Why Derivative Roots Control Function Roots
A function can cross the horizontal axis several times only if its graph changes direction or shape between successive crossings. Rolle’s Theorem makes this idea precise: between two distinct roots of a continuous function, its derivative has a root. Applying the theorem again to the derivative gives a further constraint, and repeating the argument creates bounds on how many distinct roots the original function can have.
These bounds do not usually locate every root. Instead, they let us rule out the possibility of “too many” roots by counting roots of a derivative, which may be easier to analyze.
The Root-Counting Result
A root of \(g\) means a number \(x\) in the specified domain for which \(g(x)=0\). In this tutorial, roots are counted as distinct points: if a polynomial has a repeated root at one point, that point still contributes one distinct root to the count.
The hypotheses ensure that Rolle’s Theorem can be applied at each stage. In particular, continuity on \([a,b]\) includes the endpoints, where roots of \(f\) are allowed. The roots guaranteed for the derivatives, however, lie strictly inside \((a,b)\).
The theorem counts the roots we can guarantee, not necessarily every root of the derivative. The derivative might have additional roots, and a guaranteed derivative root can sometimes also be a root of \(f\). What matters is that the roots counted at each stage are distinct.
Why Repeated Applications Give the Bound
First consider \(k=1\). List \(n\) distinct roots of \(f\) in increasing order as \(x_1<x_2<\cdots<x_n\). For each adjacent pair, \(f(x_i)=f(x_{i+1})=0\). Rolle’s Theorem applied on \([x_i,x_{i+1}]\) guarantees a point \(c_i\in(x_i,x_{i+1})\) with \(f'(c_i)=0\). The intervals between adjacent roots do not overlap, so these \(n-1\) points are distinct and lie in \((a,b)\).
To repeat the reasoning, apply Rolle’s Theorem to \(f'\) between consecutive roots just found. This gives at least \(n-2\) distinct roots of \(f''\). Each further application reduces the guaranteed count by one. The following induction makes that reasoning precise.
Choose \(n\) distinct roots \(x_1<\cdots<x_n\) in \([a,b]\). Applying Rolle’s Theorem between each adjacent pair gives \(n-1\) distinct roots of \(f'\) in \((a,b)\).
If \(f^{(j)}\) has at least \(n-j\) distinct roots in \((a,b)\), order those roots. Rolle’s Theorem between consecutive ones gives at least \(n-j-1\) distinct roots of \(f^{(j+1)}\), also in \((a,b)\).
The differentiability assumption permits these applications through \(f^{(k)}\). After \(k\) steps, the guaranteed number is \(n-k\).
For the induction step, if \(f^{(j)}\) has \(n-j\) ordered roots \(y_1<\cdots<y_{n-j}\), apply Rolle’s Theorem to \(f^{(j)}\) on each \([y_i,y_{i+1}]\). The resulting roots of \(f^{(j+1)}\) lie in the disjoint intervals \((y_i,y_{i+1})\), so they are distinct. Since the \(y_i\) are interior points and the new roots lie between them, the new roots are also in \((a,b)\). This proves the claimed count at every step.
The condition \(n>k\) ensures that the stated lower bound \(n-k\) is positive. If there are infinitely many distinct roots of \(f\), the same argument can be applied to any finite selection of them. Consequently, for each fixed \(k\), if the stated differentiability conditions hold, \(f^{(k)}\) must also have infinitely many distinct roots in \((a,b)\).
Turning a Derivative Bound into a Function Bound
Indeed, if \(f\) had \(r+2\) distinct roots, Rolle’s Theorem would give at least \(r+1\) distinct roots of \(f'\) in \((a,b)\), contradicting the assumed bound of \(r\). The same logic works for higher derivatives: if \(f\) is continuous on \([a,b]\), \(k\) times differentiable on \((a,b)\), and \(f^{(k)}\) has at most \(r\) distinct roots in \((a,b)\), then \(f\) has at most \(r+k\) distinct roots in \([a,b]\).
To see the higher-order conclusion, suppose instead that \(f\) had more than \(r+k\) roots. Then it would have at least \(r+k+1\) distinct roots. The theorem would force \(f^{(k)}\) to have at least \(r+1\) distinct roots, contrary to the bound. This argument also rules out infinitely many roots of \(f\) when \(f^{(k)}\) has only finitely many.
Worked Examples
Worked Example: Matching Root Counts Through Two Derivatives
Let \(f(x)=x^3-3x\) on \([-2,2]\). Find the distinct roots of \(f\), \(f'\), and \(f''\), and compare their counts.
Factor \(f\):
Thus the three distinct roots are \(-\sqrt{3}\), \(0\), and \(\sqrt{3}\). They all lie in \([-2,2]\), since \(\sqrt{3}<2\).
Differentiate:
The two distinct roots of \(f'\) are \(-1\) and \(1\), both in \((-2,2)\). Differentiating again gives
so \(f''\) has the single root \(0\) in \((-2,2)\). The counts are \(3\), \(2\), and \(1\), respectively, matching the lower bounds \(n-1=2\) and \(n-2=1\).
Check the roots by substitution: \(f(0)=0\), while \(f(\sqrt{3})=\sqrt{3}(3-3)=0\) and \(f(-\sqrt{3})=-\sqrt{3}(3-3)=0\). Also \(f'(\pm1)=3-3=0\) and \(f''(0)=0\). These calculations confirm the stated roots.
Worked Example: Zero Roots of the Derivative Give a One-Root Bound
Show that \(g(x)=x^3+3x\) has at most one distinct root on \([-1,1]\), then verify its root count.
The derivative is
For every real \(x\), \(x^2\geq 0\), so \(x^2+1\geq 1>0\). Therefore \(g'(x)>0\) throughout \((-1,1)\), and \(g'\) has no roots there. The corollary with \(r=0\) shows that \(g\) has at most one distinct root in \([-1,1]\).
Direct substitution gives \(g(0)=0^3+3(0)=0\). Thus \(0\) is a root, and the upper bound shows it is the only distinct root on the interval.
Check the derivative calculation: differentiating \(x^3+3x\) gives \(3x^2+3\). Since this derivative is never zero, it is consistent with Rolle’s Theorem that \(g\) cannot have two distinct roots in \([-1,1]\): two such roots would force a derivative root between them.
Worked Example: Bounding Four Roots Using the Second Derivative
Let \(p(x)=x^4-5x^2+4\) on \([-2,2]\). Use the second derivative to obtain a bound, then check the actual root counts.
Differentiate twice:
The equation \(p''(x)=0\) becomes \(12x^2-10=0\), so \(x^2=5/6\) and \(x=\pm\sqrt{5/6}\). Both roots lie in \((-2,2)\). Thus \(p''\) has two distinct roots there. The second-derivative bound with \(r=2\) and \(k=2\) shows that \(p\) has at most \(2+2=4\) distinct roots in \([-2,2]\).
Now factor \(p\):
Its roots are \(-2,-1,1,2\), all distinct and in \([-2,2]\), so the upper bound of four is attained. Check by substitution: at \(x=\pm1\), the factor \(x^2-1\) is zero; at \(x=\pm2\), the factor \(x^2-4\) is zero.
For a derivative check, \(p'(\pm\sqrt{5/2})=2(\pm\sqrt{5/2})(2(5/2)-5)=0\), and \(p'(0)=0\). Thus \(p'\) has the three distinct roots \(0,\pm\sqrt{5/2}\) in the interval, consistent with Rolle’s bound from four roots of \(p\). The second derivative has the two roots already calculated.
Endpoint Roots and a Common Mistake
When \(k\geq 1\) and \(f\) has more than \(k\) distinct roots, the theorem guarantees the stated \(n-k\) roots of \(f^{(k)}\) in the open interval \((a,b)\). Do not make the same location claim for \(k=0\): the original roots of \(f\) are only known to lie in the closed interval \([a,b]\), and may include endpoints.
Worked Example: Why the Open-Interval Claim Needs \(k\geq 1\)
Let \(f(x)=x\) on \([0,1]\). The function has one distinct root, \(x=0\), and \(f^{(0)}\) means \(f\) itself. Here \(n=1\) and \(k=0\), so \(n>k\), but the root is not in \((0,1)\). It is an endpoint root.
The correction is to distinguish the original function from its derivatives. For \(k=0\), the \(n\) chosen roots are in \([a,b]\). For \(1\leq k\leq n-1\), the theorem guarantees \(n-k\) roots of \(f^{(k)}\) in \((a,b)\). In this particular example, \(f'(x)=1\), so there is no derivative root to confuse with the endpoint root of \(f\).
Check: \(f(0)=0\), while \(f(x)=x>0\) for every \(x\in(0,1)\). Also \(f'(x)=1\) for all \(x\), confirming that the open-interval root claim cannot be extended to \(k=0\).
A related error is to count a repeated root multiple times when applying the theorem. Rolle’s Theorem requires two distinct points where the function has the same value. A root of multiplicity two at a single point does not supply two distinct points. Root multiplicity may give additional information in other settings, but the root-counting result here concerns distinct roots only.
What the Bounds Do—and Do Not—Say
A derivative-root bound is useful because it can settle a question about the original function without solving its equation. If a derivative has no roots on the interval, the original function has at most one root there. If a second derivative has at most two roots, the original function has at most four. Such conclusions are upper bounds, not promises that the maximum number of roots occurs.
Rolle’s Theorem also does not guarantee that the original function has any roots. In the second example, the root at zero was verified by substitution; the derivative argument alone supplied only the upper bound. Always separate the two questions: how many roots can there be, and does a root actually exist?
Check Your Understanding
Use distinct-root counts, and pay attention to whether the interval is open or closed.
- If \(f\) has six distinct roots in \([a,b]\), how many distinct roots of \(f''\) in \((a,b)\) does the theorem guarantee?
- If \(f'\) has at most three distinct roots in \((a,b)\), what is the largest possible number of distinct roots of \(f\) in \([a,b]\)?
- If \(f'''\) has at most one distinct root in \((a,b)\), what upper bound follows for the number of distinct roots of \(f\) in \([a,b]\)?
- Why must the roots of the derivatives guaranteed by Rolle’s Theorem lie in open intervals between roots?
- Can a root at \(a\) be counted among roots of \(f\) in \([a,b]\)? Does the theorem guarantee that root lies in \((a,b)\) when \(k=0\)?
