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.

74 Chapter 1 PRIMES!<br />

10 10 . The method in [Bernstein 1998] might be used in such an endeavor. Note<br />

that the probability predicted by Theorem 1.4.9 is ρ(10) ≈ 2.77 × 10 −11 .<br />

1.74. What is the approximate probability that a random integer (but of<br />

size x, say) has all but one of its prime factors not exceeding B, witha<br />

single outlying prime in the interval (B,C]? This problem has importance<br />

for factoring methods that employ a “second stage,” which, after a first stage<br />

exhausts (in a certain algorithm-dependent sense) the first bound B, attempts<br />

to locate the outlying prime in (B,C]. It is typical in implementations of<br />

various factoring methods that C is substantially larger than B, for usually<br />

the operations of the second stage are much cheaper. See Exercise 3.5 for<br />

related concepts.<br />

1.75. Here is a question that leads to interesting computational issues.<br />

Consider the number<br />

c =1/3+<br />

1<br />

1/5+ 1<br />

1/7+···<br />

where the so-called elements of this continued fraction are the reciprocals of<br />

all the odd primes in natural order. It is not hard to show that c is welldefined.<br />

(In fact, a simple continued fraction—a construct having all 1’s in<br />

the numerators—converges if the sum of the elements, in the present case<br />

1/3 +1/5 +···, diverges.) First, give an approximate numerical value for<br />

the constant c. Second, provide numerical (but rigorous) proof that c is not<br />

equal to 1. Third, investigate this peculiar idea: that using all primes, that is,<br />

starting the fraction as 1/2+ 1<br />

1/3+··· , results in nearly the same fraction value!<br />

Prove that if the two fractions in question were, in fact, equal, then we would<br />

have c = 1+ √ 17 /4. By invoking more refined numerical experiments, try<br />

to settle the issue of whether c is actually this exact algebraic value.<br />

1.76. It is a corollary of an attractive theorem in [Bredihin 1963] that if n<br />

is a power of two, the number of solutions<br />

N(n) =#{(x, y, p) : n = p + xy; p ∈P; x, y ∈ Z + }<br />

enjoys the following asymptotic relation:<br />

N(n)<br />

n<br />

105<br />

∼ ζ(3) ≈ 0.648 ....<br />

2π4 From a computational perspective, consider the following tasks. First, attempt<br />

to verify this asymptotic relation by direct counting of solutions. Second, drop<br />

the restriction that n be a power of two, and try to verify experimentally,<br />

theoretically, or both that the constant 105 should in general be replaced by<br />

315 <br />

p∈P, p|n<br />

(p − 1) 2<br />

p 2 − p +1 .<br />

,

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

Saved successfully!

Ooh no, something went wrong!