1 / 30

Four approaches to Shor

Four approaches to Shor. A mixture of a few. David Poulin LITQ Université de Montréal Supervisor Gilles Brassard ( SAWUNEH ma y 2001). Summary. Shor’s entire algorithm formally Probability analysis Phase estimation Shor as phase estimation Quantum circuit for QFT

uri
Download Presentation

Four approaches to Shor

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. Four approaches to Shor A mixture of a few David Poulin LITQ Université de Montréal Supervisor Gilles Brassard (SAWUNEH may 2001)

  2. Summary • Shor’s entire algorithm formally • Probability analysis • Phase estimation • Shor as phase estimation • Quantum circuit for QFT • Semi-classical circuit for QFT • Single qubit phase estimation • Mixed state quantum computing

  3. A bit of number theory... • Theorem • If a  b (mod N) but a2  b2 (mod N) • Then gcd(a+b,N) is a factor of N. • Proof • a2 - b2 0 (mod N) •  (a - b)(a+b) 0 (mod N) • ( t) [ (a - b) (a+b) = tN ]  gcd(a+b, N) is a non trivial factor of N. uN vN

  4. Easy Easy Easy Easy Easy Shor’s entire algorithm • N is to be factored: • Choose random x: 2  x  N-1. • If gcd(x,N)  1, Bingo! • Find smallest integer r : xr 1 (mod N) • If r is odd, GOTO 1 • If r is even, a = xr/2 (mod N) • If a = N-1 GOTO 1 • ELSE gcd(a+1,N) is a non trivial factor of N. Hard Easy

  5. Add this step to Shor’s algorithm: 0. -Test if N=N’2l and apply Shor to N’ -Compute for 2  j  ln2N. If one of these root is integer, apply Shor to this root.  Probability of success  ½. Success probability Theorem If N has k different prime factors, probability of success for random x is  1- 1/2k-1. Easy

  6. F F-1 Modular exponentiation EN,x HA dim= HB dim= Order finding Quantum Fourier transform = 2n n = 2lnN

  7. For sake of analysis! m Order finding A B C D m Hn F-1 |0 EN,x |0 Bit bucket r : xr 1 (mod N)

  8. EN,x B The second register is r-periodic since xnr+bmodN = xbmodN Step by step Hn A |0|0

  9. Step by step m on the second register and obtain y a power of x. What is left in the first register is an equal superposition of everything consistent with y. y xs xs+r xs+jrmodN m C

  10. Step by step Quantum Fourier transform F-1 D

  11. What we want is r : xr 1 (mod N) ! Consider a c :  t integer with 0 rc-t  r/2 t  rc  t  +r/2 t j  jrc /   tj +rj/2   tj +1/2 0 jrc /  1/2 plus a integer! Measure the first register: m “c” with probability |c|2 = What’s that probability?

  12. Length of the arc: Length of the cord: |c|2  What’s that probability?

  13. What’s that probability? If 0 rc-t  r/2 then |c|2  #{c : 0 c  -1 and (t)[0 rc-t  r/2 ]}  r  Pr( getting a good c )  What the heck is so special about those c ?

  14. Assume there is another t’/r’ satisfying this condition: Since Hence tr’ – t’r =0  t and r are unique. They can be found by continuous fraction algorithm!!! Continuous fractions The condition can be written c/ is the best n bits estimation of t/r.

  15. That was Shor’s algorithm formally. Now I’ll show what Shor’s algorithm really is. Do you need a break?

  16. Interference |0  |0 |1  ei |1 m  |0 H H |0  |0+|1  |0+ei |1  (1+ ei )|0+ (1- ei )|1 Pr(“0”)=cos2(/2) Pr(“1”)=sin2(/2)

  17. Phase kick back The previous dynamics can be simulated by: Same state as previous slide! |0 H |u |u U Apply U if top wire is 1 Bit bucket Where |u is an eigenstate of U: U|u = ei |u |0|u  (|0+ |1)|u = |0|u+|1|u  |0|u +ei |1|u  (|0+ei |1)|u

  18. 4 |0+ei2  |1  Hn | |0 |0+ei |1 2 3 4 U2 |u |u U2 U2 U2 U Phase estimation In Deutsch’s problem, we were able to determine whether  was 0 or . Q: Can me determine any  ? A: We can get the best n bit estimation of /2.

  19. F So applying F-1 to | will yield |x that is the best n bit estimation of /2. Phase estimation (binary extension of x/ - integer)

  20. UN,a Multiplication Consider UN,a : |x  |ax mod N. Then, for k = 1,...,r are eigenstates of UN,a with eigenvalues If we could prepare such a state, we could obtain an estimation of k/r hence of r. It requires the knoledge of r.

  21. Multiplication Consider the sum Since The state |1 is easy to prepare. In what follows, we show that it can be used to get an estimation of k/r for random k.

  22. m F-1 m m Make measurement here to collapse the state to a random |k : get an estimation of k/r for random k. This measurement commutes with the Us so we can perform it after. This measurement is useless! Phase estimation Hn |0 2 3 4 U2 |1 U2 U2 U2 U N,a N,a N,a N,a N,a

  23. F-1 |x0 H Qubit n-1 depends on x0 with a phase 0 or -/2 and on x1 with a phase 0 or - |x1 R1 H |x0 H QFT circuit Qubit n is |0+ |1 if x0 is |0 and |0- |1 if x0 is |1. (x0 with a phase 0 or -)

  24. H R1 R2 H R1 H Rk  Note 2: Rk QFT circuit We define the gate Rk as a -/2k phase gate. |x3 R1 R2 R3 H |x2 |x1 |x0 Note 1: H = R0

  25. Measurements! H Semi-classical QFT |x3 H |x2 R1 H |x1 R1 R2 H |x0 R3 R2 R1 All controlled phase gates are now classically controlled!

  26. Single qubit phase estimation … … |0+ |1 H   … |0+ |1 H … R1 |0+ |1 H … R1 |0+ |1 R2 H … … n-1 |1 2 1 0 … U2 U2 U2 U2 Bit bucket

  27. Almost anything will do the job!!! Single qubit phase estimation |0+ |1 |0+ |1 |0+ |1 … |0+ |1 Rn-2 H H Rn-1 Rn H H … … |1 n-1 2 … U2 U2 1 0 U2 U2 The are measurements. The Rk are phase gates with an angle 0.b1b2...bk-1 where bj is the classical outcome of the jth measurement.

  28. The other eigenvectors of UN,aare of the form: Where gd are solutions of gar-g 0 mod N and rd is the period of the period x  gdaxmod N. Mixed state computing Maximally mixed state: Independent of the basis |k. The |kk=1,2,...,r are orthogonal, but do not form a complete basis since r < .

  29. We express the maximally mixed state as a mixture of the eigenvalues of UN,a. The output of the algorithm will then be the best n bit estimation of jd/rd for d and jd chosen at random. The result is useful if gd=1: Prob = Mixed state computing Theorem: Given q and p : N = pq, then gar-g 0 mod N for at most p+q-1 values of g.

  30. Mixed state computing Since I is independent of the basis, we can input anything in the bottom register and it will work pretty well. In particular, this is useful for NMR computing. (it’s impossible to prepare a pure state)

More Related