The ancient problem of solving polynomial equations by radicals — expressing roots using only the four arithmetic operations and the extraction of $n$-th roots — runs into a wall at degree five. The quadratic formula has been known since antiquity. The cubic and quartic formulas, discovered in sixteenth-century Italy, required enormous ingenuity. But every attempt to find a similar formula for the general quintic failed. By the early nineteenth century, Abel had proved that no such formula exists, and Galois had uncovered the reason: whether a polynomial can be solved by radicals depends on a certain group of symmetries attached to it. That group is the Galois group. The insight driving Galois's theory is that the roots of a polynomial are not isolated numbers; they are related by algebraic dependencies, and any algebraic relation that holds among the roots must be respected by any automorphism that fixes the ground field. The Galois group captures exactly these symmetries. It is a finite group that encodes the arithmetic of the splitting field, and the interplay between its subgroups and the intermediate fields of the extension is one of the deepest correspondences in all of algebra. [example: Roots of Unity] Before introducing the formalism, consider the polynomial $f(x) = x^4 - 1$ over $\mathbb{Q}$. Its roots are $1, -1, i, -i$. The splitting field is $\mathbb{Q}(i)$, since $i$ generates all four roots: $1 = i^4$, $-1 = i^2$, $-i = i^3$. Any automorphism $\sigma$ of $\mathbb{Q}(i)$ fixing $\mathbb{Q}$ must send $i$ to another root of $x^2 + 1$, which means either $\sigma(i) = i$ or $\sigma(i) = -i$. There are exactly two such automorphisms: the identity and complex conjugation. The Galois group $\operatorname{Gal}(\mathbb{Q}(i)/\mathbb{Q})$ is therefore a group of order two, isomorphic to $\mathbb{Z}/2\mathbb{Z}$. Notice that this group also permutes the four roots of $x^4 - 1$: the identity fixes all four, while complex conjugation swaps $i \leftrightarrow -i$ and fixes $1$ and $-1$. Any algebraic relation among $1, -1, i, -i$ with rational coefficients — such as $i \cdot (-i) = 1$ — is preserved by both automorphisms. [/example] ## Definition The key objects in Galois theory are field extensions, and the Galois group is defined for a specific class of them. To understand why the definition takes its particular form, we need to understand what can go wrong with arbitrary extensions. Consider the extension $\mathbb{Q}(\sqrt[3]{2})/\mathbb{Q}$. An automorphism fixing $\mathbb{Q}$ must send $\sqrt[3]{2}$ to another cube root of $2$. The real cube roots of $x^3 - 2$ are just $\sqrt[3]{2}$ itself, so no nontrivial automorphism of $\mathbb{Q}(\sqrt[3]{2})$ fixes $\mathbb{Q}$. The group of such automorphisms has order one, far too small to contain the information one would want. The problem is that $\mathbb{Q}(\sqrt[3]{2})$ does not contain the other two cube roots of $2$ — namely $\omega\sqrt[3]{2}$ and $\omega^2\sqrt[3]{2}$, where $\omega = e^{2\pi i/3}$ — so there are no automorphisms to permute them. The correct setting requires a splitting field, and the extension must also be separable. When both conditions are met, the extension is called Galois. [definition: Galois Extension] Let $k$ be a field and $K/k$ a finite field extension. The extension $K/k$ is **Galois** if it is both normal and separable: - $K/k$ is **normal** if $K$ is the splitting field of some polynomial $f \in k[x]$. - $K/k$ is **separable** if every element $\alpha \in K$ has a separable minimal polynomial over $k$, meaning $\operatorname{min}(\alpha, k)$ has no repeated roots in $\overline{k}$. [/definition] Normality has an equivalent reformulation: $K/k$ is normal if and only if every irreducible polynomial in $k[x]$ that has at least one root in $K$ splits completely in $K[x]$. This reformulation makes clear that normality is a global condition — it is not enough for $K$ to contain one root of an irreducible polynomial; it must contain all of them. Separability is automatic over fields of characteristic zero (such as $\mathbb{Q}$, $\mathbb{R}$, $\mathbb{C}$) and over finite fields, so in practice most extensions encountered in a first course in Galois theory are automatically separable. Normality is the structural requirement: it ensures that the splitting field contains all the conjugates of every element. With the notion of a Galois extension in place, the Galois group is the automorphism group of the extension. [definition: Galois Group] Let $K/k$ be a Galois extension. The **Galois group** of $K/k$, denoted $\operatorname{Gal}(K/k)$, is the group of all field automorphisms of $K$ that fix $k$ pointwise: \begin{align*} \operatorname{Gal}(K/k) := \{ \sigma \in \operatorname{Aut}(K) : \sigma(\alpha) = \alpha \text{ for all } \alpha \in k \}. \end{align*} The group operation is composition of automorphisms. [/definition] [remark: Fixing the base field] The condition $\sigma(\alpha) = \alpha$ for all $\alpha \in k$ means that $\sigma$ is a $k$-algebra automorphism of $K$: it is a ring automorphism of $K$ that is $k$-linear. Since every ring automorphism of a field is injective and $K$ is finite-dimensional over $k$, this forces $\sigma$ to be bijective as well. So elements of $\operatorname{Gal}(K/k)$ are automatically $k$-linear bijections $K \to K$ that respect the field operations. [/remark] The most important structural fact about the Galois group is its size. [quotetheorem:3325] This theorem is the foundation of everything that follows. It says that the Galois group is exactly as large as the degree of the extension — large enough to "see" every dimension of $K$ as a $k$-vector space. The proof proceeds via the primitive element theorem (which guarantees $K = k(\alpha)$ for some $\alpha$, since the extension is separable and finite) and the observation that automorphisms are determined by where they send $\alpha$, which must go to a root of $\operatorname{min}(\alpha, k)$, a separable polynomial whose degree equals $[K:k]$. ## The Fundamental Theorem of Galois Theory The real power of the Galois group emerges from a perfect correspondence between its subgroups and the intermediate fields of the extension. This correspondence is not merely a coincidence of cardinality; it is a deep structural duality that reverses inclusion. To state the correspondence, we need to define the fixed field of a subgroup. The Galois correspondence works by reversing the direction of inclusion: large subgroups of $G$ fix fewer elements and so correspond to small intermediate fields, while small subgroups fix more and correspond to large intermediate fields. To make this precise, we ask: given a subgroup $H \le \operatorname{Gal}(K/k)$, which elements of $K$ survive every symmetry in $H$? These are precisely the elements that $H$ cannot distinguish from their neighbors in $K$ — the elements $\alpha$ such that every $\sigma \in H$ sends $\alpha$ back to itself. Collecting them gives the fixed field, the largest intermediate field on which $H$ acts as the identity. [definition: Fixed Field] Let $K/k$ be a Galois extension and $H \le \operatorname{Gal}(K/k)$ a subgroup. The **fixed field** of $H$ is \begin{align*} K^H := \{ \alpha \in K : \sigma(\alpha) = \alpha \text{ for all } \sigma \in H \}. \end{align*} [/definition] One checks that $K^H$ is indeed a subfield of $K$ containing $k$: if $\alpha, \beta \in K^H$ and $\sigma \in H$, then $\sigma(\alpha + \beta) = \sigma(\alpha) + \sigma(\beta) = \alpha + \beta$ and $\sigma(\alpha\beta) = \sigma(\alpha)\sigma(\beta) = \alpha\beta$, so $K^H$ is closed under the field operations. Now we have two maps: one that sends intermediate fields to subgroups (take the automorphisms fixing the field), and one that sends subgroups to intermediate fields (take the fixed field). The Fundamental Theorem says these maps are inverse bijections. [illustration:galois-correspondence] [quotetheorem:1274] The inclusion-reversal is essential: larger subgroups correspond to smaller intermediate fields, and smaller subgroups correspond to larger intermediate fields. In the extreme cases, the trivial subgroup $\{e\}$ corresponds to $K$ itself (only the identity fixes all of $K$), and the full group $G$ corresponds to $k$ (every automorphism fixes the base field). Part (3) is the statement that Galois extensions correspond to normal subgroups. When $H \trianglelefteq G$, the fixed field $K^H$ is "Galois over $k$" in the same sense as $K$ is, and its Galois group is the quotient $G/H$. This is the algebraic mechanism behind the analysis of solvability: to solve a polynomial by radicals, one builds a tower of extensions, each obtained by adjoining a root of unity or an $n$-th root, and this tower corresponds to a subnormal series of the Galois group with abelian factors — that is, to solvability of the group. [example: Galois Group of a Cyclotomic Extension] Let $n \ge 1$ be a positive integer and let $\zeta_n = e^{2\pi i/n}$ be a primitive $n$-th root of unity. The cyclotomic field is $\mathbb{Q}(\zeta_n)$, and $\mathbb{Q}(\zeta_n)/\mathbb{Q}$ is a Galois extension. The minimal polynomial of $\zeta_n$ over $\mathbb{Q}$ is the $n$-th cyclotomic polynomial $\Phi_n(x)$, whose degree is $\varphi(n)$ (Euler's totient), so $[\mathbb{Q}(\zeta_n) : \mathbb{Q}] = \varphi(n)$. An automorphism $\sigma \in \operatorname{Gal}(\mathbb{Q}(\zeta_n)/\mathbb{Q})$ is determined by $\sigma(\zeta_n)$, which must be another primitive $n$-th root of unity. The primitive $n$-th roots of unity are exactly $\zeta_n^a$ for $a \in (\mathbb{Z}/n\mathbb{Z})^\times$, the units modulo $n$. So we obtain a map \begin{align*} \operatorname{Gal}(\mathbb{Q}(\zeta_n)/\mathbb{Q}) &\to (\mathbb{Z}/n\mathbb{Z})^\times \\ \sigma_a &\mapsto a, \end{align*} where $\sigma_a(\zeta_n) = \zeta_n^a$. To see this is a group homomorphism, note that \begin{align*} \sigma_a(\sigma_b(\zeta_n)) = \sigma_a(\zeta_n^b) = (\zeta_n^b)^a = \zeta_n^{ab} = \sigma_{ab}(\zeta_n), \end{align*} so $\sigma_a \circ \sigma_b = \sigma_{ab}$, and composition of automorphisms corresponds to multiplication in $(\mathbb{Z}/n\mathbb{Z})^\times$. This map is an isomorphism: \begin{align*} \operatorname{Gal}(\mathbb{Q}(\zeta_n)/\mathbb{Q}) \cong (\mathbb{Z}/n\mathbb{Z})^\times. \end{align*} In particular, the Galois group is abelian. For $n = p$ prime, $(\mathbb{Z}/p\mathbb{Z})^\times \cong \mathbb{Z}/(p-1)\mathbb{Z}$, which is cyclic of order $p-1$. As a specific case, take $n = 8$. Then $\varphi(8) = 4$ and $(\mathbb{Z}/8\mathbb{Z})^\times = \{1, 3, 5, 7\}$. This group satisfies $3^2 = 9 \equiv 1$, $5^2 = 25 \equiv 1$, $7^2 = 49 \equiv 1 \pmod 8$, so every non-identity element has order two. Thus $(\mathbb{Z}/8\mathbb{Z})^\times \cong \mathbb{Z}/2\mathbb{Z} \times \mathbb{Z}/2\mathbb{Z}$, and $\operatorname{Gal}(\mathbb{Q}(\zeta_8)/\mathbb{Q}) \cong \mathbb{Z}/2\mathbb{Z} \times \mathbb{Z}/2\mathbb{Z}$. [/example] ## Galois Groups as Permutation Groups Every Galois group acts naturally on the roots of the polynomial that generated the splitting field. This action embeds the Galois group as a subgroup of a symmetric group and provides the most concrete way to compute Galois groups in practice. Let $f \in k[x]$ be a separable polynomial of degree $n$, and let $K$ be the splitting field of $f$ over $k$. Write the roots as $\alpha_1, \dots, \alpha_n \in K$. Any $\sigma \in \operatorname{Gal}(K/k)$ permutes these roots: since $f(\alpha_i) = 0$ and $\sigma$ fixes the coefficients of $f$ (which lie in $k$), we have \begin{align*} f(\sigma(\alpha_i)) = \sigma(f(\alpha_i)) = \sigma(0) = 0, \end{align*} so $\sigma(\alpha_i)$ is also a root of $f$. This gives a group homomorphism $\operatorname{Gal}(K/k) \to S_n$. [quotetheorem:3326] Injectivity follows from the fact that an automorphism of $K/k$ is completely determined by its values on the roots of $f$: since $K$ is generated over $k$ by $\alpha_1, \dots, \alpha_n$, any $\sigma$ fixing all roots must be the identity. The image of $\operatorname{Gal}(K/k)$ in $S_n$ is a transitive subgroup whenever $f$ is irreducible: given any two roots $\alpha_i$ and $\alpha_j$, there exists $\sigma \in \operatorname{Gal}(K/k)$ with $\sigma(\alpha_i) = \alpha_j$, because $\alpha_i$ and $\alpha_j$ are both roots of the same irreducible polynomial and so have the same minimal polynomial over $k$. [example: Splitting Field of $x^3 - 2$] Let $f(x) = x^3 - 2 \in \mathbb{Q}[x]$. The roots are $\alpha_1 = \sqrt[3]{2}$, $\alpha_2 = \omega\sqrt[3]{2}$, and $\alpha_3 = \omega^2\sqrt[3]{2}$, where $\omega = e^{2\pi i/3}$. The splitting field is $K = \mathbb{Q}(\sqrt[3]{2}, \omega)$. We compute the degree: $[\mathbb{Q}(\sqrt[3]{2}) : \mathbb{Q}] = 3$ since $x^3 - 2$ is irreducible over $\mathbb{Q}$ by Eisenstein at $p = 2$. Then $[\mathbb{Q}(\sqrt[3]{2}, \omega) : \mathbb{Q}(\sqrt[3]{2})]$ equals either $1$ or $2$: it is $1$ only if $\omega \in \mathbb{Q}(\sqrt[3]{2})$, but $\mathbb{Q}(\sqrt[3]{2}) \subset \mathbb{R}$ while $\omega \notin \mathbb{R}$, so the degree is $2$. Therefore $[K : \mathbb{Q}] = 6$. By the fundamental theorem, $|\operatorname{Gal}(K/\mathbb{Q})| = 6$. Since $\operatorname{Gal}(K/\mathbb{Q})$ embeds in $S_3$ and $|S_3| = 6$, the embedding must be surjective. So $\operatorname{Gal}(K/\mathbb{Q}) \cong S_3$. Explicitly, the six automorphisms are determined by their action on $\sqrt[3]{2}$ and $\omega$: | Automorphism | $\sqrt[3]{2} \mapsto$ | $\omega \mapsto$ | |---|---|---| | $e$ | $\sqrt[3]{2}$ | $\omega$ | | $\sigma$ | $\omega\sqrt[3]{2}$ | $\omega$ | | $\sigma^2$ | $\omega^2\sqrt[3]{2}$ | $\omega$ | | $\tau$ | $\sqrt[3]{2}$ | $\omega^2$ | | $\tau\sigma$ | $\omega\sqrt[3]{2}$ | $\omega^2$ | | $\tau\sigma^2$ | $\omega^2\sqrt[3]{2}$ | $\omega^2$ | Here $\sigma$ has order $3$ and $\tau$ has order $2$. One verifies that $\tau\sigma\tau^{-1} = \sigma^{-1} = \sigma^2$, so the group has the presentation $\langle \sigma, \tau \mid \sigma^3 = \tau^2 = e, \, \tau\sigma\tau^{-1} = \sigma^2 \rangle$, which is the dihedral group $D_6 \cong S_3$. The subgroup $\langle \sigma \rangle \cong \mathbb{Z}/3\mathbb{Z}$ is normal in $S_3$ (it has index $2$, hence is normal), and by the Fundamental Theorem it corresponds to the fixed field $K^{\langle \sigma \rangle}$. The automorphism $\sigma$ fixes $\omega$ (since $\sigma(\omega) = \omega$), so $\mathbb{Q}(\omega) \subset K^{\langle \sigma \rangle}$. Since $[K^{\langle \sigma \rangle} : \mathbb{Q}] = [S_3 : \langle \sigma \rangle] = 2 = [\mathbb{Q}(\omega) : \mathbb{Q}]$, we get $K^{\langle \sigma \rangle} = \mathbb{Q}(\omega)$. Moreover, $\mathbb{Q}(\omega)/\mathbb{Q}$ is Galois with group $S_3/\langle\sigma\rangle \cong \mathbb{Z}/2\mathbb{Z}$, consistent with the fact that $\omega$ satisfies $x^2 + x + 1 = 0$ and this polynomial has two roots $\omega, \omega^2$ that are exchanged by $\tau$. [/example] [illustration:s3-subfield-lattice] ## Solvability by Radicals The original motivation for Galois theory was the question of solvability of polynomial equations by radicals. Galois's answer connects the structure of the Galois group directly to whether such a formula can exist. [definition: Radical Extension] A field extension $K/k$ is a **radical extension** if $K = k(\alpha_1, \dots, \alpha_r)$ where each $\alpha_i$ satisfies $\alpha_i^{n_i} \in k(\alpha_1, \dots, \alpha_{i-1})$ for some positive integer $n_i$. [/definition] In other words, $K$ is built from $k$ by successively adjoining $n$-th roots. [definition: Solvable by Radicals] A polynomial $f \in k[x]$ is **solvable by radicals** if its splitting field is contained in some radical extension of $k$. [/definition] To connect this to group theory, we need to isolate the structural property of the Galois group that governs solvability. The key observation is that adjoining an $n$-th root to a field produces an extension whose Galois group is cyclic (hence abelian), at least once the field contains a primitive $n$-th root of unity. Building a radical tower therefore corresponds to breaking the Galois group into successive abelian layers — abelian quotients at each step. This is the defining property of a solvable group, and it is the right notion because it exactly captures when a group can be dismantled by a sequence of abelian extensions. [definition: Solvable Group] A group $G$ is **solvable** if there exists a chain of subgroups \begin{align*} \{e\} = G_0 \trianglelefteq G_1 \trianglelefteq \cdots \trianglelefteq G_r = G \end{align*} such that $G_i \trianglelefteq G_{i+1}$ and $G_{i+1}/G_i$ is abelian for each $i \in \{0, 1, \dots, r-1\}$. [/definition] Abelian groups are solvable (take a single-step series $\{e\} \trianglelefteq G$). Symmetric groups $S_n$ are solvable for $n \le 4$: for example, $S_4$ has the chain $\{e\} \trianglelefteq V_4 \trianglelefteq A_4 \trianglelefteq S_4$ where $V_4 = \{e, (12)(34), (13)(24), (14)(23)\}$ is the Klein four-group. The alternating group $A_5$ is simple (has no proper normal subgroups), and since $|A_5| = 60$ and $A_5$ is not abelian, the chain $\{e\} \trianglelefteq A_5$ does not work. This non-solvability of $A_5$ is the algebraic obstruction to solving the general quintic. [quotetheorem:3327] The content of this theorem is remarkable: a purely algebraic condition on the symmetry group of the polynomial — which knows nothing about radicals on the face of it — exactly characterizes when a formula by radicals can exist. The proof in the "if" direction constructs a radical tower from the subnormal series of the Galois group, using the fact that cyclic extensions of degree $n$ (over a field containing a primitive $n$-th root of unity) are generated by adjoining $n$-th roots. The "only if" direction shows that any radical extension has a solvable Galois group over the ground field. [example: Unsolvability of the General Quintic] Let $k = \mathbb{Q}(t_1, \dots, t_5)$ be the field of rational functions in five indeterminates and consider the general quintic $f(x) = x^5 - t_1 x^4 + t_2 x^3 - t_3 x^2 + t_4 x - t_5$, whose coefficients are the elementary symmetric polynomials in five variables. The splitting field of $f$ over $k$ is $K = k(\alpha_1, \dots, \alpha_5)$ where $\alpha_1, \dots, \alpha_5$ are the five roots. The Galois group $\operatorname{Gal}(K/k)$ is isomorphic to $S_5$: since the $t_i$ are algebraically independent and the $\alpha_i$ are transcendental over $k$, any permutation of the $\alpha_i$ extends to a field automorphism of $K$ fixing $k$. Therefore $\operatorname{Gal}(K/k) \cong S_5$. Now $S_5$ is not solvable. Its derived series is \begin{align*} S_5 \supset A_5 \supset A_5 \supset \cdots, \end{align*} which stabilizes at $A_5$ because $A_5$ is perfect ($[A_5, A_5] = A_5$: every element of $A_5$ is a product of commutators). Since $A_5$ is non-abelian and simple, the derived series never reaches $\{e\}$. By Galois's criterion, the general quintic is not solvable by radicals. This does not mean no quintic is solvable — for instance $x^5 - 2$ has Galois group isomorphic to the Frobenius group of order $20$, which is solvable. The theorem says the general quintic (with generic, not special, coefficients) cannot be solved. [/example] ## Computing Galois Groups Knowing the theoretical framework is one thing; computing specific Galois groups is another, and it requires a toolkit of methods. The discriminant and resolvent polynomials are among the most useful. The discriminant of a polynomial $f$ with roots $\alpha_1, \dots, \alpha_n$ is \begin{align*} \Delta_f = \prod_{i < j} (\alpha_i - \alpha_j)^2. \end{align*} The discriminant lies in $k$ (it is a symmetric function of the roots, hence a polynomial in the coefficients of $f$). Its square root $\delta_f = \prod_{i < j}(\alpha_i - \alpha_j)$ lies in the splitting field $K$ and is related to even permutations of the roots. [quotetheorem:1325] For cubics, this theorem combined with the fact that transitive subgroups of $S_3$ are either $\mathbb{Z}/3\mathbb{Z}$ or $S_3$ gives a complete classification: an irreducible cubic $f \in \mathbb{Q}[x]$ has Galois group $\mathbb{Z}/3\mathbb{Z}$ if $\Delta_f$ is a perfect square in $\mathbb{Q}$, and $S_3$ otherwise. [example: Galois Group of a Cubic] Let $f(x) = x^3 - 3x + 1 \in \mathbb{Q}[x]$. We compute the discriminant. For a depressed cubic $x^3 + px + q$, the discriminant is $\Delta_f = -4p^3 - 27q^2$. Here $p = -3$ and $q = 1$, so \begin{align*} \Delta_f = -4(-3)^3 - 27(1)^2 = -4(-27) - 27 = 108 - 27 = 81. \end{align*} Since $\Delta_f = 81 = 9^2$ is a perfect square in $\mathbb{Q}$, the Galois group satisfies $G \le A_3 = \mathbb{Z}/3\mathbb{Z}$. Now $f$ is irreducible over $\mathbb{Q}$: it has no rational roots (by the rational root theorem, the only candidates are $\pm 1$, and $f(1) = -1 \ne 0$, $f(-1) = 3 \ne 0$). An irreducible polynomial of degree $3$ has a Galois group that acts transitively on its three roots, and the only transitive subgroup of $S_3$ contained in $A_3$ is $A_3 = \mathbb{Z}/3\mathbb{Z}$ itself. Therefore $\operatorname{Gal}(K/\mathbb{Q}) \cong \mathbb{Z}/3\mathbb{Z}$. This means the splitting field has degree $3$ over $\mathbb{Q}$, and indeed all three roots are real (since $\Delta_f > 0$ for a cubic corresponds to three distinct real roots). In fact, $x^3 - 3x + 1$ factors in $\mathbb{Q}(\zeta_9)$ as its roots are $2\cos(2\pi/9)$, $2\cos(8\pi/9)$, $2\cos(14\pi/9)$. [/example] For quartics, the analysis involves the **resolvent cubic**, which is a degree-three polynomial whose roots are related to the three ways of pairing the four roots of the quartic. The Galois group of the quartic can be $S_4$, $A_4$, $D_8$ (dihedral of order $8$), $V_4$ (Klein four-group), or $\mathbb{Z}/4\mathbb{Z}$, and the discriminant and resolvent cubic together determine which case applies. [quotetheorem:3328] ## Infinite and Profinite Galois Groups The discussion so far has focused on finite Galois extensions. But Galois theory extends naturally to infinite extensions, and the correct framework for capturing the structure of an infinite Galois group requires topology. Suppose $k$ is a field and $k^{\mathrm{sep}}$ is its separable closure — the compositum inside $\overline{k}$ of all finite separable extensions of $k$. The extension $k^{\mathrm{sep}}/k$ is an algebraic extension, but in general infinite. The **absolute Galois group** $\operatorname{Gal}(k^{\mathrm{sep}}/k)$ is the group of all automorphisms of $k^{\mathrm{sep}}$ fixing $k$. For an infinite Galois extension $K/k$, the naive version of the Fundamental Theorem fails: there are subgroups of $\operatorname{Gal}(K/k)$ with no corresponding intermediate field. The problem is that an infinite group can have "too many" subgroups — subgroups that are not built from any finite piece of the extension and so cannot correspond to any field sitting between $k$ and $K$. What goes wrong concretely is this: a subgroup $H$ carved out by some abstract set-theoretic construction may not arise as $\operatorname{Gal}(K/F)$ for any intermediate field $F$. The fix is to topologize $G$ so that the "good" subgroups — those arising from intermediate fields — are precisely the closed ones. The Krull topology is the coarsest topology on $G$ that makes it into a topological group and captures exactly these subgroups: its open sets are built from the finite Galois subextensions, which are the pieces of the infinite extension that one can actually see. [definition: Krull Topology] Let $K/k$ be a (possibly infinite) Galois extension. The **Krull topology** on $G = \operatorname{Gal}(K/k)$ is the topology whose open sets are unions of cosets of the form $\sigma \operatorname{Gal}(K/F)$, where $F/k$ ranges over all finite Galois subextensions of $K/k$ and $\sigma \in G$. [/definition] Equivalently, the collection $\{\operatorname{Gal}(K/F) : F/k \text{ finite Galois}\}$ forms a neighborhood base at the identity $e \in G$: a subgroup of $G$ is open if and only if it contains $\operatorname{Gal}(K/F)$ for some finite Galois $F/k$. Under this topology, $G$ becomes a topological group, and the subgroups $\operatorname{Gal}(K/F)$ for $F/k$ finite Galois are exactly the open subgroups of finite index. The Krull topology makes $G$ into a profinite group: it is compact, Hausdorff, and totally disconnected, and it is the inverse limit \begin{align*} \operatorname{Gal}(K/k) \cong \varprojlim_{F} \operatorname{Gal}(F/k), \end{align*} where the inverse limit is taken over all finite Galois subextensions $F/k$, ordered by inclusion. [quotetheorem:2387] The restriction to closed subgroups is necessary: for an infinite profinite group, there exist non-closed subgroups with no corresponding intermediate field. This is not a pathology but a feature — it reflects the richness of infinite Galois groups and is central to modern number theory, where the absolute Galois group $\operatorname{Gal}(\overline{\mathbb{Q}}/\mathbb{Q})$ is one of the central objects of study. [remark: The Absolute Galois Group of $\mathbb{Q}$] The absolute Galois group $G_{\mathbb{Q}} = \operatorname{Gal}(\overline{\mathbb{Q}}/\mathbb{Q})$ is extraordinarily complex: it is not even known whether $G_{\mathbb{Q}}$ is determined up to isomorphism by any explicit presentation. By the Infinite Galois Correspondence, understanding $G_{\mathbb{Q}}$ is equivalent to understanding all algebraic extensions of $\mathbb{Q}$. The inverse Galois problem — which asks which finite groups occur as Galois groups of extensions of $\mathbb{Q}$ — remains open for many classes of groups. [/remark] ## References Emil Artin, *Galois Theory* (1942, revised 1944). The classic lecture notes that established the modern framework, organized around the automorphism group rather than the original permutation-theoretic approach. Serge Lang, *Algebra* (3rd ed., 2002). Chapters V and VI give a complete development of Galois theory in the modern style, including the infinite case and profinite groups. Ian Stewart, *Galois Theory* (4th ed., 2015). A thorough and accessible treatment emphasizing the solvability problem and computational aspects. David Cox, *Galois Theory* (2nd ed., 2012). Detailed treatment of explicit computations, discriminants, and resolvent polynomials. Jean-Pierre Serre, *Local Fields* (1979). Chapter II contains the Krull topology and profinite Galois groups in their arithmetic context.