Mathematics desk
< August 10	<< Jul \| August \| Sep >>	August 12 >

Welcome to the Wikipedia Mathematics Reference Desk Archives
The page you are currently viewing is an archive page. While you can leave answers for any questions shown below, please ask new questions on one of the current reference desk pages.

August 11

Definition of modulo congruence on rational numbers

The previous question elicited a link Wolstenholme_prime#Definition via harmonic numbers that seems to use the concept of modulo arithmetic applied to rational numbers without defining the concept or providing a suitable link to clarify this (the cited reference Zhao (2007) is opaque to me in this regard). Intuitively it does not seem unreasonable to map rational numbers ~~Q → Z/Z_n~~ Q → Z/nZ : p/q ↦ [p]_n/[q]_n where gcd(p,q) = 1. Could someone point me to a suitable article that treats this topic? — Quondum^☏ 13:36, 11 August 2012 (UTC)[reply]

No, there is no such map. In the case of the harmonic numbers, the numbers

1,2,\dots ,p-1

are units modulo

p^{3}

, so the quotients

1,1/2,1/3,\dots ,1/(p-1)

are in

Z/p^{3}Z

. Sławomir Biały (talk) 13:41, 11 August 2012 (UTC)[reply]

I may be missing something here. I understand what you are saying as defining the harmonic numbers in modulo arithmetic as

[H_{n}]_{m}=[1]_{m}+[2]_{m}^{-1}+[3]_{m}^{-1}+\cdots +[n]_{m}^{-1}=\sum _{k=1}^{n}[k]_{m}^{-1}

for the chosen modulus m (in this case m = p³). Without the map I suggested (which, as you say, may not exist) or the explicit indication of a prior mapping of integers Z → Z/mZ, this interpretation is less than obvious. Am I wrong? — Quondum^☏ 16:32, 11 August 2012 (UTC)[reply]

This only makes sense if the residues [1],...,[n] are units modulo m. If so, then the harmonic number H_n is uniquely defined. Sławomir Biały (talk) 16:40, 11 August 2012 (UTC)[reply]

The closest thing to the map you want uses localizations of the integers, which are contained in Q . The map you give doesn't make sense over all of Q , e.g. for p/q if when p & q are reduced to lowest terms, when q & n have a gcd > 1. Particularly: 1/n or 1/q when q divides n, "reduced mod n" doesn't make sense. If R is the subring of Q formed by adjoining inverses of all integers relatively prime to n, then this is the localization of Z at the ideal nZ, and is the largest subring where the "reduce mod n" map makes sense.John Z (talk) 21:06, 11 August 2012 (UTC)[reply]

Perhaps I should rephrase my question. My map was an attempt at interpretation, which perhaps we should forget (and thus can ignore its being a partial function). How am I to interpret the first-linked section, which appears to refer to reducing a non-integer modulo an integer? — Quondum^☏ 21:22, 11 August 2012 (UTC)[reply]

Well as Sławomir Biały indicated, these particular non-integers are reducible mod m = p³, as they are in the "reduce mod m" partial function's domain, this localization subring of Q. Put all the fractions in lowest terms & they all make sense in, are calculable in Z/mZ. Not much else to say.John Z (talk) 23:18, 11 August 2012 (UTC)[reply]

Thanks. It seems I was not so far off in my initial interpretation, though there seems to me Wikipedia is a bit weak on this aspect. — Quondum^☏ 00:01, 12 August 2012 (UTC)[reply]

Right, you weren't so far off.John Z (talk) 03:28, 12 August 2012 (UTC)[reply]