Derivative

Table of Contents

1 Definition

1.1 Formal Definition

\(f'(x) = \lim\limits_{h \to 0} \frac {f(x + h) - f(x)} {h}\)

  • equivalent : \(f'(x) = \lim\limits_{b \to x} \frac {f(b) - f(x)} {b-x}\)
  • Notations : \(f'(x)\), \(\frac{dy}{dx}\), \(\frac{d}{dx} f(x)\)

The process of calculating derivatives is called differentiation.

1.2 Find Derivatives by Definition

1.2.1 Steps

  1. Find \(f(x + h)\)
  2. Find and simplify \(f(x + h) - f(x)\)
  3. divide by \(h\) to get \(\frac{f(x + h) - f(x)} {h}\)
  4. Let \(h\to 0\) : \(f'(x) = \lim\limits_{h \to 0} \frac {f(x + h) - f(x)} {h}\), if the limit exists.

1.2.2 Examples

  • Example 1 : Constant Function

    \(f(x) = 10\). Find \(f'(x)\).

    \(f'(x) = \lim\limits_{h \to 0} \frac {f(x + h) - f(x)} {h}\)

    \(= \lim\limits_{h \to 0} \frac {10 - 10} {h}\)

    \(= \lim\limits_{h \to 0} 0\)

    \(= 0\)

  • Example 2 : Polynomial Function

    \(f(x) = 3x^2 + 5x - 1\). Find \(f'(3)\).

    \(f'(x) = \lim\limits_{h \to 0} \frac {f(x + h) - f(x)} {h}\)

    \(= \lim\limits_{h \to 0} \frac {[3(x+h)^2 + 5(x+h) - 1] - (3x^2 + 5x - 1)} {h}\)

    \(= \lim\limits_{h \to 0} \frac {[3(x^2 + h^2 + 2xh) + 5(x+h) -1] - (3x^2 + 5x - 1)} {h}\)

    \(= \lim\limits_{h \to 0} \frac {3x^2 + 3h^2 + 6xh + 5x + 5h -1 - 3x^2 - 5x + 1} {h}\)

    \(= \lim\limits_{h \to 0} \frac {3h^2 + 6xh + 5h} {h}\)

    \(= \lim\limits_{h \to 0} 3h + 6x + 5\)

    \(= 6x + 5\)

    Note:

    To find \(f'(a)\), first find \(f'(x)\), and then substitute \(x\) by \(a\) in \(f'(x)\). The order can't be changed.

    • RIGHT

      \(f'(x) = 6x + 5\), so subsitute \(x\) by \(3\) and get \(f'(3) = 6\cdot 3 + 5 = 23\)

    • WRONG

      \(f(3) = 3\cdot 3^2 + 5\cdot 3 - 1 = 41\), which is a constant, so \(f'(3) = 0\).

  • Example 3 : Sqrare Root Function : Using Conjugate

    \(f(x) = \sqrt{2x - 1} - 7\). Find \(f'(x)\).

    \(f'(x) = \lim\limits_{h \to 0} \frac {f(x + h) - f(x)} {h}\)

    \(= \lim\limits_{h \to 0} \frac {(\sqrt{2(x + h) - 1} - 7) - (\sqrt{2x - 1} - 7)} {h}\)

    \(= \lim\limits_{h \to 0} \frac {\sqrt{2(x + h) - 1} - \sqrt{2x - 1}} {h}\)

    \(= \lim\limits_{h \to 0} \frac {\sqrt{2(x + h) - 1} - \sqrt{2x - 1}} {h} \cdot \frac {\sqrt{2(x + h) - 1} + \sqrt{2x - 1}} {\sqrt{2(x + h) - 1} + \sqrt{2x - 1}}\)

    \(= \lim\limits_{h \to 0} \frac {[2(x + h) - 1] - (2x - 1)} {h} \cdot \frac {1} {\sqrt{2(x + h) - 1} + \sqrt{2x - 1}}\)

    \(= \lim\limits_{h \to 0} \frac {2x + 2h - 1 - 2x + 1} {h} \cdot \frac {1} {\sqrt{2(x + h) - 1} + \sqrt{2x - 1}}\)

    \(= \lim\limits_{h \to 0} \frac {2h} {h} \cdot \frac {1} {\sqrt{2(x + h) - 1} + \sqrt{2x - 1}}\)

    \(= \lim\limits_{h \to 0} \frac {2} {\sqrt{2(x + h) - 1} + \sqrt{2x - 1}}\)

    \(= \frac{2}{\sqrt{2x - 1} + \sqrt{2x - 1}}\)

    \(= \frac{2}{2\sqrt{2x - 1}}\)

    \(= \frac{1}{\sqrt{2x - 1}}\)

    Note:

    • Conjugate of \(\sqrt{2(x + h) - 1} - \sqrt{2x - 1}\) is \(\sqrt{2(x + h) - 1} + \sqrt{2x - 1}\).
    • \(\sqrt{2(x + h) - 1} - \sqrt{2x - 1} \neq \sqrt{2h}\)
  • Example 4 : Fraction : Using Common Denominator

    \(f(x) = \frac{2}{3x+1}\). Find \(f'(x)\).

    \(f'(x) = \lim\limits_{h \to 0} \frac {f(x + h) - f(x)} {h}\)

    \(= \lim\limits_{h \to 0} \frac {\frac{2}{3(x+h)+1} - \frac{2}{3x+1}} {h}\) \(= \lim\limits_{h \to 0} \frac {\frac{2\cdot (3x+1)}{[3(x+h)+1] \cdot (3x+1)} - \frac{2\cdot[3(x+h)+1]}{[3(x+h)+1] \cdot (3x+1)}} {h}\) \(= \lim\limits_{h \to 0} \frac {\frac {2\cdot (3x+1) - 2\cdot[3(x+h)+1]} {[3(x+h)+1] \cdot (3x+1)}} {h}\)

    \(= \lim\limits_{h \to 0} \frac{2\cdot (3x+1) - 2\cdot[3(x+h)+1]} {h\cdot [3(x+h)+1] \cdot (3x+1)}\) \(= \lim\limits_{h \to 0} \frac{6x + 2 - 6x - 6h - 2} {h\cdot [3(x+h)+1] \cdot (3x+1)}\) \(= \lim\limits_{h \to 0} \frac{- 6h} {h\cdot [3(x+h)+1] \cdot (3x+1)}\)

    \(= \lim\limits_{h \to 0} \frac{- 6} {[3(x+h)+1] \cdot (3x+1)}\) \(= \frac{- 6} {(3x+1) \cdot (3x+1)}\) \(= \frac{- 6} {(3x+1)^2}\)

2 Rules

2.1 Arithmetic Rules

2.1.1 Addition & Subtraction

If \(f(x) = g(x) \pm h(x)\) and if \(g'(x)\) and \(h'(x)\) both exist, then \[ f'(x) = g'(x) \pm h'(x) \]

2.1.2 Scalar Multiplication

If \(f(x) = k g(x)\) where \(k\) is a real number and \(g'(x)\) exists, then \[ f'(x) = k g'(x) \]

2.1.3 Product Rule

If \(f(x) = g(x) \cdot h(x)\) and if \(g'(x)\) and \(h'(x)\) both exist, then \[ f'(x) = g'(x) \cdot h(x) + g(x) \cdot h'(x)\]

  • Example

    \(f(x) = g(x) \cdot h(x)\) where \(g(x) = 2x^3 + 1\) and \(h(x) = 7x^2 - 2\)

    \(g'(x) = 2\cdot 3\cdot x^2 = 6x^2\)

    \(h'(x) = 7\cdot 2\cdot x = 14x\)

    \(f'(x) = g'(x) \cdot h(x) + g(x) \cdot h'(x)\)

    \(= 6x^2 (7x^2 - 2) + (2x^3 + 1) \cdot 14x\)

    \(= 42 x^4 - 12 x^2 + 28 x^4 + 14x\)

    \(= 70 x^4 - 12 x^2 + 14x\)

2.1.4 Quotient Rule

If \(f(x) = \frac {g(x)} {h(x)}\) and if \(g'(x)\) and \(h'(x)\) both exist, then \[f'(x) = \frac {g'(x) \cdot h(x) - g(x) \cdot h'(x)} {[h(x)]^2} \]

  • Example

    \(f(x) = \frac{g(x)} {h(x)}\) where \(g(x) = 2x^3 + 1\) and \(h(x) = 7x^2 - 2\)

    \(g'(x) = 2\cdot 3\cdot x^2 = 6x^2\)

    \(h'(x) = 7\cdot 2\cdot x = 14x\)

    \(f'(x) = \frac {g'(x) \cdot h(x) - g(x) \cdot h'(x)} {[h(x)]^2}\)

    \(= \frac {6x^2 \cdot (7x^2 - 2) - (2x^3+1) \cdot 14x} {(7x^2-2)^2}\)

    \(= \frac {6x^2 (7x^2 - 2) - (2x^3 + 1) \cdot 14x} {(7x^2-2)^2}\)

    \(= \frac {42 x^4 - 12 x^2 - 28 x^4 - 14x} {(7x^2-2)^2}\)

    \(= \frac {14 x^4 - 12 x^2 - 14x} {(7x^2-2)^2}\)

2.2 Chain Rule for Composite Function

2.2.1 Statement

For a composite function \(f(g(x))\), if we regard the function \(g(x)\) as a single variable \(g\) by letting \(g = g(x)\), then \[ \frac{df(g(x))}{dx} = \frac{df(g)}{dg} \cdot \frac{dg(x)}{dx} \]

2.2.2 Remarks

  • \(g\) and \(g(x)\)
    • when writting write \(g\), we mean a variable \(g\), just like \(x, u\) and \(v\)
    • when writing \(g(x)\), we mean a function of \(x\)
  • \(\frac{df(g)}{dg}\) means
    • the variable of function \(f(g)\) is \(g\)
    • The derivative of \(f(g)\) is with respect to the variable \(g\)
  • In short notation, we may memorize the chain rule by (plausibly, you get the LHS by cancelling \(dg\) on the RHS) \[ \frac{df}{dx} = \frac{df}{dg} \cdot \frac{dg}{dx} \]

2.2.3 Example

Let \(f(x) = (2x^2 + 1)^3\). Find \(\frac{df(x)}{dx}\).

Clearly \(f(x)\) is a composite function, where

  • The inner function is \(g(x) = 2x^2 + 1\).
  • The outer function is \(f(g) = g^3\) where \(g=g(x)\).

By Chain Rule \(\frac{df(g(x))}{dx} = \frac{df(g)}{dg} \cdot \frac{dg(x)}{dx}\), to find \(\frac{df(g(x))}{dx}\), we need \(\frac{df(g)}{dg}\) and \(\frac{dg(x)}{dx}\).

  • By regarding \(g\) as a single variable, we find \(\frac{df(g)}{dg} = \frac{d(g^3)}{dg} = 3 g^2\) by power rule
  • By regarding \(x\) as the variable, we find \(\frac{dg(x)}{dx} = \frac{d(2x^2 + 1)}{dx} = 4x\) by power rule
  • Putting them into the formula of chain rule, we get \(\frac{df(g(x))}{dx} = 3 g^2 \cdot 4x = 12xg^2\)
  • Here \(g\) is a variable that is not given by the question, so we have to replace it. Since \(g = g(x)\), we just replace \(g\) with the expression of \(g(x)\), which is \(2x^2 + 1\) here. Thus the final answer is \(\frac{df(g(x))}{dx} = 12x(2x^2+1)^2\).

2.3 Rules for Special Functions

2.3.1 Constant Rule

If \(f(x) = k\) where \(k\) is a real number, then \(f'(x) = 0\).

2.3.2 Power Rule

If \(f(x) = x^n\) where n is a real number, then \(f'(x) = n\cdot x^{n-1}\)

/Remark: / Whenever you meet something like \(\frac{1}{x}\) or \(\sqrt{x}\), rewrite them as \(x^{-1}\) or \(x^{\frac{1}{2}}\) respectively.

2.3.3 Polynomial Rule

By the power rule and addition rule, we get that for a polynomial \(f(x) = a_0 + a_1 x + a_2 x^2 + \dots + a_p x^p\) where \(a_0, a_1, a_2, \dots, a_p\) are constants, the derivative is

\(f'(x) = (a_0 + a_1 x + a_2 x^2 + \dots + a_p x^p)'\)

\(= (a_0)' + (a_1 x)' + (a_2 x^2)' + \dots + (a_p x^p)'\)

\(= 0 + a_1 + a_2 \cdot 2x + a_3 \cdot 3x^2 + \dots + a_p \cdot px^{p-1}\)

/Example: / \((5x^3 + 7x^2 - 3x + 2)' = 5 \cdot 3 x^2 + 7 \cdot 2x - 3\).

2.3.4 Exponential Rule

For any constant \(a\) such that \(a>0\) and \(a\neq 1\), \[ \frac{d}{dx} a^x = lna \cdot a^x \] \[ \frac{d}{dx} e^x = e^x \]

  • Examples
    • \(\frac{d}{dx} 3^x = ln3 \cdot 3^x\)
    • \(\frac{d}{dx} 3^{2x^3} =\) ?

      Let \(f(x) = 3^{2x^3}\), then we can see it is a composite function where

      • the inner function \(g(x) = 2x^3\), with \(\frac{dg(x)}{dx} = 6x^2\)
      • the outer function \(f(g) = 3^g\), with \(\frac{df(g)}{dg} = ln3 \cdot 3^g\) where \(g = g(x)\)

      so we should apply the Chain Rule and get

      \(\frac{df(g(x))}{dx} = \frac{df(g)}{dg} \cdot \frac{dg(x)}{dx} = ln3 \cdot 3^g \cdot 6x^2 = 6 \cdot ln3 \cdot x^2 \cdot 3^{2x^3}\)

    • \(\frac{d}{dx} e^{4x^3 + 2} =\) ?

      Let \(f(x) = e^{4x^3 + 2}\), then we can see it is a composite function where

      • the inner function \(g(x) = 4x^3 + 2\), with \(\frac{dg(x)}{dx} = 4 \cdot 3x^2 + 0 = 12 x^2\)
      • the outer function \(f(g) = e^g\), with \(\frac{df(g)}{dg} = e^g\) where \(g = g(x)\)

      so we should apply the Chain Rule and get

      \(\frac{df(g(x))}{dx} = \frac{df(g)}{dg} \cdot \frac{dg(x)}{dx} = e^g \cdot 12x^2 = e^{4x^3 + 2} 12x^2\)

    • \(\frac{d}{dx}e^2=\) ?

      Caution : \(e^2\) is a constant, so \(\frac{de^2}{dx} = e^2\) is WRONG. The exponential rule does not apply because \(2\) is not the variable \(x\). Instead, we should use the constant rule to get \(\frac{de^2}{dx} = 0\).

2.3.5 Logarithmic Rule

For any constant \(a\) such that \(a>0\) and \(a\neq 1\), \[ \frac{d}{dx} log_ax = \frac{1}{x \cdot lna} \] \[ \frac{d}{dx} lnx = \frac{1}{x} \]

  • Examples
    • \(\frac{d}{dx} log_2 x = \frac{1}{x \cdot ln2}\)
    • \(\frac{d}{dx} log_2 x^3 =\) ?

      Let \(f(x) = log_2 x^3\), then we can see it is a composite function where

      • the inner function \(g(x) = x^3\), with \(\frac{dg(x)}{dx} = 3x^2\)
      • the outer function \(f(g) = log_2 g\), with \(\frac{f(g)}{dg} = \frac{1}{g \cdot ln2}\)

      so we should apply the Chain Rule and get

      \(\frac{df(g(x))}{dx} = \frac{df(g)}{dg} \cdot \frac{dg(x)}{dx} = \frac{1}{g \cdot ln2} \cdot 3x^2 = \frac{3x^2}{x^3 \cdot ln2}\)

    • \(\frac{d}{dx} ln(3x^2 - 1)=\) ?

      Let \(f(x) = ln (3x^2 - 1)\), then we can see it is a composite function where

      • the inner function \(g(x) = 3x^2 - 1\), with \(\frac{dg(x)}{dx} = 3\cdot 2x - 0 = 6x\)
      • the outer function \(f(g) = ln(g)\), with \(\frac{f(g)}{dg} = \frac{1}{g}\)

      so we should apply the Chain Rule and get

      \(\frac{df(g(x))}{dx} = \frac{df(g)}{dg} \cdot \frac{dg(x)}{dx} = \frac{1}{g} \cdot 6x = \frac{6x}{3x^2 - 1}\)

    • \(\frac{d}{dx}ln2=\) ?

      Caution : \(ln2\) is a constant, so \(\frac{dln2}{dx} = \frac{1}{2}\) is WRONG. The logarithmic rule does not apply because \(2\) is not the variable \(x\). Instead, we should use the constant rule to get \(\frac{dln2}{dx} = 0\).

  • Remark

    Comparison between Power rule and Exponential Rule

    • Power Rule : \((x^a)' = a\cdot x^{a-1}\) requires variable \(x\) be the base and constant \(a\) be the power
    • Exponential Rule : \(a^x = lna \cdot a^x\) requires constant \(a\) \((a>0\) and \(a\neq 1)\) be the base and variable \(x\) be the power

3 Technique : Rewriting the Function

Sometimes it is wise to rewrite the function prior to applying the complicated product rule/quotient rule/etc.

3.1 Examples

3.1.1 Example 1 : Using Logarithmic Identities

For example, if \(f(x) = ln \left( \frac{e^{3x^2 + 1} \cdot (2x+5)^2 \cdot \sqrt{3x-1}}{(4x+1)^3} \right)\), then we may use the logarithmic property to rewrite it as

\(f(x) = ln \left( \frac{e^{3x^2 + 1} \cdot (2x+5)^2 \cdot \sqrt{3x-1}}{(4x+1)^3} \right)\)

\(= ln \left( \frac{e^{3x^2 + 1} \cdot (2x+5)^2 \cdot \sqrt{3x-1}}{(4x+1)^3} \right)\)

\(= ln(e^{3x^2 + 1}(2x+5)^2 \cdot \sqrt{3x-1}) - ln((4x+1)^3)\)

\(= ln(e^{3x^2 + 1}) + ln((2x+5)^2) + ln(\sqrt{3x-1}) - ln((4x+1)^3)\)

\(= (3x^2 + 1) lne + 2ln(2x+5) + \frac{1}{2} ln(3x-1) - 3ln(4x+1)\)

\(= 3x^2 + 1 + 2ln(2x+5) + \frac{1}{2} ln(3x-1) - 3ln(4x+1)\)

Now it is much easier for us to find the derivative.

Using relevant rules, we get

\(\frac{d}{dx} f(x)\)

\(= \frac{d}{dx} [3x^2 + 1 + 2ln(2x+5) + \frac{1}{2} ln(3x-1) - 3ln(4x+1)]\)

\(= 3 \frac{d}{dx} x^2 + \frac{d1}{dx} + 2 \frac{d}{dx} ln(2x+5) + \frac{1}{2} \frac{d}{dx} ln(3x-1) - 3 \frac{d}{dx} ln(4x+1)\)

\(= 3\cdot 2x + 0 + 2 \cdot \frac{2}{2x+5} + \frac{1}{2} \cdot \frac{3}{3x-1} - 3 \cdot \frac{4}{4x+1}\)

\(= 6x + \frac{4}{2x+5} + \frac{3}{2(3x-1)} - \frac{12}{4x+1}\)

3.1.2 Example 2 : Splitting Long Numerators

For example, it is not a good idea to directly use quotient rule when we find the derivative of \(f(x) = \frac{x^3 + 2x^2 + 3x + 4}{5x^2}\). Instead, we may split the long numerator and then simplify each term:

\(f(x) = \frac{x^3 + 2x^2 + 3x + 4}{5x^2}\)

\(= \frac{x^3}{5x^2} + \frac{2x^2}{5x^2} + \frac{3x}{5x^2} + \frac{4}{5x^2}\)

\(= \frac{x}{5} + \frac{2}{5} + \frac{3}{5}x^{-1} + \frac{4}{5}x^{-2}\)

In this way, we immediately get the the derivative of \(f(x)\) as

\(f'(x)\)

\(= \left( \frac{x}{5} \right)' + \left( \frac{2}{5} \right)' + \left( \frac{3}{5} x^{-1} \right)' + \left( \frac{4}{5} x^{-2} \right)'\)

\(= \frac{1}{5} + 0 + \frac{3}{5} \cdot (-1) x^{-2} + \frac{4}{5} \cdot (-2) x^{-3}\)

\(= \frac{1}{5} - \frac{3}{5} x^{-2} - \frac{8}{5} x^{-3}\)

4 High Order Derivative

\(f''(x) = [f'(x)]' =\) derivative of \(f'(x)\)

\(f'''(x) = [f''(x)]' =\) derivative of \(f''(x)\)

\(\dots\)

\(f^{(n)}(x) = [f^{(n-1)}(x)]' =\) derivative of \(f^{(n-1)}(x)\)

5 Application : Increasing/Decreasing Intervals of Function

5.1 Definition

Let \(f\) be defined on some interval. Then for any two numbers \(x_1\) and \(x_2\) in the interval,

\(f\) is increasing on the interval if \[ f(x_1) < f(x_2) \text{ whenever } x_1 < x_2, \] and \(f\) is decreasing on the interval if \[ f(x_1) > f(x_2) \text{ whenever } x_1 < x_2. \]

5.2 Critical Number and Critical Point

5.2.1 Definitions

The critical numbers for a function \(f\) are those numbers \(c\) in the domain of \(f\) for which

  • \(f'(c) = 0\) or
  • \(f'(c)\) does not exist

A critical point is a point whose \(x\)-coordinate is the critical number \(c\) and whose \(y\)-coordinate is \(f(c)\).

5.2.2 Example

For \(f(x) = \frac{x^2 + 1}{2x - 1}\), by quotient rule we have

\(f'(x)\)

\(= \left( \frac{x^2 + 1}{2x - 1} \right)'\)

\(= \frac{(x^2 + 1)'(2x - 1) - (x^2 + 1)(2x - 1)'}{(2x - 1)^2}\)

\(= \frac{2x(2x - 1) - (x^2 + 1) \cdot 2}{(2x - 1)^2}\)

\(= \frac{4x^2 - 2x - 2x^2 - 2}{(2x - 1)^2}\)

\(= \frac{2x^2 - 2x - 2}{(2x - 1)^2}\)

  • \(f'(x)\) is undefined at \(x = \frac{1}{2}\), but \(\frac{1}{2}\) is NOT a critical number of \(f(x)\) because it is NOT in the domain
  • Letting \(f'(x) = 0\) means letting \(2x^2 - 2x - 2 = 0\). By formula for quadratic roots (the roots of \(ax^2 + bx + c = 0\) are \(x = \frac{-b \pm \sqrt{4ac}}{2a}\)), we know

    \(\frac{-(-2) \pm \sqrt{(-2)^2 - 4 \cdot 2 \cdot (-2)}}{2\cdot 2} = \frac{2 \pm \sqrt{20}}{4} = \frac{2 \pm 2 \sqrt{5}}{4} = \frac{1 \pm \sqrt{5}}{2}\) are 2 critical numbers.

  • Hence, only \(\frac{1 + \sqrt{5}}{2}\) and \(\frac{1 - \sqrt{5}}{2}\) are critical numbers of \(f(x)\), and \(\left( \frac{1 + \sqrt{5}}{2}, f(\frac{1 + \sqrt{5}}{2}) \right)\) and \(\left( \frac{1 - \sqrt{5}}{2}, f(\frac{1 - \sqrt{5}}{2}) \right)\) are critical points. (Calculate \(f(\frac{1 - \sqrt{5}}{2})\) and \(f(\frac{1 + \sqrt{5}}{2})\) by yourself. The same for all following examples.)

5.3 For a Given Interval, How to Determine Whether Function \(f(x)\) is Increasing or Decreasing?

Suppose a function \(f\) has a derivative at each point in an open interval. Then

  • if \(f'(x) > 0\) for each \(x\) in the interval, \(f\) is increasing on the interval
  • if \(f'(x) < 0\) for each \(x\) in the interval, \(f\) is decreasing on the interval
  • if \(f'(x) = 0\) for each \(x\) in the interval, \(f\) is constant on the interval

5.4 For a Given Function \(f(x)\), How to Find All Intervals on Which It Is Increasing or Decreasing?

5.4.1 Procedure

  • Find all critical numbers of \(f(x)\)
  • Separate the whole domain of \(f(x)\) into several intervals by the critical numbers
  • Within each interval, choose a value of \(x\), say \(a\), and see whether \(f'(a) > 0, < 0\) or \(= 0\).
  • Decide where \(f(x)\) is increasing or decreasing on each interval
    • If \(f'(a) > 0\), then \(f(x)\) is increasing on that interval
    • If \(f'(a) < 0\), then \(f(x)\) is decreasing on that interval
    • If \(f'(a) = 0\), then \(f(x)\) is constant on that interval

5.4.2 Example

  • We have already found that the \(f(x) = \frac{x^2 + 1}{2x - 1}\) is defined on \((-\infty, \frac{1}{2}) \cup (\frac{1}{2}, \infty)\) and critical numbers of \(f(x)\) are \(\frac{1 + \sqrt{5}}{2}\) and \(\frac{1 - \sqrt{5}}{2}\).
  • These critical numbers separate the domain of \(f(x)\) into the 1st row of intervals in the following table. Now we have to fill in the 2nd and the 3rd rows.
  • Within each interval, we choose a number \(a\) and evaluate \(f'(a)\) to see its sign
  • Based on the sign of \(f'(a)\), we determine whether the function is increasing or decreasing on that interval

    \(x\) \((-\infty\), \(\frac{1 - \sqrt{5}}{2})\) \((\frac{1 - \sqrt{5}}{2}, \frac{1}{2})\) \((\frac{1}{2}, \frac{1 + \sqrt{5}}{2})\) \((\frac{1 + \sqrt{5}}{2}, \infty)\)
    \(f'(x)\) \(f'(-10) > 0\) \(f'(0) < 0\) \(f'(1) < 0\) \(f'(10) > 0\)
    \(f(x)\) increasing desreasing decreasing increasing

6 Application : Relative Extrema

6.1 Definition

Let \(c\) be a number in the domain of a function \(f\). Then \(f(c)\) is a relative (or local) maximum for \(f\) if there exists an open interval \((a, b)\) containing \(c\) such that \[ f(x) \le f(c) \] for all \(x\) in \((a, b)\).

Likewise, \(f(c)\) is a relative (or local) minimum for \(f\) if there exists an open interval \((a, b)\) containing \(c\) such that \[ f(x) \ge f(c) \] for all \(x\) in \((a, b)\).

A function has a relative (or local) extremum at \(c\) if it has either a relative maximum or a relative minimum there. If \(c\) is an endpoint of the domain of \(f\), then we only consider \(x\) in the half-open interval that is in the domain.

6.2 How to Determine Whether A Number Gives a Relative Extremum

6.2.1 First Derivative Test

Let \(c\) be a critical number for function \(f\). Suppose that \(f\) is continous on \((a, b)\) and differentiable on \((a, b)\) except possibly at \(c\), and that \(c\) is the only critical number for \(f\) in \((a, b)\). Then

  • \(f(c)\) is a relative maximum of \(f\) if \(f'(x) > 0\) in the interval \((a, c)\) and \(f'(x) < 0\) in the interval \((c, b)\)
  • \(f(c)\) is a relative minimum of \(f\) if \(f'(x) < 0\) in the interval \((a, c)\) and \(f'(x) > 0\) in the interval \((c, b)\)

6.2.2 Second Derivatitve Test

Suppose \(f''(x)\) exists on some open interval containing \(c\) except possibly at \(c\) itself, and suppose \(f'(c) = 0\).

  • If \(f''(c) > 0\), then \(f(c)\) is a relative minimum
  • If \(f''(c) < 0\), then \(f(c)\) is a relative maximum
  • If \(f''(c) = 0\) or \(f''(c)\) does not exist, then the test gives no information about extrema, so use the first derivative test. (Remark : so, if possible, always use First Derivative Test.)

6.3 How to Find All Relative Extrema of a Function

6.3.1 Procedure

  • Find all critical numbers of function \(f(x)\), and separate the domain into several intervals by these critical numbers
  • At each critical number, use First Derivative Test or Second Derivative Test to determine whether it gives a relative maximum or minimum or no information.

6.3.2 Example

  • Using First Derivative Test

    For function \(f(x) = \frac{x^2 + 1}{2x - 1}\), we already know the critical numbers are \(\frac{1 + \sqrt{5}}{2}\) and \(\frac{1 - \sqrt{5}}{2}\). Also, we have already got the following table, so we can directly apply the First Derivative Test.

    \(x\) \((-\infty\), \(\frac{1 - \sqrt{5}}{2})\) \((\frac{1 - \sqrt{5}}{2}, \frac{1}{2})\) \((\frac{1}{2}, \frac{1 + \sqrt{5}}{2})\) \((\frac{1 + \sqrt{5}}{2}, \infty)\)
    \(f'(x)\) \(f'(-10) > 0\) \(f'(0) < 0\) \(f'(1) < 0\) \(f'(10) > 0\)
    \(f(x)\) increasing desreasing decreasing increasing
    • On \((-\infty\), \(\frac{1 - \sqrt{5}}{2})\), \(f'(x) > 0\) and \(f(x)\) is increasing. On \((\frac{1 - \sqrt{5}}{2}, \frac{1}{2})\), \(f'(x) < 0\) and \(f(x)\) is decreasing. Thus \(f(\frac{1 - \sqrt{5}}{2})\) is a relative maximum.
    • On \((\frac{1}{2}, \frac{1 + \sqrt{5}}{2})\), \(f'(x) < 0\) and \(f(x)\) is decreasing. On \((\frac{1 - \sqrt{5}}{2}, \infty)\), \(f'(x) > 0\) and \(f(x)\) is increasing. Thus \(f(\frac{1 + \sqrt{5}}{2})\) is a relative minimum.
    • \(f(\frac{1}{2})\) is ignored as it is undefined.
  • Using Second Derivative Test

    For function \(f(x) = \frac{x^2 + 1}{2x - 1}\), we already know \(f'(x) = \frac{2x^2 - 2x - 2}{(2x - 1)^2}\), so the second derivative is

    \(f''(x) = [f'(x)]'\)

    \(= \left( \frac{2x^2 - 2x - 2}{(2x - 1)^2} \right)'\)

    \(= \frac{(2x^2 - 2x - 2)'(2x-1)^2 - (2x^2 - 2x - 2)[(2x-1)^2]'}{(2x - 1)^4}\)

    \(= \frac{(4x - 2)(2x-1)^2 - (2x^2 - 2x - 2)[2\cdot 2 \cdot (2x-1)]}{(2x - 1)^4}\)

    \(= \frac{(4x - 2)(2x-1)^2 - 4(2x^2 - 2x - 2) (2x-1)}{(2x - 1)^4}\)

    \(= \frac{(4x - 2)(2x-1) - 4(2x^2 - 2x - 2)}{(2x - 1)^3}\) [by cancelling \((2x-1)\)]

    \(= \frac{8x^2 - 4x - 4x + 2 - 8x^2 + 8x + 8}{(2x - 1)^3}\)

    \(= \frac{10}{(2x-1)^3}\)

    Thus we have

    • \(f'(\frac{1 - \sqrt{5}}{2}) = 0\) and \(f''(\frac{1 - \sqrt{5}}{2}) < 0\), so \(f(\frac{1 - \sqrt{5}}{2})\) is a relative maximum.
    • \(f'(\frac{1 + \sqrt{5}}{2}) = 0\) and \(f''(\frac{1 + \sqrt{5}}{2}) > 0\), so \(f(\frac{1 + \sqrt{5}}{2})\) is a relative minimum.

7 Application : Concavity of Function

7.1 Definition

  • A function is concave upward on an interval \((a, b)\) if the graph of the function lies above its tangent line at each point of \((a, b)\).
  • A function is concave downward on an interval \((a, b)\) if the graph of the function lies below its tangent line at each point of \((a, b)\).

7.2 For a Given Interval, How to Determine Whether Function \(f(x)\) Is Concave Upward or Downward?

Let \(f\) be a function with derivatives \(f'\) and \(f''\) existing at all points in an interval \((a, b)\). Then

  • \(f\) is concave upward on \((a, b)\) if \(f''(x) > 0\) for all \(x\) in \((a, b)\)
  • \(f\) is concave downward on \((a, b)\) if \(f''(x) < 0\) for all \(x\) in \((a, b)\)

7.3 For a Given Function \(f(x)\), How to Find All Intervals on Which It Is Concave Upward or Downcard?

7.3.1 Procedure

  • Find all values of \(x\) such that \(f''(x) = 0\) or \(f''(x)\) does not exist (these values are called possible inflection points, see next section)
  • Separate the whole domain of \(f(x)\) into several intervals by these numbers
  • Within each interval, choose a value of \(x\), say \(a\), and see whether \(f''(a) > 0, < 0\) or \(= 0\).
  • Decide where \(f(x)\) is concave upward or downward on each interval
    • If \(f''(a) > 0\), then \(f(x)\) is concave upward on that interval
    • If \(f''(a) < 0\), then \(f(x)\) is concave downward on that interval

7.3.2 Example

For function \(f(x) = \frac{x^2 + 1}{2x - 1}\), we already know \(f''(x) = \frac{10}{(2x - 1)^3}\).

  • Find all points where \(f''(x) = 0\) or \(f''(x)\) is undefined.
    • \(f''(x) = 0\) never holds because the numerator \(10\) can never equal \(0\).
    • \(f''(x)\) is undefined at \(x=\frac{1}{2}\)
    • Therefore, \(\frac{1}{2}\) is the only possible inflection point
  • The domain of the original function \(f(x)\) is \((-\infty,\frac{1}{2}) \cup (\frac{1}{2},\infty)\), so the possible inflection point \(x=\frac{1}{2}\) doesn't separate the domain any more. As a result, the first row of the following table contains only two intervals.
  • Calculate \(f''(x)\) at some special point within each interval.
  • Based on the sign of the second row of the table, we can determine the concavity of the function \(f(x)\) as follows.

    \(x\) \((-\infty, \frac{1}{2})\) \((\frac{1}{2}, \infty)\)
    \(f''(x)\) \(f''(0) < 0\) \(f''(1) > 0\)
    \(f(x)\) concave downward concave upward

8 Application : Point of Inflection

8.1 Definition

A point where a graph changes concavity is called an inflection point.

8.2 How to Determine Whether a Point is an Inflection Point

Suppose \(c\) is a point on interval \((a, b)\), and suppose \(f''(x)\) exists on \((a, b)\) except possibly at \(c\).

  • If \(f''(x) < 0\) on \((a, c)\) and \(f''(x) > 0\) on \((c, b)\), then \(c\) is an inflection point
  • If \(f''(x) > 0\) on \((a, c)\) and \(f''(x) < 0\) on \((c, b)\), then \(c\) is an inflection point
  • Otherwise, \(c\) is not an inflection point.

8.3 How to Find All Inflection Points of a Function

8.3.1 Procedure

  • Find all possible inflection points (candidates for further investigation). A point \(c\) is said to be a possible inflection point if \(f''(c) = 0\) or \(f''(c)\) is not defined.
  • At each possible inflection point, determine by the above method whether it is truly an inflection point.

8.3.2 Example

For function \(f(x) = \frac{x^2 + 1}{2x - 1}\), we already know \(f''(x) = \frac{10}{(2x - 1)^3}\). \(f''(x)\) is undefined at \(x=\frac{1}{2}\), and \(f''(x) = 0\) can not be achieved for any value of \(x\).

  • Thus \(\frac{1}{2}\) is the only possible inflection points.
  • At \(\frac{1}{2}\), the concavity changes, so \(\frac{1}{2}\) is an inflection point.

9 Application : More Examples

9.1 Example 1

Function \(f(x) = x^4 - x^3 - \frac{x^2}{2} + 11\).

  • First Derivative

    \(f'(x) = 4 x^3 - 3 x^2 - \frac{2x}{2} = 4 x^3 - 3 x^2 - x\)

  • second Derivative

    \(f''(x) = [f'(x)]' = (4x^3 - 3 x^2 - x)' = 12x^2 - 6x - 1\)

  • Critical Number
    • Let \(f'(x) = 0\). Then we have \(4x^3 - 3x^2 - x = x (4x^2 -3x - 1) = 0\), which means \(x=0\) or \(4x^2 -3x - 1 = 0\). Solve \(4x^2 -3x - 1 = 0\) by quadratic formula we have \(x = \frac{-1}{4}\) or \(x=1\). Hence, \(0\), \(\frac{-1}{4}\) and \(1\) are numbers such that \(f'(x) = 0\).
    • \(f'(x)\) is a polynomial and is defined everywhere, so there is no point \(c\) such that \(f'(c)\) is undefined.
    • Therefore, critical numbers of \(f(x)\) are \(0\), \(\frac{-1}{4}\) and \(1\).
  • Increasing and Decreasing Intervals
    • Separate the domain of \(f(x)\) by all critical numbers. The domain of \(f(x)\) is \((-\infty, \infty)\) and critical numbers of \(f(x)\) are \(0\), \(\frac{-1}{4}\) and \(1\), so we have the following table.
    • Evaluate \(f'(x)\) at some specific number within each interval and see the sign of \(f'(x)\).
    • By the sign of \(f'(x)\), determine whether \(f(x)\) is increasing or decreasing on that interval.

      \(x\) \((-\infty, \frac{-1}{4})\) \((\frac{-1}{4}, 0)\) \((0, 1)\) \((1, \infty)\)
      \(f'(x)\) \(f'(-10) < 0\) \(f'(-0.1) > 0\) \(f'(0.5) < 0\) \(f'(10) > 0\)
      \(f(x)\) decreasing increasing decreasing increasing
  • Relative Maximum or Minimum

    Using First Derivative Test,

    • since \(f'(x) < 0\) on \((-\infty, \frac{-1}{4})\) and \(f'(x) > 0\) on \((\frac{-1}{4}, 0)\), we know \(f(\frac{-1}{4})\) (calculate it by yourself) is a local minimum
    • similarly, \(f(0)\) is a local maximum, and
    • \(f(1)\) is a local minimum

    Alternatively, using Second Derivative Test,

    • \(f'(\frac{-1}{4}) = 0\) and \(f''(\frac{-1}{4}) = 12 \left( \frac{-1}{4} \right)^2 - 6 \cdot \left( \frac{-1}{4} \right) > 0\), so \(f(\frac{-1}{4})\) is a relative minimum
    • similarly, \(f'(0) = 0\) and \(f''(0) < 0\), so \(f(0)\) is a relative maximum
    • similarly, \(f'(1) = 0\) and \(f''(1) > 0\), so \(f(1)\) is a relative minimum
  • Possible Inflection Points

    We know \(f''(x) = 12x^2 - 6x - 1\).

    • Let \(f''(x) = 12x^2 - 6x - 1 = 0\), then we have \(x = \frac{3 + \sqrt{21}}{12}\) or \(x = \frac{3 - \sqrt{21}}{12}\).
    • \(f''(x)\) is a polynomial and is thus defined everywhere, so there exists no \(c\) such that \(f''(c)\) is undefined.
    • Therefore, \(\frac{3 + \sqrt{21}}{12}\) and \(\frac{3 - \sqrt{21}}{12}\) are the only two possible inflection points.
  • Concavity and Inflection Points
    • Separate the domain of \(f(x)\), which is \((-\infty, \infty)\), by all possible inflection points, and we get the following table
    • Evaluate \(f''(x)\) at some specific number within each interval and see the sign of \(f''(x)\).
    • By the sign of \(f''(x)\), determine whether \(f(x)\) is concave upward or concave downward on that interval.

      \(x\) \((-\infty, \frac{3 - \sqrt{21}}{12})\) \((\frac{3 - \sqrt{21}}{12}, \frac{3 + \sqrt{21}}{12})\) \((\frac{3 + \sqrt{21}}{12}, \infty)\)
      \(f''(x)\) \(f''(-10) > 0\) \(f''(0) < 0\) \(f''(10) > 0\)
      \(f(x)\) concave upward concave downward concave upward

9.2 Example 2

Function \(f(x) = e^{1-x^2}\)

  • First Derivative

    \(f'(x) = e^{1-x^2} (-2x)\)

  • second Derivative

    \(f''(x) = [f'(x)]' = (e^{1-x^2} (-2x))'\)

    \(= (e^{1-x^2})' (-2x) + e^{1-x^2} (-2x)'\)

    \(= e^{1-x^2} (-2x) (-2x) + e^{1-x^2} (-2)\)

    \(= e^{1-x^2} (4x^2 - 2)\)

    \(= 2 e^{1-x^2} (2x^2 - 1)\)

  • Critical Number
    • Let \(f'(x) = 0\). Then we have \(e^{1-x^2} (-2x) = 0\), which means \(e^{1-x^2} = 0\) or \(-2x = 0\). We know \(e^{1-x^2} = 0\) has no solution since \(e^{1-x^2} > 0\) always holds, while \(-2x = 0\) implies \(x=0\).
    • \(y = e^{1-x^2}\) is defined everywhere and \(y = -2x\) is defined everywhere, so \(f'(x)\) is defined everywhere and there is no point \(c\) such that \(f'(c)\) is undefined.
    • Therefore, \(x=0\) is the only critical numbers for \(f(x)\).
  • Increasing and Decreasing Intervals
    • Separate the domain of \(f(x)\) by all critical numbers. The domain of \(f(x)\) is \((-\infty, \infty)\) and critical numbers of \(f(x)\) is \(0\), so we have the following table.
    • Evaluate \(f'(x)\) at some specific number within each interval and see the sign of \(f'(x)\).
    • By the sign of \(f'(x)\), determine whether \(f(x)\) is increasing or decreasing on that interval.

      \(x\) \((-\infty, 0)\) \((0, \infty)\)
      \(f'(x)\) \(f'(-1) > 0\) \(f'(1) < 0\)
      \(f(x)\) increasing decreasing
  • Relative Maximum or Minimum

    Using First Derivative Test,

    • since \(f'(x) > 0\) on \((-\infty, 0)\) and \(f'(x) < 0\) on \((0, \infty)\), we know \(f(0) = e\) is a local maximum

    Alternatively, using Second Derivative Test,

    • \(f'(0) = 0\) and \(f''(0) = - 2 e < 0\), so \(f(0) = e\) is a relative maximum
  • Possible Inflection Points

    We know \(f''(x) = 2 e^{1-x^2} (2x^2 - 1)\).

    • Let \(f''(x) = 2 e^{1-x^2} (2x^2 - 1) = 0\), then we have \(e^{1-x^2} = 0\) (no solution) or \(2x^2 - 1 = 0\) (which implies \(x=\frac{\sqrt{2}}{2}\) or \(x=\frac{-\sqrt{2}}{2}\))
    • \(y = e^{1-x^2}\) is defined everywhere and \(y = 2x^2 - 1\) is define everywhere, so \(f''(x)\) is defined everywhere, and thus there exists no \(c\) such that \(f''(c)\) is undefined.
    • Therefore, \(x=\frac{\sqrt{2}}{2}\) and \(x=\frac{-\sqrt{2}}{2}\) are the only two possible inflection points.
  • Concavity and Inflection Points
    • Separate the domain of \(f(x)\), which is \((-\infty, \infty)\), by all possible inflection points, and we get the following table
    • Evaluate \(f''(x)\) at some specific number within each interval and see the sign of \(f''(x)\).
    • By the sign of \(f''(x)\), determine whether \(f(x)\) is concave upward or concave downward on that interval.

      \(x\) \((-\infty, \frac{-\sqrt{2}}{2})\) \((\frac{-\sqrt{2}}{2}, \frac{\sqrt{2}}{2})\) \((\frac{\sqrt{2}}{2}, \infty)\)
      \(f''(x)\) \(f''(-10) > 0\) \(f''(0) < 0\) \(f''(10) > 0\)
      \(f(x)\) concave upward concave downward concave upward

10 References

  1. Lial, M. L., Greenwell, R. N., Ritchey, N. P. (2011). Calculus with Applications (10th Edition). Boston, MA : Pearson.

Yunfei Wang

2015-11-04 Wed 21:58