To find the domain of a function, start with all real numbers and remove every x-value that would force a division by zero, an even root of a negative number, or the log of a non-positive number. That "start broad, then subtract" approach handles nearly every function you'll meet in algebra, precalculus, and calculus. The final answer is an interval or a union of intervals, not a single number. Three restriction families cover routine problems; two more, inverse trig and piecewise assembly, catch the rest. Work them in a fixed order and you won't miss one.
Why the Domain Is the Set of Inputs That Survive Every Operation
The domain is the complete set of input values for which the function produces a real output. Each operation inside the formula carries its own rule about which inputs it accepts. Addition, subtraction, and multiplication accept every real number, so they impose no restriction at all.
Division is the first real constraint. A denominator cannot equal zero, so every zero of the bottom expression leaves the domain. This is why f(x) = 1/(x β 3) excludes exactly x = 3 and nothing else, while f(x) = (xΒ² β 4)/(x β 2) excludes only x = 2 even though the numerator also vanishes there.
Even roots come second. A square root, fourth root, or any even-index radical needs a non-negative argument, so you solve an inequality rather than a single equation. That difference matters: a zero denominator removes isolated points, but a negative radical argument can wipe out an entire ray of the number line.
Odd roots, cubes and fifth roots and so on, accept negative inputs without complaint. Students who reflexively restrict every radical lose points on cube-root problems. The rule depends on the index being even, and nothing else.
Check the Five Restriction Types Before You Write a Single Interval
Run through this checklist in order. Each line tells you what to solve and what kind of answer you get.
| Restriction type | Trigger in the formula | What you solve | Effect on the domain |
|---|---|---|---|
| Division | Variable in a denominator | Denominator = 0 | Removes isolated points |
| Even root | β, β΄β, or any even index | Radicand β₯ 0 | Keeps an interval or ray |
| Logarithm | log, ln of a variable expression | Argument > 0 | Keeps an interval or ray |
| Inverse trig | arcsin or arccos of an expression | β1 β€ argument β€ 1 | Keeps a bounded interval |
| Context or assembly | Real-world units, or a piecewise definition | Stated conditions | Restricts to stated values |
Most textbook problems trigger one line. Harder ones stack two, and the answer is the overlap. A function like f(x) = β(x β 1)/(x β 4) needs the radical rule and the division rule at once, giving x β₯ 1 with x = 4 removed.
The order matters less than completeness. Work top to bottom, note each restriction separately, then intersect the results at the end. If you combine as you go, you'll forget which constraint came from where.
Apply the Division Rule: Factor the Denominator and Remove Each Zero
Set the denominator equal to zero and solve. That gives you the excluded values. For f(x) = (x + 1)/(xΒ² β 5x + 6), factoring xΒ² β 5x + 6 gives (x β 2)(x β 3), so the domain is all reals except 2 and 3, written (ββ, 2) βͺ (2, 3) βͺ (3, β).
Quadratics that don't factor cleanly still produce excluded values. For f(x) = 1/(xΒ² + 2x β 7), the quadratic formula gives x = β1 Β± 2β2, roughly 1.83 and β3.83. Two ugly numbers, two holes in the domain. Report them in exact form.
Higher-degree denominators can hide repeated roots. If the denominator is (x β 1)Β²(x + 4), the root x = 1 has multiplicity two, but it still removes only the single point x = 1 from the domain. Multiplicity affects graph behavior, not domain membership.
Canceled factors create removable holes, not domain gaps in the simplified function. For f(x) = (xΒ² β 9)/(x β 3), the expression simplifies to x + 3 for every x except 3. The domain of the original function still excludes 3, and that hole is what a graph will show.
Solve Radical and Fractional-Exponent Domains With One Inequality Each
For an even root, write the radicand as an inequality greater than or equal to zero and solve. For f(x) = β(4 β 2x), the inequality 4 β 2x β₯ 0 gives x β€ 2, so the domain is (ββ, 2]. Check the endpoint: x = 2 makes the radicand zero, and β0 = 0 is a real number, so the endpoint stays.
Quadratics under a radical require a sign chart. For g(x) = β(xΒ² β 9), factor to (x β 3)(x + 3) β₯ 0. The product is non-negative outside the interval between the roots, so the domain is (ββ, β3] βͺ [3, β). Test a point in each region to confirm; x = 0 gives β9, which is negative, so the middle region fails.
Fractional exponents follow the same logic with one twist. The expression x^(1/2) means βx, so the base must be non-negative. But x^(1/3) means the cube root, which accepts any real base. Rewrite the exponent as a root before deciding, and match the parity of the denominator.
Nested radicals need the inner one solved first. For h(x) = β(βx β 2), the inner βx needs x β₯ 0, and the outer needs βx β 2 β₯ 0, which means βx β₯ 2 and x β₯ 4. The tighter condition wins, so the domain is [4, β). Stacked inequalities always resolve to the most restrictive one.
Handle Logarithm Domains by Requiring a Strictly Positive Argument
A logarithm accepts only positive arguments. For f(x) = ln(x β 5), the condition x β 5 > 0 gives x > 5, so the domain is (5, β) with an open bracket at 5. The boundary is excluded, unlike the radical case.
When the log argument is a polynomial with real roots, the domain is where that polynomial is positive, not non-negative. For f(x) = log(xΒ² β 4), factor to (x β 2)(x + 2) > 0, giving (ββ, β2) βͺ (2, β). Both endpoints drop out here because log(0) is undefined.
Bases don't change the domain rule as long as the base is positive and not equal to 1. ln, log base 10, and log base 2 all require a positive argument. Base restrictions matter for solving the equation, not for finding the domain.
Composite logs stack constraints. For g(x) = ln(ln(x)), the outer log needs ln(x) > 0, which means x > 1. The inner log already demands x > 0, but the outer condition is stricter, so the domain is (1, β). Always take the intersection of every log requirement.
Find Domains of Trig, Inverse Trig, and Exponential Functions From Their Ranges
Sine and cosine accept every real number, so their domains are all reals, often written β. Tangent is different: tan(x) = sin(x)/cos(x), and cos(x) = 0 at x = Ο/2 + kΟ for every integer k. The tangent domain excludes all those points, and the answer is usually stated as x β Ο/2 + kΟ.
Exponential functions like 2^x or e^x accept every real input because a positive base raised to any real power stays real. That's why the domain of e^x is β while its range is (0, β). Students often swap these two; keep them separate.
Inverse trig functions carry bounded domains by definition. arcsin(x) and arccos(x) require β1 β€ x β€ 1, while arctan(x) accepts every real number. This is the one place where a domain is bounded above and below without any denominator or radical in sight.
Inverse trig of an expression needs a compound inequality. For f(x) = arcsin(2x β 1), solve β1 β€ 2x β 1 β€ 1. Add 1 to all three parts to get 0 β€ 2x β€ 2, then divide by 2 for 0 β€ x β€ 1. The domain is [0, 1], a closed interval on both ends.
Build Piecewise and Real-World Domains From Their Own Conditions
A piecewise function's domain is the union of the intervals assigned to each piece, provided those intervals cover or touch without weird gaps. For f(x) = xΒ² for x < 0 and f(x) = x + 1 for x β₯ 0, every real number fits one branch, so the domain is β.
When piecewise intervals skip values, the domain must reflect the gap. If one branch covers x < 2 and the next starts at x > 3, the interval [2, 3] produces no output, and the domain is written as the union of the two branches only.
Real-world context adds constraints the formula itself doesn't show. Area as a function of side length A(s) = sΒ² technically accepts any real s, but geometry limits the domain to s > 0. Selling shirts for revenue R(n) = 12n restricts n to non-negative integers.
Volume problems often produce domains bounded on both ends. A box made from a 12-inch square sheet by cutting corners of length x has volume V(x) = x(12 β 2x)Β², and the domain is 0 < x < 6 because the cut cannot exceed half the side. Context restrictions are physical, not algebraic; always check what the variables represent.
Work the Numbers: Three Examples From Start to Finish
Example 1: f(x) = (x β 1)/(xΒ² β 4). Set xΒ² β 4 = 0 and factor to (x β 2)(x + 2) = 0. The excluded values are x = 2 and x = β2. No radical, no log, no other restriction. Domain: (ββ, β2) βͺ (β2, 2) βͺ (2, β).
Example 2: f(x) = β(x + 3)/(x β 5). The radical requires x + 3 β₯ 0, so x β₯ β3. The denominator requires x β 5. Combine the two: x β₯ β3 with the point 5 removed. Domain: [β3, 5) βͺ (5, β).
Example 3: f(x) = ln(xΒ² β 1). The log argument needs xΒ² β 1 > 0, which factors to (x β 1)(x + 1) > 0. Both endpoints are excluded. Domain: (ββ, β1) βͺ (1, β).
Every one of these follows the same path: identify the restriction, solve it, then combine. The variety lies in the algebra, not the strategy.
The Domain Mistakes That Cost the Most Points
Over-restricting radicals is the most common error. A cube root like β(x + 2) accepts x values that make the radicand negative, because the cube root of a negative number is a real negative number. Students who apply the even-root rule to every radical lose valid inputs every time.
Dropping the endpoint on logs is the second. Writing x β₯ 3 instead of x > 3 for ln(x β 3) includes the value that makes the argument zero, and log(0) doesn't exist. Closed brackets on log boundaries are always wrong.
Forgetting to factor fully is the third. A denominator of xΒ² β 4 that's left unfactored hides both roots, and the domain looks unrestricted when it isn't. Factor completely, including higher-degree cases, before deciding.
Missing composite restrictions is the fourth. A function with both a denominator and a log needs the intersection, not just one rule. Take every restriction and overlap them. If the results conflict, the domain can even be empty, which is a valid answer.
How the Same Function Shifts Domain When the Input Variable Changes
Replacing x with a linear expression like 2x β 1 transforms the domain by inverse operations. If f(x) = βx has domain [0, β), then g(x) = β(2x β 1) needs 2x β 1 β₯ 0, which gives x β₯ 1/2. The domain shifts and scales along with the input substitution.
The transformation rule: for g(x) = f(ax + b), take each restriction on the input t = ax + b and solve the inequality for x. If f(t) requires t β₯ c, then ax + b β₯ c gives x β₯ (c β b)/a when a > 0, or x β€ (c β b)/a when a < 0. The direction flips.
Graph transformations confirm the same result. Compressing the graph of βx horizontally by a factor of 1/2 and shifting right by 1/2 moves the left endpoint of the domain from x = 0 to x = 1/2. The algebra and the geometry agree.
Nonlinear substitutions break this simple rule. For f(x) = β(xΒ² β 1), there's no single a or b; the input is quadratic. You have to solve the inequality from scratch rather than shift an existing answer. Recognize the difference before reaching for a formula.
Verify Your Answer by Testing Points Just Inside and Just Outside
Pick values on either side of every claimed boundary and plug them in. If your answer excludes x = 2, test x = 1.9 and x = 2.1 in the original formula. A negative under an even root or a zero in a denominator confirms the exclusion was needed.
A worked check for f(x) = β(x β 4)/(x β 1): the domain should be [4, β) minus the point 1, but 1 isn't in [4, β) anyway, so the point removal is redundant and the domain is just [4, β). Testing x = 1, x = 3, x = 4, and x = 5 shows which inputs survive.
Graphing supports the check. Every excluded value surfaces as a hole, asymptote, or the edge of the function's drawn curve. If your interval claims the graph extends past x = 4 but the plotted curve stops there, one of the two is wrong.
Simplify only as a final check, and don't let simplification erase exclusions. The formula (xΒ² β 1)/(x β 1) simplifies to x + 1 with a hole at x = 1, and the domain of the original still excludes 1. Simplified form loses the hole; the domain must remember it.
FAQ
Can a domain be written as a single number?
Only for a function defined at exactly one input, which is rare but possible. For f(x) = β(1 β xΒ²) + β(xΒ² β 1), both radicals force xΒ² = 1 simultaneously, so the domain is exactly {β1, 1}, a set with two numbers. Everywhere else, a domain is an interval, a union of intervals, or occasionally an isolated set of points from integer constraints.
What's the domain of f(x) = xβ°?
The domain of xβ° is all real numbers except 0, because 0β° is undefined in standard convention. For every other x, xβ° = 1, so the function is constant apart from the hole at zero. Writing β without the exclusion is a common slip.
Why does the domain of 1/x differ from the domain of 1/xΒ²?
Both exclude x = 0, so their domains are actually identical: (ββ, 0) βͺ (0, β). The graphs differ near zero, since 1/x flips sign across the origin and 1/xΒ² doesn't, but the domain of each is the same set. Students who expect a different domain for 1/xΒ² are confusing graph behavior with input restriction.
Does every function with a square root have a domain that ends at one side?
Yes, for a single square root: an even root produces a domain that is either a ray or a closed interval, never a union of two separate pieces. Matters change when you layer a square root with another restriction. A rational expression inside the root, like β((x β 1)/(x β 3)), splits into two intervals because the quotient is positive on two disjoint ranges.
How do log and square-root domains differ at the boundary?
A square root accepts the endpoint where its radicand equals zero, so [0, β) includes 0. A logarithm rejects the endpoint where its argument equals zero, so (0, β) excludes 0. That single difference between non-negative and positive turns a closed bracket into an open one every time a log appears.
Can I use interval notation with the infinity symbol as a closed endpoint?
No. Infinity is not a real number, so every interval involving β or ββ uses an open parenthesis next to it. Writing [5, β] is nonstandard and incorrect. Use [5, β) instead, and reserve closed brackets for actual finite endpoints.
What's the domain of arcsin(x) + arcsin(2x)?
Both inputs must fall within [β1, 1]. The first requires β1 β€ x β€ 1, and the second requires β1 β€ 2x β€ 1, which is β1/2 β€ x β€ 1/2. The intersection is [β1/2, 1/2], the stricter of the two. That's narrower than either individual inverse trig domain, so combine before writing the answer.
How does the domain of f(g(x)) differ from the domains of f and g alone?
The composite f(g(x)) needs two conditions: x must be in g's domain, and g(x) must be in f's domain. For f(x) = βx and g(x) = x β 5, the composite β(x β 5) requires x β₯ 5, which is the overlap. The outer function's restriction narrows the inner function's domain after composition.