was used. The blocks were then clamped in a Teflon jig to maintain alignment (fig. 4). A palette knife was used to apply the mix so that the gap was overfilled. After 24 hours the blocks were removed from the jig and the additional tape across each pair of blocks peeled away. The blocks were placed upside down and left to ensure complete curing of the adhesive. Excess filler was pared away carefully with a chisel and the protective tape removed. In some samples it was found that air bubbles formed beneath the excess filler. These corresponded to the samples with the greatest initial quantities of water, fig. 5 Photograph of Instron and four-point bend jig. such as the low viscosity Resin W filler mixtures. Blocks were conditioned at 55% RH and 20oC for a minimum of two weeks before testing. For compari- son samples were made with 2 mm gaps filled with Tests adhesive without the addition of filler (see Table 3). With the exception of the Araldite paste, in practice The bending tests were performed on an Instron the low viscosity of a pure adhesive is unsuitable for 4301 test machine at 55% +/- 3% RH and 20 +/- this application and good fills were hard to obtain. 2oC. The four-point bend jig consisted of two sets of rollers, one set fixed to the lower static crosshead of the test machine and the upper set attached to TABLE 3List of Pure Adhesive Joins the top-moving crosshead (fig. 5). The top and bottom rollers were at different separations (100 5ml of adhesive Gap size mm and 150 mm). The samples were supported on Hide glue Hairline the bottom rollers of the four-point bend jig. The Extramite Hairline and 2mm moving crosshead was then lowered at a speed of 2 mm/min, so that the upper rollers pressed down on Resin W Hairline and 2mm the top face of the sample, creating a bending Araldite 2014 Hairline and 2mm moment. The displacement of the top rollers, the compressive force, the temperature and relative 88 | NATIONAL GALLERY TECHNICAL BULLETIN VOLUME 23 The Mechanical Behaviour of Adhesives and Gap Fillers for re-joining Panel Paintings fig. 6 Summary bar chart of peak bending loads and stiffness. humidity were logged during the test. A CCD an inherent weakness or defect at such points and so camera with 50 mm macro lens and fibre optic illu- the measured stiffness did not necessarily represent mination was used to capture real-time video of the the true stiffness of the wood or the adhesive. These tests. The video was used to confirm the mode of results have therefore not been included in the failure and aid in the post-failure analysis. analysis. The averaged peak load for this type of failure is 53N and is shown in fig. 6 as a vertical line. Results For samples where the wood failed at high loads, Depending on the ease at which good-quality joints away from the joint in the bulk of the wood, the could be repeated for each type of adhesive/filler stiffnesses have been calculated and included in the between three to eight samples were tested. A average for each type of joint. The averaged peak summary of the results is given in fig. 6. This gives load for this type of wood failure is 139N and is the average stiffness and peak loads at bending fail- shown in fig. 6as a vertical line. The range of these ure for all the samples. The stiffness of the samples values is shown as a shaded region. was calculated by taking the gradient of the initial For samples where failure occurred along the linear section of load-displacement curves. For some adhesive interface with the wood (adhesive failure) of the very flexible adhesives, it is not possible from or within the body of the adhesive (cohesive failure) these tests to distinguish the transition from elastic the stiffness and peak load values have been calcu- to plastic deformation (the yield point). There was lated and averaged for each type of joint. These no distinct linear region of the load-displacement values are shown as a bar chart in fig. 6. Wood curve and so the measurement of stiffness is less ‘removal’ is indicative of damage to the wood at the reliable. wood/adhesive interface. Such damage will have In a number of tests failure occurred in the wood occurred if any wood is visible on the fracture at a very low load, usually along the grain and close surfaces of the adhesive. to one of the rollers. This indicated that there was NATIONAL GALLERY TECHNICAL BULLETIN VOLUME 23 | 89 Christina Young, Paul Ackroyd, Roger Hibberd and Stephen Gritt fig. 7 Typical load-displacement curves for pure adhesives with hairline joints. calculated from the low load gradients of these Hairline joints joints. These gave a value of 41 Nmm-1; this stiff- ness is equivalent to a Young’s Modulus 0.98 GPa Pure adhesives which is relatively high for an oak stiffness modulus Typical load-displacement curves for these tests are (typically 0.6to 1.1GPa) taken across the grain. given in fig. 7. In a hairline joint the adhesive layer Araldite 2014 is designed as a strong structural is very thin and therefore any flexibility within the adhesive and as expected the wood failed in all the elastic region of the adhesive should make a negligi- Araldite samples. In two out of the six samples the ble contribution to the change in the stiffness of the fracture appeared to have initiated at the adhesive sample. Thus, for a good-quality joint the measured interface and then ran into the wood. In the four stiffness of the sample may be considered to be the other samples the wood failed away from the joint, same as a continuous piece of wood. In the case of leaving it intact with no sign of cracking. The tests Extramite and Araldite 2014 the wood failed before of the Extramite joints all resulted in brittle cohe- the joint and hence the stiffness for the wood was sive and adhesive failures and in all cases there was plate2 Failed joint – Araldite 2014. plate3 Failed joint – Extramite. 90 | NATIONAL GALLERY TECHNICAL BULLETIN VOLUME 23