Principal Ideals and Ideal Generation

Principal and finitely generated ideals are the computational backbone of commutative algebra and algebraic geometry: an affine variety is cut out by a finitely generated ideal of polynomials, and Hilbert's basis theorem (that polynomial rings over Noetherian rings are Noetherian) guarantees finite generation is always available even when it is not obvious. In algebraic number theory, the failure of principal ideals to capture every ideal in a ring of integers measures the ideal class number, a fundamental invariant controlling factorization and appearing in class field theory and the study of Diophantine equations. In cryptography, ideal lattices — ideals in polynomial quotient rings such as $\mathbb{Z}[x]/(x^n+1)$ — underlie ring-based lattice cryptosystems used in post-quantum key exchange. In computer algebra, Gröbner basis algorithms are precisely machinery for finding well-behaved finite generating sets of ideals in polynomial rings, enabling automated solving of polynomial systems. Even in linear algebra, the minimal polynomial of a matrix arises as the unique monic generator of a principal ideal inside $F[x]$, the ideal of polynomials annihilating the matrix.

:::quiz[Why is $(2,x)$ not principal in $\mathbb{Z}[x]$?]

• Because $2$ and $x$ have no common factor besides units, yet no single polynomial generates exactly the "even constant term" ideal ✓
• Because $\mathbb{Z}[x]$ has no ideals at all
• Because $2$ is not a unit in $\mathbb{Z}[x]$
• Because $x$ is not invertible

Ask the tutor: How do sums and products of ideals interact with their generating sets, and why does $(a)+(b)=(a,b)$ while $(a)(b)=(ab)$ in a commutative ring?

Think of an ideal as a "closed room" in the ring: once you are inside it, adding two elements of the room, or multiplying a room-element by any ring element, keeps you inside the room. Given a handful of elements $a_1,\dots,a_n$, the ideal they generate is the smallest such room that contains them — you build it by throwing in $a_1,\dots,a_n$, then throwing in everything that closure forces you to add (all ring multiples, all sums), and stopping the instant no more elements are forced in. There are two equivalent ways to picture this construction, and the theorem below proves they agree. The top-down picture: intersect every ideal that contains $S$; since an intersection of ideals is again an ideal (a fact one checks directly from the definition of ideal), this produces an ideal, and it is visibly the smallest one containing $S$, since it sits inside every candidate. The bottom-up picture: start with $S$ and mechanically close it up — form all finite $R$-linear combinations $r_1a_1+\cdots+r_na_n$ — and check that this set is already an ideal. The theorem shows these two constructions yield the same set, which is why we are entitled to compute with generators concretely instead of reasoning abstractly about intersections of unnamed ideals. Principal ideals are the simplest nontrivial case: the "room" generated by one element $a$ is just all its ring-multiples, $Ra$. In $\mathbb{Z}$, $(a)$ is the set of all multiples of $a$, so $(3) = \{\dots,-6,-3,0,3,6,\dots\}$. When more generators are allowed, the room can only grow (or stay the same): $(a_1,\dots,a_n) \supseteq (a_1)$. Whether a multi-generator ideal secretly collapses to a principal one — as $(6,10)$ collapses to $(2)$ in $\mathbb{Z}$ — is a deep structural question, and rings where this always happens are the principal ideal domains foreshadowed at the end of this article.

Example 1: $(6,10)$ in $\mathbb{Z}$ collapses to a principal ideal. Work in $R = \mathbb{Z}$, commutative with identity $1$. By the theorem,

We claim $(6,10) = (2)$. $(6,10) \subseteq (2)$: Every element $6r+10s$ is even, since $6r$ and $10s$ are both even and a sum of two even integers is even. Hence $6r+10s \in (2)$ for all $r,s\in\mathbb{Z}$. $(2) \subseteq (6,10)$: We must express $2$ itself as an integer combination $6r+10s$. By the Euclidean algorithm on $6$ and $10$: $10 = 1\cdot 6 + 4$, then $6 = 1\cdot 4 + 2$, then $4 = 2\cdot 2 + 0$, so $\gcd(6,10)=2$. Back-substituting:

So $2 = 6\cdot 2 + 10\cdot(-1) \in (6,10)$. Since $2 \in (6,10)$ and $(6,10)$ is an ideal, every multiple $2n = n\cdot 2$ is absorbed into $(6,10)$, i.e. $(2) = \mathbb{Z}\cdot 2 \subseteq (6,10)$. Both containments give $(6,10) = (2)$. This illustrates the general fact (a preview of Principal Ideal Domains) that in $\mathbb{Z}$, every finitely generated ideal is principal, generated by the greatest common divisor of the given generators: $(a_1,\dots,a_n) = (\gcd(a_1,\dots,a_n))$.

Example 2: $(2,x)$ in $\mathbb{Z}[x]$ is not principal. Work in $R = \mathbb{Z}[x]$, the ring of polynomials in one variable over $\mathbb{Z}$, again commutative with identity. Let

First, note explicitly what $I$ looks like: an element $2f(x)+xg(x)$ has constant term $2f(0)$ (since $xg(x)$ contributes $0$ to the constant term), which is always even. Conversely, any polynomial $p(x) = c_0 + c_1x + c_2x^2+\cdots$ with $c_0$ even can be written as $2f(x) + xg(x)$: take $f(x) = c_0/2$ (constant) and $g(x) = c_1+c_2x+\cdots$, so that $2f(x) = c_0$ and $xg(x) = c_1x+c_2x^2+\cdots$, summing to $p(x)$. Hence

We now show $I$ is not principal, i.e. there is no single $h(x) \in \mathbb{Z}[x]$ with $I = (h)$. Suppose toward contradiction that $I = (h)$ for some $h \in \mathbb{Z}[x]$. Since $2 \in I = (h)$, we can write $2 = h(x)k(x)$ for some $k \in \mathbb{Z}[x]$. Comparing degrees (degree is additive for products of nonzero polynomials over the integral domain $\mathbb{Z}[x]$), $\deg h + \deg k = \deg 2 = 0$, forcing $\deg h = \deg k = 0$; so $h$ and $k$ are nonzero integer constants with $hk=2$. Thus $h \in \{1,-1,2,-2\}$.

• If $h = \pm 1$, then $(h) = \mathbb{Z}[x]$ (since $1 \in (h)$ forces every polynomial, including $x$ and constant $1$, to lie in $(h)$), but $1 \notin I$ because $1$ has odd constant term. Contradiction.
• If $h = \pm 2$, then $(h) = 2\mathbb{Z}[x]$, the set of polynomials all of whose coefficients are even. But $x \in I$ (constant term $0$ is even) while $x \notin 2\mathbb{Z}[x]$ (its coefficient of $x$ is $1$, odd). Contradiction.

In every case we reach a contradiction, so no such $h$ exists: $I=(2,x)$ is finitely generated (by two elements) but not principal. This shows $\mathbb{Z}[x]$ is not a principal ideal domain, even though $\mathbb{Z}$ is — generation by two elements can be genuinely necessary.