Rationalizing and Radical Conjugates

The move you already made

Back in the complex numbers you met the conjugate $\overline{z} = a - bi$ of $z = a + bi$, and the one identity that made it useful:

The cross terms $-abi$ and $+abi$ cancelled, and the only surviving appearance of $i$ was through $i^2 = -1$, which is real. So $z\overline{z}$ is a real number, and since $a$ and $b$ are real, $a^2 + b^2$ is zero only when $a = b = 0$, that is, only when $z = 0$ itself.

That is exactly what let you divide. To evaluate $\frac{1}{3 + 2i}$ you multiplied by $\frac{3-2i}{3-2i}$, which is $1$, so the value did not change, and the denominator turned real:

Look at what that computation actually proved, because it is more than an answer. It proved that the reciprocal of a complex number is again a complex number in standard form $a + bi$. The conjugate is what carried out the proof: it converted a division by an awkward object into a division by an ordinary real number, $13$.

The same move over the rational numbers

Now fix a positive integer $d$ that is not a perfect square, so that $\sqrt{d}$ is irrational, and look at the numbers you can build from it:

You have met this number already. In Complex and Irrational Roots you gave the surd partner of $\alpha$ its own symbol,

precisely because the bar was taken by complex conjugation. There the swap $\sqrt{d} \mapsto -\sqrt{d}$ paired up the roots of a rational polynomial. Here it is going to clear a denominator. Same map, new job, and this lesson calls it the radical conjugate.

One practical note before the proof, because the sign can trip you. The swap flips the surd term and leaves the rational term alone, so applied to $\sqrt{3} - 1$, which is really $-1 + \sqrt{3}$, it returns $\alpha^{*} = -1 - \sqrt{3}$. In practice nobody multiplies by that. They flip the sign between the two terms and multiply by $\sqrt{3} + 1$, which is exactly $-\alpha^{*}$. Both choices rationalize, and both give the same final answer, because the extra factor of $-1$ shows up in the numerator and the denominator at once and cancels. The second choice simply spares you a negative denominator to tidy afterwards, so it is the one used throughout this lesson whenever a denominator is written surd-first.

Both halves of that proof are load-bearing. The first half is what clears the radical; the second half is what guarantees you are not dividing by zero while you do it. Put them together and the reciprocal falls out:

Read the right-hand side slowly, because it says far more than “the radical is gone.” It says the reciprocal of a number of the form $a + b\sqrt{d}$ is another number of exactly that form. The family is closed under division. Rationalizing a denominator is not tidying up after yourself; it is the proof of that closure, and the proof is constructive: it hands you the answer.

The same cancellation handles a denominator built from two square roots, like $\sqrt{5} - \sqrt{2}$, which is not of the shape $a + b\sqrt{d}$ with $a$ rational. Multiply it by $\sqrt{5} + \sqrt{2}$ and the cross terms cancel exactly as before, leaving $(\sqrt{5})^2 - (\sqrt{2})^2 = 5 - 2 = 3$. In general, for positive rationals $m \ne n$, the pair $\sqrt{m} \pm \sqrt{n}$ multiplies to $m - n$, which is rational and nonzero. (The condition $m \ne n$ matters, because at $m = n$ this particular pairing collapses: the $\sqrt{m} - \sqrt{n}$ partner of $\sqrt{2} + \sqrt{2}$ is $\sqrt{2} - \sqrt{2} = 0$, and multiplying by that form of “1” is really multiplying by $\frac{0}{0}$. Nothing is wrong with the number $2\sqrt{2}$ itself; it is a single radical, so you clear it the single-radical way, by multiplying by $\sqrt{2}$.)

One principle, three jobs

Set the two computations next to each other and the shared skeleton is impossible to miss.

Why does a mere sign flip do so much work in both rows? Because in both rows the new object satisfies a relation in which it appears squared: $i^2 = -1$ and $(\sqrt{d})^2 = d$. Squaring is blind to sign, so flipping the sign of the new object leaves that defining relation untouched. In the product, every term carrying an odd power of the new object shows up twice with opposite signs and cancels. In the same product, every term carrying an even power was already back in the base system. Nothing irrational, and nothing imaginary, can survive.

There is a third job you have already seen this same sign flip do. When you studied the roots of polynomials, you proved that a real polynomial carries its nonreal roots in conjugate pairs. You also proved that a rational polynomial carries its surd roots in pairs the same way: if $a + b\sqrt{d}$ is a root, so is $a - b\sqrt{d}$. The engine of that proof was that conjugation slides through sums and products, so applying it to $P(\alpha) = 0$ term by term leaves the coefficients alone and turns $\alpha$ into $\alpha^{*}$. That is the same map as the one you are using here, doing a different job. Pairing roots and clearing denominators look like unrelated chores; they are one symmetry, seen from two sides.

Where the sign flip stops working

A student who memorised “flip the sign in the middle” is about to be stranded. A student who learned “find the multiplier that completes a power and lands the product back in the rational numbers” is not. Here is the cliff.

A single cube root. Take $\frac{1}{\sqrt[3]{2}}$. There is no sign to flip, because there is only one term. And look again at what you actually did for $\frac{1}{\sqrt{2}}$: multiplying $\sqrt{2}$ by $\sqrt{2}$ works because it completes the square, not because it flips anything. So complete the cube instead. Rational exponents make the bookkeeping obvious:

The general rule follows from the same exponent arithmetic. For $a > 0$, $a^{1/n}\cdot a^{(n-1)/n} = a^{1} = a$, so

You need whatever power is missing to fill the root up to a whole one, which is the single idea behind every “multiply by $\sqrt{2}$” you have ever done.

A cube-root binomial. Now take $\frac{1}{1 + \sqrt[3]{2}}$ and try the memorised sign flip:

Still irrational. You have not removed a radical; you have swapped one for another. The reason is worth stating exactly, because it tells you what to do instead. A square root has only one irrational power below the point where it turns rational: $\sqrt{d}$ is irrational and $(\sqrt{d})^2 = d$ is rational. So a square root leaves exactly one bad term to kill, and one sign flip kills it. A cube root has two: both $\sqrt[3]{2}$ and $(\sqrt[3]{2})^2 = \sqrt[3]{4}$ are irrational, and only $(\sqrt[3]{2})^3 = 2$ is rational. A single sign flip cannot cancel two independent bad terms. You need a multiplier with three terms, engineered so that everything except the cubes cancels.

That multiplier is the one hiding inside the sum-of-cubes identity. Expand it and watch the middle collapse:

Every intermediate term appears twice with opposite signs. Only the cubes survive, and cubes are exactly what turn a cube root rational. Put $x = 1$ and $y = \sqrt[3]{2}$:

The minus case runs on the twin identity $(x - y)(x^2 + xy + y^2) = x^3 - y^3$, with every sign in the multiplier now positive. In full generality, for rational $a$ and $b$ and any $d$,

and the right-hand side is rational. The second half of the square-root proof carries over as well, so you are still safe from dividing by zero. If $a$ and $b$ are rational and $\sqrt[3]{d}$ is irrational, then $a^3 + b^3 d = 0$ forces $a = b = 0$: were $b$ nonzero, we could write $d = \left(-\frac{a}{b}\right)^3$. And because every real number has exactly one real cube root, that would make $\sqrt[3]{d} = -\frac{a}{b}$ rational, a contradiction. So $b = 0$, and then $a^3 = 0$ gives $a = 0$. A nonzero denominator produces a nonzero rational to divide by, exactly as before.

The principle did all the work: ask which multiplier makes every radical term cancel, and the algebra answers. The sign flip was never the point; it was just what the answer happened to look like when the root was a square root.

Rationalizing the numerator

The principle says “clear the awkward part out of the way,” and it never said the awkward part has to be downstairs. Sometimes the numerator is the problem, and then you aim the same move at it.

How big is $\sqrt{101} - \sqrt{100}$? Subtracting two nearly equal numbers is a terrible way to find out by hand. The leading digits destroy each other, so you would need many digits of $\sqrt{101}$ before even the first digit of the answer settled down. Rationalize the numerator instead, by multiplying top and bottom by $\sqrt{101} + 10$:

The delicate cancellation is gone. In its place is a division by a number you barely need to know, and a crude bound on $\sqrt{101}$ is now enough to pin the answer tightly. The same rewrite in general,

also explains something you may have noticed on a number line. Consecutive square roots crowd closer and closer together as $n$ grows, because the denominator on the right grows without bound while the numerator stays at $1$.

Notice what this does to the school rule. Here the useful form is the one with the radical in the denominator, and the form to be avoided is the one with the radical upstairs. “Rationalize the denominator” is a presentation convention, not a law of nature.

Why bother, when a calculator exists

This deserves an honest answer rather than a slogan, because the usual justification stopped being true around 1975.

The historical reason was hand computation. Suppose a table tells you $\sqrt{2} = 1.41421356$ and you want $\frac{1}{\sqrt{2}}$. Taken literally, that is a long division of $1$ by an eight-digit decimal, done by hand, and it is miserable. Rationalize it first and the same table entry gives the answer by halving:

One is a page of long division, the other is a line of halving. Multiply that saving across every computation in a book of tables and you see why nineteenth-century textbooks drilled the rule until it felt like a moral principle. The calculator quietly removed that reason, and the rule stayed.

The reason that survives is canonical form. Rationalizing puts every number of this kind into one standard shape, $p + q\sqrt{d}$ with $p$ and $q$ rational, and that shape is unique.

That theorem is what makes the convention worth keeping. Two expressions can look nothing alike and be the same number, and until you reduce them to canonical form, “they look different” is evidence of nothing. Consider these three:

Rationalize the first: $\frac{4}{3 + \sqrt{5}} = \frac{4(3 - \sqrt{5})}{9 - 5} = 3 - \sqrt{5}$, which is the second exactly. Rationalize the third: $\frac{2}{\sqrt{5} + 1} = \frac{2(\sqrt{5} - 1)}{5 - 1} = \frac{\sqrt{5} - 1}{2} = -\frac{1}{2} + \frac{1}{2}\sqrt{5}$. Now compare canonical forms. The first two are both $3 + (-1)\sqrt{5}$, so they are equal. The third has $p = -\frac{1}{2}$ and $q = \frac{1}{2}$, which differ from $p = 3$ and $q = -1$, so by the theorem it is a different number. And no amount of algebra will ever turn one into the other. Canonical form did not merely tidy the expressions. It settled a question.

Common mistakes