Logarithmic differentiation

$\newenvironment {prompt}{}{} \newcommand {\accordion }[0]{} \newcommand {\ungraded }[0]{} \newcommand {\xmDefaultPreamble }[0]{xmPreamble.tex} \newcommand {\xmDefaultPrintstyle }[0]{xmPrintstyle.sty} \newcommand {\todo }[0]{} \newcommand {\oiint }[0]{{\large \bigcirc }\kern -1.56em\iint } \newcommand {\mooculus }[0]{\textsf {\textbf {MOOC}\textnormal {\textsf {ULUS}}}} \newcommand {\tkzTabSlope }[1]{\foreach \x /\y /\z in {#1}{\node [left = 3pt] at (Z\x 1) {\scriptsize $\y $}; \node [right = 3pt] at (Z\x 1) {\scriptsize $\z $}; }} \newcommand {\RR }[0]{\mathbb R} \newcommand {\R }[0]{\mathbb R} \newcommand {\N }[0]{\mathbb N} \newcommand {\Z }[0]{\mathbb Z} \newcommand {\sagemath }[0]{\textsf {SageMath}} \newcommand {\d }[0]{\,d} \newcommand {\l }[0]{\ell } \newcommand {\ddx }[0]{\frac {d}{\d x}} \newcommand {\zeroOverZero }[0]{\ensuremath {\boldsymbol {\tfrac {0}{0}}}} \newcommand {\inftyOverInfty }[0]{\ensuremath {\boldsymbol {\tfrac {\infty }{\infty }}}} \newcommand {\zeroOverInfty }[0]{\ensuremath {\boldsymbol {\tfrac {0}{\infty }}}} \newcommand {\zeroTimesInfty }[0]{\ensuremath {\small \boldsymbol {0\cdot \infty }}} \newcommand {\inftyMinusInfty }[0]{\ensuremath {\small \boldsymbol {\infty - \infty }}} \newcommand {\oneToInfty }[0]{\ensuremath {\boldsymbol {1^\infty }}} \newcommand {\zeroToZero }[0]{\ensuremath {\boldsymbol {0^0}}} \newcommand {\inftyToZero }[0]{\ensuremath {\boldsymbol {\infty ^0}}} \newcommand {\numOverZero }[0]{\ensuremath {\boldsymbol {\tfrac {\#}{0}}}} \newcommand {\dfn }[0]{\textbf } \newcommand {\unit }[0]{\mathop {}\!\mathrm } \newcommand {\eval }[1]{\bigg [ #1 \bigg ]} \newcommand {\seq }[1]{\left ( #1 \right )} \newcommand {\epsilon }[0]{\varepsilon } \newcommand {\phi }[0]{\varphi } \newcommand {\iff }[0]{\Leftrightarrow } \DeclareMathOperator {\arccot }{arccot} \DeclareMathOperator {\arcsec }{arcsec} \DeclareMathOperator {\arccsc }{arccsc} \DeclareMathOperator {\si }{Si} \DeclareMathOperator {\scal }{scal} \DeclareMathOperator {\sign }{sign} \newcommand {\arrowvec }[1]{{\overset {\rightharpoonup }{#1}}} \newcommand {\vec }[1]{{\overset {\boldsymbol {\rightharpoonup }}{\mathbf {#1}}}\hspace {0in}} \newcommand {\point }[1]{\left (#1\right )} \newcommand {\pt }[1]{\mathbf {#1}} \newcommand {\Lim }[2]{\lim _{\point {#1} \to \point {#2}}} \DeclareMathOperator {\proj }{\mathbf {proj}} \newcommand {\veci }[0]{{\boldsymbol {\hat {\imath }}}} \newcommand {\vecj }[0]{{\boldsymbol {\hat {\jmath }}}} \newcommand {\veck }[0]{{\boldsymbol {\hat {k}}}} \newcommand {\vecl }[0]{\vec {\boldsymbol {\l }}} \newcommand {\uvec }[1]{\mathbf {\hat {#1}}} \newcommand {\utan }[0]{\mathbf {\hat {t}}} \newcommand {\unormal }[0]{\mathbf {\hat {n}}} \newcommand {\ubinormal }[0]{\mathbf {\hat {b}}} \newcommand {\dotp }[0]{\bullet } \newcommand {\cross }[0]{\boldsymbol \times } \newcommand {\grad }[0]{\boldsymbol \nabla } \newcommand {\divergence }[0]{\grad \dotp } \newcommand {\curl }[0]{\grad \cross } \newcommand {\lto }[0]{\mathop {\longrightarrow \,}\limits } \newcommand {\bar }[0]{\overline } \newcommand {\surfaceColor }[0]{violet} \newcommand {\surfaceColorTwo }[0]{redyellow} \newcommand {\sliceColor }[0]{greenyellow} \newcommand {\vector }[1]{\left \langle #1\right \rangle } \newcommand {\sectionOutcomes }[0]{} \newcommand {\luastring }[1]{"\luatexluaescapestring {#1}"} \newcommand {\luastringO }[1]{\luastring {\unexpanded \expandafter {#1}}} \newcommand {\luastringN }[1]{\luastring {\unexpanded {#1}}} \newcommand {\RestoreMathJaxEnvironment }[1]{\expandafter \let \csname #1\expandafter \endcsname \csname mathjax-#1\endcsname \expandafter \let \csname end#1\expandafter \endcsname \csname mathjax-end#1\endcsname } \newcommand {\DeclareMicrotypeVariants }[1]{} \newcommand {\DeclareMicrotypeAlias }[2]{} \newcommand {\LoadMicrotypeFile }[1]{} \newcommand {\DeclareMicrotypeFilePrefix }[1]{} \newcommand {\DeclareMicrotypeBabelHook }[2]{} \newcommand {\microtypesetup }[1]{} \newcommand {\microtypecontext }[1]{} \newcommand {\textmicrotypecontext }[2]{#2} \newcommand {\leftprotrusion }[1]{#1} \newcommand {\rightprotrusion }[1]{#1} \newcommand {\noprotrusionifhmode }[0]{} \newcommand {\lsstyle }[0]{} \newcommand {\lslig }[1]{#1} \newcommand {\maketitle }[0]{} \newcommand {\problem }[0]{ } \newcommand {\exercise }[0]{ } \newcommand {\question }[0]{ } \newcommand {\exploration }[0]{ } \newcommand {\hint }[0]{ } \newcommand {\ConfigureTheoremEnv }[1]{\renewenvironment {#1}[1][]{\refstepcounter {problem}\ifthenelse {\equal {###1}{}}{}{\HCode {<span class="theorem-like-title">}###1\HCode {</span>}}}{} \ConfigureEnv {#1}{\stepcounter {identification}\ifvmode \IgnorePar \fi \EndP \HCode {<div class="theorem-like problem-environment #1" id="problem\arabic {identification}" titletext=" \GetTranslation {#1}">}}{\ifvmode \IgnorePar \fi \EndP \HCode {</div>}\IgnoreIndent }{}{}} \newcommand {\dialogue }[0]{\begin {description}} \newcommand {\foldable }[0]{\stepcounter {identification}\ifvmode \IgnorePar \fi \EndP \HCode {<div id="foldable\arabic {identification}" class="foldable">}} \newcommand {\expandable }[0]{\stepcounter {identification}\ifvmode \IgnorePar \fi \EndP \HCode {<div data-original="expandable" id="foldable\arabic {identification}" class="foldable">}} \newcommand {\unfoldable }[1]{\HCode {<span class="unfoldable">}#1\HCode {</span>}} \newcommand {\selectAll }[0]{\refstepcounter {problem}} \newcommand {\freeResponse }[0]{\refstepcounter {problem}} \newcommand {\javascript }[0]{\NoFonts } \newcommand {\sageCell }[0]{\NoFonts } \newcommand {\sageOutput }[0]{\NoFonts } \newcommand {\sagesilent }[0]{\NoFonts } \newcommand {\ungraded }[0]{\ifvmode \IgnorePar \fi \EndP \HCode {<div class="ungraded">}\IgnoreIndent } \newcommand {\accordion }[0]{\ifvmode \IgnorePar \fi \EndP \HCode {<div class="accordion">} } \newcommand {\paragraph }[1]{\HCode {<span class="paragraphHead">}#1\HCode {</span>}\par \IgnorePar } \newcommand {\subparagraph }[1]{\HCode {<span class="subparagraphHead">}#1\HCode {</span>}\par \IgnorePar } \newcommand {\outcome }[1]{\HCode {<span class="learning-outcome">#1</span>} } \newcommand {\javascriptCode }[0]{\NoFonts } \newcommand {\GetTranslation }[1]{#1} \newcommand {\d }[0]{\,d} \newcommand {\ref }[1]{\HCode {<a class="reference" href="\##1">#1</a>}}$

We use logarithms to help us differentiate.

Logarithms were originally developed as a computational tool, mainly due to their famous property:

\[ \log _b(xy) = \log _b(x)+\log _b(y). \]

Before the days of calculators and computers, this was critical knowledge for anyone in a computational discipline.

Compute $138\cdot 23.4$ using logarithms.

Start by writing both numbers in scientific notation

\[ \left (1.38\cdot 10^{2}\right )\cdot \left (2.34 \cdot 10^{1}\right ). \]

Next we use a log-table, which gives $\log _{10}(N)$ for values of $N$ ranging between $0$ and $9$. We’ve reproduced part of such a table below.

From the table, we see that

\[ \log _{10}(1.38) \approx 0.1399\qquad \text {and}\qquad \log _{10}(2.34)\approx 0.3692 \]

Add these numbers together to get $0.5091$. Essentially, we know the following at this point:

[Picture]

Using the table again, we see that $\log _{10}(3.23)\approx 0.5091$. Since we were working in scientific notation, we need to multiply this by $10^3$. Our final answer is

\[ 3230 \approx 138\cdot 23.4 \]

Since $138\cdot 23.4 = 3229.2$, this is a good approximation.

The moral is:

Logarithms allow us to use addition in place of multiplication.

1 Logarithmic differentiation

When taking derivatives, both the product rule and the quotient rule can be cumbersome to use. Logarithms will save the day. A key point is the following

\[ \ddx \ln (f(x)) = \frac {1}{f(x)}\cdot f'(x) = \frac {f'(x)}{f(x)} \]

which follows from the chain rule. Let’s look at an illustrative example to see how this is actually used.

Compute:

\[ \ddx \frac {x^9e^{4x}}{\sqrt {x-4}} \]

Recall the properties of logarithms:

$\log _b(xy) = \log _b(x) + \log _b(y)$
$\log _b(x/y) = \log _b(x) - \log _b(y)$
$\log _b(x^y) = y\log _b(x)$

While we could use the product and quotient rule to solve this problem, it would be tedious. Start by taking the logarithm of the function to be differentiated.

\begin{align*} \ln \left (\frac {x^9e^{4x}}{\sqrt {x-4}} \right ) &= \ln \left (\answer [given]{x^9e^{4x}}\right ) - \ln \left (\answer [given]{\sqrt {x-4}}\right )\\ &= \ln \left (x^9\right )+\ln \left (e^{4x}\right ) - \ln \left ((x-4)^{1/2}\right )\\ &= \answer [given]{9}\ln (x)+4x - \answer [given]{\frac {1}{2}}\ln (x-4). \end{align*}

Setting $f(x) = \frac {x^9e^{4x}}{\sqrt {x-4}}$, we can write

\[ \ln (f(x)) = 9\ln (x)+4x - \frac {1}{2}\ln (x-4).\]

Differentiating both sides, we find

\[ \frac {f'(x)}{f(x)} = \answer [given]{\frac {9}{x}+4} - \frac {1}{2(x-4)}.\]

Finally we solve for $f'(x)$, write

\[f'(x) = \left (\frac {9}{x}+4 - \frac {1}{2(x-4)}\right )\left (\answer [given]{\frac {x^9e^{4x}}{\sqrt {x-4}}}\right ). \]

The process above is called logarithmic differentiation. It starts by taking the natural logarithm of the function, then using the properties of logarithms to simplify. Logarithmic differentiation allows us to compute new derivatives, too.

For $x>0$, compute:

\[ \ddx x^x \]

The function $x^x$ is tricky to differentiate. We cannot use the power rule, as the exponent is not a constant; the function is not an exponential function either, since the base is not a constant. However, if we set $f(x) = x^x$ we can write

\begin{align*} \ln (f(x)) &= \ln \left (x^x\right )\\ &=x\ln (x). \end{align*}

Differentiating both sides, we find

\[ \frac {f'(x)}{f(x)} = \answer [given]{1 + \ln (x)}. \]

Now we can solve for $f'(x)$,

\[ f'(x) = x^x + x^x\ln (x). \]

Alternative way: We can write

\[ x^x= e^{\ln (x^x)}=e^{x\ln (x)}. \]

Therefore,

\[ \ddx x^x=\ddx e^{x\ln (x)}=e^{x\ln (x)}\left (\ln (x)+1\right ), \]

by the Chain Rule and the Product Rule.

Now, it follows that

\[ \ddx x^x=x^x\left (\answer [given]{\ln (x)+1}\right ). \]

The function from the previous example $x^x$ is a type of function that we call a tower function. Tower functions are functions written as $f(x)^{g(x)}$, an exponential where both the base and exponent depend on the variable. This logarithmic differentiation is one method of differentiating a tower function. However we can also write

\[ f(x)^{g(x)} = e^{\ln \left ( f(x)^{g(x)} \right )} = e^{g(x)\ln ( f(x) ) }. \]

By rewriting like this we can differentiate tower functions using chain rule and product rule. In the previous example, the derivative of $x^x$ was calculated in both of these methods. Let’s differentiate one more tower functions using both methods.

Calculate the derivative of $f(x) = x^{\cos (x)}$ in two ways: logarithmic differentiation and using chain rule and product rule.

We start with the logarithmic differentiation method.

\begin{align*} \ln ( f(x) ) &= \ln \left ( x^{\cos (x)} \right ) \\ &= \cos (x) \ln (x) \end{align*}

Differentiating both sides gives:

\begin{align*} \dfrac {f'(x)}{f(x)} &= -\sin (x)\ln (x) + \dfrac {\cos (x)}{x} \\ f'(x) &= f(x)\left ( -\sin (x)\ln (x) + \dfrac {\cos (x)}{x}\right )\\ &= x^{\cos (x)} \left ( -\sin (x)\ln (x) + \dfrac {\cos (x)}{x}\right ) \end{align*}

Let’s differentiate $f$ one more time, but using chain rule and product rule. We begin by rewriting the tower function as an exponential with constant base $e$.

\[ f(x) = x^{\cos (x)} = e^{\ln \left (x^{\cos (x)} \right )} = e^{\cos (x) \ln (x)} \]

Using chain rule we find:

\begin{align*} f'(x) &= e^{\cos (x)\ln (x)}\left ( \cos (x)\ln (x)\right )'\\ &= e^{\cos (x)\ln (x)}\left ( -\sin (x) \ln (x) + \dfrac {\cos (x)}{x}\right ) \end{align*}

Rewriting $e^{\cos (x)\ln (x)}$ as $x^{\cos (x)}$, we see that both methods gave the same derivative $f'$.

Compute:

\[ \ddx \ln {(|x|)} \]

The function $|x|$ is a piecewise defined function. If $x>0$

\begin{align*} \ddx \ln {(|x|)} &= \ddx \ln {(\answer [given]{x})}\\ &=\frac {1}{\answer [given]{x}}. \end{align*}

The function $\ln {(|x|)}$ is not defined at $x=0$. If $x<0$

\begin{align*} \ddx \ln {(|x|)} &= \ddx \ln {(\answer [given]{-x})}\\ &=\frac {-1}{\answer [given]{-x}}\\ &=\answer [given]{\frac {1}{x}}. \end{align*}

The next example will be useful when we want to use logarithmic differentiations for functions that assume negative values.

Compute:

\[ \ddx \ln {(|f(x)|)} \]

By the previous example and the Chain Rule

\begin{align*} \ddx \ln {(|f(x)|)} &=\answer [given]{\frac {1}{f(x)}}\cdot f'(x).\\ &=\frac {f'(x)}{\answer [given]{f(x)}} \end{align*}

Compute:

\[ \ddx \left ( 6\sin ^2(x) + 3(x^2+1)^x(\cos (2x) + 3)^5\right ) \]

Notice the function here is the sum of two terms. The logarithm properties listed above show that logarithms of products can be simplified, but logarithms of sums cannot be. Instead of attempting logarithmic differentiation, let us use the Sum Rule for derivatives to take the derivative of each term separately.

Call $f(x) = 6\sin ^2(x) + 3(x^2+1)^x(\cos (2x) + 3)^5$, and the individual terms $g(x) = 6\sin ^2(x)$ and $h(x) = 3(x^2+1)^x(\cos (2x) + 3)^5$. That means $f(x) = g(x) + h(x)$. The Sum Rule says $\ddx f(x) = \ddx g(x) + \ddx h(x)$, so we differentiate each term separately.

To calculate $g'(x)$ we use Chain Rule:

\[ g'(x) = 12\sin (x)\cos (x). \]

To calculate $h'(x)$ we use logarithmic differentiation (since $h(x)$ is a product of factors each of which are raised to powers, logarithms will help simplify this).

\begin{align*} \ln ( h(x) ) &= \ln \left ( 3(x^2+1)^x(\cos (2x) + 3)^5 \right ) \\ &= \ln \left ( 3\right ) + \ln \left ((x^2+1)^x\right ) + \ln \left ( (\cos (2x) + 3)^5 \right ) \\ &= \ln \left ( 3\right ) + x \ln (x^2+1) + 5 \ln (\cos (2x) + 3) \end{align*}

Then

\[ \ddx \ln ( h(x) ) = \ddx \left ( \ln \left ( 3\right ) + x \ln (x^2+1) + 5 \ln (\cos (2x) + 3) \right ) \]

which we calculate as:

\begin{align*} \dfrac {h'(x)}{h(x)} &= 0 + \left ( 1\cdot \ln (x^2+1) + x\cdot \dfrac {2x}{x^2+1}\right ) + 5 \cdot \dfrac {-2\sin (2x)}{\cos (2x)+3}\\ &= \ln (x^2+1) + \dfrac {2x^2}{x^2+1} -10 \dfrac {\sin (2x)}{\cos (2x)+3}\\ h'(x) &= h(x)\left (\ln (x^2+1) + \dfrac {2x^2}{x^2+1} -10 \dfrac {\sin (2x)}{\cos (2x)+3}\right )\\ &= \left (3(x^2+1)^x(\cos (2x) + 3)^5\right )\left (\ln (x^2+1) + \dfrac {2x^2}{x^2+1} -10 \dfrac {\sin (2x)}{\cos (2x)+3}\right ) \end{align*}

The derivative of the entire function is $f'(x) = g'(x) + h'(x)$:

\begin{align*} &12\sin (x)\cos (x) + \\ &\left (3(x^2+1)^x(\cos (2x) + 3)^5\right )\left (\ln (x^2+1) + \dfrac {2x^2}{x^2+1} -10 \dfrac {\sin (2x)}{\cos (2x)+3}\right ). \end{align*}

2 Proof of the power rule

Finally, recall that previously we only proved the power rule for positive integer exponents. Now we’ll use logarithmic differentiation to give a proof for all real-valued exponents. We restate the power rule for convenience sake:

Power Rule For any real number $n$,

\[ \ddx x^n = n x^{n-1}. \]

We will use logarithmic differentiation. Set $f(x) = x^n$. Write

\begin{align*} \ln (|f(x)|) &= \ln \left (|x|^n\right ) , x\ne 0\\ &= n\ln (|x|). \end{align*}

Now differentiate both sides, and solve for $f'(x)$

\begin{align*} \frac {f'(x)}{f(x)} &= \frac {n}{x}\\ f'(x) &=\frac {n f(x)}{x}\\ &= \answer [given]{n x^{n-1}}. \end{align*}

Whenever $x=0$ is in the domain of $f$, the definition of the derivative at $x=0$ implies that $f'(0)=0$, if $n\ne 1$ and $f'(0)=1$, if $n=1$ , so the power rule applies to $x=0$, too.

Thus we see that the power rule holds for all real-valued (nonzero) exponents.

While logarithmic differentiation might seem strange and new at first, with a little practice it will seem much more natural to you.