Quantum computers: how do they work?

Uni­ver­si­ties and lab­o­ra­to­ries around the world, as well as the big tech com­pa­nies, Google, IBM, Intel and Microsoft, are study­ing and devel­op­ing quan­tum com­put­ers at break­neck speed. The stakes are high. But what exact­ly is this technology?

A quan­tum com­put­er is not real­ly a ‘com­put­er’ as such, but rather a super-cal­cu­la­tor, capa­ble of run­ning spe­cif­ic pow­er­ful quan­tum algo­rithms much faster than an ordi­nary proces­sor. It does this using the prin­ci­ples of quan­tum mechan­ics, which are respon­si­ble for the behav­iour of ele­men­tary par­ti­cles such as pho­tons, elec­trons and atoms, but also of larg­er sys­tems such as super­con­duct­ing circuits.

Such sys­tems allow for the imple­men­ta­tion of quan­tum bits (‘qubits’): two-state quan­tum sys­tems that rep­re­sent the fun­da­men­tal com­pu­ta­tion­al bricks of quan­tum information.

Unlike con­ven­tion­al com­put­ers, which encode infor­ma­tion in bina­ry form, qubits are not lim­it­ed to ‘0’ and ‘1’ but can be in any com­bi­na­tion (or ‘super­po­si­tion’) of the two. This abil­i­ty, com­bined with the fact that N qubits can be com­bined or ‘entan­gled’ to rep­re­sent 2^N states simul­ta­ne­ous­ly, allows cal­cu­la­tions to be per­formed in par­al­lel on a mas­sive scale. A quan­tum com­put­er could there­fore, in prin­ci­ple, out­per­form a clas­si­cal com­put­er for some impor­tant tasks, such as sort­ing large unsort­ed lists or for ‘prime fac­tori­sa­tion’. The lat­ter forms the basis of most encryp­tion algo­rithms in use today – in par­tic­u­lar for bank­ing operations.

Quan­tum com­put­ers look more like large cans, sus­pend­ed from the ceil­ing, cooled to near absolute zero with hun­dreds of cables hang­ing from them.

Copié !

Quan­tum com­put­ers don’t look any­thing like their clas­si­cal coun­ter­parts either. Cur­rent mod­els look more like large cans that are sus­pend­ed from the ceil­ing, cooled to near absolute zero (-273.14°C), with hun­dreds of cables hang­ing from them.

Any­one wish­ing to build a quan­tum com­put­er today must first over­come a big prob­lem: the fact that qubits are extreme­ly frag­ile. Almost any inter­ac­tion with exter­nal ‘noise’ in their envi­ron­ment can cause them to col­lapse like a souf­flé and lose their quan­tum nature in a destruc­tive process known as deco­her­ence. If this hap­pens before an algo­rithm has fin­ished run­ning, the result is a com­plete mess (and not the result of a cal­cu­la­tion) because any infor­ma­tion stored in the qubit is lost (imag­ine a com­put­er that has to restart every sec­ond). And, the more qubits a quan­tum com­put­er has, the more dif­fi­cult it is to keep them coher­ent.  So much so that even the most advanced quan­tum proces­sors today strug­gle to sur­pass 60 phys­i­cal qubits. A real device would require sev­er­al thou­sands, however.

Because of this prob­lem, a quan­tum com­put­er must be well iso­lat­ed from its envi­ron­ment. This requires very strin­gent con­di­tions: sim­ple and very cold sys­tems, pro­tect­ed from the out­side world. Such extreme con­fine­ment cre­ates a para­dox­i­cal sit­u­a­tion, how­ev­er, because the more iso­lat­ed the com­put­er is, the more dif­fi­cult it is for us to actu­al­ly com­mu­ni­cate with it (to access the results of its cal­cu­la­tions) and con­trol what it does.

In recent years, qubits have been made from a num­ber of iso­lat­ed sys­tems that remain coher­ent while algo­rithms are being run. These include trapped ions (atoms with an elec­tron removed or added), ultra-cold ‘Ryd­berg’ atoms, and pho­tons (light particles).

A trapped-ion based com­put­er stores infor­ma­tion in the ener­gy lev­els of indi­vid­ual ions. This is how qubits are formed in this sys­tem. Infor­ma­tion is shared between these qubits, and laser puls­es are then used to manip­u­late their states and cre­ate entan­gle­ment between them. This tech­nol­o­gy is quite advanced and researchers have recent­ly suc­ceed­ed in cre­at­ing a ful­ly entan­gled ‘24-qubit-GHZ’ (GHZ for ‘Green­berg­er-Horne-Zeilinger’) state using cal­ci­um ions¹.  

Oth­er sys­tems are based on sol­id-state mate­ri­als that might be inte­grat­ed into tra­di­tion­al elec­tron­ic devices. These micron-sized struc­tures are small on every­day scales, but large com­pared with atoms and can behave like quan­tum par­ti­cles, such as elec­trons or atoms. Such ‘arti­fi­cial quan­tum objects’ include ‘quan­tum dots’ (tiny pieces of semi­con­duct­ing mate­r­i­al), super­con­duct­ing cir­cuits, and dia­monds con­tain­ing spe­cial type of defects called ‘nitro­gen vacan­cies’. A quan­tum com­put­er based on super­con­duct­ing qubits, for exam­ple, is cooled down to mil­likelvin tem­per­a­tures (which is cold­er than inter­stel­lar space) and is con­trolled using microwaves².

Researchers are also try­ing to work out which of these sys­tems would make the best qubits. An impor­tant para­me­ter is, of course, a qubit’s resis­tance to deco­her­ence, which can be assessed in terms of the ‘fideli­ty’ of a quan­tum oper­a­tion. While the fideli­ty does not have to be a per­fect 100%, any­thing low­er will even­tu­al­ly lead to errors after mul­ti­ple oper­a­tions; most of today’s quan­tum com­put­ers are very sus­cep­ti­ble to errors.

Although ‘quan­tum error cor­rec­tion’ pro­to­cols can mit­i­gate deco­her­ence, these are cost­ly from a hard­ware point of view and any prac­ti­cal sys­tem must have a suf­fi­cient­ly high fideli­ty to begin with. Researchers are mak­ing progress in this area, how­ev­er, and recent work has shown that a two-qubit gate can be made from two sil­i­con quan­tum dots³. This gate can achieve 98% fideli­ty for the CROT oper­a­tion, which an essen­tial com­po­nent of a quan­tum computer.

The idea of a quan­tum com­put­er was first put for­ward by the late physi­cist and Nobel lau­re­ate Richard Feyn­man in the 1980s to sim­u­late the com­pli­cat­ed equa­tions of quan­tum mechan­ics, which take too long to solve on a clas­si­cal com­put­er. Today, appli­ca­tion areas are much more var­ied: cryp­tog­ra­phy; sim­u­lat­ing the prop­er­ties of mate­ri­als (with a view to improv­ing them); solv­ing dif­fer­en­tial equa­tions extreme­ly quick­ly; and opti­miz­ing machine learn­ing. Progress has been impres­sive and researchers have gone from being able to entan­gle just three qubits to more than 50 qubits in recent years, with error rates of 1 in 1000⁴.

Although it is dif­fi­cult to pre­dict what the future holds, a ful­ly func­tion­al, com­mer­cial­ly viable quan­tum com­put­er is unlike­ly to see the day of light any­time soon. The same goes for any kind of ‘per­son­al quan­tum com­put­er’. A quan­tum device will more like­ly be used for fun­da­men­tal research, R&D or gov­ern­ment and mil­i­tary pur­pos­es to begin with.

Today, researchers in this field are not only advanc­ing hard­ware (i.e. the machines them­selves), but are also devel­op­ing inno­v­a­tive soft­ware with new types of algo­rithms espe­cial­ly adapt­ed to quan­tum computing.

While quan­tum com­put­ing relies on the prin­ci­ples of fun­da­men­tal physics, it is a tremen­dous oppor­tu­ni­ty for sci­en­tists from many fields – includ­ing com­put­er sci­ence, math­e­mat­ics, mate­ri­als sci­ence and engi­neer­ing – to work togeth­er. The road to a real-world quan­tum com­put­er will cer­tain­ly be long, but there will be many excit­ing dis­cov­er­ies along the way. 

Interview by Isabelle Dumé