I was reading my book on Elementary Algebra and saw this theorem:

Suppose that $F$ is a finite field of order $q$, then the group $F^*$ is a cyclic group of order $q-1$.

I don't understand the transition from a field to a group. How come the theorem says $F^*$ is a group when the element $0$ is omitted?

  • $\begingroup$ Please clarify: are you asking why it is a (cyclic) group, or why $0$ had to be omitted? $\endgroup$ – Bill Dubuque Jan 22 at 16:17

Well, $\mathbb{F}^*$ is the group $(\mathbb{F}\setminus \{0\},\cdot)$ hence you are only considering multiplication as operation, in particular, you need to throw out $0$. In more general algebra (Rings) this still holds, but there you throw out every element which is not invertible. hence you can imagine $\mathbb{F}^*$ as the multiplicative group of invertible elements.


$F^\ast$ is a set with a binary operation (the multiplication operation on $F$) which is associative (since it's associative on all of $F$) and in which all elements have inverses.

That's a group.

It wasn't a group before when you had all of $F$ since $0$ did not have an inverse. $F$ was a monoid though, with its multiplication operation.


If $F^{*}$ were to be a field, it would need an additive identity. But this would have to be the additive identity of the original field $F$, which as you pointed out, is omitted from $F^{*}$ since it isn't a unit. So $F^{*}$ is just a group.


Every field is a group if you remove the $0$ element and restrict it to just the multiplication operation because:

The definition of a group is:

A set of elements $G$ and an associated binary operation, $*$, so that:

1) $*$ is associative.

2) There is an element $e\in G$ so that for any $g\in G$ we have $e*g=g*e = g$.

3) For every $g \in G$ there is an element $g'$ so that $g*g' = g'*g = e$.

The definition of a field is

A set of elements $F$ and two associated binary operations, $+, \cdot$ so that

Addition axioms.

A1) $+$ is associative.

A2) $+$ is commutative.

A3) There is an element, $0 \in F$ so that for all $q\in F$ that $q+0=q$.

A4) For every $q \in F$ there is an element $-q\in F$ so the $q+ (-q) = 0$.

[Notice: A1, A3, A4, are the same axioms (with different notation) as the axioms for a group. Thus we can say a Field $F$ with only the addition operation considered forms a Group (with an addition condition that it is commutative).]

Multiplication Axioms

M1: $\cdot$ is associative.

M2: $\cdot$ is commutative.

M3: There exists an element $1 \in F$ so that for all $q \in F$, $1\cdot q = q$.

[Some text explicitly state as an axiom that $1 \ne 0$. Other texts will prove that if $1=0$ then then Field only has one element and is therefore trivial and not worth considering (The field would just be the set $\{a\}$ with the properties $a+a=a$ and $a\cdot a = a$ and there is absolutely nothing more to say about it). We'll take it for granted that $1 \ne 0$]

M4: For every element $q\in F$ so that $q\ne 0$ there is an element $q^{-1} \in F$ so that $q*q^{-1} = 1$.

[The axiom doesn't state that $q^{-1} \ne 0$. However it can be proven easily that $q*0 = 0 \ne 1$ so we can assume that $q^{-1} \ne 0.]

Note: M1, M3, M4 when applied to the set $F\setminus \{0\}$ are the same axioms as the axioms of a group with only different notation. So we can state that $F^*$, the set $F\setminus \{0\}$ associated with the binary operation $\cdot$, is a group (with the additional condition $\cdot$ is commutative).

The final axiom:

D: For all, $q,r,s \in F$, $q\cdot (r + s) = (q\cdot r) + (q\cdot s)$.


What the theorem is saying is that if $F$ is finite then the group is cyclic. Which is not so trivial.


Once you remove the $0$ of the field, you have a group under multiplication (the same multiplication as defined on the original field, that is), assuming your field is not the trivial field (i.e. a field with $0$ as its only element).

This is because, in a field, every non-zero element is invertible, so you can reinterpret $F\backslash\{0\}$ in a more useful light- as being the set of all invertible elements in $F$. If $a$ is invertible and $b$ is invertible, $ab$ is also invertible, as $(ab)^{-1}=b^{-1}a^{-1}$- this means that $F$ is closed with respect to products. Also, if $a$ is invertible, so is $a^{-1}$ (since $((a^{-1})^{-1}=a$). Also, multiplication on any field is defined to be associative so it remains associative in $F\backslash\{0\}$. Finally, since $F$ is not the trivial field, $F$ has at least one non-zero element, so $F\backslash\{0\}$ is non-empty. Suppose it has some element, $x$, in it. By the arguments already outlined, it must be the case that $x^{-1}\in F$ and, then, that $x*x^{-1}=1\in F$, so $F\backslash\{0\}$ has a neutral element (again, as the multiplication on $F\backslash\{0\}$ is the same on that on $F$, $1$ is the neutral element of $F\backslash\{0\}$).

(the assumption that $F$ is finite will come in handy in the next part of the argument)

So $F\backslash\{0\}$ is a group. But why a cyclic one?

You can actually make a more general claim than the one for which you ask for a proof in your question- that is, that any subset of a finite field, $F$, that acts as a group under multiplication (like $F\backslash\{0\}$) is in fact a cyclic group under multiplication (but this will require a smidge of information from the theory of rings of polynomials).

The proof (at least, the one that I've seen) is as follows. Suppose we have a subset, $G$, of a field, $F$, that is a group under multiplication. Since $F$ is finite, this group will also be finite. Let its number of elements be denoted by $n$ (some finite number).

Notice that the polynomial equation $x^k-1=0$ (as an element of $F[X]$) can have no more than $k$ distinct solutions as its degree is $k$ (this was the thing that needed a smidge of polynomial ring knowledge).

We also know that any solution, $y$, to this equation will have a multiplicative order that divides $k$, as $y^k-1=0\implies y^k=1$ (and, as mentioned before, $1$ must be the neutral element of any group under multiplication that is a subset of $F$).

But, note that if $k\not | n$, then, $x^k-1=0$ will have no solutions in $G$ as every element in $G$ must have an order that divides $n$ (as the order of any element will equal the order of its corresponding cyclic group, which, given its status as a subgroup of $G$, must, by Lagrange's theorem must have an order dividing $G$'s).

So, suppose for some divisor, $d$, of $n$, we have at least one element, $a$, in $G$ with order $d$. Then, we must have at least $d$ distinct elements in $G$ with an order dividing $d$ as the list of elements $(1,a,a^2,a^3,...,a^{d-1})$ is a list of distinct elements (given $a$'s order) and each element's order divides $d$ as, for any $r$, $(a^r)^d=a^{rd}=a^{dr}=(a^d)^r=1^r=1.$ But, on the other hand,(given what was said above about equations like $x^d-1=0$), there can be no more than $d$ elements in $G$ whose orders divide $d$. So, if there is at least one element in $G$ of order $d$ (where $d|n$), there are exactly $d$. Otherwise, there are $0$.

What remains now is simple- note that, if $\mathbb Z_{<d>}$ is the cyclic subgroup of $\mathbb Z_n$ consisting of all elements with additive order dividing $d$, then, treating $\mathbb Z_n$ as a group under addition (talking about the same $a$ as in the last paragraph), $<a>\cong \mathbb Z_{<d>}$ (forgive the loose notation). This means that the number of elements in $<a>$ with order equal to exactly $d$ (and note that every element in $G$ with order exactly $d$ must be in $<a>$) is equal to the number of elements in $\mathbb Z_{<d>}$ with order exactly $d$.

(in case you're wondering, $\mathbb Z_{<d>}$ looks like $\{0,\frac{n}{d},\frac{2n}{d},...,\frac{(d-1)n}{d}\}$)

So, denoting the number of elements in $\mathbb Z_n$ with order exactly $d$ (where $d|n$) with $\mathbb Z(d)$ and the number of elements in $G$ with order exactly $d$ with $G(d)$, we have either $\mathbb Z(d)=G(d)$ or $G(d)=0$, or, put another way $G(d)\leq \mathbb Z(d)$ always holds.

Now, what we argue is that we can partition $G$ into equivalence classes of elements based on their order. We can do the same with $\mathbb Z_n$. In each equivalence class in $G$ consisting of all elements of order $d$, we have $G(d)$ elements so the total number of elements across all equivalence classes must be $\sum_{d|n}G(d)$. Similarly, in $\mathbb Z_n$, the total number of elements in all equivalence classes must be $\sum_{d|n}\mathbb Z(d)$. Using the previous paragraph, then $$\sum_{d|n}G(d)\leq \sum_{d|n}\mathbb Z(d)$$

But, recall that these equivalence classes are disjoint sets whose union add up to $G$ or $\mathbb Z_n$ respectively so the total number of elements in either will be $\Big|G\Big|=\Big|\mathbb Z_n\Big|=n$.

So if for a single $d$, we have $G(d)<\mathbb Z(d)$, then we'll have $$\sum_{d|n}G(d)< \sum_{d|n}\mathbb Z(d)\\\implies n<n$$

A contradiction. So $G(d)=\mathbb Z(d)$ for all divisor of $n$ (even, non-divisors technically). In particular, $G(n)=\mathbb Z(n)$, i.e., there are as many elements in $G$ with order $n$ as there are in $\mathbb Z_n$- but there is always at least one element of additive order $n$ in $\mathbb Z_n$, i.e., $1$, so there is at least one element of order $n$ in $G$, say, $b$. Since $<b>\subset G$, but also $\operatorname{ord}(b)=n=\Big|G\Big|$, we must have $<b>=G$ and, so, $G$ is cyclic.

(please edit in or comment for any mistakes/ corrections)


By definition, a field is a commutative ring in which every nonzero element is a unit.

Thus if you consider the set of nonzero elements,it is a group with the multiplication operation as operation.

It is cyclic, as is the multiplicative group of any finite field.


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.