Let $R$ be a ring and let $n\in\mathbb{Z}$.

Given $a\in R$, I've seen $na$ defined as $$ na:=\begin{cases}0&\text{if }n=0,\\\underbrace{a+a+\cdots+a}_{n\text{ times}}&\text{if }n>0,\\\underbrace{-a-a-\cdots-a}_{|n|\text{ times}}&\text{if }n<0.\end{cases}\tag{1} $$ With this definition, I've even read that $$ (m+n)a=ma+na\tag{2} $$ and $$ (mn)1=(m1)(n1)\tag{3} $$ are obvious (and I do agree with this view).


(i) Isn't $(1)$ informal? If so, would $na$ be formally defined, for $n\geq0$, recursively via the equations \begin{align*} 0a&=0\tag{4}\\ (n+1)a&=na+a\tag{5} \end{align*} and for $n<0$ by $na:=-(-n)a$ ?

(ii) Would formal proofs of $(2)$ and $(3)$ be as given below?

I'm asking since I've never seen this done...

Proof of $(2)$: For fixed $m$, proceed by induction on $n$ for $n\geq-m$. Then, for $n<-m$, $$ (m+n)a=-(-m-n)a=-((-m)a+(-n)a)=-(-m)a-(-n)a=ma+na $$

Proof of $(3)$: First show the result for $n\in\mathbb{Z}$ and $m\geq0$. To do so, prove by induction on $m$ for fixed $n$ that $(mn)1=m(n1)$ (making use of $(2)$) and then that this equals $(m1)(n1)$, by induction on $m$ again. Finally, for $m<0$, $$ (mn)1=(-(-m)n)1=-((-m)n)1=-((-m)1)(n1)=(-(-m)1)(n1)=(m1)(n1) $$


Your comment in the answer to @lhf is correct. This is formalized in set theory with the concept of recursive definition, in this case, the recursive definition of a homomorphism $f_a : \mathbb{Z} \to R$ with the property that $f_a(1)=a$.

Informally it goes just like what you wrote: first define $f_a(1)$ which has already been done; then define $f_a(n)$ by a formula expressed in terms of the values of $f_a(1),…,f_a(n-1)$, in this case $f_a(n)=f_a(n-1)+a$. The principal of recursive definition then guarantees the existence of this function $f_a$ on the natural numbers, and then you extend to zero and negative numbers in the obvious way.

This informal way of doing a recursive definition is almost universally acceptable, because a complete and formal recursive definition is often quite tedious and without much mathematical interest. There are definitely exceptions, though, where one really ought to write out the details of a recursive definition.

Once you've proved the existence of $f_a$, you still want to prove that it is a homomorphism, and this will be a proof by induction.

And as long as we've gone this far, you may also wish to formally verify, using the axioms of set theory, that the formula $a \mapsto f_a$ defines a function from $R$ to the set of homomorphisms $\mathbb{Z} \to R$. This way one can completely formalize the definition of $n \cdot a$ to be $n \cdot a = f_a(n)$.


$na$ is simply $a^n$ in the additive group of $R$, written additively of course.

In other words, since $\mathbb Z$ is a cyclic group generated by $1 \in \mathbb Z$, there is a unique homomorphism $\mu: \mathbb Z \to R^+$ of additive groups such that $\mu(1)=a$. Then $na$ is simply $\mu(n)$.

  • $\begingroup$ But before defining a cyclic group, don't you have to define $a^n$ where the same considerations occur? To prove that $\mu$ is compatible with the addition in $R$, don't you have to define $na$ (or $a^n$) first? $\endgroup$ – Guest Dec 17 '15 at 2:13
  • $\begingroup$ You only have to do this for $\mathbb Z$, where the properties are a consequence of Peano's axioms. Then $na$ is defined to be $\mu(n)$. $\endgroup$ – lhf Dec 17 '15 at 8:49
  • $\begingroup$ I mean, how would you show that $\mu$ exists? Here, for example, they use $h^n$ to do so. $\endgroup$ – Guest Dec 17 '15 at 16:52
  • $\begingroup$ I think you define $\mu$ by induction: $\mu(1)=a$, $\mu(n+1)=a\mu(n)$ and extend to negatives. The point is that this definition is forced onto you if you want a homomorphism. Hence uniqueness. $\endgroup$ – lhf Dec 17 '15 at 17:21
  • $\begingroup$ Well, then it seems to me like you just use another notation. Instead of writing $na$ as in my question, you write $\mu(n)$ and your recursive definition of $\mu$ corresponds to the one in my question, part (i). It seems to me like you would still have to prove that $\mu(nm)=\mu(n)\mu(m)$ using induction (as in my question, part (ii)). $\endgroup$ – Guest Dec 17 '15 at 18:05

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.