Tutorials › Calculus II › Exponential Denominators and Geometric Bounds

Integral and Direct Comparison · Tutorial 616 of 998

Exponential Denominators and Geometric Bounds

Use exponential terms to construct valid geometric upper bounds, even when polynomial factors appear in the numerator or denominator.

Intermediate 10 min read

What You'll Learn

  • Bound reciprocals of sums containing a positive exponential term.
  • Use the geometric-series formula to justify convergence from an explicit upper bound.
  • Prove an exponential bound for polynomial numerators.
  • Handle added polynomial terms without reversing a denominator inequality.
  • Identify a common comparison error and correct it.

Why Exponential Denominators Suggest Geometric Bounds

An exponential term such as \(4^n\) grows by a fixed factor each time \(n\) increases by one. Its reciprocal, \(1/4^n=(1/4)^n\), is therefore a geometric term. This makes exponential denominators natural candidates for direct comparison with a geometric series.

The comparison must have the correct direction. If a positive denominator is at least \(a^n\), then its reciprocal is at most \(1/a^n\). That gives an upper bound, which is the direction needed to prove convergence by direct comparison. Added positive terms in a denominator can make the original term smaller, not larger.

Definition: A geometric series has terms \(c r^n\), where \(c\) is constant and \(r\) is the common ratio. When \(0\le r<1\), the series converges. In particular, for \(a>1\), \(\sum_{n=1}^{\infty} a^{-n}\) converges because its common ratio is \(1/a\), which is less than \(1\).

The Geometric Bound and Direct Comparison

The formula for a finite geometric sum explains why a geometric upper bound is useful. For \(0\le r<1\), let \(S_N=r+r^2+\cdots+r^N\). Multiplying by \(r\) gives \(rS_N=r^2+\cdots+r^{N+1}\). Subtracting the second equation from the first leaves \(S_N-rS_N=r-r^{N+1}\), so

$$ S_N=\frac{r(1-r^N)}{1-r}. $$

As \(N\) increases, \(r^N\) tends to zero when \(0\le r<1\). Thus the partial sums approach \(r/(1-r)\), a finite value. This proves convergence of the geometric series starting at \(n=1\), and gives its sum.

The direct comparison rule says that if \(0\le u_n\le v_n\) for every sufficiently large \(n\), and \(\sum v_n\) converges, then \(\sum u_n\) converges as well. A finite number of initial terms cannot change whether a series converges. For series with exponential denominators, the main task is therefore to establish a correct upper bound by geometric terms.

Rule: If \(a>1\), \(P(n)\ge0\), and \(Q(n)\ge0\), then \(a^n+Q(n)\ge a^n\), and hence \[ \frac{1}{a^n+Q(n)}\le \frac{1}{a^n}. \] This simple bound applies directly when the numerator is constant. A polynomial numerator requires an additional bound.

Worked Example: A Polynomial Added to the Denominator

Worked Example: Comparing \(1/(5^n+n^2)\) with a Geometric Series

Determine whether \(\sum_{n=1}^{\infty}1/(5^n+n^2)\) converges.

For every positive integer \(n\), \(n^2\ge0\). Therefore \(5^n+n^2\ge5^n>0\). Taking reciprocals reverses the inequality:

$$ 0<\frac{1}{5^n+n^2}\le\frac{1}{5^n}=\left(\frac15\right)^n. $$

The comparison series has common ratio \(1/5\), and \(0<1/5<1\). Its sum, starting at \(n=1\), is \((1/5)/(1-1/5)=1/4\). The direct comparison rule now proves that the given series converges.

Check: The original term is positive and no larger than the comparison term because its denominator is larger. The comparison ratio \(1/5\) is less than \(1\), and the resulting geometric sum is finite. Thus both the inequality and the convergence conclusion have the required direction.

Polynomial Numerators Can Also Be Bounded

A term such as \(n^2/4^n\) is not bounded above by \(1/4^n\), because \(n^2\) can exceed \(1\). The needed idea is that a fixed power of \(n\) can itself be bounded by an exponential with a smaller base. Then the remaining quotient is geometric.

Fact: If \(m\) is a fixed positive integer and \(c>1\), there are a constant \(C>0\) and an integer \(N\) such that \(n^m\le Cc^n\) for all \(n\ge N\).

To justify the fact, choose an integer \(N\) large enough that \((1+1/n)^m\le c\) whenever \(n\ge N\). Such an \(N\) exists: by the binomial theorem, \((1+1/n)^m\) is a finite sum whose first term is \(1\) and whose other terms tend to zero as \(n\) grows. Set \(C=N^m/c^N\). At \(n=N\), \(n^m=Cc^n\). If \(n^m\le Cc^n\) for some \(n\ge N\), then

$$ (n+1)^m =n^m\left(1+\frac1n\right)^m \le Cc^n c =Cc^{n+1}. $$

Induction proves the stated bound for every \(n\ge N\). In particular, given \(a>1\), choose \(c\) with \(1<c<a\). Then \(Cc^n/a^n=C(c/a)^n\) is a constant times a convergent geometric sequence. The fact explains why a fixed polynomial numerator does not prevent convergence when the denominator grows exponentially. In applications, it is best to choose a convenient explicit inequality whenever possible.

Worked Example: A Power in the Numerator

Worked Example: Comparing \(n^2/4^n\) with a Geometric Series

Determine whether \(\sum_{n=1}^{\infty}n^2/4^n\) converges, using an explicit bound.

First we prove that \(n^2\le2^n\) for all \(n\ge4\). At \(n=4\), both sides equal \(16\). Suppose \(n^2\le2^n\) for some \(n\ge4\). Since \((n+1)^2\le2n^2\) is equivalent to \(n^2-2n-1\ge0\), and \(n^2-2n-1\ge16-8-1=7>0\) for \(n\ge4\), we obtain

$$ (n+1)^2\le2n^2\le2\cdot2^n=2^{n+1}. $$

Induction establishes the bound. Therefore, for every \(n\ge4\),

$$ 0\le\frac{n^2}{4^n} \le\frac{2^n}{4^n} =\left(\frac12\right)^n. $$

The comparison series has ratio \(1/2\), so its tail from \(n=4\) converges. More explicitly, that tail sums to \((1/2)^4/(1-1/2)=1/8\). The first three terms of the original series are \(1/4\), \(4/16=1/4\), and \(9/64\), which are finite. Adding these finitely many terms to a convergent tail preserves convergence. Hence the given series converges.

Check: The induction starts with \(4^2=2^4\), and the induction step establishes the next case. Dividing the proven upper bound \(n^2\le2^n\) by the positive quantity \(4^n\) gives the stated comparison with \((1/2)^n\). The geometric ratio is strictly less than \(1\), as required.

Worked Example: A Polynomial Numerator and a Polynomial Addition

Worked Example: Comparing \((3n+1)/(2^n+n^3)\) with a Geometric Series

Determine whether \(\sum_{n=1}^{\infty}(3n+1)/(2^n+n^3)\) converges.

For \(n\ge8\), we claim that \(3n+1\le(3/2)^n\). At \(n=8\), the left side is \(25\), while the right side is \(6561/256\); the inequality holds because \(25\cdot256=6400\le6561\). For the induction step, when \(n\ge8\),

$$ \frac{3(n+1)+1}{3n+1} =\frac{3n+4}{3n+1} \le\frac32, $$

because this inequality is equivalent to \(6n+8\le9n+3\), or \(5\le3n\). If \(3n+1\le(3/2)^n\), it follows that

$$ 3(n+1)+1 \le\frac32(3n+1) \le\left(\frac32\right)^{n+1}. $$

Thus the claim holds for every \(n\ge8\). Also, \(2^n+n^3\ge2^n>0\). Combining these inequalities gives

$$ 0\le\frac{3n+1}{2^n+n^3} \le\frac{3n+1}{2^n} \le\frac{(3/2)^n}{2^n} =\left(\frac34\right)^n \qquad(n\ge8). $$

The comparison series has ratio \(3/4<1\), so its tail converges. The original terms for \(n=1,\ldots,7\) are finite, since each denominator is positive. Therefore the full series converges.

Check: The base case compares \(6400\) with \(6561\), and the induction step uses a factor no larger than \(3/2\). The denominator bound gives the first upper inequality; the numerator bound gives the second. Their combination yields the geometric ratio \(3/4\), which is less than \(1\).

A Common Mistake: Reversing the Reciprocal Inequality

When comparing positive quantities, taking reciprocals reverses the inequality. For example, from \(5^n+n^2\ge5^n\), it is incorrect to conclude that

$$ \frac{1}{5^n+n^2}\ge\frac{1}{5^n}. $$

The denominator on the left is at least as large as the denominator on the right, so its reciprocal is no larger. The correct calculation is

$$ 5^n+n^2\ge5^n>0 \quad\Longrightarrow\quad \frac{1}{5^n+n^2}\le\frac{1}{5^n}. $$

The incorrect inequality would claim that adding \(n^2\) to the denominator makes the term larger. That is precisely backwards. A lower bound on the terms is not what establishes convergence by direct comparison; the useful geometric comparison here is an upper bound by a convergent series.

Choosing and Checking a Bound

When a term has an exponential denominator, separate the task into two questions: what can be said about the denominator, and what remains in the numerator after that bound? If the denominator is \(a^n+Q(n)\) with \(Q(n)\ge0\), then \(a^n\) is a lower bound for the denominator. This immediately handles a constant numerator. If the numerator contains a fixed power or polynomial, bound that factor by a smaller exponential and check that the quotient of the two exponential bases is less than \(1\).

1
Check positivity.
Ensure the terms and denominators are positive on the indices being considered.
2
Write the inequality in the useful direction.
For convergence, seek an upper bound on the terms by a known convergent series.
3
Check the geometric ratio.
A geometric comparison proves convergence only when the absolute value of its ratio is less than \(1\).
4
Account for initial terms.
If the bound begins only after some index, verify that the omitted finite set consists of finite terms.

These checks matter because an expression that looks exponential may not automatically give a valid comparison. A denominator must remain positive before reciprocals can be compared, any numerator factor must be included in the bound, and a geometric ratio greater than or equal to \(1\) cannot supply the required convergence conclusion. When each condition is checked explicitly, the comparison is short and reliable.

Check Your Understanding

For each question, focus on the direction of the inequality and the geometric ratio.

  1. For \(n\ge1\), what geometric series gives an upper bound for \(1/(3^n+n)\)?
  2. Why does \(a^n+Q(n)\ge a^n\), with \(a>1\) and \(Q(n)\ge0\), give an upper bound on the reciprocal?
  3. What condition on \(r\) ensures convergence of \(\sum_{n=1}^{\infty}r^n\)?
  4. For \(n\ge4\), use \(n^2\le2^n\) to give a geometric upper bound for \(n^2/8^n\).
  5. Explain why a bound that holds only for \(n\ge N\) can still establish convergence of the full series.