St John the Evangelist, Howsham (Yorkshire): Further Research into ‘Crizzling’
The phenomenon often referred to as ‘crizzling’ is usually associated with vessel glass, and its occurrence in stained glass has not been studied extensively to date. Historically, it was thought that in windows the problem only occurred with green glass, leading to the name ‘green glass disease’, but we now know that is also occurs in other forms of glass. Purple crizzled glass is found in the windows of the nineteenth-century church of St John the Evangelist in the small North Yorkshire hamlet of Howsham [Fig. 1]. The present author selected this church and its windows as the primary case study for her recent MA dissertation. Here she provides an introduction to crizzling, considers the case at Howsham (and the wider implications for our understanding of glass made in Britain in the mid-nineteenth century), and discusses the conservation issues around crizzling. Finally, she proposes a method for treating crizzled glass that would retain the material in situ whilst greatly decreasing its rate of deterioration.
Merlyn Griffiths (York Glaziers Trust)
The deterioration process that results in ‘crizzling’ (most commonly recognized by a network of fine cracks on the glass surface) is caused by a chemical imbalance in the glass itself – when the material’s composition is characterized by an excess of alkali and insufficiency of calcium oxide. The process of deterioration occurs in a number of stages. Generally speaking, the glass first appears to ‘sweat’ or ‘weep’; this is caused by the appearance of alkali on the surface of the glass, either as droplets (in higher relative humidity) or as crystals (in lower relative humidity). The glass can then become cloudy and opaque, and develop a fine silvery network of cracks, a stage known as ‘incipient crizzling’. These cracks become deeper and the surface of the glass begins to spall away, eventually leading to full fragmentation of the glass. Although medieval examples of the phenomenon may be found in Germany, almost all instances of crizzling in Britain date to c.1860, and instances of such deterioration are found in windows produced by many of the major glass firms of the Gothic Revival.
In the mid-nineteenth century, a renaissance in the production of glass was under way. Martin Harrison describes how the ‘thin and watery’ glass available in the early nineteenth century – used in windows by manufacturers such as William Wailes – was supplanted by a higher-quality product, based on medieval methods of manufacture. The latter was produced only after a great deal of experimentation, and it is therefore understandable that some discrepancies occurred, leading to batches of crizzle-prone glass. Today, crizzling poses a real problem for conservators, yet there is no remedy that is both readily available and cost- and time-effective. The most popular treatment in the past has been the removal and replacement of affected glass, but the resultant loss of many examples makes it difficult to identify windows that once contained crizzled glass and assess the scale of the problem.
Because it relates to a chemical imbalance, the potential for crizzling is inherent in some batches of glass from the moment of manufacture. However, the process of deterioration is triggered and accelerated by moisture and the effects of fluctuating relative humidity. Indications of the problem are therefore not usually evident until after some years of exposure to the environment. While the trigger is no different to that by which all glass corrosion is initiated, with crizzling the lack of sufficient stabilizer – calcium oxide – means that the glass is much more prone to the leaching of alkaline compounds. Many examples of stable glass contain approximately 10% calcium oxide, but crizzled glass can contain as little as half of this amount.
There are also terminological problems. Many terms are associated with the phenomenon of crizzling, such as devitrification, sugaring, sick glass, and so on. Confusingly, ‘crizzling’ is also used to describe the surface not just of damaged glass paint, but also of enamels and ceramic glazes. As noted above, the word only describes the visual characteristics of a late stage in the deterioration process (the appearance of cracks), and earlier stages in the process, where these fault lines are not yet visible, are often described as something else. An umbrella term to describe the overall process, and not just a single symptom, is required. ‘Glass disease’, as it is sometimes called, is not specific enough, so the present author has coined the expression ‘low lime glass degradation’ (LLGD for short).
The church of St John the Evangelist in Howsham, including its glass and interior fittings, dates to 1859–60 and was designed by George Edmund Street (1824–1881), the renowned Gothic Revival architect [Fig. 2]. Despite its rural location, the church contains a full scheme of windows made by the prolific London company of Clayton & Bell, a firm still in its infancy at the time. The majority of the glass has not been altered from its original condition, and despite not having protective glazing, remains in an excellent state. Within two windows however (sVI and wI) a small number of pieces of a purple glass have suffered from LLGD. The Howsham glass is an example of the problem at an advanced stage: the glass has browned and become opaque, and developed micro-cracks across both interior and exterior surfaces; holes have also formed, which will expand as the deterioration continues, until the glass disappears almost entirely [Figs 3–4]. These windows form an ideal case study: the fact they have remained untouched means that the full effects of the degradation process can be seen at first hand. Window sVI was chosen for study, as it was more accessible and contained a slightly larger amount of the affected glass.
In the early stages of research, Dr. Manfred Torge of the Bundesanstalt für Materialforschung und -prüfung (BAM) in Berlin analyzed a fragment of the Howsham glass, and the results were revealing. They confirmed that the elevated levels of alkali and insufficient levels of calcium oxide made this form of degradation inevitable. Energy-dispersive X-ray analysis (EDX) conducted with an environmental scanning electron microscope (ESEM) did not pick up any calcium oxide in the core of the glass (known as the ‘bulk’, essentially the area as yet unaffected by the corrosion process), which was particularly surprising, and further highlighted the instability of this material.
The two major types of glass made in the Middle Ages were soda glass and potash glass. The former was made with a sodium-oxide flux obtained from kelp, and is less susceptible to corrosion that the latter. Potash glasses were made with a potassium-oxide flux obtained from wood ash, and became the most common type of glass from around 1000 AD. It was also prevalent in stained glass after the Gothic Revival, and the Howsham sample is most certainly made from potash glass. Wood ash however cannot have been a component of the purple LLGD sample, since calcium oxide, a natural component of wood ash, would have been present. This may be explained by the fact that English nineteenth-century glasshouses used synthetic ingredients and generally acquired their potash from chemical manufacturers, who supplied them with dry potassium carbonate. This had been purified, so calcium oxide was absent, and therefore also absent from the finished glass itself unless added separately. Calcium oxide could be added in the form of lime, but in the case of the Howsham glass it was left out, either accidentally, or because glasshouses were still experimenting with ingredients at the time of manufacture.
In window sVI, only certain pieces of the purple glass have been affected, whilst others of the same colour have remained intact, demonstrating that only some batches were problematic. Interestingly, a few pieces of the same LLGD purple can be found in Clayton & Bell windows in the churches at Scrayingham and Whitwell-on-the-Hill, both within a few miles of Howsham, and also dating to c.1860. Again, not all of the purple glass is affected in these places, showing that the firm was likely to have been picking glass from batches of differing dates, thus making the material in these churches difficult to date precisely.
The loss of Clayton & Bell’s records and the absence of documentation in the archives of the patrons has rendered it impossible to identify the supplier of the glass for these commissions. It has also been difficult to pinpoint the time at which the glass was bought for the windows at Howsham, other than to confirm 1860 as the terminus ante quem. The firm is known however to have purchased glass from James Powell & Sons of Whitefriars, London, and it is likely that they did so until the early 1860s. Powells began producing its own ‘muff’ (blown cylinder) glass in 1845, and throughout the early 1850s was guided by Charles Winston, who was commissioning the analysis of medieval glass at the Royal College of Chemistry. He forwarded the results to Powells, where the glass was reproduced and became known as ‘Winston’s Antique’. The company was known for its high-quality window glass and for a long time had the edge over its competitors. The location of the glasshouse in the capital meant that the firm was well placed for the increasing demand for glass during the Victorian period. Almost all of the examples of LLGD window glass from the 1850s and 1860s are found in windows made by London firms, especially in those by Lavers & Barraud, who were known to be Powells customers. It is therefore possible that the Howsham purple was made in the Powells glasshouse, and there is certainly sufficient evidence to suggest that at the very least, Powells was producing glass prone to LLGD [Fig. 5].
Winston is known to have relied on the glass-making instructions provided by Theophilus in his twelfth-century treatise, and he was likely to have imposed these guidelines on Mr Green, his assigned labourer at Powells. Theophilus advised the use of only two ingredients when making glass: beech ash and sand. In the Middle Ages, these would have been sourced in their raw form, and the ash would have contained a substantial percentage of calcium oxide, obviating the need for extra lime; this would explain why so few medieval examples of potash glass have been affected by LLGD. If the nineteenth-century manufacturer followed Theophilus’s instructions, but using the ingredients with which they were familiar (purified silica and purified potash), they would have manufactured a product without calcium oxide. A letter from Winston to Powell in 1853 suggests that he start using slaked lime in glass-making experiments, implying that lime had not featured in the experiments before this point, or at least not in any great quantity.
LLGD presents a range of conservation issues. As the rate at which this form of deterioration proceeds is governed by climate, the available conservation literature should be considered with care, because it focuses primarily on glass kept and treated within tightly regulated museum conditions. Howsham holds fewer than twenty services a year, and the heating system is only turned on for autumn and winter services; fluctuations in the temperature and relative humidity are therefore mostly caused by the natural rises and falls of the daily climate cycle. In order to measure this, a Lascar data logger was placed at base of window sVI, and readings of temperature and relative humidity were taken every 30 minutes from 8 February until 7 August 2015. Whilst not a full annual cycle, the results showed that the average relative humidity inside the church was 72.6%, with a peak of 88% in February and a low of 38.5% in April. As expected, the relative humidity increased as the temperature dropped, with the lowest recorded temperature being 2.5° Celsius, at the exact point at which the relative humidity peaked at 88%. Koob recommends that glass suffering from LLGD should ideally be stored at 40–45% relative humidity, at which level the deterioration process would virtually cease. We can therefore see why deterioration of the glass at Howsham has reached such an advanced stage: although the readings taken there were typical for this kind of building, the fluctuations were enough to accelerate LLGD.
The readings showed that no conservation treatment could decrease the rate of degradation unless the climate were first regulated. Regulation of a church’s climate, for example through the installation of a protective-glazing system, is not always feasible. In smaller rural churches the budget is often incredibly tight, and it is within these buildings that many examples of LLGD glass are found. Protective glazing necessitates the removal of the glass from the stone, and this may cause more immediate damage than the cumulative effects of direct exposure to the environment. Each window needs to be assessed individually.
A study of other past and current conservation options sufficed to underline the lack of available choices for the treatment of LLGD glass in many cases, and demonstrated why the replacement of this glass is often the preferred option. As the British examples of this problem are limited to Victorian glass, they are generally treated differently to the examples of LLGD glass found in Continental medieval windows; for example, it is unlikely that medieval glass would be removed without other methods that retained the glass in situ being tested first.
The holes in the glass at Howsham mean that the function of the window as a weather shield is compromised and the aesthetic appearance of the glass somewhat altered. The LLGD glass is however confined to small areas of the window, and the rest of the glass is in excellent condition, making it difficult to justify the installation of an expensive and potentially disruptive protective-glazing system at this point. It became apparent that it was necessary to formulate an alternative treatment method that did not employ inappropriate materials, did not necessitate the removal of the window, was relatively simple to undertake, and was both cost- and time-effective. Any alternative treatment also needed to comply with the Corpus Vitrearum code of good practice. (It must be noted that the proposal outlined here has not yet been trialled on LLGD glass, so is only discussed from a theoretical position.)
In recent years, Paraloid B72 has been trialled and adopted as a means of consolidating LLGD glass by Koob at the Corning Museum, due to its ‘moderate strength and high permeability to water vapour’. Although these qualities are particularly attractive, Paraloid B72 has a poor resistance to humidity. This is a big problem considering the environmental conditions of a church like Howsham and the self-imposed criterion for this location of not using protective glazing. Discussions with colleagues at the York Glaziers Trust resulted in the formulation of a concept that would allow the use of Paraloid B72 without the need for a full protective-glazing system. The method adopted to ensure sufficient ventilation between the historic glass and the protective glazing in the canopy heads of the Great East Window of York Minster could be adapted to the specific requirements of LLGD glass. During the reglazing of the minster panels, a small number of pieces was left to one side, and wire mesh was installed in their places. This was undertaken in order to allow the free flow of air between the historical panels and the protective glazing behind them, for they had no other means of ventilation. Lead was then wrapped around the edges of the historic pieces and they were tack-soldered onto the main lead matrix in front of the corresponding mesh inserts. This enabled the internal ventilation of these panels and had minimal visual impact. Even on the bench, the difference was barely noticeable [Fig. 6].
The present author has therefore proposed that a similar system be used for the treatment of LLGD glass, as a means of retaining the LLGD glass in situ, whilst vastly decreasing the amount of condensation that would form on its surface, and also removing its function as a weather shield. The system in the canopy heads was designed to allow ventilation with minimal overall aesthetic impact, but the principle of tacking a piece of historic glass to the surrounding lead matrix so it sits inwards of the rest of the panel can be adapted as a conservation treatment for LLGD glass. As this LLGD glass does not have a protective-glazing system installed behind it, it therefore does not require ventilation in the same way as the aforementioned examples of the York Minster panels. Thus, rather than using wire mesh, which was appropriate for the minster panels to ensure ventilation but is not required here, the affected piece of glass could be replaced with a glass infill of the same colour as the historic glass. If it were replaced with clear glass, any holes or fissures in the LLGD piece would transmit daylight, which would be disguised if a matching colour of glass was used. Then, the edges of the LLGD glass would be surrounded by lead and tack-soldered onto the panel in front of the new infill, which would leave a slight interspace, whilst also appearing as though the historic glass was still in its original location. This would create a localized protective-glazing system that would allow for the use of a product like Paraloid B72. Despite the slight alteration to the lead, this approach would be consistent with the concept of minimal intervention, as it would be ‘as much as necessary, but as little as possible’.
Although Paraloid B72 would work well for securing breaks, it would do little to secure the glass if further fissures opened up in future. The solution could be a combination of a consolidant and a secondary material acting as a physical support preventing any further loss of glass. To this end, fabric repairs have been successfully been carried out in the stained-glass conservation studios at Cologne Cathedral and the Burrell Collection, and so were also trialled for the present author’s research.
In order to insert the infill, the LLGD glass would first need to be removed from its location in a window. In the more extreme cases, a stabilizer would be needed to hold the glass together during removal, which would then be taken off in order to treat the glass with Paraloid B72. The present author has recently demonstrated the usefulness of cyclododecane as a means of temporarily stabilizing breaks in glass, and found Japanese tissue paper to be useful for securing broken glass during transport when it was impregnated with a cyclododecane melt. For the purpose of use with LLGD glass in situ, Japanese tissue paper could be impregnated with a cyclododecane melt in advance, and then remelted onto the affected LLGD glass in situ with a heated spatula in order to secure it. Once removed from the window, the affected glass could be treated with Paraloid B72 on the opposite side whilst the cyclododecane melt sublimed.
For the trial, a crack pattern was created on model glass. Windscreen glass was chosen for the model glass, since it consists of a layer of gel laminated between the two glass sheets, and when damaged the broken pieces of glass are held together. 10cm-square samples were hit with a hammer on both sides in order to shatter the glass but without fully separating the pieces; this would have been impossible on regular glass. Paraloid B72 was chosen as the consolidant. This was readily available, and recent research has demonstrated its suitability for use on LLGD glass; it is also not organic, so decreasing the risk of microbial growth. (An alternative option would be Ormocer, which has been used with great success in conservation of LLGD glass in Cologne.) Concentrations of 10% and 30% Paraloid B72 in acetone were trialled on the model glass in conjunction with three types of fabric, including glass-fibre cloth (as used in Cologne); Japanese paper (as it was the best carrier for cyclododecane); and Polyester Stabiltex (as used at the Burrell Collection) [Fig. 7].
The tests showed that all of the fabrics had their merits and were sufficiently strong when applied with Paraloid B72. The glass-fibre cloth was the thickest and therefore the most visually disruptive, though it can also be purchased in a finer gauge that was not trialled. Japanese paper was successful, and almost invisible against the glass, but its woven fibres caught in the uneven surface of the shattered glass. It was also more likely to hold moisture against surface of glass, which would be especially damaging to LLGD examples. The author found that the best option for this study to be Polyester Stabiltex, as it was sufficiently strong without being visually intrusive [Fig. 8].
Paraloid B72 is known to leave a somewhat shiny layer on the surface of glass, and stronger concentrations can cause white blooms. It may therefore be preferable to use a weaker-percentage solution. In the test series, the 10% solution caused the glass-fibre cloth and Japanese paper to lift away from the glass surface as the Paraloid B72 dried, which would necessitate additional coats in practice; the 30% concentration worked well with both supports, especially if smoothed with a spatula when wet. There was not much difference between the two concentrations of Paraloid B72 when used alongside Polyester Stabiltex, but repeated applications of the 10% solution would ensure a better infiltration of the cracks. It could be argued that the use of Paraloid B72 would affect the reversibility of the treatment, but this might be outweighed by the benefit of retaining original glass within a window; realistically however no other consolidant would be any more reversible.
Every practitioner has different ideas about what should be done with LLGD glass. There are certainly merits in attempting to find an alternative method that retains original material for as long as possible rather than replacing it, but in some situations replacement may be the better option, either because the glass is extremely degraded, or because there are incredibly strict time and financial pressures. If replacement of the glass is warranted, this should be fully justified, and any insertion should be identifiable in accordance with the CVMA guidelines. Thorough documentation should also be undertaken, and glass that has been removed should be retained. It is unlikely that extreme examples of LLGD glass where the glass itself has crumbled to minor fragments will ever be returned to a window, but the material itself is incredibly useful, and can be used for sampling and further investigation. Indeed, during the course of research, the author was fortunate enough to be invited to a number of stained-glass conservation studios, including Lincoln Cathedral and Barley Studio in York. Both of these studios adhere to documentation and retention protocols, which enabled the author to learn from the fragments in their archives and their corresponding documentation.
Documentation can be as simple as a high-resolution photograph of the glass before, during and after treatment. Photographs are also a useful way to monitor the rate of degradation, such as the propagation of cracks or formulation of holes. Photographs taken of window sVI at Howsham in 2012 showed the state of the glass at that time, and were particularly useful when compared with the glass at present. The comparison showed that in the short space of three years, the holes in the glass had grown, something that would have otherwise been impossible to measure.
1. Stephen Koob, Conservation and Care of Glass Objects, London, 2006, pp. 117–18.
2. Ibid., p. 118.
3. Ibid., p. 118.
4. Ibid., pp. 120–21.
5. Martin Harrison, Victorian Stained Glass, London, 1980, p. 23.
6. Koob, Conservation and Care of Glass Objects (as n. 1), p. 118.
7. Sandra Davison, Conservation and Restoration of Glass, Second Edition, Oxford, 2006, p. 75.
8. J. W. Smedley, C. M. Jackson and C. A. Booth, ‘Back to the Roots: The Raw Materials, Glass Recipes, and Glassmaking Practices of Theophilus’, in Patrick McRay and W. D. Kingery (eds), The Prehistory and History of Glassmaking Technology, Westerville, 1998, Ceramics and Civilisation VIII, pp. 145–65, here p. 149.
9. Georges Bontemps, Bontemps on Glass Making: The Guide du Verrier of Georges Bontemps, trans. Michael Cable, Society of Glass Technology, 2008, p. 54.
10. Harrison, Victorian Stained Glass (as n. 5), p. 10.
11. Tony Benyon, ‘The Development of Antique and Other Glasses Used in 19th- and 20th-Century Stained Glass’, Journal of Stained Glass, 29, 2005, pp. 184–98, here pp. 188 and 190.
12. Marta Galicki, Victorian and Edwardian Stained Glass: The Work of Five London Studios 1855-1910, Bristol, 2001, p. 31.
13. Henry J. Powell, Glass-Making in England, London, 1923, p. 118.
14. Charles Winston, An Enquiry into the Difference of Style Observable in Ancient Glass Paintings Especially in England with Hints on Glass Painting, London, 1867, p. 358.
15. Museum of London, 80.547/3173, letter from Winston to Powell.
16. Koob, Conservation and Care of Glass Objects (as n. 1), p. 127.
17. Stephen Koob, ‘An Experimental Treatment for Severely Crizzled Glass’, in Hannelore Roemich (ed.), Glass and Ceramics Conservation 2010: Interim Meeting of the ICOM-CC Working Group, October 3–6, 2010, Corning, New York, New York, 2010, pp. 128–32, here pp. 129–30.
18. Ulf Korn, ‘“As much as necessary, as little as possible”: Notes on the Protection and Restoration of Medieval and Renaissance Stained Glass’, in E. Bacher (ed.), Stained Glass: Conservation of Monumental Stained and Painted Glass, ICOMOS General Assembly, Colombo, Sri Lanka, Central Cultural Fund Publication 123, Colombo, 1993, pp. 112–26, here p. 115.
19. Marie Stumpff, ‘Repairing Fractures II’, Boppard Conservation Project, 25 June 2015; https://boppardconservationproject.wordpress.com/2014/06/25/, accessed 13 April 2016. Ulrike Brinkmann et al., ‘Anwendungen innovativer Restaurierungsmaterialien und -methoden zur Sicherung craquelierter Glasmalereien: Modellhafte Anwendung an Glasfenstern des Kölner Domes (Weltkulturerbe)’, Förderprojekt der Deutschen Bundesstiftung Umwelt, Abschlußbericht AZ 24583‐45, Osnabrück, 2013, p. 70.
20. CVMA, Guidelines for the Conservation and Restoration of Stained Glass, 2004.