To prove that both distinct real roots of a quadratic lie strictly inside a specified interval, checking the discriminant alone is not enough. You must also establish that the parabola’s axis of symmetry lies within that interval and that both endpoint values are on the outside-of-roots side of the graph.
For an upward-opening quadratic f(x) = ax² + bx + c, a > 0, the exact tests for both roots to lie in (L,U) are: discriminant D > 0, vertex input L < −b/(2a) < U, and f(L) > 0 and f(U) > 0. For a downward-opening quadratic, use f(L) < 0 and f(U) < 0 instead. This guide derives the conditions and solves an exact parameter example in which both roots must lie between zero and five.
Two real roots are not enough information
A quadratic may have two distinct real roots but place both of them outside the interval named in the question.
For example, f(x) = (x − 6)(x − 7) has real roots 6 and 7. Both lie outside the interval (0,5), even though its discriminant is positive.
Therefore D > 0 answers the question “are there two distinct real roots?” but not “are both of them inside the requested interval?”
A complete interval-root problem needs information about root location. The most efficient way is often to combine the discriminant, the vertex and the signs of the quadratic at both interval endpoints instead of explicitly calculating complicated roots.
Define the interval and strictness precisely
Suppose the question requires two distinct real roots r₁ and r₂ such that L < r₁ < r₂ < U, with fixed real bounds L < U.
Every inequality is strict. Neither root may equal L or U, and the roots cannot coincide.
If a question says “both roots belong to [L,U]”, its endpoint inclusion is different. If it says “at least one root in the interval”, that is a weaker requirement.
Read these distinctions before writing a parameter inequality. A correct discriminant calculation can still lead to an incorrect final range if a boundary root is included despite strict wording.
Factorisation explains the endpoint signs
Write f(x) = a(x − r₁)(x − r₂), with a ≠ 0 and r₁ < r₂.
When x lies to the left of both roots, both bracket factors are negative, so their product is positive. When x lies to the right of both, both are positive, so their product is again positive.
Therefore f(x)/a is positive at both outside regions and negative between the two roots.
If both roots lie inside (L,U), then L is to their left and U to their right. Consequently f(L)/a > 0 and f(U)/a > 0. These are necessary endpoint tests, regardless of whether the parabola opens upward or downward.
Why the vertex must lie inside
The axis of symmetry of f(x) = ax² + bx + c is x = h = −b/(2a).
For two distinct roots r₁ and r₂, their midpoint is h = (r₁ + r₂)/2. If both roots belong to (L,U), then their midpoint also belongs to (L,U).
Thus the necessary vertex condition is L < h < U. It is not optional because positive values at both endpoints can occur even when the two roots lie entirely to one side of the interval.
The vertex positions the centre of the root pair. The endpoint signs then determine whether the pair fits inside rather than spilling beyond one of the interval boundaries.
Why the discriminant must be positive
The discriminant is D = b² − 4ac. A genuine real quadratic has two distinct real roots exactly when D > 0.
For an upward-opening parabola, the vertex value is f(h) = −D/(4a). If a > 0 and D > 0, the vertex lies below the x-axis.
For a downward-opening parabola with a < 0 and D > 0, the vertex lies above the x-axis. In either case, the vertex is on the between-roots side of zero.
This condition supplies the required pair of genuine crossings. Without it, a quadratic could have positive endpoint values and a vertex inside yet never touch the x-axis at all.
State the complete general test
For a ≠ 0 and L < U, both distinct real roots of ax² + bx + c lie strictly inside (L,U) exactly when three conditions hold.
First D = b² − 4ac > 0. Second L < h = −b/(2a) < U. Third f(L)/a > 0 and f(U)/a > 0.
For a > 0, the third condition reads f(L) > 0 and f(U) > 0. For a < 0, it reads f(L) < 0 and f(U) < 0.
The sign-normalised form f(endpoint)/a > 0 is useful when the sign of a depends on a parameter. Keep a ≠ 0 explicit so that the quadratic does not degenerate into a linear equation.
Why these conditions are sufficient
Assume a > 0, D > 0 and h lies strictly between L and U. The quadratic has its minimum below zero at x = h.
If f(L) > 0, continuity forces a zero between L and h, because the graph changes from positive at L to negative at h. If f(U) > 0, it forces another zero between h and U.
A quadratic has at most two roots, so these are exactly its two distinct real roots, both strictly inside the interval.
For a < 0, the same reasoning uses negative endpoint values and a positive vertex maximum. This establishes the sufficiency of the test, not just a convenient set of necessary-looking rules.
Set up the parameter example
Let f_k(x) = x² − (k + 2)x + 2k − 3, where k is real. Find k such that both distinct roots lie in the interval (0,5).
The leading coefficient is a = 1, so the parabola opens upward. Therefore both endpoint values must be strictly positive.
We will calculate the discriminant, the vertex input and the two endpoint values separately, then intersect the resulting parameter conditions.
This sequence avoids manipulating the two roots in a formula containing √D and the parameter k. The root-location tests are simpler and provide a clearer explanation of why the final parameter range works.
Check the discriminant of the example
Here a = 1, b = −(k + 2) and c = 2k − 3.
The discriminant is D = (k + 2)² − 4(2k − 3). Expand it to k² + 4k + 4 − 8k + 12 = k² − 4k + 16.
Complete the square: D = (k − 2)² + 12. This is strictly positive for every real k.
Therefore the requirement of two distinct real roots places no further restriction on k in this particular example. The remaining work concerns their location inside the specified interval.
Check that the vertex lies between zero and five
The axis of symmetry is h = −b/(2a) = (k + 2)/2.
Require 0 < (k + 2)/2 < 5. Multiply through by the positive number 2 to get 0 < k + 2 < 10.
Subtract 2 to obtain −2 < k < 8.
This is a broad necessary condition. It ensures that the midpoint of the two real roots lies inside the interval, but it does not yet guarantee that both roots fit within its boundaries. The endpoint sign conditions provide that additional restriction.
Check the left endpoint
Evaluate f_k(0) = 0 − 0 + 2k − 3 = 2k − 3.
Since the parabola opens upward, require f_k(0) > 0 for zero to lie strictly to the left of both roots.
Thus 2k − 3 > 0, giving k > 3/2.
If k = 3/2, then x = 0 becomes a root, violating the requirement that both roots be strictly greater than zero. The inequality must remain strict.
Check the right endpoint
Evaluate f_k(5) = 25 − 5(k + 2) + 2k − 3.
Simplify: 25 − 5k − 10 + 2k − 3 = 12 − 3k.
Require f_k(5) > 0, because x = 5 must lie strictly to the right of both roots for an upward-opening parabola.
Thus 12 − 3k > 0, giving k < 4. At k = 4, x = 5 becomes a root and is therefore not strictly inside the open interval.
Intersect all the parameter restrictions
We found D > 0 for every real k. The vertex condition gives −2 < k < 8.
The left endpoint requires k > 3/2. The right endpoint requires k < 4.
The intersection is 3/2 < k < 4. This is the complete answer.
The endpoint inequalities are stronger than the vertex interval in this example, so the final range automatically satisfies −2 < k < 8. Do not report the broader vertex interval by itself. The answer must satisfy all three logical tests simultaneously.
Verify an interior parameter exactly
Choose k = 2, which lies between 3/2 and 4. Then f_2(x) = x² − 4x + 1.
Its roots are x = [4 ± √(16 − 4)]/2 = [4 ± 2√3]/2 = 2 ± √3.
Because 0 < 2 − √3 < 2 + √3 < 5, both roots lie strictly inside (0,5).
The direct root calculation confirms the result for one sample parameter. The general interval proof remains the discriminant, vertex and endpoint sign analysis, which applies to every k between the two bounds.
Understand the lower boundary k = 3/2
At k = 3/2, the polynomial becomes x² − (7/2)x + 0.
Factorise as x(x − 7/2). Its roots are x = 0 and x = 7/2.
One root lies exactly on the left endpoint zero, so the strict condition 0 < r₁ fails.
This explains why the lower bound is excluded. The failure is not a repeated root or lack of real roots; it is a valid root at an inadmissible boundary position. Boundary interpretation should be tied to the original interval requirement.
Understand the upper boundary k = 4
At k = 4, the polynomial is x² − 6x + 5.
Factorise as (x − 1)(x − 5). The roots are 1 and 5.
The second root equals the right interval endpoint. Therefore the strict requirement r₂ < 5 fails even though the roots are real and distinct.
The upper endpoint is also excluded. These two boundary cases make the strict parameter interval 3/2 < k < 4 geometrically meaningful rather than an unexplained result of algebraic manipulation.
An outside parameter can place one root beyond an endpoint
Take k = 0, which fails the left condition k > 3/2. The polynomial becomes x² − 2x − 3 = (x − 3)(x + 1).
Its roots are 3 and −1. The first is inside (0,5), but the second lies to the left of zero.
Take k = 5, which fails the right condition k < 4. The polynomial becomes x² − 7x + 7. Its larger root is (7 + √21)/2, greater than 5.
These examples show that failing one endpoint sign test may leave one root inside while the other escapes. The question requires both roots inside; having at least one inside is insufficient.
A downward-opening example
Consider g(x) = −x² + 4x − 3 and the interval (0,4). It factorises as −(x − 1)(x − 3).
Its roots are 1 and 3, both inside the interval. The leading coefficient is a = −1.
At the endpoints, g(0) = −3 and g(4) = −3, both negative, matching the sign of a. Its vertex lies at x = 2, and the discriminant is 4 > 0.
Thus the general test applies exactly as written: g(0)/a > 0 and g(4)/a > 0, h inside and D > 0. Requiring positive endpoint values here would be the wrong sign condition.
Endpoint signs alone are not sufficient
Take f(x) = (x − 6)(x − 7) and the interval (0,5). Both f(0) and f(5) are positive.
The discriminant is positive, so the quadratic has two distinct real roots. But the vertex is halfway between them at x = 6.5, outside (0,5).
Therefore the endpoint signs and discriminant do not ensure both roots are inside. The vertex condition detects that the complete root pair lies to the right.
This counterexample is worth remembering because endpoint checks can look convincing on a rough graph. The midpoint of the two roots must also belong to the requested interval.
The vertex condition alone is not sufficient
Consider f(x) = (x + 1)(x − 3) = x² − 2x − 3 on (0,4).
The roots are −1 and 3, so one lies outside and one inside. Their midpoint is x = 1, which lies within (0,4).
The discriminant is positive, yet f(0) = −3 is negative. Therefore the left endpoint is actually between the roots, not outside both.
The endpoint sign check detects the problem. Knowing that the axis of symmetry lies inside does not guarantee that the complete distance from the midpoint to either root fits within the interval.
A minimum above zero gives no real roots
Take f(x) = (x − 2)² + 1 on (0,4). Its vertex lies at x = 2 inside the interval.
The endpoint outputs are positive: f(0) = 5 and f(4) = 5. But the minimum value is 1, which remains above the x-axis.
Its discriminant is negative, so the quadratic has no real roots at all.
This shows why all three tests are necessary. Vertex location and positive endpoints alone are not enough; the curve must actually cross the axis twice between them.
A compact alternative using normalised coefficients
Divide the quadratic by its nonzero leading coefficient a. This gives the monic polynomial q(x) = x² + (b/a)x + c/a.
The roots are unchanged by this division. The monic polynomial opens upward, so both roots lie strictly inside (L,U) exactly when q(L) > 0, q(U) > 0, the vertex is inside and the discriminant is positive.
Writing q(L) > 0 is equivalent to f(L)/a > 0. The normalisation makes the sign logic uniform across originally upward- and downward-opening quadratics.
When a itself depends on a parameter, however, do not silently divide before excluding a = 0. Work in parameter cases or use af(L) > 0 and af(U) > 0 while keeping a nonzero.
A symmetric interval example
Suppose q(x) = x² − px + q, with real parameters p and q, has both distinct roots in (−1,1).
Its vertex input is h = p/2, so −1 < p/2 < 1, giving |p| < 2.
The endpoint conditions are q(−1) = 1 + p + q > 0 and q(1) = 1 − p + q > 0. Together they give q > |p| − 1.
The discriminant condition is p² − 4q > 0, so q < p²/4. Therefore |p| < 2 and |p| − 1 < q < p²/4 are the exact conditions for both distinct roots to lie strictly inside (−1,1).
Check the symmetric example with specific parameters
Take p = 0 and q = −1/4. The conditions become |p| = 0 < 2, q = −1/4 > −1 and q < 0.
The quadratic is x² − 1/4 = 0, with roots −1/2 and 1/2. Both belong to (−1,1).
At p = 0, choosing q = 1/4 would violate q < p²/4 = 0 and yield x² + 1/4 = 0, with no real roots.
These examples illustrate how an apparently complicated two-parameter region reflects the same three geometric tests: actual crossings, vertex position and endpoint signs.
What changes for a closed interval?
If both roots may equal endpoints, replace the strict location tests with conditions appropriate to the inclusive interval. For two distinct roots in [L,U], D remains strictly positive, while the endpoint signs can allow equality.
The vertex of two distinct roots in a closed interval still lies strictly between L and U, because their midpoint is strictly inside even if one or both roots equal an endpoint.
For example, (x − 0)(x − 4) has roots 0 and 4 in [0,4] but not in (0,4). Both endpoint outputs are zero.
Always distinguish the requirement of two distinct roots from the requirement of strict placement. Changing bracket notation can change which parameter boundary values are allowed.
Why direct root formulas can be less efficient
One could solve x = [−b ± √D]/(2a) and impose L < [−b − √D]/(2a) and [−b + √D]/(2a) < U, with careful attention to the sign of a.
When a or b depends on a parameter, those inequalities may involve radicals and uncertain denominator signs, leading to multiple cases.
The vertex and endpoint approach replaces them with simpler polynomial inequalities in many examples. The exact same geometry is being described without explicitly writing both roots.
The quadratic formula remains a valid verification method for a sample parameter or for problems that explicitly request the roots themselves. Choose the representation that best matches the requested information.
A concise exam solution routine
First ensure the expression is genuinely quadratic, with a ≠ 0. Define the desired interval L < x < U.
Second require D > 0 for two distinct real roots. Third compute the axis h = −b/(2a) and require L < h < U.
Fourth calculate f(L) and f(U). Require each to have the same sign as a, using f(L)/a > 0 and f(U)/a > 0.
Fifth intersect all parameter restrictions and check boundary cases in the original equation. This is a complete logical chain: existence, midpoint location and placement of both endpoint values outside the root interval.
Practice questions with checked answers
Question one: f_k(x) = x² − (k + 2)x + 2k − 3. Find k such that both roots are strictly in (0,5). Answer: 3/2 < k < 4.
Question two: for k = 2 in Question one, find the roots. Answer: 2 − √3 and 2 + √3.
Question three: for k = 3/2, explain why strict inclusion fails. Answer: one root equals 0.
Question four: for k = 4, explain why strict inclusion fails. Answer: one root equals 5.
Question five: g(x) = −x² + 4x − 3. Are both roots inside (0,4)? Yes: they are 1 and 3.
Question six: (x − 6)(x − 7) has positive values at 0 and 5. Why are its roots not both in (0,5)? Its vertex lies at 6.5 outside the interval.
For each, explain the sign criterion and boundary check without immediately invoking the quadratic formula.
Learning transfer and syllabus references
Quadratic roots within a specified interval connect discriminants, turning points, quadratic inequalities and parameter conditions. The full test can be useful enrichment for students moving from basic root counting to rigorous root localisation.
For Singapore, refer to the relevant SEAB syllabus pages for 2026 O-Level Additional Mathematics, 2027 SEC G3 or 2027 SEC G2, as appropriate.
Related guides include How to Find When Two Quadratic Roots Straddle a Given Number, How to Find Parameter Values for Two Positive Quadratic Roots, How to Find the Range of a Quadratic on a Restricted Interval and the Additional Mathematics archive.
Families may enquire about suitable eduKate Singapore small-group Mathematics tutorials near Sixth Avenue MRT. The enduring skill is proving that both roots fit inside a specified window, not merely proving that two real roots exist. Properly taught kids shine a bright light into the future.
