How Old is That Silk Artifact?
A chemist from the Textile Museum is perfecting a new technique for understanding the past
- By Joseph Stromberg
- Smithsonian magazine, February 2012, Subscribe
People have been weaving silk into fabric for at least 5,000 years. The delicate material, made from the threads silkworms excrete to create their cocoons, has been used for everything from the robes of Byzantine emperors to the parachutes of World War II paratroopers. In ancient China, the birthplace of silk, it became luxury paper, a medium for paintings and even a form of currency; for centuries, laws forbade anyone except the emperor and other dignitaries from wearing it as clothing. Silk has been found in the tombs of Egyptian mummies, and in ancient Rome it was blamed for making young women promiscuous.
To historians, silk artifacts reflect the trade and social customs of past cultures. Researchers have long sought a surefire method that measures the age of silk for which there is no continuous historical record and uses just a tiny sample of the material. Now Mehdi Moini, a chemist at Smithsonian’s Museum Conservation Institute, has developed a technique to date silk based on its chemical composition. His technique serves as a kind of clock, and he is testing and calibrating it with silk of known vintage in the Institution’s collections. “Making a clock is easy,” he says. “Calibrating the clock is difficult.”
Silk proteins are made of amino acids, small molecules with a three-dimensional structure. Each amino acid has two possible variants, exact mirror images of each other: left-handed, known as “L” amino acids, and right-handed, referred to as “D.” The amino acids produced by most living things—like silkworms—are left-handed. The key to the dating process, Moini says, is that as silk proteins age, some of the amino acids rearrange themselves into the D variant. He can tell how old a silk thread is by looking at the ratio of D to L amino acids. At year zero, all will have the L structure; given enough time, there will eventually be equal parts of both. Researchers have been using this approach on various proteins for decades, but Moini is the first to apply it to very small samples of silk, says Darrell Kaufman, a geologist at Northern Arizona University.
Moini and his colleagues borrowed a Chinese textile more than 2,000 years old from the Metropolitan Museum of Art and a flag from the Smithsonian that was used in 1846 in the Mexican War. Sumru Krody, senior curator at the Textile Museum in Washington, D.C., let him take samples from an ancient Egyptian tiraz, a ceremonial turban band with an inscription dating it precisely to A.D. 993. “It’s very hard to find textiles with an accurate date on them,” she says. “But sometimes you get lucky.”
Moini’s method uses a minuscule amount of material, something that appeals to curators of priceless fabrics. “If you go to a museum and say, ‘I want five milligrams of this precious silk,’ silk is very light, so five milligrams is a lot of compound,” Moini says. Previous techniques, such as carbon dating, consumed several milligrams of silk for each test; by contrast, he needs just one-hundredth that amount. “If there’s just one tiny speck of fabric,” he says, “it’s enough for us to do the analysis.”
To determine the ratio of D to L amino acids in each piece of silk, Moini and his team dissolve the fabric in hydrochloric acid, put the liquid in a thin glass tube and apply an electric field. Because amino acids have a slight electrical charge, they are pulled through the tube. The tube also contains a substance that chemically attracts the D amino acids. “Think of it like this: You have a bunch of hands, left and right hands, and then we fill the [tube] with only right gloves,” Moini says. “Only the right hand goes into the right glove and gets caught, and the left hand cannot fit into the right glove, so it goes through faster.”
After analyzing a variety of silk samples, Moini and his team determined that 50 percent of the amino acids change from L to D forms after 2,500 years.
The technique is more efficient and more precise than previous methods. It will soon be put to use to date artifacts of unknown age. “After this technique is fully developed, it’s then the fun part will start,” says Krody. She may use the dating method on items such as the Buyid silks, a group of artifacts that may have originated in Iran sometime between the 8th and 12th centuries.
Subscribe now for more of Smithsonian's coverage on history, science and nature.
Comments (5)
Quibble? First, I am an honest person who has zero motivation to evade truth here. I'm just curious. Second, knowing the accuracy of a test is not trivial, it's imperative. As a chemist, I personally like to summarize the accuracy of a test method in the form of +/- x at a confidence level of y, over a range of z. For example, +/- 5 years, at a confidence level of 90%, between a range of 100 and 1000 years. Of course raw test data, methods, and graphs would be expected in a complete answer on test accuracy. But the form I suggest may be a polite answer to a casual question.
Posted by Wilson on April 4,2012 | 12:53 PM
I surmise that the D/L ratio over time follows an asymptotic curve, increasing linearly from time zero (when the silk is made), then 'flatting out' at a ratio of 1 (50:50 D to L) near 2,500 years. In this case the _rate_ of D generation is a maximum at time zero, decreasing to 0 rate of change at around 2,500 years. If we have a few points along this curve mostly before the horizontal part, and we assert loudly that the asymptotic form is valid, then a half dozen points will get the single key parameter to describe it. At this point the 'clock' has been calibrated, and as Sumru Krody says in the article, "the fun part will start."
The sensitivity of the age estimate (its precision) is a function of the slope, or rate of D generation. Thus the precision will be a maximum for recent silks, and rather poor somewhere out past 2,500 years. But this age effect is not the only factor in the accuracy of the age estimate. A lot will depend on the precision of that key constant.
Nonetheless, what Dr. Moini has developed is a great step forward. I'm sure that in time we'll have the data to nail this D/L ratio shift. I refuse to quibble over precision now.
Posted by Jay Warner on February 13,2012 | 01:12 PM
My question isn't whether the test covers the entire period of silk use. Sounds like it does. The question is how much does the test accuracy change over the entire period of silk use because it looks to me that the test accuracy cannot be linear. My wild screaming guess, but I don't know, is that the accuracy decreases a large amount near the oldest end of the period, due to a presumably asymtotic behavior of the Levo to Dextro ratio. I also wondered how often calibration samples need to be used to prevent accuracy drift.
Posted by Wilson on February 1,2012 | 09:15 AM
Your question is generally acceptable, however, if it is even acceptable for only 1sigma then it covers the entire history of human use of silk.
Posted by godanov on January 26,2012 | 09:50 PM
Since the ratio of L to D approaches 50 percent after 2,500 years, then, predictions of dates should be less and less accurate as the ratio becomes asymtotic at 2,500. Does the know the behavior of the graph at the asymtotic points of 0 years and 2,500 years? It would have been interesting to see a graphy. How can he 'calibrate the clock' or speak confidently about its accuracty without a great number of calibration samples? How does he prevent drift of the test method without continued use of test samples, particularly near the extremes?
Posted by Wilson on January 25,2012 | 05:35 AM