When someone first tells you that a^{1/2} means "the square root of a", the natural reaction is to nod and memorise. A fraction in the exponent? Sure. Why that fraction should correspond to a root — why the \tfrac{1}{2} produces a \sqrt{\phantom{a}} rather than, say, a halving — is usually left unexplained. It sits there as a convention, something you learn to compute with but never really see.
There is a picture, though, and once you see it the convention stops feeling like a convention. It becomes the only sensible thing the exponent could mean.
Start with what you already trust. The expression a^2 is the area of a square whose side is a. If the side is 3, the area is 9. The exponent 2 is the number of dimensions the side gets repeated across — height times width — and the result is an area. Similarly a^3 is the volume of a cube whose side is a: a three-dimensional box, side times side times side. The exponent is counting how many copies of the side you multiply together.
Now go the other way. Suppose someone hands you a square whose area is a, and asks you for the side. You do not know the side directly; you only know what it produces when squared. The side must be the number that, multiplied by itself, gives a. That number has a name: \sqrt{a}. And by the same logic, if someone hands you a cube of volume a and asks for the side, the answer is the cube root, \sqrt[3]{a}.
That is where the fractional exponent comes in. The notation a^{1/2} is not a new operation invented by mathematicians to confuse you. It is the geometric inverse of a^2 — the side of a square of area a. The notation a^{1/3} is the side of a cube of volume a. The fractional exponents let you read the area-and-volume picture backwards, and the fractions \tfrac{1}{2} and \tfrac{1}{3} are forced on you the moment you insist that the laws of exponents should keep holding.
The widget
Drag the slider to choose a value of a. On the canvas you will see either a square of area a or a cube of volume a, depending on which exponent you pick. The side of that shape is labelled, and its numeric value is shown below — that is a^{1/2} or a^{1/3}. Try to watch two things at once: how the side grows much more slowly than a, and how the side of a perfect square (a = 4, 9, 16) comes out to a whole number, while the side of a = 2 comes out irrational.
The square is not a metaphor. Its area, measured in whatever unit you like, really is a. The side, measured in the same unit's root, really is a^{1/2}. That is the whole claim of the fractional-exponent notation — the claim is geometric, and the widget lets you watch it.
Try these
Play with the slider and the exponent menu until each of these feels obvious.
- a = 4, exponent \tfrac{1}{2}. Side = 2. A square of area 4 has side 2, because 2 \times 2 = 4. The numeric output confirms 4^{1/2} = 2.
- a = 9, exponent \tfrac{1}{2}. Side = 3. Another perfect square: 3 \times 3 = 9.
- a = 8, exponent \tfrac{1}{3}. Side = 2. Switch to the cube view. A cube of volume 8 has side 2, because 2 \times 2 \times 2 = 8. So 8^{1/3} = 2.
- a = 2, exponent \tfrac{1}{2}. Side \approx 1.414. The square of area 2 exists — you can draw it — but its side is an irrational number, \sqrt{2}. The widget shows a perfectly reasonable square; the number underneath just never terminates.
- a = 4, exponent \tfrac{3}{2}. Value = 8. This is \left(\sqrt{4}\right)^3 = 2^3 = 8. The widget draws the root as a small square, then shows that root cubed.
Every one of these is the same picture read at a different a and a different exponent. The geometry is constant; only the numbers change.
Why "fractional" has to equal "root"
The fractional exponent is not a guess. It is forced by the product rule, a^m \cdot a^n = a^{m+n}. If that rule is to keep working when m and n are fractions, then a^{1/2} \cdot a^{1/2} must equal a^{1/2 + 1/2} = a^1 = a.
Read that carefully. You have a number, call it a^{1/2}, which — when multiplied by itself — gives a. That is the definition of \sqrt{a}: the number whose square is a. So the product rule, applied to fractional exponents, forces a^{1/2} = \sqrt{a}. No new rule, no new postulate — just consistency with what the integer-exponent laws already say.
The same argument gives a^{1/3}. You want a^{1/3} \cdot a^{1/3} \cdot a^{1/3} = a^{1/3 + 1/3 + 1/3} = a^1 = a. That makes a^{1/3} the number whose cube is a — the cube root. And for any positive integer q, you get a^{1/q} \cdot a^{1/q} \cdots a^{1/q} (q copies) = a, so a^{1/q} is the q-th root of a.
The general case a^{p/q}
Once you accept a^{1/q} = \sqrt[q]{a}, a general fractional exponent p/q is just a \tfrac{1}{q} followed by a p. Use the power-of-a-power law:
Or, equivalently, do the operations the other way around:
Both orders give the same answer, for a > 0. So a^{3/2} means "take the square root of a, then cube it" — which is what the widget shows when you pick the 3/2 option. For a = 4, that is \sqrt{4}^3 = 2^3 = 8. For a = 9, it is \sqrt{9}^3 = 3^3 = 27. You can verify either of these on the widget's readout.
Geometrically, a^{p/q} is the p-th power of the side of a q-dimensional hypercube whose content is a. That is a mouthful, but the 2D and 3D cases are where you will spend all your time — and in those cases, you can literally draw the picture.
Irrational exponents, briefly
What about a^{1.41421\ldots} = a^{\sqrt{2}}? There is no "\sqrt{2}-th root" picture to draw — \sqrt{2} is not a fraction, so there is no integer dimension to take a hypercube in. The geometric story breaks down.
Algebraically, though, everything is fine. You approximate \sqrt{2} by a sequence of rationals — 1.4, 1.41, 1.414, 1.4142, \ldots — each of which has a well-defined fractional-exponent meaning. The values a^{1.4}, a^{1.41}, a^{1.414}, \ldots form a sequence that converges to a single limit, and you define a^{\sqrt{2}} to be that limit. The product rule and power rule still hold, by continuity. The picture is gone; the laws survive.
Domain caveat
One warning: everything above assumes a > 0. For negative a, fractional exponents misbehave. The expression (-4)^{1/2} is asking for a real number whose square is -4, and there is no such real number — 2 \times 2 = 4, not -4, and (-2) \times (-2) = 4 too. So (-4)^{1/2} is not defined in the real numbers; you need complex numbers for it to mean anything.
Cube roots of negatives are fine — (-8)^{1/3} = -2, because (-2)^3 = -8. But mixed cases like (-4)^{3/2} run into trouble because the square-root step fails. Safe rule: restrict a to positive numbers when you work with fractional exponents, unless you are explicitly in a context that handles complex numbers. The widget stays in the a \geq 1 range for exactly this reason.
Closing
Fractional exponents are not a new rule bolted onto the laws of exponents. They are the geometric inverse of integer exponents — the sides of squares and cubes and their higher-dimensional cousins — and the notation a^{1/q} is forced on you the moment you insist the product rule should keep working for fractions. Once you see the square, the cube, and the side labelled with its fractional exponent, the convention stops feeling like a convention. It becomes the only label the picture could carry.