Available online at

www.sciencedirect.com

Journal of Cultural Heritage 9 (2008) e73ee80

Original article

A new cleaning method for historic stained glass windows

Sonia Murcia-Mascaros a,*, Paola Foglia b, M. Laura Santarelli b, Clodoaldo Roldan a,Rafael Iba~nez a, Alfonso Mu~noz c, Pablo Mu~noz c

a Materials Science Institute of Valencia University, IUCMUV. Valencia, Spainb Interdepartmental Center for the Science and Technology of Historical and Architectural Conservation, CISTeC.

‘‘La Sapienza’’ University, Rome, Italyc Vetraria Mu~noz de Pablos S.L. Ctra Arevalo Km 13, Segovia, Spain

Received 30 March 2008; accepted 5 August 2008

Abstract

Historical stained glass has a clear tendency to form a crusted layer on its surface due to the environmental exposure. One of the most delicateaspects to be faced during the restoration of historic glass windows is the cleaning of these thick corrosion crusts.

For several centuries, stained glass windows have been cleaned using damaging mechanical (scalpel) and chemical (high acidic or alkalinesolutions) methods. Today’s understanding of the cleaning process comprises two complementary aims: improving the readability of the glassand reducing the weathering process of the historical glass. The act of removing deposits and encrustations resulting from corrosion should notendanger the artwork itself. Mechanical methods, cleaning solutions or gel pads are now being developed. However, these methods could presentfurther problems.

In this study, we examine a new cleaning method that can be employed to remove encrustations in a quick and efficient way. Results up tonow, obtained on specific stained glass windows are promising; further researches are in process for other cases. We propose an optimizedsolution to dissolve calcium carbonates and lead sulphates from Avila Cathedral glass windows crust. This system is tailored to control pH,temperature, conductivity and concentration of Ca2þ. Continuous on-line analysis of these parameters allows us to monitor the cleaning process.In particular, the Ca2þ concentration in the cleaning solution is controlled by means of a Ca2þ ion selective analyzer.� 2008 Elsevier Masson SAS. All rights reserved.

Keywords: Glass cleaning; Glass crust; Accelerated glass weathering; Grisaille; Laminar flow; Crust solubility

1. Introduction

In the mid XII century, some glass workshops emergedwithin Spanish cathedrals such as Leon and Toledo. Subse-quently, independent workshops were developed in Burgosand Seville. Stained glass had been imported into Spain, whichexplains the presence of German, Flemish and French glassartists who worked and taught their craft in these workshops[1]. Nowadays, Spain preserves relevant artworks made bygreat Flemish artists, although in some cases, their state ofconservation is precarious due to the fragility of both the glassand its pictorial layers.

* Corresponding author. Tel.: +34963544547; fax: +34963543633

E-mail address: sonia.mascaros@uv.es (S. Murcia-Mascaros).

1296-2074/$ - see front matter � 2008 Elsevier Masson SAS. All rights reserved.

doi:10.1016/j.culher.2008.08.008

Glass decoration is accomplished by applying a mixturecalled grisaille which consists of a low-melting glass powder,a binding agent and colouring pigments of iron and copper.Grisaille is used not only to paint details on faces, hands, clothsor patterns, but also to modify glass tonalities. After applyingthe mixture, the glass piece is again heated at around 600 �C. Ifthe firing and/or the composition are not appropriate, tensionswill arise, causing the detachment of this pictorial layer [2].

One of the most delicate stages in the restorationeconser-vation of stained glass is the cleaning process, whereby bothslightly-adhered debris (usually, inside painted glass faces)and strongly-fixed crusts (usually, outside on non-paintedglass) of crystallized materials should be removed from theglass surface without irreparably damaging it or its grisaille.The aim of this removal of surface crust is to avoid further

e74 S. Murcia-Mascaros et al. / Journal of Cultural Heritage 9 (2008) e73ee80

artwork deterioration and to allow its iconographic reading.Nevertheless, this process should not be aimed at giving shineor transparency to the glass, since this would ‘‘strip’’ theweathered protective layer surface of all the protection,making the glass piece more vulnerable to future attack andremove possible thin original grisaille layers [3]. Furthermore,original stained glass is likely to have been almost completelypainted with more or less dense cold-applied patinas whichmodulate light, thus avoiding glare.

The durability of a glass and, subsequently, the crusts thatwill form on its surface depend on a series of factors such as:glass composition (relationship between formers, modifiersand stabilizers), production process; environmental parame-ters; previous restorations, and biological attack.

The chemical stability of a glass depends on its composi-tion, the presence of water on its surface and on the pH of themedium where it is located. In an acidic medium, the alkaliions migrate to the solution (dealkalinisation or leaching) andform leached hydrated silica glass layer. The alkaline mediumis much more aggressive because the break down of the glassnetwork occurs very quickly [4].

Historical stained glass windows are in permanent contactwith water. Condensation of environmental humidity generallytakes place in the form of dew drops on the inside glass face,whereas the outer side of the glass is affected also by rainwashings. The state of conservation that each side presents isusually very different. Also the paint layers are very different.The most of the paint is on the inside of the windows and thecrusts are mainly outside. This directs cleaning actions: trans-parency can be achieved by cleaning the crusts outside but moregentle cleaning actions are required for cleaning painted areas.Apart from that, crusts or dirt inside are never so dense anddifficult to clean, so that different cleaning methods could be usedon cleaning inside and outside surfaces. The reason for thisdifference in the degree of surface alteration is based on ionexchange reaction which occur between the water and the glass,generating significant changes in pH and favouring the processesof dealkalinisation and break down of the glass network.However, a shorter time of wetness does not allow a chemicalbalance reducing glass leaching [5]. In fact, by using a convenientsolution volume during a short time, insoluble salts could besolved. Based on this, we designed a system to remove corrosioncrusts consisting of the use of an over saturation volume ofneutral or slightly acid solutions, allowing them to flow on oneside of the glass to avoid saturation and changes in surface pH.

Methods for cleaning stained glass should be chosen accord-ing to the composition of both the glass and its crusts. Most of thereported crusts have a composition consisting of silica (30e50%),SO3 (17e28%), CaO (9e15%) and K2O (10%) [6]. Poorlysoluble salts, which darken the glass, are mainly SiO2 xH2O(hydrated silica), CaSO4 2H2O (gypsum), CaSO4 1/2H2O(basanite), K2SO4 (arcanite), K2Ca(SO4)2 H2O (syngenite),K2Pb(SO4)2 (palmierite), PbSO4 (anglesite) and CaCO3 (calcite).Nitrates, other carbonates and other sulphates are more solubleand do not form crusts [7]. Grisailles also develop crusts, gener-ally consisting of CaCO3, CaSO4 and/or PbSO4. If the grisailleis also detached, the cleaning process can be very difficult.

In any case, a previous crust analysis is needed for everyconservationerestoration on glass windows. A careful adap-tation of cleaning conditions has to be made for every case.

Once the state of conservation of glass and adherence of thegrisaille has been established, the least fixed debris can beremoved using very soft natural bristle brushes. The mostadhered deposits must be removed using carefully selected‘‘chemical’’ methods as complexing solutions (EDTA, thio-sulphates, citrates, etc.) or gel pads, which allow greatercontrol over the application areas. This method is highlyeffective for cleaning very small and damaged areas [8].However, the application conditions must be carefullycontrolled to avoid dissolution of lead and iron cations thatcould affect the grisaille or the calcium in the glass. Gel padsand complexing solutions may leave some debris that mightend up damaging the glass in the long term [9,3]. Also lasercleaning has been studied, concluding that cleaning of stainedglass windows with lasers is possible within limits [10].

The choice of a cleaning method also depends on the fact ifthe window has to be dismantled to replace the lead cames. Inthis case the glass pieces can be cleaned before placing themback in a new lead structure. Conservation ethics require,however, that the lead cames shall be kept as long as possible andfor these cleaning campaigns, a method which allows cleaning‘‘in situ’’ of the glass-lead composition has to be applied.

2. Research aims

In view of the difficulties to fix grisailles, and the currentlack of effective and guaranteed methods to do it, the methodproposed here allows one to clean only unpainted sides of thestained glass being restored, avoiding thus any intervention onthe damaged grisaille.

Application conditions were optimized using current glasspieces and grisailles with diverse manufactures and composi-tions which had previously undergone an accelerated weath-ering process. The optimized method was subsequentlyapplied to clean the outside of stained glass window fromAvila Cathedral which is currently being restored by the studioVetraria Mu~noz de Pablos, S.L. This method requires that theglass to be taken out from its lead structure.

The relative composition of all materials, as well as thecrystalline phases of the corrosion products were determinedusing portable and/or non-destructive methods for analysiswithout sampling, so that they could be applied to real samplesof cultural heritage artworks. The techniques used in this workmeet several of these requirements. We applied portable X-rayfluorescence analysis (EDXRF) and focused-XRD diffractionsystems. Some destructive analyses (SEM-EDX) of model glasshave been done to prove the absence of leaching after cleaning.

3. Experimental section

3.1. Instrumental techniques

Non-destructive focused diffraction analyses (f-XRD) werecarried out using a SEIFERT XRD 3003 TT diffractometer,

e75S. Murcia-Mascaros et al. / Journal of Cultural Heritage 9 (2008) e73ee80

with BrageBrentano geometry, ThetaeTheta configuration,CuKa1/a2 single point configuration tube, working at 40 kV and40 mA. Focussing was carried out on a 0.6e0.8 mm area bymeans of a 210 mm collimator and an XYZ automatic table. Aposition-sensitive detector MBraun PSD-500 was placed at185 mm in the secondary axis. This set up allowed us toperform non-destructive focused-XRD scans on differentpoints of interest in the same specimen.

A portable-XRF system with a Pd tube operated at 35 kVe250 mA and a Peltier cooled Si-PIN detector (FWHM¼ 190 eVat 5.9 keV) was used for the non-destructive analysis of samples.The XRF spectra were accumulated over 300 s and the WinQ-XAS (version 1.2, IAEA, 2000) code was used for the spectraanalysis. Semiquantitative analysis was done using relativeamount of Ca and K content.

HITACHI mod. S-4100 Scanning Electronic Microscopywith a BSE AUTRATA detector, BRUKER Esprit 1.8 imagecapture and RONTEC camera analysis system was used tocheck the integrity of grisaille and glass after cleaning.

3.2. Analytical study of glass samples, grisaillesand crusts

Before removing the crusts from the glass, the first stepconsisted on the analysis of all the materials in order to knowtheir composition and decide which substance was the mostappropriate to dissolve these encrustations.

The optimized method was applied to the dense crustedglasses of the stained glass window HI, one of the oldest inAvila Cathedral (late 14th Ceearly 15th C) [11,12]. Due to theinstability of the lead frame, the conservators decided that itneeded replacement.

For this study, we have used test specimens of glass(0.06� 0.05 m). These glasses, usually employed in therestoration of stained glass windows, were facilitated byVetraria Mu~noz de Pablos. Shown in Table 1 are their originand manufacture process.

Lead salts had been found on HI glass crust. Usually, thissalt appears only over the grisalle but in this case we found italso on glass; probably the lead originates from the lead frame.We used weathering of actual grisalle to simulate a crustsimilar to the one found on HI glass crust.

Only one of the original glasses analysed contained a highamount of potassium, the rest being soda glass samples. On

Table 1

Description of model glass samples

Sample Colour Type Origin

M-0 Transparent Float Spain

M-3 Red Double-rolled USA

M-4 Blue Rolled plate USA

M-9 Blue Blown France

M-11 Purple Rolled plate USA

M-12 Yellow Double-rolled France

M-14 Peach Blown France

M-18 Green Blown France

M-21 Grey Blown France

one side of the model samples, Vetraria Mu~noz de Pablos hadpainted and heat-treated at 620 �C two lines with currentlyused grisailles: 49-R/3 (49R3), of English origin, and Debitusnumber 2 (Dbn2), of French origin.

3.3. Accelerated weathering process on current glass

The accelerated weathering process was designed to repro-duce the formation of the most insoluble salts found on historicalstained glass windows in Avila: calcium carbonate and leadsulphate. To form these salts, a specifically-designed chamberwas used under the following conditions: humidity z100%,environmental temperature (20e23 �C), pressure 2 atm,300 ppm CO2, and 300 ppm SO2. We carried out three cycles of15 min each 10 min in a SO2-saturated and wet atmosphere,followed by three CO2 cycles under the same conditions. Thetotal treatment time for each sample was 48 h, with an exposuretime of 18 h and 200 ml/min to each gas [13].

We will take into account only the most insoluble salts foundon HI Avila glasses, calcium carbonate (Ksp¼ 3.7� 10�9 at25 �C) and lead sulphate (Ksp¼ 2.53� 10�8). Two factorsaffecting the solubility of CaCO3, are a decreasing temperatureand an increasing carbon dioxide pressure. When CO2 partialpressure increases, the pH drops, and much of the carbonate ionis converted to bicarbonate ion, which results in a higher solu-bility of Ca2þ. The concentration of carbonate drops by abouta factor of 3 for a pH drop of 0.5. Also lead sulphate is poorlysoluble in pure water, but in this case, solubility increases withincreasing temperature and decreasing pH. All of these factorsdetermine the working conditions for their dissolution, sinceboth salts could be dissolved in water and other glass-savingappropriate solutions.

3.4. Method design

The proposed cleaning solution contains controlled deion-ised water and NaNO3. The deionised water used in thepreparation of solutions has a slightly acid pH (5.5e6.1), lowionic conductivity (2e6 mS/cm) and [Ca]¼ 0.4e0.6 (ppm). Inthe presence of 5 ppm of NaNO3, conductivity increases up to20e35 mS/cm.

The system works at atmospheric pressure because underthese conditions, the amount of CO2 in equilibrium with air isthe appropriate to avoid changes in pH. In the absence of air,16 mg/l of calcium carbonate are dissolved, whereas in contactwith the atmosphere (PCO2 ambient air¼ 3.5� 10�4 atm)40 mg/l are dissolved. The solution was air conditioned until itreached constant pH and conductivity values. The amount ofenvironmental CO2 was thus kept constant during the cleaningprocess and did not affect the monitoring measurements of thesalts dissolution.

The choice of temperature depends on the salts contained ein this case, CaCO3 and PbSO4 presented opposite solubilitybehaviours in relation to temperature (CaCO3: 70 mg/l at 0 �C,50 mg/l at 15 �C and PbSO4: 33 mg/l at 0 �C, 44.5 mg/l at25 �C). Since the amount of carbonate was higher than that ofsulphate, we decided to fix temperature at 15 �C.

Fig. 1. Cleaning system prototype.

e76 S. Murcia-Mascaros et al. / Journal of Cultural Heritage 9 (2008) e73ee80

The last consideration refers to the recirculation of thesolution. Glass leaching will be taking place as long as theglass is in contact with the aqueous medium, even wetatmosphere [14], but recirculation avoids the damagingchanges in pH which results from leaching into a small finitevolume of water (such as a droplet of condensed water), andminimises overall treatment time by constantly refreshing thereagent in contact with the glass surface to avoid depletion inthe solution. Consequently, we decided to design a systemwhich kept the washing solution in motion, simulating theeffect of rain on the outer side of the stained glass window [7].The solution moved on one side of the glass surface subjectedto laminar flow to avoid turbulence and damage on the paintedside. Thus, the pH of the solution in contact with the glass wasalways the same, pH between 5.5 and 6.1.

K (p

eak area)

0

5

10

15

20

25

30

0 5 10 15 20 25 30 35

Ca (peak area)

HI M-0 M-11 M-12 M-14M-18 M-21 M-3 M-4 M-9

Fig. 2. Fluorescence peak areas of potassium versus those for calcium of Avila

Cathedral HI stained glass windows and model glasses by portable-XRF

analysis.

3.5. System design

Based on the parameters described above, we designeda system consisting of the following parts:

- A bath in which to position the glass piece. The materialchosen to build it was poly(methyl methacrylate)(PMMA), because it has good mechanical resistance; it islight, transparent, economical and does not contribute ionsto the medium. Dimensions: 0.30 m long� 0.21 mwide� 0.14 m high, with a volume of 0.09 m3.

- Bolts on which to rest the glass piece, also built withPMMA. These are keel-shaped, so that they do not inter-fere with the flow, and mobile and their height depend onthe volume of solution and on the glass thickness.

- Cryogenic bath with ethylene glycol at 5 �C, which coolsthe solution to 15 �C.

- A pump to the recirculation of the solution within the bath.- A flowmeter (Aalborg Instruments & Controls) which

allows regulation of the flow at 300 ml/min.- Electrodes to continuously measure Ca concentration (ISE

electrode 9660 Crison), pH (GLP 22 Crison), ionicconductivity (GLP 32 Crison) and temperature.

Table 2

Components of the crust layers on HI glass outside pieces by portable-XRF

and focused-XRD analysis

Glass colour Compound

Black SiO2

Blue SiO2

Brown K2SO4, SiO2

Brown K2SO4, CaCO3

Green K2SO4

Green K2SO4, CaCO3

Red CaCO3, CaSO4

Magenta SiO2

White K2SO4, CaCO3, SiO2

Yellow CaCO3, PbSO4

Fig. 3. Focused-XRD diffraction patterns of (a) HI glass crust (anhydrite: CaSO4, arcanite: K2SO4, aragonite: CaCO3) and (b) weathered model glass crust over the

grisaille (anhydrite: CaSO4, anglesite: PbSO4, hematite: Fe2O3).

e77S. Murcia-Mascaros et al. / Journal of Cultural Heritage 9 (2008) e73ee80

The volume of solution wetting the glass surface is 3 l(Fig. 1).

To avoid wetting of the original painted surface, the fluidsurface must remain still. For these reason, glass pieces rest onthe bolts and laminar flow is created by a distribution systemconsisting of two superimposed PMMA layers. The fluxshould have a rate of flow of 5� 10�6 m3/s and a surface of7.1� 10�4 m2, the flow rate of the solution at the end of the

0

1

2

3

4

5

6

7

0 5 10 15 20 25time (h)

pH

0

0.5

1

1.5

2

[C

a] p

pm

a

Fig. 4. (a) pH and Ca2þ concentration values vs. time and (b) conductivity

pipe was 0.7� 10�2 m/s, so Reynolds number was 184.37(Re< 2100 laminar flow).

4. Results and discussion

The glass composition of most of the glass from the HIstained glass windows from Avila Cathedral and glass modelswere obtained by XRF measurements. Fig. 2 shows peak

0

2

4

6

8

10

12

0 5 10 15 20 25time (h)

Co

nd

uctivity [m

s cm

-1]

b

vs. time, of non-weathered model glass in contact with distilled water.

0

10

20

30

40

50

0 10 20 30

Co

nd

uctivity [m

s cm

-1]

0

0.5

1

1.5

2

2.5

3

[C

a] p

pm

a

e78 S. Murcia-Mascaros et al. / Journal of Cultural Heritage 9 (2008) e73ee80

fluoresce areas of potassium versus the peak areas fromcalcium. The K and Ca composition of the model glasssamples is in the same range as the HI glasses.

The grisailles used on model glass contained iron, copper,lead and silica. Grisaille Dbn2 contained cobalt, chromiumand zinc as well. Focused-XRD analysis evidenced the pres-ence of hematite, magnetite and lead trioxide in grisaille 49R3,whereas only hematite was detected in grisaille Debitus.

The composition of the corrosion products on HI glasspieces is reported in Table 2. As can be observed, crustscontain potassium, calcium and lead sulphates, and calciumcarbonate. The presence of K2SO4 is caused by the potassiumcomposition of the glass. The major compounds are K2SO4

and CaCO3, PbSO4 is also present, whereas CaSO4 can onlybe found in some of the red glass pieces.

As a result of the weathering treatment, diverse crustsappeared on the model glass samples. These crusts were denseon grisaille Dbn2, light on the glass and little obvious ongrisaille 49R3. Focused-XRD analysis determined that thecrusts on grisaille Dbn2 consisted of anglesite (PbSO4). Thecrusts formed on the glass pieces varied according to theircomposition. As example: on M-3 potassium sulphate isdetected, on M-12 and M-4 we found anglesite and on M-21,calcium sulphate and carbonate. In Fig. 3a and b, two exam-ples of diffraction patterns from these crusts are shown.Focused-XRD analysis of cleaned surfaces showed theabsence of any of these crystalline phases.

System operation was optimized using two kinds of modelglass samples: glass samples that mainly presented calciumcarbonate crusts and glass samples that mainly presented leadsulphate crusts. In this last case, the grisaille face of the modelglass was used for the cleaning method optimization.

Initially, we carried out continuous measurements of pH,conductivity and Ca2þ concentration on both sides of the non-weathered model glass samples, with and without grisaille,

0

1

2

3

4

5

6

7

0 5 10 15 20 25time (h)

pH

0

0.5

1

1.5

2

[C

a] p

pm

Fig. 5. pH and Ca2þ concentration values during cleaning of M-11 (z4 mg

CaCO3).

using only distilled water under the aforementionedconditions.

Results are shown in Fig. 4a and b. Whereas pH and Ca2þ

concentration values do not vary noticeably, conductivityincreases indefinitely. This variation is higher on the grisailleside. These measures show both the widely-known glassleaching [15] e even under controlled conditions e and thefragility of the glass pictorial layer. To avoid leaching of Na,K, Fe or Pb ions, increasing of solution ionic strength isneeded. To this aim NaNO3 solution is used because it alsoincrease lead sulphate dissolution [16]. This salt is highlywater-soluble (81.5 g/100 ml at 15 �C), does not contributeions alien to the glass and does not cause significant changesin pH. The optimized NaNO3 concentration had been found on5 ppm. Initial values of conductivity and pH are slightlysuperior: 35 mS/cm and 5.8, respectively. Under these

time (h)

0

10

20

30

40

50

60

0 20 40 60time (h)

Co

nd

uctivity [m

s cm

-1]

0

0.5

1

1.5

2

2.5

[C

a] p

pm

b

Fig. 6. Conductivity and Ca2þ concentration values during cleaning of (a) M-18

(z12 mg CaCO3) glass and (b) M-4 (PbSO4) glass.

0

15

30

45

60

75

0 5 10 15 20 25time (h)

Co

nd

uctivity [m

s cm

-1]

0

5

10

15

20

25

[C

a] p

pm

Fig. 7. Conductivity and Ca2þ concentration values during cleaning of ori-

ginal glass piece from the stained glass windows of Avila Cathedral

Vcleaning solution ¼ 51.

e79S. Murcia-Mascaros et al. / Journal of Cultural Heritage 9 (2008) e73ee80

conditions, the increase in conductivity is due only to the ioncontribution originated by crust dissolution.

In Fig. 5 pH and [Ca] values during cleaning of the non-painted face of M-11 glass (z4 mg CaCO3) are shown: pHreached a constant value after 2 h of washing, whereas all thecarbonate was dissolved after 20 h. Under these conditions, thepH measurement does not give any information about crustdissolution because the volume of solution was very high inrelation to the amount of salt-crust dissolved. However, initialand final pH in the washing solution was constantly measuredto check such stability.

Consequently, the continuous measurements of conduc-tivity and Ca2þ concentration were the best parameters valid

Fig. 8. Glass piece of the Virgin Maria face from th

to control the cleaning process of these glass pieces. Once theaforementioned parameters were established, we proceeded toclean the non-painted face of M-18 glass, which containedz12 mg of calcium carbonate crust (Fig. 6a) and the paintedface of M-4 glass, which had a lead sulphate crust on itsgrisaille (Fig. 6b). As can be observed, both Ca2þ concentra-tion and conductivity were stabilized after 20e25 h of contacttime with the cleaning solution for glass M-18. Glass M-4shows minimal variation in [Ca], with a possible presence oftraces of calcium carbonate and a light leaching, whereasconductivity increases, reaching a plateau after 50e60 h ofcleaning. Under these conditions of temperature, lead sulphateis more insoluble than carbonate, so it demands a longercontact time. In this case, only conductivity monitoring indi-cates the end of the cleaning process.

To sum up, we can hypothesize that, if we were able tocompletely dissolve encrustations of the two most insolublesalts under these conditions in approximately 20-50 h, weshould be able to dissolve more soluble salts such as gypsum,syngenite or potassium sulphate in the same time and underthe same conditions.

Moreover, the SEM-EDX microanalysis over the glasssurface and in the fresh cut bulk proved that the crust had beeneliminated. The amount of ions leached from the glass is sosmall that SEM analysis cannot detect a difference betweencomposition of surface before and after cleaning. Furtheranalyses are needed to detect also the thickness of thedegradation layer by SEM cross section. For the time being,the described method gives good results on the typology of thestudied glasses and is encouraging to carry out furtherresearch.

To prove our proposed methodology on the laboratory, weused an original unmatched glass piece from the stained glasswindows of Avila Cathedral. Focused-XRD analysis of itsdense crust revealed, in this case, the presence of calcite andpotassium sulphate. Conductivity and Ca2þ concentrationvalues reached their maximum after just 10 h of contact with

e HI stained glass window of Avila Cathedral.

e80 S. Murcia-Mascaros et al. / Journal of Cultural Heritage 9 (2008) e73ee80

the cleaning solution (Fig. 7), indicating the major presence ofCaCO3.

The laboratory prototype of this system was scaled up toenable the cleaning of outside faces of stained glass in theworkshop of Vetraria Mu~noz de Pablos. As an example of themethod’s efficiency on a greater scale, we can mention thecleaning of the outside faces of stained glass window HIof Avila Cathedral. Monitoring of conductivity and Ca2þ

concentration values allowed to completely clean the glassin less than 24 h of contact with the cleaning solution (Fig. 8).

5. Conclusions

The need to find new efficient and accurate methods toremove crusts from non-painted faces of historical stainedglass windows has led us to optimize a system that meets theserequirements. The reported system was designed to cleanstained glass pieces which present K2SO4, CaCO3 and PbSO4

encrustations, among others, and which have to be removedfrom the lead frame due to other requirements. Its applicationparameters were optimized in the laboratory using weatheredmodel glass samples. Finally, these parameters were applied toa workshop-scaled system and successfully used to cleanstained glass windows from Avila Cathedral. This proposedmethodology has to be followed for each restoration campaignbecause glass and crust composition depend on glass compo-sition and atmospheric environment and thus the methodmight need adaptation.

We employed diluted solutions of NaNO3 in deionisedwater with a pH between 5.5 and 6.1 at 15 �C subjected toa laminar rate of flow of 5� 10�6 m3/s on one of the sides ofthe glass piece, which had been laid horizontally in an opensystem designed to dissolve crusts. Continuous measurementof conductivity and Ca2þ concentration allowed us to establishthe removal time of these crusts between 24 and 50 h. In anycase, the maximum cleaning time could be decided also by thechange in the Ca2þ concentration slope. In the early part of thecleaning process the solution gathers calcium especially fromthe crusty deposits, after all the deposits are dissolved, thecalcium uptake could be due only to the glass surface e hencethe change in slope. To minimise the damage to the glass, thispoint indicates the best time to stop washing.

The system is highly versatile and its parameters can beeasily modified according to the kind of salts present in theencrustations, allowing us to work with different temperatures,solutions or laminar flows, and to immerse the glass piececompletely or clean just one of its sides. Furthermore, it ispossible to simultaneously measure the presence of diverseions which allow us to determine the end of the salt dissolutionprocess.

Acknowledgments

Our acknowledgements go to the Spanish Ministerio de Edu-cacion y Ciencia for economical support (Project Ref. MAT2006-04072). S.M.M. is funded by the program Ramon y Cajal. Theauthors wish to thank Pablo Pardo for the XRD analysis and to theICMUV technical staff for the execution of system prototypes.

References

[1] L. Perez-Bueno, La vidriera artıstica en la arquitectura cristina espa~nola,

in: Vidrios y vidrieras (Ed.), Alberto Martın, 1942, pp. 187e200.

[2] O. Schalm, K. Janssens, J. Caen, Characterization of the main causes of

deteroration of grisaille paint layers in 19th century stained-glass

windows by J.-B. Capronnier, Spectrochimica Acta, Part B 58 (2003).

[3] Guidelines for the Conservation and Restoration of Stained Glass, in:

International Committee for the Conservation of Stained Glass of

International Corpus Vitrearum at the XXII Colloquium of CVMA,

Nuremberg (September 2004).

[4] T.M. ElShamy, J. Lewis, R.W. Douglas, The dependence on the pH of the

decomposition of glasses by aqueous solutions, Glass Technology 13

(1972) 81e87.

[5] M. Verita, Modern and ancient glass: nature, composition and deterio-

ration mechanisms, in: R. Lefevre (Ed.), The Materials of Cultural

Heritage in their Environment, Edipuglia-Bari, 2006, pp. 119e132.

[6] H. Romich, E. Jagers, M. Torge, W. Muller, K. Adam, Cleaning:

a balancing act, in: A. Wolff (Ed.), Restaurierung und Konservierung

historischer Glasmalereien, Mainz, 2000, pp. 101e128 Version in

English: www.cvma.ac.uk/conserv/tl/cleaning.html.

[7] J.M. Fernandez Navarro, Constitucion quımica de las vidrieras y metodos

para su analisis y para el estudio de sus alteraciones, in: Proceedings of

International Seminar ‘‘Conservacion de vidrieras historicas: Analisis y

diagnostico de su deterioro. Restauracion’’, The Getty Conservation

Institute, Santander, 1994, pp. 85e113.

[8] A. Corallini, V. Bertuzzi, Il restauro dellevetrate, Nardini, Cap. 6, pp. 131e141.

[9] C. Altavilla, E. Ciliberto, S. La Delfa, S. Panarello, A. Scandurra, The

cleaning of early glasses: investigation about the reactivity of different

chemical treatments on the surface of ancient glasses, Applied Physics A:

Materials Science and Processing 92 (2008) 251e255.

[10] H. Romich, K. Dickmann, P. Motter, J. Hildenhagen, E. Muller, Laser

cleaning of stained glass windows e final results of a research project,

Journal of Cultural Heritage 4 (2003) 112e117.

[11] M. Melcher, M. Schreiner, Leaching studies on naturally weathered potash-

lime-silica glasses, Journal of Non-crystalline Solids 352 (2006) 368e379.

[12] J.R. Duralde, Restauracion de la Catedral de �Avila: pasos actuales de un

largo camino, Restauracion y rehabilitacion, Revista internacional del

patrimonio historico 80 (2003) 24e31.

[13] C. Giavarini, M.L. Santarelli, R. Natalini, F. Freddi, A nonlinear model

of sulphation of porous stones: numerical simulations and preliminary

laboratory assessments, Journal of Cultural Heritage 9 (2008) 14e22.

[14] S. Fearn, D.S. McPhail, V. Oakley, Moisture attack on museum glass

measured by SIMS, Physics and Chemistry of Glasses B 46 (2005) 505e511.

[15] R. Conradt, Chemical durability of oxide glasses in aqueous solutions:

a review, Journal of American Ceramic Society 91 (2008) 728e735.

[16] I.M. Kolthoff, R.W. Perlich, D. Weiblen, The solubility of lead sulfate

and of lead oxalate in various media, Journal of Physical Chemistry 46

(1942) 561e570.