Problems & Puzzles: Puzzles

Puzzle 192. Reversible & Digit Complementary Prime Pairs (*)

Two numbers (A,B) are said to be a 'Digit Complementary Prime Pair' (DCPP from now on) if:

a) A & B are primes
b) digits in corresponding positions sum to "10" or "0"

Heinz example for K=4 digits in A & B:

A = 4721
B = 6
389

If we rule an additional condition for A & B, to be a reversible prime pair, Heinz provide the following example:

A = 3467
B = 7
643

This is the kind of numbers we will ask for in this puzzle: DCCP & reversible

Questions:

Find the smallest K digits example for:

a) K=50
b) K=500
c) K=1000
 

_____
*
The definition goes up to Charles W. Trigg, JRM 22:2, 1990,  p 95-97, according to one of the prime patterns interesting page by Harvey Heinz.


Solution:

Sudipta Das and Jens Kruse Andersen got both the same solution for K=50 & K=500. Jens Kruse Andersen got the solution for K=1000. Here are the solutions:

K A & B
50 A=1*10^49+188229*10^27+9
B=9*10^49+922881*10^27+1
500 A=1*10^499+278661944238*10^244+9 B=9*10^499+832449166872*10^244+1
1000 A=1*10^999+411369147996*10^494+9 B=9*10^999+699741963114*10^494+1

***

This is the approach by Andersen:

I assume the "smallest" solution means the smaller of the two primes is as small as possible. Then it must have the decimal form A=10...0D0...09 where the number D (not a digit) is centered, digit complementary with itself, and as small as possible.

My algorithm: Test all D from 0 to the first solution, i.e. the first time both A and B=reverse(A) are primes. I wrote a C program with the Miracle big integer library. When the first half of D's digits are chosen, the other half is given. A and therefore D always has an even number of digits for the solutions asked. I computed A and B and first trial factored them in parallel: Test if a prime divides any of them, then try the next prime. This prevents "wasted" time on one number when the other has a small factor. If no factor below 65536 is found, I make a single probabilistic test on A, and then B if A is a probable prime. When a probable solution is found I make more tests to give a high certainty. All probable primes were finally proven prime with Primo. The primality certificates were validated by Cert_Val.

The trial factoring part could be much faster by only computing D mod p for each D and prime p and then look up in a precomputed table whether that modulo means A is divisible by p. If not then compute reverse(D) mod p and look up in another table whether p divides B. Assume we want a 50-digit solution and D has 6 digits. Then A=10^49+D*10^27+9. To precompute the tables use this: A mod p = ( (10^49 mod p) + (D mod p)*(10^27 mod p) + 9 ) mod p. B=9*10^49+reverse(D)*10^27+1 is similar. I estimated the simple trial division would find the 1000-digit solution within a few hours and did not bother to implement this fast version. The probable solution took around 1 hour. Primality proofs took 2.5 hours.

Smallest 50-digit solution:

A=1*10^49+188229*10^27+9

B=9*10^49+922881*10^27+1

Decimal expansions: A=10000000000000000000001882290000000000000000000009

B=90000000000000000000009228810000000000000000000001

Smallest 500-digit solution:

A=1*10^499+278661944238*10^244+9 B=9*10^499+832449166872*10^244+1

Smallest 1000-digit solution:

A=1*10^999+411369147996*10^494+9 B=9*10^999+699741963114*10^494+1

It would be relatively easy to compute probable solutions with 2000 or more digits but the time required for primality proofs would grow fast.

***

 

 



Records   |  Conjectures  |  Problems  |  Puzzles