FOR several years , physicists , mathematicians and computer scientists have been worrying at the problem of how to make a quantum computer - - a device that would out - perform conventional ( or " classical " ) computers as thoroughly as conventional computers have beaten the abacus . Some think that , because such a machine would rely on those most fragile of phenomena - - the quantum uncertainties of the microscopic world - - it may never be possible to construct one . Now , as reported at theAAAS by Neil Gershenfeld of the Massachusetts Institute of Technology , quantum computing may have been rescued by a chemists ' trick .

A classical computer is , indeed , little more than a speeded - up abacus . It stores and shuffles binary numbers ( those containing only the digits 1 and 0, which it remembers as the " on " and " off " positions of tiny switches , or " bits " ) . By contrast , the switches in a quantum computer would be far tinier , and thus able ( thanks to the strange rules of quantum theory ) to be both " on " and " off " simultaneously . Each switch ( known in this context as a " qubit " ) could then , in effect , be doing two calculations at once . Two qubits could thus do four things at once , three qubits could do eight , and so on . A quantum computer with fairly few components could therefore crack vast problems that a classical computer could not solve before the end of time .

---

But there is a catch . Anything in such a quantum combination of states is in a highly delicate balance , and any disturbance - - including that needed to read the result of a calculation from such a computer - - upsets it . It is said to " decohere " : to fall completely into one or another of its possible simultaneous states , to the exclusion of the others . This wipes out most of the information it contains .

Solving the problem of decoherence is thus essential if a quantum computer is ever to be built . It is toward this end that Dr Gershenfeld and his colleague Isaac Chuang , of Los Alamos National Laboratory in New Mexico , believe they have taken a significant step .

They have done it by abandoning a crucial assumption about quantum devices: that only something tiny , and isolated from almost everything else , could hover in multiple states and so act as a qubit . They worked out that the same effect could be achieved by using the total opposite: a sea of molecules , jostling and tumbling like a band of excitable schoolchildren .

---

Molecules contain atoms , and the nuclei of atoms act like tiny magnets . A property called " spin " indicates which way an atomic magnet points . A single nucleus can therefore act as a qubit , its spin pointing perhaps up for " off " and down for " on " . A bunch of connected atoms - - a molecule - - can thus be a quantum computer , capable of as many simultaneous calculations as there are ways of arranging the spins in it . The nuclei , cocooned within the molecule , make far more stable qubits than the other kinds tried so far . And there is , crucially , a way of manipulating the qubits and reading the result of the calculation .

---

This tool is nuclear magnetic resonance NMR spectroscopy - - a technique used by chemists to analyse molecules . It involves putting the molecules in a magnetic field and hitting them with pulses of radio waves , to which they respond with signals that depend on how their spins are arranged .

But these pulses do make the spins decohere . The researchers ' answer is to use back - up copies , by having not one molecule , but a small cup containing about a hundred billion trillion of them .

---

This was where previous assumptions had to be dropped . If it is so hard to manipulate an isolated qubit or a single molecule without its decohering , then how , when huge numbers of them are colliding with each other , can you tell which ones are doing what , and keep them controlled ? It would be like trying to pick out a few children walking due north , and then making sure they kept going in exactly the same direction , in a vast playground jam - packed with kids running around everywhere .

However , NMR works because in any given bunch of molecules there are slightly more with spins pointing in one direction rather than another . The signals from this small surplus ( about one part in a million ) stand far enough out from the background noise for the NMR equipment to pick them up . The crucial thing that Dr Chuang and Dr Gershenfeld realised was that the background noise would average out , thus leaving the quantum states of this small surplus of molecules undisturbed. ( If the kid going due north is buffeted equally in every direction , he will still end up going due north ) . So , unlike a single qubit , which is prey to any passing distraction , they could sit in the molecular equivalent of a boiling sea and not lose their calm . These molecules could thus act as multiple copies of a quantum computer . Each radio pulse used to instruct or interrogate them would unbalance a few . But enough would be left acting in concert to last for several dozen computational steps .

---

The final hurdle was getting the right answer out at the end . A quantum computation usually gives a variable result which has to be corrected by a final , classical calculation . Lots of quantum computers would produce lots of different answers . However , a quantum computer can do the corrective calculation too ; in fact , it turns out , it can do the calculation on itself . Dr Chuang and Dr Gershenfeld managed to apply this technique to all their molecules , so that they gave the correct answer .

Since writing a paper on the subject ( published a few weeks ago in the AAAS 's journal , Science ) , Dr Chuang and Dr Gershenfeld have been putting this into practice . Using the carbon atoms in a molecule called alanine as qubits , they have built a computer that can add one and one and give the result . This may look trivial , but it is way ahead of other approaches . And , by using more complex molecules - - the caffeine in a cup of coffee is apparently a candidate - - Dr Chuang and Dr Ger shenfeld hope by the end of the year to make a ten - qubit device which could divide the number 15 into its factors . For still larger problems , perhaps cinnamon coffee , in which Seattle abounds , will be suitable .

---