麻豆传媒

ON A SUNNY DAY, the atrium of the DG Bank building near the Brandenburg Gate
in Berlin is awash with light. It can be so bright that office workers seek
shade when they eat lunch here. Fortunately, 30 clouds floating a few metres
above the ground offer some protection. They are puffy, curvaceous and a
luminous white. And like clouds they appear weightless. Yet at two and a half
tonnes, they make up one of the largest all-glass sculptures in the world.

The sculpture is actually a unique chandelier that filters a mix of sunlight
and artificial light into the cafe and conference hall below. It was
commissioned in 1996 by the bank鈥檚 architect Frank O. Gehry, and is due to be
completed early next year when the final four panels are hung. You might not
think that creating big things out of glass would take rocket science, but the
chandelier at Pariser Platz 3 is different.

From the outset, the chandelier鈥檚 designer, San Francisco artist Nikolas
Weinstein, was in uncharted territory. 鈥淣o one had ever tried to put glass
together in this kind of way,鈥� he says. 鈥淭echnically, as well as aesthetically,
there was really not a lot of precedent for us to look to.鈥� The danger that
invisible cracks or a dropped wrench could shatter the glass into millions of
razor-sharp shards turned the chandelier from an ordinary sculpture into a
four-year odyssey that has challenged some of the world鈥檚 foremost experts on
glass.

The team Weinstein recruited included a man who designed windows for the
space shuttle, a builder of mirrors for the Hubble Space Telescope and a
physicist who models random processes. By the end, they had designed a brand new
kiln, developed a software package to assemble the 鈥渃louds鈥�, and broken an awful
lot of glass.

Walk into Pariser Platz 3 and you pass through a hallway into a breathtaking
atrium six storeys high. During the day it鈥檚 flooded with sunlight and, just as
at Gehry鈥檚 Guggenheim Museum in Bilbao, unusual shapes abound. At one end of the
atrium a billowy white shell that resembles a bed sheet blowing in the wind
shrouds an elevated meeting space. At the other end, an open-sided, arched
latticework of steel beams and triangular panes of glass defines a conference
hall and cafeteria within which the sculpture hangs.

Weinstein wanted his chandelier to complement this open design. He imagined
voluminous, curved panels swooshing skywards like cirrus clouds. But melting
huge pieces of solid glass to make the flowing shapes was out of the question.
The glass would probably cool unevenly in the kiln and crack, and even if it
survived that, such large pieces of glass might not support their own weight
once hung.

Instead, Weinstein decided to melt together a raft of long, hollow glass
tubes into a flat panel and then warp the whole thing over a curved mould. As a
test, he baked a few lengths of glass tube into a briefcase-sized sample. Then a
few months later, Tim Eliassen of TriPyramid Structures鈥攖he firm that
engineered the glass pyramids at the Louvre鈥攁rranged a meeting with
Weinstein and Graham Dodd, a glass safety expert with consulting engineers Ove
Arup and Partners in London.

At a hotel bar in downtown San Francisco, Weinstein pulled out the sample. He
was a little embarrassed, since by then it had developed some bad cracks, but to
Eliassen and Dodd these imperfections were very revealing. The cracks ran
halfway through the sample and then stopped. It looked as if Weinstein had
unwittingly created something akin to safety glass. But would larger panels
behave the same way?

To construct these panels, Weinstein planned to lay the tubes alongside each
other and heat them to 675 掳C鈥攈ot enough to make the glass tacky. When
the assembly cooled, the tubes would fuse, forming a panel that could then be
heated again and bent to shape over a curved mould.

But Weinstein quickly realised that this two-stage approach would never work.
Each heating cycle introduces more opportunity for cracking, and the mould
creates uneven cooling. Every panel he made broke into fragments.

His solution was a unique, 4-metre-long kiln that could fuse and shape the
tubes in one go. It was built by Seattle kiln maker Fred Metz and shipped to
Weinstein鈥檚 studio behind a laundromat in San Francisco鈥檚 Mission district.
Inspired by pinscreens, the toys that record impressions of a face or hand in a
layer of movable pins, the bed of the kiln is dotted with discs that can be
raised or lowered on rods to shape the softened glass.

To create a panel, Weinstein makes a scale model of it and pushes it into a
pinscreen the size of a briefcase. This creates a scale-model relief map of the
panel. He then measures the length of each pin, and uses it as a guide to the
height to which each disc in the oven must rise in order to shape the glass
correctly. 鈥淚t鈥檚 an old-world version of digitisation,鈥� he says.

With all the rods lowered, the kiln bed is flat, and Weinstein and his team
can lay out the tubes for the panel horizontally on the floor. Once the tubes
are in place, he shuts the heavy door and the lorry-sized kiln starts to heat up
very slowly to around 700 掳C.

When the glass glows orange, the tubes begin to slump like cooked macaroni.
At this point, a metal plate which all the rods run through creeps slowly up
below the kiln. Each rod has an adjustable collar that makes contact with the
plate. As the plate rises, it pushes the collars and the rods with it. After up
to 60 minutes of slow movement the rods have moulded the glass panel to its
final shape and the kiln begins to cool very slowly. As far as Weinstein knows,
it鈥檚 the only kiln in the world with a bed that morphs at temperature.

So would the curved panels be safe to hang above people鈥檚 heads? To find out,
Weinstein called Corning in New York, among the world鈥檚 leading specialists in
heat-resistant glass. Corning suggested he should get in touch with glass expert
Herb Miska, who designs high-impact glass for aircraft. He also helped make
windows for the space shuttle that would not burst on lift-off or re-entry.

To Miska it was immediately obvious that the panels could not be strengthened
with standard techniques such as heat or chemical tempering (New
Scientist, 11 February 1995, p 23). And unlike aircraft glass, the stresses
couldn鈥檛 be worked out mathematically. 鈥淭hese panels had a very strange
geometry, so it was very difficult to quantify anything,鈥� says Miska. There was
nothing for it. They would simply have to test the panels to see what kind of
stress they could take.

So Miska put one on the floor of his garage and dropped a wrench on it.
Incredibly, the panel didn鈥檛 shatter. The only damage was around the site of
impact and never spread beyond it. So he dropped more tools, large bolts,
anything that might fall on it during installation or cleaning. Always the
damage was local. 鈥淭he thing had so many walls and surfaces, it was very
effective at absorbing energy,鈥� he says. 鈥淪ometimes things I dropped even got
stuck in it.鈥� So why didn鈥檛 it shatter?

Miska believes the glass panels behave like the walls of the old Spanish
forts in Florida. Just about the only building material available to the Spanish
in the Caribbean was coral. Not only was it light and porous, but they soon
realised it was great at absorbing the impact of cannonballs. All the energy of
the collision was dissipated in tiny breaks rather than radiating outwards as it
might in stone. Weinstein had unwittingly created a sculpture with fortress-like
strength. So the panels wouldn鈥檛 fail catastrophically in a construction
accident.

But what about the ravages of time? Over the centuries the Spanish forts have
crumbled, so would the chandelier weaken as the years went by? Tiny scratches on
the surface of glass create extremely high local stresses. As moisture in the
air reacts with the silicon, it weakens the chemical bonds in the glass and the
scratches turn into cracks.

To make the panels as strong as possible, Miska encouraged Weinstein to build
each one from a variety of tube sizes. By mixing smaller-diameter tubes with
larger ones, many of the gaps between the tubes could be filled, and the number
of places where tubes touched and bonded would increase. Weinstein liked the
idea, too, because a random distribution of tube sizes improved the panels鈥�
appearance.

Each panel of the sculpture contains tubes ranging from 3 to 10 centimetres
in diameter. But this created a whole new problem. For the best strength, each
tube has to touch all of its neighbours. Yet to lay them out like this by trial
and error and in a way that looks random would take forever. So late last year,
Weinstein commissioned his long-time friend Jamie Bernardin, a physicist and
simulation expert, to write a computer program that could do it in minutes.

Bernardin鈥檚 usual computer simulations use random numbers to model complex
phenomena such as financial markets. For Weinstein, he wrote a program that
would add tubes at random positions and then jiggle them all a bit until they
were as closely packed as possible. The program calculates the 鈥渆nergy state鈥� of
each design, a number that gets smaller as more contacts are made within the
panel. If an extra tube makes the energy state too high, it is rejected and
tried elsewhere. The panel鈥檚 structure evolves by something like natural
selection. The artist can also move tubes if they don鈥檛 look right, and the
program will shift the tube into its best fit at the new position. This way,
Weinstein could give each panel a different look, without compromising its
strength.

So much for the theory. In practice, things didn鈥檛 go entirely according to
plan. Just one year ago, the project was in deep trouble. Despite all the
engineering, and with only two months to go before the panels were due to start
arriving in Germany, Weinstein had not made a single one that was
crack-free.

The problem was occurring in the kiln during annealing, the final stages of
cooling. So on Miska鈥檚 advice Weinstein called retired Corning guru Hank Hagy at
his home in New York. Hagy had been responsible for the annealing of large
telescope mirrors, such as those in the Palomar observatory and the Hubble Space
Telescope. The first thing he wanted to see were the cracks.

Cracks in glass are packed with information. They reveal where the fracture
started and how much stress was involved. Hagy could see that the cracks all
originated at the end of a joint between two tubes, where one tube ended and the
other continued. Staggering the ends of the tubes gave the clouds a curved
profile, but it also meant that if the shorter tube cooled faster than the
longer tube, it would pull along the joint. Where that joint ended, the forces
were magnified 20 to 30 times.

The solution was to cool each sculpture more evenly. But Weinstein鈥檚 kiln
wasn鈥檛 really designed for that, Hagy says. A kiln for a telescope mirror would
have heating elements all around to control the temperature precisely, but
Weinstein鈥檚 kiln only had elements on the top. 鈥淚 said, `Jesus, you haven鈥檛 got
a kiln, you鈥檝e got a broiler鈥�,鈥� Hagy recalls.

Running short of time, they added insulation and a few thermocouples that
allowed them to monitor the temperature of the glass directly, rather than just
the temperature of the kiln. Then Hagy slowed down the annealing time from hours
to days.

Soon the panels were rolling out of the kiln at a rate of nearly one a week.
That left the final challenge: 鈥淭he thing that made me nervous about it is that
the strength of glass is low under long-term load,鈥� Dodd says. 鈥淚f you load the
thing up with its own weight, you have to ask yourself how long it鈥檚 going to
hang out before it breaks.鈥�

When Dodd and the engineers from TriPyramid delved into all the research they
could find on 鈥渟tatic fatigue鈥濃�攖he effects of long-term stress on
glass鈥攖hey discovered a magic number: one million. If glass under a
constant load can hold up for one million seconds, all the evidence suggests it
will hold up forever. So once each panel has been hung at the bank, it is loaded
with lead sheets to increase the weight. After a million seconds鈥攁bout
eleven days鈥攖he sheets are removed and if the panel is intact it is deemed
safe. So far, not a single panel has failed the test.

The chandelier is a feat of engineering, to be sure, but not in the modern
sense of number crunching and predictive design. 鈥淲e did a fairly ancient kind
of engineering where you try a few things to get in the right ballpark, and then
you go ahead and build it and see if it actually works,鈥� Dodd says. 鈥淚t鈥檚
reminiscent of the old European cathedrals, with flying buttresses and high,
vaulted roofs. All those things were developed by trial and error. Nobody had a
technique for calculating stress or load.鈥� Old-fashioned it may be, but it
works. And with Chartres pushing 800 years, the diners at Pariser Platz 3 have
nothing to fear from the glass over their heads. For a few centuries, at least.