Sum of all integer digits of all primes between 1 an n

AntonVrba · 2006-09-20, 12:26

With the discovery of M44 (congratulations GIMPS) I pondered over the thought if it is possible to count/calculate all the ones of all the primes between 2 and M44. It is a bit off-topic but the result is interesting.

To calculate or estimate the number of ones, I set about as follows.

Define $c_b(n)$ as the sum of all base-b integer digits between 1 and n and can be expressed as.
$c_b(n)=\sum_{k = 0}^{m}\,\,{d_k}\,\left( b^k\, \frac{{d_k} + k\, \left( b - 1\right) - 1}{2} + {mod (n,\,{b^k})} + 1\right)$

Above has the spot values $c_b\left( b^m-1 \right) =\frac{m\left( b-1\right)b^m}{2}$

Now assume $m$ to be large then $m=log_b(b^m-1)=ln(b^m-1)/ln(b)$ and proportioning $c_b(b^m-1)$ to the number of primes between $1$ and $b^m-1$ which is approximated in the Prime Number Theorem as $(b^m-1)/ln(b^m-1)$ we obtain the unexpected result that the cumulative sum of all base-b integer digits of all the primes between $1$ and prime $b^m-1$ approximates to

$\frac{(b-1)(b^m-1)}{2\, ln(b)}\,$

Conjecture
The ratio $\kappa_b$ defined as "the sum of all base-b digits of all the primes between 1 and n" to "n", converges to the constant $\frac{b-1}{2\, ln(b)$ for increasing n.

$\kappa_2=0.7213476$ and $\kappa_{10}=1.9543251$

A computation check confirms above tendency already at relatively small values of n.

Is above already known or have I introduced a new constant?

In parctice how will the constant depart from above definition?

Regards
Anton Vrba

AntonVrba · 2006-09-20, 12:34

For the Mathematica users

Code:

CumSumDigits[n_, b_] := Module[
    {ss, p, d0, d1, m} ,
    ss = p = d0 = 0 ;
    m = n ;
While[m > 0, {
If[(d1 = Mod[m, b]) != 0, {
   ss += d1( b^p(b - 1) p + b^p (d1 - 1) + 2(d0 + 1))/2,
   d0 += b^p d1}],
p += 1,
m = IntegerPart[m/b],
}] ;
Return [ss ] ;
    ]

and can be checked by evaluating

Code:

CumSumDigits[(10^123 - 1)/9, 10] - CumSumDigits[(10^123 - 1)/9 - 1, 10]

AntonVrba · 2006-09-20, 17:20

here is a base-10 evaluation of $\kappa_{10}$ counting up to

Code:

Prime	   Actual      Calculated    Actual/Calculated
99999989     2.09217    2.07412       1.0087
999999937    2.07641    2.05933       1.0083
9999999967   2.06389    2.00702       1.0078   
99999999977  2.05366    2.03844       1.00747

Actual cumulative digit count of all primes 2 to p divided by the last prime p (C program)

Calculated value $\frac{\pi(p)\, c_{10}(p)}{p^2}$

$\pi(p)$ being the prime counting function

An on the same basis a base-2 evaluation for $\kappa_2$

Code:

32749     0.845614   0.804225   1.05146
65521     0.837533   0.798769   1.04853
131071    0.832198   0.794553   1.04738
262139    0.825139   0.789684   1.0449
524287    0.820511   0.78624   1.04359
1048573   0.815104   0.782262   1.04198
2097143   0.810823   0.779118   1.04069
4194301   0.806539   0.776155   1.03915
8388593   0.802665   0.773417   1.03782
16777213   0.798989   0.770954   1.03636
33554393   0.795861   0.768785   1.03522
67108859   0.792815   0.766688   1.03408
134217689   0.790022   0.764787   1.033
268435399   0.787444   0.763058   1.03196
536870909   0.785043   0.76144   1.031
1073741789   0.782845   0.75996   1.03011
2147483647   0.780761   0.758568   1.02926

2006-09-20, 12:26	#1
AntonVrba Jun 2005 1100010₂ Posts	Sum of all integer digits of all primes between 1 an n With the discovery of M44 (congratulations GIMPS) I pondered over the thought if it is possible to count/calculate all the ones of all the primes between 2 and M44. It is a bit off-topic but the result is interesting. To calculate or estimate the number of ones, I set about as follows. Define $c_b(n)$ as the sum of all base-b integer digits between 1 and n and can be expressed as. $c_b(n)=\sum_{k = 0}^{m}\,\,{d_k}\,\left( b^k\, \frac{{d_k} + k\, \left( b - 1\right) - 1}{2} + {mod (n,\,{b^k})} + 1\right)$ Above has the spot values $c_b\left( b^m-1 \right) =\frac{m\left( b-1\right)b^m}{2}$ Now assume $m$ to be large then $m=log_b(b^m-1)=ln(b^m-1)/ln(b)$ and proportioning $c_b(b^m-1)$ to the number of primes between $1$ and $b^m-1$ which is approximated in the Prime Number Theorem as $(b^m-1)/ln(b^m-1)$ we obtain the unexpected result that the cumulative sum of all base-b integer digits of all the primes between $1$ and prime $b^m-1$ approximates to $\frac{(b-1)(b^m-1)}{2\, ln(b)}\,$ Conjecture The ratio $\kappa_b$ defined as "the sum of all base-b digits of all the primes between 1 and n" to "n", converges to the constant $\frac{b-1}{2\, ln(b)$ for increasing n. $\kappa_2=0.7213476$ and $\kappa_{10}=1.9543251$ A computation check confirms above tendency already at relatively small values of n. Is above already known or have I introduced a new constant? In parctice how will the constant depart from above definition? Regards Anton Vrba

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
k*b^n+/-c where b is an integer greater than 2 and c is an integer from 1 to b-1	jasong	Miscellaneous Math	5	2016-04-24 03:40
All top 5000 primes will have the same number of digits in 2014 ;-)	Batalov	Conjectures 'R Us	10	2013-03-29 01:29
Always an integer.	mfgoode	Puzzles	18	2007-07-13 18:03
get all the 5000 biggest primes above 100K digits	jasong	jasong	1	2007-06-09 22:51
Primes for a mersenne integer DWT FNT	gbvalor	Math	1	2003-09-08 16:05