Motivation
The nonzero residues modulo seven form a multiplicative group of order six. Listing powers of three gives every nonzero residue, but listing powers of two reaches only three of them. There is a structural question behind this calculation: why does some suitable element reach the whole multiplicative group of a finite field?
Finiteness alone is insufficient. The group of units modulo eight has four elements, all with square one, so no element generates it. The extra information comes from polynomial arithmetic over a field: a nonzero polynomial cannot have more distinct roots than its degree. We will combine this restriction with a carefully constructed element of large order. The proof supplies a local prerequisite for the earlier classification of groups of order two distinct primes, and it also turns equations such as into linear congruences.
Use the order and cyclic-group rules and the polynomial root bound.
Definitions
Throughout, is a field and is its group under multiplication. A finite subgroup contains one and is closed under multiplication and inverses. In particular, every element of has a finite positive order. The identity has order one.
Definition 4.2.1
(Exponent and primitive multiplicative element)
For a finite group , its exponent is the positive integer
If a finite field has elements, a primitive multiplicative element of is an element for which . Such an element must have order . This definition does not yet assert that one exists.
The least common multiple is taken over a finite, nonempty set. It is therefore defined even for the trivial group, whose exponent is one. Every satisfies , but it is a separate question whether some element has order exactly . For example, the exponent of the nonabelian group is six, yet it has no element of order six. We need commutativity to establish the existence assertion below.
Theorems
Theorem 4.2.2
(Finite subgroups of field units are cyclic)
Let be a finite subgroup of . There is an element of order , and hence is cyclic. In particular, if is any finite field with elements, a primitive multiplicative element exists and is cyclic of order .
Theorem 4.2.3
(Power equations in a finite cyclic group)
Let have order , let , and write . Put . The equation in is solvable exactly when . When it is solvable, it has exactly distinct solutions. Equivalently,
Consequently, in a finite field with elements, has exactly distinct roots. For , the equation has exactly the one solution .
The first assertion concerns an existing finite field. It neither constructs fields of every prime-power size nor classifies their isomorphism types. The power-equation theorem counts solutions; finding a generator or a discrete logarithm is an additional computational task.
Concept lensCombinatorial
Cyclicity as a root-count condition
The theorem has an equivalent counting form: a group of finite order is cyclic exactly when, for each divisor of , at most elements satisfy . If some has order , those solutions are exactly the elements of , so precisely elements have order . Each count is therefore or , and forces every count to be , including . In the root bound supplies the hypothesis.
Proof sketch or proof idea
Combining commuting elements of coprime orders
Proof
(Orders multiply for commuting elements)
First suppose commute and have coprime finite orders . Their cyclic subgroups have trivial intersection: the order of an element in the intersection divides both and , hence is one. Certainly . If for a positive integer , commutativity gives . This common element belongs to the trivial intersection, so . Thus and , implying . The order of is therefore exactly .
Counterexample mode
Coprime orders without commutativity
The commutativity hypothesis is doing real work. Suppose one claimed that whenever have coprime orders , the product has order . In , composing from right to left, take of order two and of order three. Then sends , and , so has order two, not six. The step that fails is , which needs ; here . The repaired statement is the one just proved: commuting elements of coprime orders have a product of order . Inside , commutativity is automatic.
Producing an element whose order is the exponent
Proof
(Assembling the prime-power parts)
Now let be a finite abelian group and put . If , every element is one, so the identity already has the required order. Otherwise, write over its distinct prime divisors. For each , the definition of least common multiple gives an element whose order contains exactly as its -part. Write that order as , where , and set .
The order of is . Indeed, exactly when , which is exactly . The elements commute, and their orders are pairwise coprime. Applying the preceding argument repeatedly shows that
This argument uses only finite abelian-group arithmetic. It does not assume the classification of finite abelian groups or a primitive-element theorem.
The root bound forces the whole group to appear
Proof
(Counting roots of )
Apply this construction to , which is abelian because field multiplication commutes. Every element of is a root of . This polynomial is nonzero of degree : its leading coefficient is one, in every characteristic. The field root bound therefore gives . On the other hand the element just constructed has distinct powers in , so . Thus and , proving cyclicity. Taking proves the finite-field consequence.
Proof X-ray
A largest order needs both algebra and a root bound
Choosing an element of maximum order does not by itself show that every other order divides it. The prime-power construction supplies an element attaining the least common multiple of all orders. The polynomial bound then forces its powers to exhaust the group. The two steps explain different features: commutativity lets orders be assembled, while the field hypothesis limits how many elements can satisfy the resulting equation. No assumption about distinct roots in a splitting field, and no derivative calculation, enters this argument.
Turning powers into a linear congruence
Proof
(Reducing to )
Every has a unique expression with taken modulo . Thus means . If this congruence holds, for an integer , so . Conversely suppose . Divide all three integers by and put , , . Bézout gives integers with . Hence solves .
Since is coprime to , every solution is . Lifting this residue class to classes modulo gives precisely
They are distinct modulo , and they give all solutions. This includes , when the congruence modulo one imposes no restriction. It also shows why dividing a congruence must change its modulus.
Finally, means , which is equivalent to . For the exponent always satisfies the condition, yielding the stated root count in . Zero is not a root of for positive , so this is also the count in the whole field. If and , multiplying by would give ; this proves the remaining assertion.
Worked examples
Worked example 4.2.4
(A generator and a fourth-power equation modulo seven)
The successive powers modulo seven are . They are distinct, so three generates . To solve , write and . Then . The gcd is two, which divides the right side. Dividing by two gives , hence . The two classes modulo six are , giving .
Direct substitution checks both: . In contrast would require , which is impossible. A nonzero right side need not be an th power.
Worked example 4.2.5
(A primitive element in a nine-element field)
The polynomial has no root in , since the only squares are zero and one, and neither becomes zero after adding one. A degree-two polynomial over a field is irreducible if it has no root. Therefore is a field, with nine elements and . Here denotes the residue class of , not an element chosen from .
Put . Then , , and . Its order divides eight but does not divide four, so it is eight: every proper positive divisor of eight divides four. Thus is primitive, while has order four and is not primitive.
To solve , note . Hence , giving . Since and , the roots are and . They are negatives of each other, as expected. Squaring either expression gives using ; the congruence argument proves that there are no additional roots.
Worked example 4.2.6
(Positive exponents and repeated roots)
In , the polynomial has distinct roots. No list of all twelve units is needed to obtain this count. The kernel of the eighth-power map has four elements, and its image has three: the twelve input elements split into equal fibers of size four over the image.
Characteristic dividing the exponent does not invalidate the count. In , , so it has the single distinct root one, in agreement with . Three is the root's multiplicity, not the number of field elements that solve the equation. Our theorem always counts distinct solutions.
Common mistakes
Common mistake
(The units of a finite ring need not be cyclic)
The units of an arbitrary finite ring need not form a cyclic group. In they are , and all four solve . Every nonidentity unit has order two, so the unit group is not cyclic. The field root bound fails here: does not force either factor to be zero in a ring with zero divisors. This identifies the broken step rather than merely presenting an exception.
Common mistake
(No claim about infinite fields)
Nor does the theorem say is cyclic for an infinite field. For example is infinite and contains of order two. An infinite cyclic group has no nonidentity element of finite order: if has infinite order and with , then and . Thus cannot be cyclic.
Common mistake
(Generators, testing, and the condition )
A primitive multiplicative element is not the same as a nonzero element, and finding one by testing candidates is not the existence proof. For a cyclic group of order , is a generator exactly when ; its order is . Also retain in the polynomial root-count formula. For , is the zero polynomial and every field element is a root. Within alone, has all solutions when and none otherwise.
Summary
A finite abelian group has an element attaining its exponent. Inside a field's multiplicative group, the polynomial root bound forces this exponent to equal the group's size, proving cyclicity. After a generator is chosen, power equations become linear congruences: check divisibility by , solve at the reduced modulus, and lift all solutions back.
Exercises
Quick check
(In , verify that two is a primitive multiplicative element. Solve and explain why the answer is complete.)
Solution · Answer
Modulo eleven, , so the order of two does not divide five; also . Its order divides ten by Lagrange, whose positive divisors are , so the order is ten. Since , the equation is . Dividing by two gives , hence . Thus modulo ten, and . There are exactly two solutions by the gcd theorem, and direct checks give .
Quick check
(Let be a finite field with sixteen elements. Without constructing , determine the numbers of solutions of and . Determine the number of fifth powers among the nonzero elements.)
Solution · Answer
The group has order fifteen. The root counts are and . The fifth-power map has a five-element kernel. Every nonempty fiber is a coset of that kernel: if , then , and the converse holds. Its image consequently has elements, so precisely three nonzero elements are fifth powers. Including zero would give four fifth powers in the whole field.
Quick check
(Prove that a finite subgroup has exactly one subgroup of order for every positive divisor of , and identify it as the set of roots of in .)
Solution · Answer
Put and choose a generator . The subgroup has order , by the cyclic order formula, and all its elements solve . That equation has exactly solutions, so its solution set is exactly . Any subgroup of order has every element satisfying by Lagrange applied within . Thus ; equal finite sizes give . This proof does not assume uniqueness before identifying the roots.
Quick check
(Suppose has order . Show that an element is primitive exactly when for every prime dividing . Use the test to check again in the nine-element field above.)
Solution
If has order , no smaller positive exponent gives one. Conversely let its order be , a divisor of . If , the integer has a prime divisor . Then is a multiple of , so , contradicting the test. Thus . In only the prime two divides , and , so is primitive. The case has only the identity, which is its own generator; the stated test excluded it to keep the prime-divisor discussion explicit.