10.12.2012 Views

Prime Numbers

Prime Numbers

Prime Numbers

SHOW MORE
SHOW LESS

You also want an ePaper? Increase the reach of your titles

YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.

8.7 Exercises 435<br />

of positive composite numbers. Next, starting from the Lagrange theorem<br />

that every positive integer is a sum of 4 squares (see Exercise 9.41), exhibit a<br />

polynomial in 8 variables with integer coefficients such that its values at all<br />

integral arguments constitute the set of positive composites.<br />

8.22. Suppose the integer n of Proposition 8.5.1 has the distinct prime<br />

factors p1,...,pk, where2 si pi − 1ands1 ≤···≤sk. Show that the relevant<br />

probability is then<br />

1 − 2 −(s1+···+sk)<br />

<br />

1+ 2s1k − 1<br />

2k <br />

− 1<br />

and that this expression is not less than 1 − 2 1−k . (Compare with Exercise<br />

3.15.)<br />

8.23. Complete one of the details for Shor factoring, as follows. We gave as<br />

relation (8.4) the probability Pc,k of finding our QTM in the composite state<br />

| c 〉| xk 〉. Explain quantitatively how the probability (for a fixed k, withc<br />

the running variable) should show spikes corresponding to solutions d to the<br />

Diophantine approximation<br />

<br />

<br />

<br />

c d<br />

− <br />

q r <br />

≤ 1<br />

2q .<br />

Explain, then, how one can find d/r in lowest terms from (measured)<br />

knowledge of appropriate c. Note that if gcd(d, r) happens to be 1, this<br />

procedure gives the exact period r for the algorithm, and we know that two<br />

random integers are coprime with probability 6/π2 .<br />

On the computational side, model (on a classical TM, of course) the<br />

spectral behavior of the QTM occurring at the end of Algorithm 8.5.2, using<br />

the following exemplary input. Take n = 77, so that the [Initialization] step<br />

sets q = 8192. Now choose (we are using hindsight here) x = 3, for which<br />

the period turns out to be r = 30 after the [Detect periodicity ...]step.Of<br />

course, the whole point of the QTM is to measure this period physically, and<br />

quickly! To continue along and model the QTM behavior, use a (classical)<br />

FFT to make a graphical plot of c versus the probability Pc,1 from formula<br />

(8.4). You should see very strong spikes at certain c values. One of these values<br />

is c = 273, for example. Now from the relation<br />

<br />

<br />

<br />

273 d<br />

− <br />

8192 r <br />

≤ 1<br />

2q<br />

one can derive the result r = 30 (the literature explains continued-fraction<br />

methods for finding the relevant approximants d/r). Finally, extract a factor of<br />

n via gcd(x r/2 − 1,n). These machinations are intended show the flavor of the<br />

missing details in the presentation of Algorithm 8.5.2; but beyond that, these<br />

examples pave the way to a more complete QTM emulation (see Exercise 8.24).<br />

Note the instructive phenomenon that even this small-n factoring emulationvia-TM<br />

requires FFT lengths into the thousands; yet a true QTM might<br />

require only a dozen or so qbits.

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!