2525

Manfred Torge, Wolfgang Müller,Karin Adam. Bundesanstalt fürMaterialforschung und -prüfung Berlin,Germany

This article is the first of two deal-ing with historical stained glass.The aim of this article is to providean overview of the composition ofhistorical glasses, to discuss theensuing problems when attempt-ing to restore medieval churchwindows, and to illustrate how sci-entific examination of glass sur-faces can be used to select an ap-propriate cleaning method. Thesecond article, which is to appearin the next issue of Structure, willdeal with the materialographicpreparation and examination of his-torical glasses and modern copiesproduced for test purposes.

Stained glassStained glass is composed ofthree basic materials: coloured,mouth-blown flat glass, the dark-brown vitreous enamel that ispainted onto the glass and thelead used to hold the pieces ofglass together. The glass is cutinto segments to match the car-toon (a full-size working drawing).A suspension of the dark vitreousenamel is applied to the glass witha brush and worked to give thedesired effect. Once dry, the glassis fired at a temperature of 600-700 oC. Usually the glass is sub-jected to repeated paint-fire cy-cles. Finally, the coloured glassesare assembled to create the de-sired form by mounting the edgesof the individual segments in flex-ible leaden cames (H-section leadstrips) which themselves becomepart of the picture being created.To improve the mechanical stabilitythe glass and lead are cementedtogether using a putty. Medievaland contemporary techniques dif-fer only in detail (see Figure 1).

Scientific Examinations of HistoricalStained Glass

In medieval Europe, glass was es-sentially made by melting sandand wood ash. The chemical com-position of this glass can be deter-mined by examining minute frag-ments of the original material usingelectron probe microanalysis(EPMA) and energy dispersive x-ray spectroscopy (EDS). Thechemical composition of these his-torical glasses is found to varywidely, depending not only on theage of the glass and the regionfrom which it came, but also onthe type of glass examined. Thesmelting and processing tech-niques used in the glass-makinghuts of the Middle Ages necessi-tated the use of relatively largeamounts of ash and lime, so thatthe chemical composition of thesehistorical glasses is very differentto the coloured glass manufac-tured today (Figure 2).

Corrosion phenomena and cor-rosion-induced crustingIn contrast to glasses manufac-tured in the 19th and 20th centu-

Fig. 1. Medieval stained glassfrom the Marienstern Cister-cian Monastery, Panschwitz-Kuckau (1370/1380)

ries, the chemicalcomposition of the col-ourless and colouredflat glasses producedin the Middle Agesmade them highly sus-ceptible to weathering.Contact with high lev-els of humidity, precipi-tation and harmfulgases emitted fromdomestic, industrialand traffic sourceshave meant that thecoloured glasses havesuffered considerablechemical transforma-tion especially on theirexterior surfaces. Themedieval stained glassin church windows,which was exposed

unprotected to the weather forcenturies, often exhibit a hardcrust on the exterior surface. Asstained glass windows displaytheir full splendour only in trans-mitted light, the presence of thiscrust, which absorbs incident light,can impair the visual appreciationof the window to such an extentthat cleaning is essential if thework of art is to be enjoyed as itwas meant to be seen.Many types of glass have becomevisibly thinner over time due to aprocess of dissolution, and inmany cases the original paintedlayer on the exterior surface hasalso been lost completely. Viewedin cross-section normal to the sur-face, badly weathered glass of thistype displays three distinct layers:an unweathered glass core (1 to4 mm thick), followed by a thin gellayer with a thickness of between0.01 and 0.1 mm, and, finally, theoutermost corrosion-induced layer.This layer, which may have adepth of several millimetres, is fre-quently highly inhomogeneous in

26

structure; it may, in parts, have al-ready flaked off, but in many casesit adheres very firmly to thesubstrate (see Figure 3).The exterior crust is composed ofthe corrosion products gypsumand syngenite, i.e. it consists ofcalcium sulphate, potassium cal-cium sulphate as well as amor-phous silica. The gel layer whichforms by potassium, calcium andother components leaching fromthe glass whilst water penetratesinto the glass is thus almost solelycomposed of silica and water. Asthis porous layer of corrosionproducts absorbs water and harm-ful substances as well as marringthe beauty of the stained glasswindow, it should be removed, atleast in part.However, in removing this layerone runs the risk of penetrating

too far into the glass, thus damag-ing the unweathered core layer.

Experimental investigations intothe cleaning of the glass surfaceBy recording an electron micro-graph at a particular location onthe edge of a glass segment be-fore and after cleaning, one canestablish whether the corrosionproducts were removed safely orwhether the cleaning procedurewas too intense, resulting in dam-age to the gel layer or even to theintact glass core. Typically, anedge that is normally concealedunder the lead came is chosen.The edge of the glass segment isprepared by grinding off a fewtenths of a millimetre normal to thesurface of the edge and then pol-ishing the ground cross-section toyield a smooth surface. The grind-ing and polishing stages must bedone in the absence of water inorder not to alter the corrosion-

sensitive glass any further. This isthe main difficulty of the method;the preparation of an undamagedsample demands a great deal ofexperience.The effects of the various cleaningregimes can be systematicallycompared with one another by di-viding the sample into severalzones, each of which has beenuniformly cleaned with a differenttool or with a different degree ofintensity, and each of which ex-tends to the edge of the glasssegment. A cleaning method is re-garded as acceptable if it suc-ceeds in reducing the thickness ofthe corrosion-induced crust with-out damaging the glass core. Thegel layer should also remain un-damaged, as it acts as a diffusionbarrier for ion transport processesthat cause further weathering. Thegel layer can thus be regarded asa natural protective layer thatguards the unweathered glass core.

OthersOthers

Fig. 3. Electron micrograph of the cross-section of a sample of weathered medievalglass and the composition of the differentregions

Glass core

Gel layer

Corrosion-induced crust

Fig. 2. Typical composition ranges formedieval glasses (left) and glasses from the19th century (right)

ESMA-Analysis on medieval glass(main components)

1 Römich, H.; Jägers E.; Torge, M.; Müller W.;Adam K.; Reinigung - eine Gratwanderung[Walking the tightrope – cleaning historical stainedglass. In German] In: Restaurierung undKonservierung historischer Glasmalereien[‘Cleaning and Conserving Historical StainedGlass’. In German] Verlag Phillipp von ZabernMainz, 2000 ISBN 3- 8053-2648-3 S. 101-127

27

A B C D E

a) b)

b) right: after partially removing the corrosivecrust with a scalpel

Fig. 4a/b. SEM image of the marked crosssection of a sample of medieval glass:

a) left:original state

Fig. 6. Texture and grain size ofthe abrasives used:A: sodium hydrogen carbonate,

~50 µm (2);B:plastic granules,

60-80 µm (3);C: instantized flour, ~80 µm (6);D:walnut husk granules from

Canterbury,~80 µm (4);E: walnut husk granules

125-400 µm (5)(Figures in brackets refer tothe cleaning zone in whichthe respective abrasive wasused, see Fig. 5)

Corrosion-induced crusts can beremoved with a variety of tools andtechniques depending on theircomposition and the strength towhich they are bonded to theglass. If the layers are loose, abrush or similar soft implement issufficient to remove them. Moretightly bound layers are very care-fully removed by experienced re-storers using a scalpel whilstmonitoring progress under astereo microscope.The scanning electron micro-graphs in Figures 4a/b show thereduction in the thickness of theexterior crust achieved using ascalpel and without damaging thegel layer. The preparation (grindingand polishing) of the edge of theglass segment was performedprior to the cleaning process. Inaddition, a force of 1 N was ap-plied to a Vickers diamondindenter (monitored under a micro-scope) to make pyramid-shapedindentations in the glass coreclose to the gel layer. The marksenable reliable assessments of thevarious cleaning procedures to bemade as the state of the surfacecan be recorded at exactly thesame location before and aftercleaning1 .

In England (Canterbury), good re-sults have been achieved by usinga microblast jet machining tech-nique (Airbrasive) to remove corro-sion-induced crusting. The effec-tiveness of the cleaning process(or, in unsuccessful attempts, theseverity of any damage done) de-pends not only on the jet param-eters (pressure, distance, type andgrain size of the abrasive, etc.),but also to a large extent on thestructure of the original sample,specifically on the structure of thecorrosion-induced crust. It istherefore important to select glasssamples that cover a range of dif-ferent structures when carrying outcomparative testing. In Germany,the removal of the exterior crustusing this microblast jet abrasionmethod has only been tested un-der laboratory conditions. Themain results and the conclusionsreached have been published else-where2 and will only be mentionedin the following for the purposes ofillustration.

Figure 5 shows a sample of redglass from the window n III(triforium, tracery) in Cologne Ca-thedral, Germany. The glass has arelatively uniform, light colouredcorrosion crust on its exterior sur-face. The corrosion layer on thissample was only moderately hardso that a pressure of at most0.6 bar was sufficient with all thejets tested to remove the layer. Adifferent abrasive was used foreach of the cleaning zones (2-6)and compared to the resultsachieved using a scalpel (zone 1).The size of the jet nozzle is chosento match the grain size of the ab-rasive used (Figure 6). The abrasi-ves used were: sodium hydrogencarbonate (2) and plastic granules(3), both delivered through a noz-zle with a diameter of 0.6 mm;English granulated walnut husk (4)and cereal flour (6) using a nozzlediameter of up to 1.2 mm; andGerman granulated walnut husk(5) delivered via a 1.4 mm nozzle.Guided by results achieved in pre-

Fig. 5. Glass sample from Window n III, Colog-ne Cathedral showing the test cleaning zones

1 2 3 4 5 6

Cleaning segments

2 Historische Glasmalereien ; Schutzverglasung-Bestandssicherung-Weiterbildung [Historical Stained Glass Windows;Protective Glazing – Conservation – Further Training. In German] Project sponsored by the Deutsche Bundesstiftung Umwelt[German Federal Environmental Foundation] Berlin-Brandenburgische Akademie der Wissenschaften [Berlin-BrandenburgAcademy of Sciences] Verlag Edition Leipzig, 1999 ISBN 3-361-00500-0

28

▲

a)

b)

a) b)

liminary investigations, other pa-rameters varied were: the distancebetween the nozzle and the target,the jet angle, the duration of abra-sion and the vibration settingwhich controls the flow of abrasiveto the nozzle opening. Figure 7shows the sharply delineated clea-ning zones on the edge of thesample. The areas cleaned withthe fine abrasive jet are clearly vis-ible. The surface profile in zone 2indicates that the thickness of thecorrosion-induced crust has beenreduced by 80–100 µm. Compa-red to the uncleaned area, thetransparency of the glass in zone 2has increased significantly. Theelectron micrographs of the edgeof the glass segment before andafter the cleaning procedure showthat the gel layer is still presenteven after the partial removal ofthe crust and that damage to theunweathered core of original mate-rial can therefore be ruled out (seeFigures 8a/b). By comparing thecleaned surfaces with the cross-sectional electron micrographs,one can determine the best meansof cleaning a particular corrosionlayer. Abrasive removal with a jetof flour or plastic granules provedto be quite coarse and irregular.Better results were achieved usinggranulated walnut husk, with themore finely ground variant beingmore effective. The best resultswere obtained with sodium hydro-gen carbonate. However, these re-sults only apply to this specificcorrosion layer. For other layershaving a different chemical com-position, preliminary tests will berequired to determine the optimummethod of cleaning the surface.Corrosion-induced crusting that isvery inhomogeneous and stronglybonded to the glass often cannotbe removed mechanically withoutdamaging the unweathered core

Fig. 8a/b. Glass sample from CologneCathedral. Electron micrograph of cross-section through cleaning zone 4 before (a: left)and after microblast jet cleaning with walnuthusk granules (b: right)

Fig. 7.Glass sample from Cologne Cathedral;cleaning zone: 2; abrasive: sodium hydrogencarbonate

Fig. 9a/b.Cross-sectional electronmicrograph recorded in1993 (at the locationmarked) of a sample ofmedieval glass from ErfurtCathedral before (a: left)and after removal of thecorrosion-induced crustwith cleaning compresses(b: right). Imagesrecorded 1993

Fig. 10.Cross-sectional electronmicrograph recorded in2000 (at the locationmarked) of the samesample of medieval glassfrom Erfurt Cathedral

29

a) b)

38

glass. As part of a joint project, achemical cleaning method was de-veloped, tested and then used torestore an original segment ofstained glass from a window inErfurt Cathedral, Germany3 . Byusing special compresses soakedin an ammonium carbonate solu-tion, the transparency of thestained glass could be improvedsignificantly. A crucial factor in thedecision to use this method wasthat no immediate or long-termdamage to the original windowmaterial was to be expected.The cleaning compresses can beused to partially remove corrosionproducts. Using this technique,unweathered glass elements be-low the weathered outer crustcould be revealed (Figure 11)Electron micrographs of theground and polished cross-sec-tion, recorded at the same pointon the sample, before cleaning(Figure 9a) and after cleaning(Figure 9b) show that the structure

of the gel layer remains unchan-ged. The Vickers indentation onthe edge of the glass enables thepoint of examination to be identi-fied years later. Figure 10 showsthe examination area seven yearsafter removal of the corrosionproducts with the chemical com-presses. The morphology of thegel layer is unchanged (c.f. Figu-res 9a/b) so that subsequent cor-rosion of the glass can be ruled out.

Today, historical stained glass win-dows are preserved behind an ex-ternal glass screen. This ‘isother-mal’ protective glazing with itsventilated air space has been in

Fig. 11.Microscopic image of thecleaned exterior surfaceshowing pitting cratersand exposedunweathered glass

Fig. 12a/b. Photograph taken with transmitted light of a segment of medieval stained glass from Erfurt Cathedral (1405) before(a: left) and after removal of the corrosion-induced crust with cleaning compresses (b: right)

Scientific Examinationsof Historical Stained Glass

use for decades in a large numberof countries, and has establisheditself as the most important meansof protecting and preservingstained glass windows4 .

4 Historische Glasmalereien ; Schutzverglasung-Bestandssicherung-Weiterbildung [Historical StainedGlass Windows; Protective Glazing - Conservation -Further Training. In German] Project sponsored bythe Deutsche Bundesstiftung Umwelt [GermanFederal Environmental Foundation]Berlin-Brandenburgische Akademie der Wissen-schaften [Berlin-Brandenburg Academy of Sciences]Verlag Edition Leipzig, 1999 ISBN 3-361-00500-0

3 Römich, H.; Jägers E.; Torge, M.; Müller W.; AdamK.; Reinigung - eine Gratwanderung [Walking thetightrope - cleaning historical stained glass. InGerman] In: Restaurierung und Konservierunghistorischer Glasmalereien ['Cleaning andConserving Historical Stained Glass'. In German]Verlag Phillipp von Zabern Mainz, 2000 ISBN 3-8053-2648-3 S. 101-127