Curiosity about the nature of primes going to infinity.

63 Views Asked by At

My question is about the nature of primes as you go to infinity.

I was watching a video about the last digits of primes and Chebeshev's bias and I had a thought about the Goldbach conjecture. If N ends with a 2, then N minus primes ending in 7 will equal a multiple of 5. The same can be done to find multiples of 7 in base 14 (minus multiples of 5 of course) and so forth.

I used this with some code to predict how many prime pairs would add up to N where N has only simple factors of 2's, 3's and primes greater than the square root of N/2 . My laptop could only handle this up to just above 1 billion.

That biggest calculation had a prediction that was about half of the actual result and smaller numbers had a smaller difference. The predictions were always smaller that the actual amount.

I know this can't hold as we go to infinity but why not? All I can think of is that Chebeshev's bias must become much much higher than 3/1000 but that contradicts the video I watched.

Here are the results just for reference.

N (multiple of) Prediction(Pre) Actual Pre/Actual
54 (3,2) 5.2 6 0.866666667
92 (Px4) 3.7 4 0.925
128 (2) 4.1 5 0.82
162 (3,2) 9.4 10 0.94
212 (Px4) 5.6 7 0.8
486 (3,2) 20.1 24 0.8375
1,024 (2) 17.1 23 0.743478261
1,458 (3,2) 44.4 48 0.925
4,096 (2) 47.8 53 0.901886792
6,088 (Px8) 64.1 71 0.902816901
39,366 (3,2) 558.7 569 0.981898067
65,536 (2) 419.6 438 0.957990868
112,396 (Px4) 655.5 672 0.975446429
524,288 (2) 2,335.9 2,372 0.984780776
1,062,882 (3,2) 8,421.2 8,607 0.97841292
1,495,636 (Px4) 5,608.2 5,711 0.981999649
4,194,304 (2) 13,319.9 13715 0.971192125
9,565,938 (3,2) 52,912.6 55,737 0.9493263
16,489,952 (Px32) 41,427.4 44,863 0.923420102
33,554,432 (2) 74,058.4 83,480 0.887139435
86,093,442 (3,2) 313,306.8 382,818 0.818422331
155,140,352 (Px256) 245,376.7 322,551 0.760737682
258,280,326 (3,2) 712,371.8 1,015,231 0.701684444
536,870,912 (2) 585,543.5 975,734 0.600105664
1,073,741,824 (2) 889,644.5 1,817,166 0.489578002