Algebraic and Transcendental Elements

The algebraic/transcendental dichotomy is the entry point to the entire theory of field extensions: every extension $K/F$ decomposes, element by element, into algebraic and transcendental parts, and the finiteness criterion proved here is the tool used to show finite extensions are algebraic. In Galois theory, algebraic elements with their minimal polynomials are the raw material for constructing splitting fields, computing Galois groups, and proving classical impossibility results (doubling the cube, trisecting the angle, and squaring the circle all reduce to algebraicity/degree arguments). In transcendence theory and Diophantine geometry, showing specific constants ($e$, $\pi$, values of the Gamma function, periods of elliptic curves) are transcendental over $\mathbb{Q}$ remains a central and often extremely hard research question. In algebraic geometry, transcendental elements generalize to transcendence bases and transcendence degree, measuring the "dimension" of a field extension the way a basis measures the dimension of a vector space. In computer algebra and cryptography, algebraic elements of finite fields (roots of irreducible polynomials over $\mathbb{F}_p$) are used to construct the finite field extensions $\mathbb{F}_{p^n}$ that underlie elliptic-curve cryptography and error-correcting codes; the finite-dimensionality criterion proved above is exactly what guarantees these constructions produce finite fields at all.

:::quiz[If $\alpha$ is algebraic over $F$ via a minimal-degree annihilating polynomial of degree $n$, what does the proof show about $\dim_F F[\alpha]$?]

• It is always exactly $n^2$
• It is infinite
• It is at most $n$, spanned by $1,\alpha,\dots,\alpha^{n-1}$ ✓
• It cannot be determined from $n$ alone

Ask the tutor: How does the minimal polynomial of an algebraic element get constructed, and why is it automatically irreducible over $F$?

Think of $F[x]$, the ring of polynomials with coefficients in $F$, as a toolbox of possible "tests" you can apply to an element $\alpha$ of a bigger field $K$. Plugging $\alpha$ into a polynomial and evaluating gives another element of $K$. The element $\alpha$ is algebraic exactly when some nontrivial test in this toolbox returns zero — when $\alpha$ satisfies some algebraic law with coefficients drawn from $F$. If $\alpha$ is algebraic, the powers $1, \alpha, \alpha^2, \alpha^3, \ldots$ cannot all be $F$-linearly independent, because the relation $f(\alpha) = 0$ is precisely a nontrivial linear dependence among these powers. Eventually the powers of $\alpha$ "fold back" onto a finite-dimensional space spanned by $1, \alpha, \dots, \alpha^{n-1}$. Geometrically, you can picture the tower of powers of $\alpha$ collapsing after finitely many steps, like a recursive sequence obeying a finite recurrence. If $\alpha$ is transcendental, no such collapse ever happens: the powers $1, \alpha, \alpha^2, \dots$ remain linearly independent forever, and $\alpha$ behaves exactly like a free indeterminate $x$. The field $F(\alpha)$ generated by $\alpha$ is then indistinguishable, as an abstract field, from the field of rational functions $F(x)$.

Example 1: an algebraic element built from nested radicals. Let $F = \mathbb{Q}$, $K = \mathbb{R}$, and $\alpha = \sqrt{2} + \sqrt[3]{5}$. We show $\alpha$ is algebraic over $\mathbb{Q}$ by exhibiting an explicit annihilating polynomial, using the finiteness criterion as a guide rather than guessing blindly. Since $\sqrt 2$ satisfies $t^2 = 2$ and $\sqrt[3]{5}$ satisfies $t^3=5$, set $\beta = \alpha - \sqrt2 = \sqrt[3]{5}$, so $\beta^3 = 5$. Substitute $\beta = \alpha - \sqrt2$:

Expand using the binomial theorem, treating $\sqrt2$ as a fixed quantity with $(\sqrt2)^2 = 2$:

Collect the terms with a factor of $\sqrt2$:

Square both sides (both sides are real numbers, so this is legitimate and reversible given we track signs are irrelevant to squaring):

Expand the left side:

and the right side:

Setting left minus right to zero:

So $\alpha = \sqrt2 + \sqrt[3]{5}$ is a root of the nonzero polynomial

Hence $\alpha$ is algebraic over $\mathbb{Q}$, and by the theorem, $\mathbb{Q}[\alpha] = \mathbb{Q}(\alpha)$ is a finite-dimensional $\mathbb{Q}$-vector space (of dimension at most $6$; one can show it is exactly $6$, but that requires irreducibility of $f$, a question for the theory of the Minimal Polynomial). This also illustrates that finding an annihilating polynomial by direct elimination, rather than by cleverness, is always possible for such nested radical expressions.

Example 2: a transcendental element in a rational function field. Let $F$ be any field, let $K = F(t)$ be the field of rational functions in an indeterminate $t$ over $F$, and consider inside $K$ the element

We ask: is $\alpha$ algebraic or transcendental over $F$? (Note $F(t)$ itself is generated by the transcendental element $t$; the question here is subtler, since $\alpha$ is not $t$ itself but a rational function of it.) Claim: $\alpha$ is transcendental over $F$, and moreover $F(\alpha) \subsetneq F(t)$ is a proper subfield, yet $F(t)$ is a finite-dimensional $F(\alpha)$-vector space of dimension $2$. Proof of transcendence. Suppose, for contradiction, that $\alpha$ is algebraic over $F$, i.e. there is a nonzero polynomial $g(x) \in F[x]$ with $g(\alpha) = 0$. Write $g(x) = \sum_{i=0}^n c_i x^i$ with $c_i \in F$, not all zero. Then in $K = F(t)$,

Multiply through by $(t+1)^n$ (nonzero in $F[t]$, and in fact $F[t]$ is an integral domain so no zero divisors are introduced):

This says a specific polynomial identity holds in $F[t]$, namely that the polynomial $P(t) := \sum_i c_i t^{2i}(t+1)^{n-i}$ is the zero polynomial. But consider the coefficient of $t^{2n}$ in $P(t)$: only the $i=n$ term, $c_n t^{2n}(t+1)^0 = c_n t^{2n}$, can contribute a term of degree $2n$ (every other term $c_i t^{2i}(t+1)^{n-i}$ with $i<n$ has degree $2i + (n-i) = n+i < 2n$ since $i < n$). So the coefficient of $t^{2n}$ in $P$ is exactly $c_n$. Since $P \equiv 0$, we need $c_n = 0$. But then repeat with $n$ replaced by the largest index $i$ with $c_i \neq 0$ (such an index exists since not all $c_i$ vanish): by the same degree-counting argument applied to the truncated sum, the coefficient of $t^{2i}$ in the (now shorter) identity is $c_i$, forcing $c_i = 0$ too — contradiction. Hence no nonzero $g$ exists, and $\alpha = t^2/(t+1)$ is transcendental over $F$, exactly as $t$ itself is. Finite degree over $F(\alpha)$. Although $\alpha$ is transcendental over the small field $F$, the element $t$ satisfies a genuine polynomial equation over the larger field $F(\alpha)$: rearranging $\alpha = t^2/(t+1)$ gives

a quadratic equation for $t$ with coefficients $1, -\alpha, -\alpha$ lying in $F(\alpha)$. So $t$ is algebraic over $F(\alpha)$ (of degree at most 2), even though $\alpha$ is transcendental over the smaller field $F$. This shows vividly that "algebraic" is relative to the base field: the same tower $F \subset F(\alpha) \subset F(t)$ has a transcendental step followed by an algebraic (in fact finite, degree $2$) step. This phenomenon — a transcendental generator of a subfield over which the original variable becomes algebraic — is the seed of Simple Extensions and F(alpha) and of transcendence-degree arguments in more advanced field theory.

:::counterexample[Why the extension $K/F$ (and not just the ring $F[x]$) is essential] The definitions of algebraic and transcendental only make sense relative to an ambient field $K$ containing $F$ — dropping this and asking "is $\alpha$ algebraic over $F$" with no ambient field specified is meaningless, and conflating $F[x]$ itself with $F[\alpha]$ is the trap this example exposes. Consider $F = \mathbb{Q}$ and the indeterminate $x$ inside the polynomial ring $\mathbb{Q}[x]$ versus a root of a polynomial inside a genuine field extension. The symbol $x$ in $\mathbb{Q}[x]$ is not an element of any field extension of $\mathbb{Q}$ a priori — $\mathbb{Q}[x]$ is a ring, not a field — so the question "is $x$ algebraic over $\mathbb{Q}$" is not directly one the definition addresses until we embed $x$ into some field $K \supseteq \mathbb{Q}$. The correct ambient field to use is $K = \mathbb{Q}(x)$, the field of rational functions, into which $\mathbb{Q}[x]$ embeds naturally. Inside this specific $K$, $x$ is transcendental over $\mathbb{Q}$: if $f(x) = \sum a_i x^i$ were a nonzero polynomial with $f(x) = 0$ as an element of $\mathbb{Q}(x)$, this would force every coefficient $a_i$ to vanish, since two polynomials define the same rational function if and only if they are equal as polynomials (an element of $F[x]$, viewed inside $F(x)$, is the zero element of $F(x)$ if and only if it is the zero polynomial — this is how the ring embeds into its fraction field). So no nonzero relation exists. Now here is the point of the counterexample: if instead we sloppily take $K$ to be, say, the field $\mathbb{F}_2(y)/(y^2+y+1)\cdot(\ldots)$ or any field where a different symbol named "$x$" happens to satisfy a relation, we could wrongly "prove" $x$ algebraic by confusing notation between rings. The moral, stated precisely: algebraicity is not an intrinsic property of an abstract symbol; it is a property of a specific element sitting inside a specific field extension $K/F$. The same abstract polynomial ring $F[x]$ can be embedded as a subring of many different fields, and whether the class of $x$ in that field is a root of a polynomial over $F$ depends entirely on which embedding — which ambient $K$ — is used. Without fixing $K/F$ first, the dichotomy algebraic/transcendental has no content.