Why does the horizontal line test matter for finding an inverse?

A horizontal line that crosses the graph more than once means two different x-values produce the same y-value. The inverse would then need to send that y-value to two different x-values, which violates the definition of a function. The horizontal line test is the graphical way to check if a function is one-to-one.

Can I choose any interval where the function is monotonic?

Yes, any interval where the function is strictly increasing or strictly decreasing works. But the choice determines the inverse function you get. For f(x) = x², restricting to [0, ∞) gives f⁻¹(x) = √x, while restricting to (-∞, 0] gives f⁻¹(x) = -√x. Both are valid, but they are different functions with different ranges.

What happens if I use an interval that is not monotonic?

Then the function still fails the horizontal line test on that interval, and the inverse will not be a function. For example, restricting f(x) = x² to [-1, 1] still contains both sides of the vertex, so the horizontal line y = 0.25 hits at x = -0.5 and x = 0.5. The restriction must be to a monotonic interval.

How do I know if my restricted domain is correct?

Check that the function is strictly monotonic on the interval and that the interval covers the full range of output values you need. Then verify by composing the function and its inverse. If f(f⁻¹(x)) = x for every x in the inverse's domain, and f⁻¹(f(x)) = x for every x in the restricted domain, the restriction is correct.

Does domain restriction always produce an inverse function?

Yes, if the function is continuous and monotonic on the restricted interval, it will be one-to-one and have an inverse function on that interval. But the inverse's domain will equal the original's range, which may be a different interval. For f(x) = x² restricted to [0, ∞), the range is [0, ∞), so the inverse's domain is [0, ∞).

Restrict the Domain Before Finding an Inverse Explained

Why Restrict the Domain Before Finding an Inverse

Most students assume that if a function has an inverse at all, that inverse is automatically a function. That is wrong. The inverse of f(x) = x² is not a function unless you restrict the domain before finding an inverse. The horizontal line test is the graphical version of the problem: any horizontal line crossing the graph more than once means the function is not one-to-one, so its inverse would assign two outputs to a single input, which breaks the definition of a function. The only fix is to chop the domain down to an interval where the function is strictly increasing or strictly decreasing.

Consider f(x) = x² on its natural domain, which is all real numbers. The horizontal line y = 4 hits the parabola at x = 2 and x = -2. If you try to write an inverse without restriction, you need an operation that gives both 2 and -2 for input 4. That operation is the square root only if you take the principal (non-negative) root, which discards -2. But discarding -2 means the inverse you wrote does not actually reverse the original function for every input. The only way to produce a true inverse function is to restrict the original domain to an interval where no horizontal line hits twice.

A Worked Example

The standard example is f(x) = x² restricted to the interval [0, ∞). On that domain f is strictly increasing and passes the horizontal line test. The inverse function is f⁻¹(x) = √x, defined only for x ≥ 0. Recompute the pair: for x = 3, f(3) = 9. Then f⁻¹(9) = √9 = 3. Compose them: f(f⁻¹(9)) = f(3) = 9, and f⁻¹(f(3)) = f⁻¹(9) = 3. Both compositions return the original input, confirming the inverse is a function.

Now test the failure case without restriction. On the natural domain (-∞, ∞), what is f⁻¹(4)? It would have to return both 2 and -2 to reverse f(2) = 4 and f(-2) = 4. No single-valued operation can do that. The square root function returns only 2, which fails to reverse f(-2). So f has no inverse function on its natural domain. This is not a quirk of quadratics; every function that fails the horizontal line test has this problem. Rational functions, absolute value functions, and sine and cosine all require domain restriction before their inverses are functions.

Calculation Steps

To find the correct restricted domain for any function, you need to identify an interval where the function is monotonic. For f(x) = x², the vertex at x = 0 splits the parabola into a decreasing half for x < 0 and an increasing half for x > 0. Either half works, but the choice determines the inverse's range. If you restrict to [0, ∞), the inverse is √x with range [0, ∞). If you restrict to (-∞, 0], the inverse is -√x with range (-∞, 0]. Both are valid inverse functions, but they are different functions.

At the point where you verify the inverse by plugging in a value, an inverse function calculator can confirm the composition returns the original input. That check is necessary because even a correctly restricted domain can produce an inverse that passes the vertical line test but fails to reverse the original function if the restriction was chosen incorrectly. The calculator shows each step: it evaluates the composition and checks that the domain of the inverse matches the range of the restricted original.

What Happens With Trigonometric Functions

Trigonometric functions are periodic, so they fail the horizontal line test infinitely many times. The standard restricted domain for sine is [-π/2, π/2], producing the inverse arcsine. Cosine uses [0, π] for arccosine. These restrictions are not arbitrary; they are the intervals where the function is monotonic and covers the full range of output values. Without them, arcsine would have no single function value for sin(π/6) = 0.5, which also equals sin(5π/6).

The Limitation of Non-Unique Restrictions

The restriction is not unique. For a quadratic, any interval that contains the vertex on one side works. For sine, any interval of length π that does not include a turning point works. The choice affects the range of the inverse, which means the inverse function you get depends on the domain you pick. This is why textbooks define the principal values for inverse trigonometric functions: to guarantee everyone uses the same inverse function.

Why the Method Works

The horizontal line issue makes domain choice a necessary prerequisite. A function must be one-to-one to have an inverse function. The horizontal line test is the graphical equivalent of the one-to-one condition. Restricting the domain eliminates the multiple intersections, making the function one-to-one on that interval. This is not an optional step; it is a requirement of the definition of a function. OpenStax Precalculus 2e, Section 3.7 'Inverse Functions' states it directly: a function must be one-to-one to have an inverse function.

Students often confuse domain restriction with the range of the inverse. The domain of the inverse equals the range of the original function. When you restrict the domain of f to [0, ∞), the range of f becomes [0, ∞) as well, so the domain of f⁻¹ is [0, ∞). If you had restricted to (-∞, 0], the range of f would still be [0, ∞), so the domain of f⁻¹ is still [0, ∞) but the inverse is -√x, which maps to (-∞, 0]. Swapping domain and range is the mechanical check: the inverse's domain must match the original's range, and the inverse's range must match the original's domain.

How to Find the Right Interval

Find the turning points of the function. For a quadratic, find the vertex. For a rational function, find the vertical asymptotes and check monotonicity between them. For sine and cosine, identify the intervals of length one period where the function is monotonic. Then test the endpoints to confirm the function covers the full range needed for the inverse. If the range of the original function is bounded, the inverse's domain will be bounded as well.

The Single Thing That Goes Wrong

The most common failure is assuming the inverse of a non-one-to-one function exists without restriction. A student writes f⁻¹(x) = √x for f(x) = x² without restricting the domain, then cannot explain why f⁻¹(f(-3)) = 3 when f(-3) = 9. The inverse function they wrote fails to reverse the original for negative inputs, proving it is not a true inverse. The fix is always the same: restrict the domain first, then find the inverse.