The Rational Root Theorem gives a list of candidates. Descartes’ Rule tells you how many positive and negative zeros to expect. The bound theorems narrow the search further by establishing a range [L,U] that contains all real zeros. Any candidate outside this range can be skipped entirely.
The Upper Bound Theorem
If you divide f(x) by (x−c) using synthetic division where c>0, and all numbers in the bottom row are non-negative (no sign changes), then c is an upper bound for the real zeros of f. No real zero of f is greater than c.
Why It Works
When you perform synthetic division of f(x) by (x−c) with c>0 and get all non-negative entries in the bottom row, the quotient polynomial has all non-negative coefficients. For any x>c, both the quotient q(x) and the divisor (x−c) are positive, so f(x)=(x−c)q(x)+r>0 (where r≥0 is the non-negative remainder). Therefore f(x)>0 for all x>c, meaning no zero exists beyond c.
Worked Example 1: Finding an Upper Bound
Find an upper bound for the zeros of f(x)=x4−5x3+3x2+7x−10.
Try c=5:
511−55030371522−10110100
Bottom row: 1,0,3,22,100 — all non-negative.
Conclusion:5 is an upper bound. No real zero is greater than 5.
Can we find a tighter bound? Try c=4:
411−54−13−4−17−43−10122
Bottom row: 1,−1,−1,3,2 — contains negative values. So 4 is NOT confirmed as an upper bound.
Since c=4 failed and c=5 succeeded, 5 is the smallest integer upper bound for this polynomial.
The Lower Bound Theorem
If you divide f(x) by (x−c) using synthetic division where c<0, and the numbers in the bottom row alternate in sign (counting zeros as either sign), then c is a lower bound for the real zeros. No real zero of f is less than c.
The alternation means: positive, negative, positive, negative, … or negative, positive, negative, positive, … (zeros can be assigned to either sign to continue the pattern).
Worked Example 2: Finding a Lower Bound
Find a lower bound for the zeros of f(x)=x4−5x3+3x2+7x−10.
All non-negative: 3 is an upper bound. Eliminate 6.
Test c=−3:
−322−1−6−7−142177−21−1464248
Alternating: +,−,+,−,+. So −3 is a lower bound. Eliminate −6.
Remaining candidates:±1,±2,±3,±21,±23.
Step 5 — Test:
f(1)=2−1−14+7+6=0. Found x=1.
122−121−141−137−13−66−60
Reduced: 2x3+x2−13x−6.
Test x=−21:
−2121−1−130−66.5
2,0,−13,0.5 — remainder 0.5=0. Not a zero.
Test x=2:
222145−1310−3−6−6−12
Remainder −12=0.
Test x=−2:
−2221−4−3−136−7−6148
Remainder 8=0.
Test x=3:
322167−13218−62418
Remainder 18=0.
Test x=−3:
−3221−6−5−13152−6−6−12
Remainder −12=0.
Test x=21:
2122112−131−12−6−6−12
Remainder −12=0.
Test x=23:
2322134−136−7−6−10.5−16.5
Not a zero. Test x=−23:
−23221−3−2−133−10−6159
Not a zero. So the cubic 2x3+x2−13x−6 has no rational zeros from our list (other than what we may have missed). Let us re-check x=−21 more carefully:
−21221−10−130−13−66.50.5
Not zero. Let me try x=−2 again on the original: f(−2)=2(16)−(−8)−14(4)+7(−2)+6=32+8−56−14+6=−24=0.
Try the original polynomial with x=−1: f(−1)=2+1−14−7+6=−12=0.
The remaining cubic 2x3+x2−13x−6 must be solved numerically or has rational zeros we have not found. Let me verify by trying to factor directly: grouping 2x3+x2−13x−6. Testing x=2 gave remainder −12, x=−2 gave 8. By the IVT, there is a zero between 2 and 3, and also between −2 and −3. These are irrational zeros. The quadratic formula finishes the job after finding one more rational zero or confirming there are none.
This example illustrates an important point: not all polynomials with integer coefficients have rational zeros, even when the Rational Root Theorem gives many candidates.
In manufacturing, a polynomial model describes how a measurement varies with a parameter x. If all real roots must lie within a tolerance band [a,b], the bound theorems confirm this algebraically. For instance, if a vibration amplitude is modeled by:
A(f)=f3−12f2+44f−48
where f is frequency in Hz, the bound theorems can verify that all resonant frequencies (zeros) lie within the operational range.
Testing c=12:
1211−1212044044−48528480
All non-negative: 12 is an upper bound.
Testing c=−1:
−111−12−1−13441357−48−57−105
Alternating signs: −1 is a lower bound.
All resonant frequencies lie in [−1,12]. (In fact, this factors as (f−2)(f−4)(f−6), with zeros at 2, 4, 6 Hz.)
Practice Problems
Test your understanding with these problems. Click to reveal each answer.
Problem 1: Use synthetic division to show that 4 is an upper bound for the zeros of f(x)=x3−3x2−2x+4.
411−341−2424812
Bottom row: 1,1,2,12 — all non-negative.
Answer: Since c=4>0 and all entries are non-negative, 4 is an upper bound.
Problem 2: Show that −3 is a lower bound for the zeros of f(x)=x3+2x2−5x−6.
−3112−3−1−53−2−660
Bottom row: 1,−1,−2,0. The signs are +,−,−,?. The zero can be assigned negative to continue: +,−,−,−. This does NOT alternate.
Actually, −3 is a zero (remainder is 0), so it IS a lower bound (trivially — it is itself a zero, and we should check if any zeros are smaller).
Let us try c=−4:
−4112−4−2−583−6−12−18
Alternating: +,−,+,−. Yes, −4 is a lower bound.
Since f(x)=(x+3)(x+1)(x−2) (zeros at −3,−1,2), −3 is the smallest zero and −4 is confirmed as a lower bound.
Answer:−4 is a lower bound (alternating signs). In fact, −3 is the smallest zero.
Problem 3: Find the smallest positive integer upper bound for the zeros of f(x)=x3−7x+6.
Try c=2: 1,2,−3,0. Has a negative, not all non-negative. Not confirmed.
Try c=3: Synthetic division:
311033−7926612
All non-negative: 3 is an upper bound.
Answer:3 is the smallest positive integer upper bound. (The zeros are 1,2,−3, confirming 3 works.)
Problem 4: Determine the bounds for zeros of g(x)=2x3+5x2−4x−3 and list all rational zeros.
Upper bound: try c=1:
122527−473−330
Bottom row all non-negative AND remainder 0. So x=1 is both a zero and an upper bound.
Quotient: 2x2+7x+3=(2x+1)(x+3).
Zeros: x=1, x=−21, x=−3.
Lower bound: try c=−4:
−4225−8−3−4128−3−32−35
Alternating: +,−,+,−. Confirmed, −4 is a lower bound.