The three broken 'vases' - a covered baluster jar and two yan yan type vases - are part of a matching set, or 'garniture', of five Fitzwilliam accession numbers:

1. C.17.1 & A-1948
2. C.17.2 & A-1948
3. C.17.3 & A-1948
4. C.17.4-1948 
5. C.17.5-1948.

The remaining pair of baluster jars, displayed in wall niches on either side of the stairs, was unharmed in the incident.

The vases were produced in late 17th or early 18th century China during the reign of Emperor Kangxi (1662-1722) of the Qing Dynasty. Intended to be decorative rather than functional, they are very large and extremely heavy. The baluster jar is 80cm high and weighs approximately 45.45 kg; it requires two people to carry it.

The yan yan vases are approximately 71cm and 72.5 cm high respectively. The ceramic walls are up to 3cm thick in places. Sets of such vases were usually made for the European export market. They were popular with a wealthy clientele for decorating the rooms and corridors of their large houses.

The damage revealed that the interiors of the vases are ridged, with a series of rings. These give clues to the way they were made: the pots were probably thrown, that is, made on a potter's wheel. The outside was then smoothed. 

It also revealed that the yan yan necks were each made as a separate piece that was then joined, or 'luted', to the sloping shoulder of the vase with 'slip', a watery paste of the same porcelain clay. These are weak points in the construction. One vase broke along these joins when it fell, detaching completely.

The lid finial was also made separately and luted to the lid. The finial is hollow, with a small air-hole cut into its base to allow moisture and gases to escape during the firing. Without this vent, it could explode in the kiln Lid finial

The yan yan vases already had some old restorations (joins, fills and repainted areas to replace the missing designs) on the bases and rims. The fills and overpaint on the rims had discoloured and yellowed with age. These yellow areas can easily be seen on the photo of the vases before they were broken. Most of the old fills broke off when the vases smashed.

The vases were given to the Fitzwilliam Museum in 1948 by Anthony de Rothschild through the National Art Collections Fund (now known as The Art Fund). They had originally been acquired from Cyril Andrade, 8 Duke Street, St James's, London, from whom purchased on 5 June 1925 by the donor.  Soon after their acquisition they were installed on the recessed windowsill of the grand staircase of the 1930s Smith and Brewer extension building, in line with the Fitzwilliam's distinctive house style. Here they remained on permanent display for decades, enjoyed by the Museum's 300,000 visitors a year.

The disaster happened on the landing of the grand stone staircase connecting the first-floor Flower Paintings gallery (Gallery 17) and ground-floor Islamic gallery (Gallery 33). The main flight of stairs rises from the ground floor to a central landing beneath a large window, then divides into two smaller flights left and right. The visitor tumbled down the right-hand flight of stairs, then along the windowsill from right to left, colliding with each vase in turn. The impact reduced them to rubble and scattered them across the landing and stairs.

The stone flags were gouged in places where he skidded on some of the hard porcelain sherds. Other fragments were crushed by staff coming to help him. The noise of the crash was immense and echoed through the galleries. Museum attendants and first-aiders quickly attended to the visitor at the scene, moments after the crash. Although an ambulance was called, he later walked away unharmed.

On Monday 30th January 2006, The Daily Telegraph newspaper published a photograph taken moments after the smash by another visitor with his mobile phone. It showed the man sitting amidst the devastation on the landing. The incident captured the public imagination and sparked a global media storm. The Museum was inundated with an unprecedented flood of queries from all over the world. The Fitzwilliam's small press team worked round the clock and cancelled leave.

The Museum has amassed several enormous files of press cuttings. The event became so well known that, by the time the restoration began, it continued to be lampooned in satirical political cartoons. References were even made to it in insurance adverts. Later, the German sculptor Thomas Demand was inspired to replicate the exact scene of the smash, including the stairs and window sill. His work was exhibited in his exhibition 'L'Esprit d'Escalier' at the Irish Museum of Modern Art in Dublin, Ireland in Spring 2007.

As the noise of the crash died away, the Museum's disaster response procedures swung into operation. While first-aiders attended to the visitor, staff sealed off the area to prevent the scattered pieces from being crushed underfoot (flecks of white porcelain are difficult to see on pale limestone stairs) and assessed the damage.

1. First the area was photographed, to record the relationships of the sherds.
2. The ground floor was examined for fragments. A few were retrieved. Then the stairs were systematically cleared, step by step, working from the outside in towards the landing, the main scene of devastation. Each step was given a number that corresponded to a tray or a bag of fragments. The landing was divided into a notional 'grid', with each stone flag given a letter. Each step or flag was cleared before moving onto the next one.
3. The pieces from each step or flag were carefully placed in padded trays labelled with the corresponding number or letter and cross-referenced to the grid.
4. It became apparent that fragments from the same vase, especially the left and centre vases, had fallen close to each other. This relationship was maintained as far as possible in the trays, to help reassembly. Particular care was needed in dismantling the jumble of fragments that had collected in the far left corner of the landing. Some of the larger sherds were stacked precariously on top of each other, imitating a giant game of jack straws.
5. Removing the wrong piece could cause a pile to collapse, causing further damage.
   
6. Where a sherd had fallen across a grid line, the photograph was annotated to show which tray it was placed in.
7. It took staff nearly three days to document the scene and retrieve every tiny fragment.
8. At the end, 24 large trays of fragments were collected and numerous small bags

The first jar to be hit, on the right of the sill, sustained the most damage and smashed into approximately 120 big sherds. The middle vase, the baluster jar, broke into around 100 large fragments, its lid into about 30 and the last vase to break into 80 main pieces. There were thousands of smaller additional fragments.

From the beginning, and despite press scepticism, the Fitzwilliam's small conservation team were confident the vases could be restored. However the Museum only has two conservators who work with 'three dimensional' objects (the others dealing with 'flat' art such as paintings, paper and manuscripts) and their work schedules were already full. This would be a long, time-consuming job. The work would need to be done by a specialist independent conservator. It would also be expensive. Happily, soon after news of the smash was reported, local firm Hewitsons Solicitors offered to sponsor the restoration.

Some press had misleadingly reported that the Museum did not know where to turn for assistance. As a result, the Museum was inundated with offers from conservators and restorers all over the world, including from Chinese conservators made through the British Embassy in Beijing. In fact, the Museum had already selected a specialist.

In March 2006 Penny Bendall, an experienced independent ceramics conservator, was appointed to carry out the conservation and restoration of all three vases. Mrs. Bendall trained at West Dean College and holds a Warrant to work on The Royal Collection. She brought 18 years experience of working on major Oriental ceramics collections around the world.

The conservator was presented with a room full of trays of sherds. She sorted out which of the larger pieces belonged to each vase by doing a draft reconstruction of all three at the same time. The shapes, curvatures and diameters of the three vases are all slightly different. So too are the underglaze and overglaze designs and the colour schemes, for example the rim and foot patterns differ on the two yan yan vases. The conservator uses these clues to match pieces and and put them together. The thicknesses also vary greatly - from 3cm near the base of the largest jar to less than 1cm near the rims. Pieces of the lid, flat bases and glazed rims were very easy to identify. The recovery system used meant that many of these were found adjacent to each other in the relevant grid squares/trays. The neck and base of the baluster vase that had sat on the left of the sill were largely intact.

The quality of the painting also varies. The painted enamels on the largest baluster jar is much coarser than that on the other pots  and the painted leaves only have one tone of green. In contrast, the yan yan vases and the baluster vase lid have both dark and pale greens. Penny occasionally referred to old photographs to examine details of the designs. Her years of experience meant she quickly mastered all three 3D jigsaws. At this stage, she also identified found the positions of many, though not all, of the smaller surface fragments.

From the reconstruction. Penny discovered that much of the surface design was intact and, except for the impact points, there were very few areas so damaged that they were beyond recovery. Some of the pieces had completely disintegrated, leaving holes. When she had finished, she was able to determine how many holes would need filling. Fortunately the reconstruction also showed that the porcelain had not 'sprung' significantly (see Springing below).

Amazingly, she was able to sort the fragments and reassemble all three vases in a single working day! The following day, the sherds were numbered in order of assembly, then the pots were dismantled for cleaning. The fragments were packed up and taken to the Bendall Ceramics Conservation studio.

Possible complications: Locking out

The temporary reconstruction also helped Penny identify the correct order of assembly for each vase. This is very important. If the fragments are put back together in the wrong sequence, some can be 'locked out' by their neighbours and can't be inserted. When this happens, the joins must be dismantled back to the point where the locked-out sherd can be included, then built back up again. The conservator constantly has to think ahead.

For large ceramics broken into many pieces, even an experienced conservator finds some tricky areas require several attempts before finding the correct sequence. You can see this happening for real in the time-lapse film of the reassembly of a yan yan vase, especially around the point of impact.

Possible complications: Springing

The temporary reassembly also confirmed that most of the vases had not 'sprung' as had been feared. Ceramics fired at high temperatures, such as porcelain, are often under tension. These stresses can be released when the object is broken. The fragments literally 'spring' apart and curves straighten out, making accurate reassembly almost impossible. Springing may affect a ceramic to such an extent that the distorted sherds cannot be realigned without applying considerable force (for example, using clamps) and sometimes not at all. This can complicate the restoration. Some conservators consider reshaping under tension unethical, because it reintroduces stresses that may cause new breaks in other areas. In these instances, the mismatched gap is filled to strengthen and stabilise the ceramic. It may be possible to disguise a narrow gap by retouching, but larger gaps will always look slightly strange.

In fact, one portion had sprung slightly. The top of one of the yan yan vases (originally positioned left on the sill) had separated as an almost complete fragment. A vertical crack down one side of the neck had sprung open by half a millimetre. Taping brought the edges of the join almost completely back together but the join was not perfect and had to be filled like other cracks.

What tape did she use to make the temporary joins?

Penny used a colourless transparent plastic self-adhesive tape, ('sticky tape') of the kind found in most stationers. This has sufficient strength and grab to temporarily support the heavy sherds.

How did she apply the tape?

It is important to position the tape properly when taping up joins or the fragments will not be supported and may sag. A ninety-degree angle across the join gives the greatest strength. The tape is pulled tight across the fragments. Tape is not applied over fragile areas of surface decoration, such as flaking enamels or gilding, although this is unavoidable for some small fragments. The tape can be peeled off cleanly from areas of intact glaze with the assistance of acetone. Any residues are also removed with acetone.

Where were the 'points of impact' on the vases? How many were there?

Areas which sustained the greatest damage, especially the 'point of impact' – the part which first hit the ground and took the weight of the fall – were often extremely fragmentary, with many losses. They are usually on the widest point of the vase, i.e. on the shoulders of the yan yan vases, the upper body of the baluster vase and the curve of the lid.

These were particularly difficult to reassemble. In some cases, for example on the baluster jar, the decorated outer surface had exploded into miniscule fragments and the thick porcelain wall splintered into wafer thin slices. Pieces of a reasonable size only survived on the inner surface. These had been fractured by typical star-shaped crack pattern that radiated out from the centre of the impact point. On the yan yan vases and the lid, even this inner wall was too fragmented to reconstruct, leaving each with a large hole in its side. The first vase that toppled (a yan yan) had two points of impact, one on either side of the body.

The fragments were soaked in a plastic bowl of hot water and washing powder for 20 minutes. The bowl was lined with a soft towel to cushion the pieces. Each fragment was then brushed with a good-quality hogs-hair stencil brush to remove the dirt. The solution must not be allowed to dry on the surface. The sherds were rinsed thoroughly in clean cold water, with a final rinse of deionised water. The process was then repeated. Finally, excess water was blotted off and the sherds left dry in a warm place. The fragments must be completely dry before reconstruction begins.

Most of the old restorations had broken away when the vases shattered. Some still had fragments of porcelain attached, here seen as the bright white fragments adhered to the underside of the restoration. The fill is an off-white colour. A few old fills still needed to be removed from fragments of porcelain, especially on the rim, or they would interfere with the new joins. Old painted fills were removed with a scalpel. After testing with various solvents, any residues were removed with dichloromethane.

Stubborn stains that resisted the washing process (such as an old discoloured hairline crack across the bottom of one of the yan yan vases and dirt on break edges) were broken down with a chemical poultice. A sausage of cotton-wool soaked in a very weak solution of a bleaching agent (hydrogen peroxide) was laid along the join and the whole area wrapped in food-grade plastic film to prevent evaporation. The poultice was left for 24 hours. Afterwards the sherds were rinsed thoroughly.

Caution! When working with chemicals the appropriate Health & Safety procedures must be followed and protective clothing worn. Observe the written COSHH procedures.

Avoid using any chemicals on porous ceramic bodies such as earthenware. Mechanical cleaning methods are preferable here.

The vases were reassembled as soon as possible after cleaning to prevent dirt settling on the cleaned break edges. The fragments were laid out in sequence so the pieces could be built upwards from the base. This time the joins had to be perfect. Although the assembly takes longer than the dry run, it is still a relatively fast process. The time-lapse in the interactive shows the actual complete reassembly of the most damaged yan yan vase (120 main fragments) from start to finish. It took 8 hours. Note how the conservator has to take some areas up and down to prevent locking out some of the fragments. The large baluster vase also took 8 hours. The assembly of the second yan yan vase (80 main pieces) only took 6 hours.

Penny used the same self-adhesive tape she used to make the draft assembly and applied it in the same way. (see Sorting). Once again she was careful to avoid any fragile or flaking surface decoration and apply the tape perpendicular to the join. Because the breaks were fresh and clean, they produced a very close fit when reassembled, rendering the joins invisible in some places. (The exception was the sprung neck section of one yan yan vase (see Sorting for information on springing) and areas of loss along some break edges. Finally, the tiniest flakes of glaze and painted enamel designs were inserted.

If the joins are not perfect there is a risk with very large heavy ceramics that the fragments will settle and move out of alignment. When this happens the whole object has to be taken apart, the tape residues removed and the pieces reassembled all over again.

Therefore, after assembly the conservator leaves them taped up for a few days before applying the adhesive. Unfortunately, this happened with the heavy baluster jar. After two days, the pieces settled and slumped and the whole vase had to be dismantled, re-cleaned and reassembled. There was no problem with the second reassembly.

The joins were made with a colourless 2-part epoxy resin (HXTAL™ NYL-1), a very fluid conservation-grade adhesive that crawls into very narrow cracks by capillary action. It does not shrink when it hardens. Epoxy resins are very strong, hard adhesives. Conservation-grade epoxies are suitable for repairing hard non-porous ceramics such as hard-paste porcelains. They are not appropriate for weaker or porous ceramics such as earthenware. HXTAL™ NYL-1 was chosen because it has better ageing characteristics than some of the other available conservation-grade epoxy resin adhesives. The cured translucent resin also has a similar refractive index to the glazes on hard-paste porcelain. It is this quality that makes cracks seem to disappear when it is applied.

The HXTAL™ NYL-1 is a 2-part liquid adhesive measured by weight in a 3:1 ratio. If not mixed accurately the adhesive will not set and the bonds may fail. The components were weighed out exactly, drop-wise on scales, before mixing thoroughly. A small metal sculptor's spatula was used to mix the adhesive and apply it to the taped cracks.

HXTAL™ NYL-1 has one principle drawback. The adhesive takes two weeks to harden. There is a risk that dirt will settle onto the wet surface, causing staining. To prevent this, the vase is covered loosely in acid-free tissue paper while it dries.

What are the desirable features in a conservation-grade adhesive?

• Bond strength appropriate to the ceramic to be repaired
• Compatibility (e.g. appropriate viscosity) with the ceramic to be repaired
• Good long term stability and reversibility, non-yellowing
• Good surface wetting (i.e. the adhesive coats ('wets') all parts of the surface to which it is appliedColourless ('water-white'), transparent
• Close match to the refractive index of any glazes
• Good working time

All adhesives will eventually discolour as they age, but conservators use those that remain stable for longer. In conservation timescales, good stability and reversibility means ideally using materials that will not discolour and remain reversible for a minimum of 100 years under museum conditions. This is not possible for most epoxy resins currently on the market. Most adhesives age and discolour faster if regularly exposed to strong light and heat, such as direct sunlight. Oddly, there is also evidence that some epoxy resins need some light to prevent them from discolouring, so repaired objects should not be stored in total dark, such as a cupboard.

The adhesive was the same 2-part epoxy resin, HXTAL™ NYL-1, used to bond the fragments (see Bonding). Fumed silica was used as the bulking agent mixed into it. Fumed silica is opaque, white, very finely ground and chemically inert.

4 powder pigments were mixed to match the glaze: titanium white, cerulean blue, raw umber and yellow ochre. Each was first mixed separately on a white tile with a small amount of the filler, then combined. The choice of pigments and the exact amounts used to achieve a colour match with the glaze are determined by experience. The colour of the glaze may vary, even on a single object. Each area must be matched using different batches of filler.

The conservator used two different types of fill on the vases. The smaller losses and cracks were all filled with HXTAL™ NYL-1 and fumed silica. However, a very few large losses that went through the thickness of the walls, such as the points of impact, required a slightly different approach.

HXTAL™ NYL-1/fumed silica filler is, by itself, too translucent and brittle to make a robust thick fill and could fail if stressed. So these fills were given a little extra strength by making an inner layer of a slightly flexible material. These were prepared in two stages. First UHU™ Plus, another epoxy resin with slightly greater flexibility and opacity, was kneaded with French chalk and titanium powder pigment to make a second filler.

A thin layer was inserted into the centre of the hole to make the core of the fill. It also acted as a backing layer for the rest of the fill. UHU™ Plus yellows faster than HXTAL, but this is not a problem as it is covered on both sides and would not be seen. Once hardened, the core fill was covered on both sides with the more translucent HXTAL™ NYL-1 filler, prepared as usual.

The following abrasives were all used on each fill, starting with the coarsest and proceeding to the finest, finishing with micro-crystalline wax. It took weeks to rubdown all the fills to produce a smooth glossy finish.

Abrasives (coarsest to finest) 

• Metal file
• Scalpel
• Flour paper (grades 0 and 00)
• Micromesh™ cloths (grades 3600 (coarse), 6000 (medium) and 12000 (fine))
• Solvol™ Autosol chrome cleaner, a fine abrasive paste
• Prelim™ Surface Cleaner, a fine abrasive paste
• Renaissance™ micro-crystalline wax

Penny Bendall used Nylon watercolour brushes (sizes 1 and 2) and the crushed pigments were mixed with Windsor and Newton™ Acrylic Glaze. The paints were a little too matt, so they were given a final coat of the same glaze to blend in with the glossy enamels.

None of the fills on the body were retouched with paint, only the missing areas of painted enamels. Although the filler for the body colour was tinted to match the colour of the glaze, variations in colour of both the glaze and the filler mean that not every fill is a perfect colour match. A few of these areas can be identified when seen close up. Some restorers completely cover their fills by spraying on the retouching paints. Usually the spray extends on either side of the fill to blend in with the original surface. This makes the repair 'invisible' but also covers up some of the original object. This method tends to disguise the restoration and may deceive the viewer. It is increasingly regarded as a less ethical approach for museum quality objects.

The artists who decorated the vase made a mistake when painting the red enamel flowers. Either the red enamel was applied too wet or it melted during firing, causing dribbling.

On 9th November 2007, all three vases went back on display in a specially designed case in the Flower Paintings gallery (Gallery 17), the upper room adjacent to the staircase where they were broken. The case allows the visitor to examine the vases closely from all sides.

The windowsill was no longer appropriate for displaying the restored vases. The location receives some direct sunlight which can accelerate the ageing and yellowing of the adhesive and fills. There is also a huge amount of visitor interest in the vases and stairs are an inappropriate place for large groups to congregate.

The remaining pair of baluster jars from the garniture can be seen to either side of the case, in the niches above the stairs. In early 2007, they were removed for safety during maintenance work on the windows. They were recently cleaned and returned to the niches.

Many of the smallest chips and fragments could not be easily incorporated into the restoration. The museum will keep these as reference material and for analysis. Technological advances in the future may mean that, when the vases next need restoring, many more of the tiny fragments can be reintegrated. Some fragments however were ground in to dust or tiny splinters and are permanently beyond reincorporation. These can however be used for analysis of the porcelain and enamels, to shed further light on the techniques used in their manufacture, such as the compounds used and the temperature they were fired at. Most modern analytical techniques require samples no larger than a pin head.

If you have ceramics that need conserving, The Conservation Register online http://www.conservationregister.com/  provides a free list of many reputable conservators throughout the U.K. Their website allows you to search for a conservator by geographic region or specialism (e.g. ceramics). 

Some conservator/restorers offer short courses for the interested amateur. You may find people offering courses in the local paper, or possibly by searching on The Conservation Register online (search under 'course' &/or 'ceramics' and contact individuals: if they do not offer course themselves they may be able to recommend someone else.

A reputable teacher will teach the basic ethics of conservation/restoration as part of the course.

The restoration had taken 7 months from start to finish. There were long intervals in between some stages, such as waiting 2 weeks for the adhesive and fills to cure (See Bonding and Filling). The baluster vase was the first to be completed. It took 4 months, from April to July 2006 and was first unveiled at the Fitzwilliam's Mission Impossible? exhibition in 2006. The retouching of the first vase took 6 days. The two yan yan vases were worked on in tandem and took 3 months.

Theoretically it would have been possible to reinstate every single fragment - except for those pieces which were ground into dust when trodden on, or exploded into so many tiny slivers they could not be reconstructed, such as the points of impact (see Sorting). The restoration would however have taken twice as long and cost twice as much, and some areas would still have required filling, including the old restorations. It was possible to replace all of the larger pieces and hundreds of the smaller ones, especially those from the inner and outer glazed surfaces. A very small proportion of the total surface area was not reinserted. Many of those fragments not replaced came from areas of the porcelain wall that had completely delaminated, i.e. separated into a series of brittle, wafer-thin slices.

If you have been inspired by this story, there are many ways in which you can help us.

Admission to the museum is free and we welcome 400,000 visitors per year through our doors, plus over 430,000 users to our website.

We are a charity and rely on supporters and donors to help us achieve our purpose: To contribute to society through the pursuit of education, learning and research at the highest levels To preserve and extend our world-class collections To offer exhibitions and public programmes that engage the widest possible audience.

Thank you for your support.

Chinese porcelain

• John Carswell, Blue and White - Chinese Porcelain Around the World (British Museum Press, 2000)
• Margaret Medley, The Chinese Potter - A Practical History of Chinese Ceramics (Phaidon Press, Oxford, 1976)

Modern ceramics conservation

• Lesley Acton and Paul McAuley, Repairing Pottery and Porcelain: A Practical Guide (2nd edition, Lyon Press, 2003)
• Lesley Acton and Natasha Smith, Practical Ceramic Conservation (The Crowwood Press, 2003)
• Susan Buys and Victoria Oakley, The Conservation & Restoration of Ceramics (Butterworth-Heinemann, 1993)
• David Everett, Manual of Pottery and Porcelain Restoration (2nd edition, Robert Hale Ltd, London, 1994)
• Nigel Williams, Porcelain Repair and Restoration - a handbook (2nd edition, The British Museum Press, 2002)

To the many Fitzwilliam colleagues who helped with the disaster response, recovery of the fragments and worked with the media, especially the front of house team, the Photographic Department, the conservators, Press and Marketing and the Applied Arts Department; and to the IT Department who worked on these webpages.

The conservation and restoration of the Qing vases was made possible thanks to the generosity of Hewitsons Solicitors, Cambridge.

License and source code

The code for this story is available under GPLv3, the content is copyright to the University of Cambridge CC-BY-SA unless attribution states otherwise. This story was generated from original HTML source code and flash files from 2006 by Daniel Pett.