There it is — born from putting −c/a and b²/(4a²) over the common denominator 4a², since −c/a = −4ac/(4a²). The discriminant is nothing more mysterious than the numerator of the right-hand side after you complete the square.
Now finish. Take the square root of both sides, remembering the ±. The bottom is easy: √(4a²) = 2|a|, and since ± already produces both signs, writing 2a instead of 2|a| gives the same pair of answers either way. So:
Now look at the formula and ask what can vary. The −b is a plain number. The 2a is a plain number. Only one thing in the whole expression can behave in more than one way, and that is the square root. So the entire question — how many real solutions does this equation have — collapses onto the sign of one number. Call it D.
If D > 0, then √D is a real number bigger than zero. Adding it and subtracting it push you to two different places on either side of −b/(2a). Two distinct real solutions.
If D = 0, then √0 = 0, and ±0 adds nothing and subtracts nothing. Both branches land on the same number, x = −b/(2a). Exactly one real solution. It is often called a double root, because the quadratic factors as a(x − r)² and the factor x − r appears twice.
If D < 0, you are asked for the square root of a negative number. Square a positive and you get a positive; square a negative and you get a positive; square zero and you get zero. No real number is left to do the job. No real solution.
The picture agrees, and it agrees for a reason. Complete the square on the function instead of the equation and you get vertex form: y = a(x + b/(2a))² − D/(4a). So the vertex sits at
The vertex’s height above or below the x-axis is controlled entirely by D. When D < 0, the height −D/(4a) takes the same sign as a: an upward parabola floats above the axis, a downward one hangs below it — and either way it opens away from the axis and never reaches it. When D = 0 the vertex height is exactly 0, so the curve rests on the axis at a single point. When D > 0 the vertex is on the opposite side of the axis from the direction the arms travel, so the curve must cross twice. Watch the dashed vertex line in the graph as you slide c: the number in the ledger and the gap on the screen are the same fact said two ways.
A second thing D measures: how far apart the roots are. Subtract the two solutions and the −b/(2a) cancels, leaving a gap of √D ÷ |a|. Big D, wide-apart roots. Shrink D toward 0 and you can watch the two crossings slide together until they collide into one. That collision is D = 0. Nothing discontinuous happens — the roots simply meet.
And a third thing, if a, b, c are integers: when D is a perfect square (0, 1, 4, 9, 16, 25, …), the radical evaporates into a whole number and both solutions come out rational — which is exactly the situation in which the quadratic factors over the integers. When D is positive but not a perfect square, the roots are real but irrational, and no amount of guessing at factors will find them. That is a genuinely useful thing to know before spending five minutes hunting for factors that do not exist.
| equation | D = b² − 4ac | real roots | what they look like |
|---|---|---|---|
| x² − 5x + 6 = 0 | 25 − 24 = 1 | two | 2 and 3 — rational |
| 2x² + 7x − 4 = 0 | 49 + 32 = 81 | two | −4 and 12 — rational |
| x² − 4x + 1 = 0 | 16 − 4 = 12 | two | 2 ± √3 — irrational |
| x² − 6x + 9 = 0 | 36 − 36 = 0 | one | 3, doubled — (x − 3)² |
| x² + x + 1 = 0 | 1 − 4 = −3 | none | curve floats above the axis |
| −x² + 2x − 5 = 0 | 4 − 20 = −16 | none | curve hangs below the axis |
The excluded value: a ≠ 0. Every line of the derivation divided by a, and the final formula divides by 2a. If a = 0 there is no x² term, the equation is not quadratic at all, and 2a = 0 makes the formula read as division by zero — undefined, not infinite, just meaningless. Worse, the discriminant does not fail loudly: for 0x² + 3x − 6 = 0 it computes 9 − 0 = 9 and cheerfully announces “two real roots,” when the honest answer is the single root x = 2. Confirm a ≠ 0 before you trust D at all — and be especially careful when a contains a parameter, as in (k − 2)x² + 3x + 1 = 0, which stops being quadratic exactly at k = 2. The fifth case in the stepper walks through this.
On extraneous solutions. Solving a quadratic that is already a polynomial equation cannot manufacture false answers: dividing by a nonzero a, adding the same amount to both sides, and completing the square are all reversible, and a polynomial is defined for every real x, so there are no excluded values to screen out. But quadratics often arrive from somewhere else — from clearing a denominator, or from squaring both sides of a radical equation — and those steps are not reversible. If your quadratic came from 1/(x − 3) + 1 = x, then x = 3 must be thrown out no matter what the discriminant says, because it was never allowed in the original. If it came from squaring, every root must be tested in the original equation, since squaring can promote a false statement into a true one. The rule: the discriminant counts the roots of the quadratic in front of you; only the original problem decides which of them you are permitted to keep.