Tutorials › Real Analysis › Bernoulli’s Inequality

Number Systems · Tutorial 123 of 1000

Bernoulli's Inequality

Learn to apply Bernoulli’s Inequality, prove it by induction, and identify when its lower bound is strict.

Beginner 9 min read

What You'll Learn

  • State Bernoulli’s Inequality for nonnegative integer exponents and identify its domain.
  • Prove the inequality by induction using a nonnegative factor.
  • Determine the equality cases and when the inequality is strict.
  • Apply the result to positive and negative increments.
  • Use a scaled form to estimate powers of sums.

A Lower Bound for Powers

The Binomial Theorem gives an exact expansion of \((1+x)^n\) when \(n\) is a nonnegative integer. Sometimes an exact expansion is more information than we need: a simple lower bound can be enough to compare quantities or estimate a power. Bernoulli’s Inequality supplies such a bound using only the constant and linear terms.

The sign of \(x\) matters. When \(x\geq0\), the inequality follows naturally from the nonnegative terms in the binomial expansion. It also remains valid for some negative values, but the induction proof must be handled carefully: multiplying an inequality by a factor preserves its direction only when that factor is nonnegative. This is why the condition \(x\geq-1\) will be part of the result.

Bernoulli’s Inequality

We use the convention that \(x^0=1\), including when \(x=0\). The exponent in the theorem is an integer; no assumption about powers with noninteger exponents is needed.

Theorem (Bernoulli’s Inequality). Let \(x\in\mathbb{R}\) satisfy \(x\geq-1\), and let \(n\geq0\) be an integer. Then $$ (1+x)^n\geq1+nx. $$

The assertion includes \(n=0\), when both sides equal \(1\). For \(n=1\), it is also an equality. For larger exponents it is a genuine lower bound, with equality conditions described below.

Proof. We use induction on \(n\), keeping the assumption \(x\geq-1\) fixed. When \(n=0\),

$$ (1+x)^0=1=1+0x. $$

Now suppose that for some integer \(n\geq0\),

$$ (1+x)^n\geq1+nx. $$

Because \(x\geq-1\), we have \(1+x\geq0\). Multiplying the induction hypothesis by \(1+x\) therefore preserves the direction of the inequality. It gives

$$ (1+x)^{n+1}\geq(1+nx)(1+x) =1+(n+1)x+nx^2. $$

Since \(n\geq0\) and \(x^2\geq0\), the remaining term satisfies \(nx^2\geq0\). Thus

$$ (1+x)^{n+1}\geq1+(n+1)x+nx^2\geq1+(n+1)x. $$

This proves the result for \(n+1\). By induction, it holds for every integer \(n\geq0\). \(\square\)

The proof identifies the main mechanism: after one more multiplication by \(1+x\), the desired linear expression appears, along with the extra term \(nx^2\). That term is nonnegative. The restriction \(x\geq-1\) ensures that the multiplication used in the induction does not reverse the inequality.

Equality and Strictness

The inequality is an equality for every \(x\) when \(n=0\) or \(n=1\). For \(n\geq2\), \(x=0\) still gives equality, since both sides are \(1\). For all other allowed \(x\), the inequality is strict.

Corollary (Equality Cases in Bernoulli’s Inequality). If \(n\geq2\) and \(x\geq-1\), then $$ (1+x)^n=1+nx $$ if and only if \(x=0\). For \(n=0\) and \(n=1\), equality holds for every \(x\geq-1\).

Proof. Define the difference between the two sides by

$$ D_n=(1+x)^n-1-nx. $$

We have \(D_1=0\). For every integer \(n\geq1\), direct rearrangement gives

$$ D_{n+1} =(1+x)^{n+1}-1-(n+1)x =(1+x)D_n+nx^2. $$

Indeed, substituting \(D_n=(1+x)^n-1-nx\) into the middle expression on the right yields \((1+x)^{n+1}-(1+x)(1+nx)+nx^2\), which simplifies to \((1+x)^{n+1}-1-(n+1)x\). For \(n=1\), this recurrence gives \(D_2=x^2\). If \(x\ne0\), then \(D_2>0\). Whenever \(n\geq2\) and \(D_n>0\), the assumptions \(1+x\geq0\), \(n>0\), and \(x^2>0\) imply

$$ D_{n+1}=(1+x)D_n+nx^2>0. $$

Thus \(D_n>0\) for every \(n\geq2\) when \(x\ne0\). If \(x=0\), then \(D_n=1-1-0=0\) for every \(n\). The cases \(n=0\) and \(n=1\) follow directly from their formulas. \(\square\)

Worked Applications

Worked Example: A Positive Increment

Use Bernoulli’s Inequality to find a lower bound for \((1+\frac{2}{5})^4\). Here \(x=\frac{2}{5}\), which satisfies \(x\geq-1\), and \(n=4\). Therefore

$$ \left(1+\frac{2}{5}\right)^4 \geq1+4\left(\frac{2}{5}\right) =1+\frac{8}{5} =\frac{13}{5}. $$

The bound is strict because \(n\geq2\) and \(x\ne0\). As a direct check, the left side is \((7/5)^4=2401/625\), while \(13/5=1625/625\). Since \(2401>1625\), the calculated value does exceed the stated lower bound.

Worked Example: A Negative Increment

Apply the inequality to \((1-\frac{1}{4})^3\). Set \(x=-\frac{1}{4}\) and \(n=3\). The domain condition holds because \(-\frac{1}{4}\geq-1\). Then

$$ \left(1-\frac{1}{4}\right)^3 \geq1+3\left(-\frac{1}{4}\right) =1-\frac{3}{4} =\frac{1}{4}. $$

This is a strict inequality: the left side is \((3/4)^3=27/64\), and the right side is \(1/4=16/64\). In this case the bound is positive and informative. If the same \(x\) is used with \(n=6\), the right side becomes \(1-6/4=-1/2\); the inequality remains correct, but a negative lower bound for a positive power is not very useful.

Worked Example: A Bound Valid for Every Positive Integer

Let \(m\) be any positive integer, and consider \((1+\frac{1}{m})^m\). The value \(x=\frac{1}{m}\) is nonnegative, so Bernoulli’s Inequality applies with exponent \(n=m\):

$$ \left(1+\frac{1}{m}\right)^m \geq1+m\left(\frac{1}{m}\right) =2. $$

Thus every one of these powers is at least \(2\). If \(m=1\), the expression is exactly \(2\). If \(m\geq2\), the equality-case result shows that it is strictly greater than \(2\), since \(x=1/m\ne0\). No expansion with \(m\) terms is needed to get this lower bound.

Worked Example: Scaling the Inequality

Suppose \(a>0\), \(b\geq-a\), and \(n\geq1\) is an integer. To estimate \((a+b)^n\), write it as \(a^n(1+b/a)^n\). Since \(a>0\), the ratio \(b/a\) is defined, and \(b\geq-a\) implies \(b/a\geq-1\). Bernoulli’s Inequality gives

$$ (a+b)^n =a^n\left(1+\frac{b}{a}\right)^n \geq a^n\left(1+n\frac{b}{a}\right) =a^n+na^{n-1}b. $$

For example, take \(a=4\), \(b=-1\), and \(n=3\). The hypotheses hold, and the scaled bound is

$$ (4-1)^3\geq4^3+3\cdot4^2(-1) =64-48=16. $$

Indeed, \(3^3=27\geq16\). The scaling requires \(a>0\): it lets us divide by \(a\) and multiply the resulting inequality by the positive number \(a^n\) without changing its direction.

Why the Domain Condition Matters

The hypothesis \(x\geq-1\) is more than a technical detail. In the induction proof, it makes \(1+x\) nonnegative, so the induction hypothesis can be multiplied by that factor without reversing its direction. If \(x<-1\), the theorem can fail. For instance, take \(x=-4\) and \(n=3\). Then

$$ (1+x)^n=(-3)^3=-27, \qquad 1+nx=1+3(-4)=-11, $$

and \(-27<-11\), contrary to the claimed lower bound. The theorem therefore cannot be used without checking its domain. A separate check of \(n\) is needed too: the result concerns nonnegative integer exponents, not arbitrary real exponents.

For \(x\geq0\), the Binomial Theorem also helps explain why the inequality is natural. The expansion begins with \(1+nx\); when \(n\geq2\), its remaining terms are nonnegative because they involve nonnegative powers of \(x\) and positive binomial coefficients. For \(-1\leq x<0\), however, the later terms can have alternating signs. The induction proof handles that case without incorrectly assuming that every term in the expansion is nonnegative.

A useful habit is to check three items before applying Bernoulli’s Inequality: the exponent is a nonnegative integer, the increment satisfies \(x\geq-1\), and the resulting lower bound is strong enough to be useful. A true lower bound may still be weak, as the negative right side in the second example illustrates.

Key takeaway. For \(x\geq-1\) and an integer \(n\geq0\), Bernoulli’s Inequality gives \((1+x)^n\geq1+nx\). Its induction proof depends on multiplying by the nonnegative factor \(1+x\). When \(n\geq2\), equality occurs exactly at \(x=0\).

Check Your Understanding

Use the domain condition, induction argument, and equality cases to answer these questions.

  1. For which real values of \(x\) does Bernoulli’s Inequality apply, and why does the induction proof need this condition?
  2. Apply the inequality with \(x=\frac{1}{3}\) and \(n=5\) to obtain a lower bound for \((1+\frac{1}{3})^5\).
  3. For \(x=-\frac{1}{2}\) and \(n=2\), is the inequality an equality or a strict inequality? Explain using the equality cases.
  4. Use the scaled form with \(a=3\), \(b=1\), and \(n=4\) to obtain a lower bound for \((3+1)^4\).
  5. Why does the induction step not establish Bernoulli’s Inequality for all \(x<-1\)?