search
Search
Login
Unlock 100+ guides
menu
menu
web
search toc
close
Comments
Log in or sign up
Cancel
Post
account_circle
Profile
exit_to_app
Sign out
What does this mean?
Why is this true?
Give me some examples!
search
keyboard_voice
close
Searching Tips
Search for a recipe:
"Creating a table in MySQL"
Search for an API documentation: "@append"
Search for code: "!dataframe"
Apply a tag filter: "#python"
Useful Shortcuts
/ to open search panel
Esc to close search panel
to navigate between search results
d to clear all current filters
Enter to expand content preview
icon_star
Doc Search
icon_star
Code Search Beta
SORRY NOTHING FOUND!
mic
Start speaking...
Voice search is only supported in Safari and Chrome.
Navigate to

Properties of derivatives

schedule Aug 12, 2023
Last updated
local_offer
Calculus
Tags
mode_heat
Master the mathematics behind data science with 100+ top-tier guides
Start your free 7-days trial now!

Almost all properties of derivatives can be derived from the formal definition of derivativeslink as well as the properties of limits.

$$\begin{equation}\label{eq:VygyD4YWBzdgmccCkPN} \begin{aligned} f'(x)&= \lim_{\Delta{x}\to0}\frac{f(x+\Delta{x})-f(x)}{\Delta{x}}\\ \end{aligned} \end{equation}$$
Theorem.

Derivative of a constant

If $k\in\mathbb{R}$ is a constant, then:

$$\frac{d}{dx}k=0$$

Proof. We define a function $f(x)=k$. Let's now use the formal definition of derivatives:

$$\begin{align*} \frac{d}{dx}k&= \frac{d}{dx}f(x)\\ &=\lim_{\Delta{x}\to0}\frac{f(x+\Delta{x})-f(x)}{\Delta{x}}\\ &=\lim_{\Delta{x}\to0}\frac{k-k}{\Delta{x}}\\ &=\lim_{\Delta{x}\to0}0\\ &=0\\ \end{align*}$$

Here, we used the fact that $f(x+\Delta{x})=k$ for the third equality. This completes the proof.

Theorem.

Derivative of a constant times a function

If $k\in\mathbb{R}$ is a constant, then:

$$\frac{d}{dx}\Big(k\cdot{f(x)}\Big) = k\cdot\frac{d}{dx}f(x)$$

Proof. We define a new function:

$$g(x)=k\cdot{f(x)}$$

We then begin with the formal definition of derivatives:

$$\begin{align*} \frac{d}{dx}\Big(k\cdot{f(x)}\Big)&=\frac{d}{dx}g(x)\\ &=\lim_{\Delta{x}\to0}\frac{g(x+\Delta{x})-g(x)}{\Delta{x}}\\ &=\lim_{\Delta{x}\to0}\frac{k\cdot{f(x+\Delta{x})}-k\cdot{f(x)}}{\Delta{x}}\\ &=\lim_{\Delta{x}\to0}k\cdot \Big(\frac{{f(x+\Delta{x})}-{f(x)}}{\Delta{x}}\Big)\\ &=k\cdot\lim_{\Delta{x}\to0}\Big(\frac{{f(x+\Delta{x})}-{f(x)}}{\Delta{x}}\Big)\\ &=k\cdot\frac{d}{dx}f(x)\\ \end{align*}$$

Here, we have taken out the constant term $k$ from the limit using the scalar product rule of limitslink. This completes the proof.

Theorem.

Sum rule

If $f(x)$ and $g(x)$ are differentiable functions, then:

$$\frac{d}{dx} \Big(f(x)+g(x)\Big)= \frac{d}{dx}f(x)+ \frac{d}{dx}g(x) $$

Proof. Let define a new function $z(x)$ like follows:

$$z(x)=f(x)+g(x)$$

We use the formal definition of derivatives:

$$\begin{align*} \frac{d}{dx} \Big(f(x)+g(x)\Big)&= \frac{d}{dx}z(x)\\ &=\lim_{\Delta{x}\to0}\frac{z(x+\Delta{x})-z(x)}{\Delta{x}}\\ &=\lim_{\Delta{x}\to0}\frac{(f(x+\Delta{x})+g(x+\Delta{x}))-(f(x)+g(x))}{\Delta{x}}\\ &=\lim_{\Delta{x}\to0}\frac{(f(x+\Delta{x})-f(x))+(g(x+\Delta{x})-g(x))}{\Delta{x}}\\ &=\lim_{\Delta{x}\to0}\frac{f(x+\Delta{x})-f(x)}{\Delta{x}}+ \lim_{\Delta{x}\to0}\frac{g(x+\Delta{x})-g(x)}{\Delta{x}}\\ &=\frac{d}{dx}f(x)+\frac{d}{dx}g(x) \end{align*}$$

This completes the proof.

Theorem.

Chain rule

Let $f(x)$ and $g(x)$ be both differentiable functions and let $F(x)$ be the composite function:

$$F(x)=(f\circ{g})(x)=f(g(x))$$

The derivative of $F(x)$ is:

$$F'(x)=f'(g(x))\cdot{g'(x)}$$

If we let $y=f(g(x))$ and $u=g(x)$, then the chain rule in Leibniz's notation is:

$$\frac{dy}{dx}= \frac{dy}{du}\cdot \frac{du}{dx}$$

Proof. Let $y=f(g(x))$. By definition of derivatives, we know that the derivative of $y$ with respect to $x$ is:

$$\frac{dy}{dx}=\lim_{\Delta{x}\to0} \frac{f(g(x+\Delta{x}))-f(g(x))}{\Delta{x}}$$

Let's now perform some algebraic manipulation:

$$\begin{equation}\label{eq:EjqqDXU29n2jd594FrM} \begin{aligned}[b] \frac{dy}{dx}&=\lim_{\Delta{x}\to0} \Big( \frac{f(g(x+\Delta{x}))-f(g(x))}{\Delta{x}} \cdot \frac{g(x+\Delta{x})-g(x)}{g(x+\Delta{x})-g(x)} \Big)\\ &= \lim_{\Delta{x}\to0} \Big( \frac{f(g(x+\Delta{x}))-f(g(x))}{g(x+\Delta{x})-g(x)} \cdot \frac{g(x+\Delta{x})-g(x)}{\Delta{x}} \Big) \end{aligned} \end{equation}$$

Here, we are assuming $g(x+\Delta{x})-g(x)$ does not equal zero - otherwise we will be dividing by zero. Most proofs of the chain rule silently make this assumption, but this is in fact a flawed assumption because $g(x+\Delta{x})-g(x)$ may actually equal zero for certain functions. For instance, consider the case when $g(x)$ is a flat horizontal line. Adding some infinitesimally small amount $\Delta{x}$ to the input $x$ clearly does not change the output, and so $g(x+\Delta{x})-g(x)=0$ in this case.

We therefore have to consider two separate cases:

  • when $g(x+\Delta{x})-g(x)=0$.

  • when $g(x+\Delta{x})-g(x)\ne0$.

For the first case, we have that $g(x+\Delta{x})=g(x)$. Therefore, we have that:

$$f(g(x))=f(g(x+\Delta{x}))$$

This means that an infinitesimally small change of $\Delta{x}$ to $x$ will not affect the output at all. In other words, the slope of $f$ is flat, that is, the rate of change of $f$ is always equal to zero:

$$\begin{equation}\label{eq:X3WMn3xuZMdfCxegruw} (f\circ{g})'(x)=0 \end{equation}$$

Similarly, we have that:

$$g(x+\Delta{x})=g(x)$$

Again, since an infinitesimally small change in $x$ does not affect the output, the slope of $g$ must also equal zero. This means that the derivative of $g$ with respect to $x$ must be equal to zero:

$$\begin{equation}\label{eq:GFnBMGxZLVkFQ1qokNp} g'(x)=0 \end{equation}$$

From \eqref{eq:X3WMn3xuZMdfCxegruw} and \eqref{eq:GFnBMGxZLVkFQ1qokNp}, we can clearly see that the Chain rule trivially holds:

$$(f\circ{g})'(x)=f'(g(x))\cdot{g'(x)}$$

Let's now consider the second case when $g(x+\Delta{x})\ne{g(x)}$. By the product rule of limitslink, we can distribute the limit in \eqref{eq:EjqqDXU29n2jd594FrM} like so:

$$\begin{equation}\label{eq:KFof5YOVEjfm0e8U5LZ} \frac{dy}{dx} = \lim_{\Delta{x}\to0} \Big( \frac{f(g(x+\Delta{x}))-f(g(x))}{g(x+\Delta{x})-g(x)} \Big)\cdot \lim_{\Delta{x}\to0} \Big(\frac{g(x+\Delta{x})-g(x)}{\Delta{x}} \Big) \end{equation}$$

Notice how the right limit is simply the definition of the derivative of $g$ with respect to $x$, that is:

$$\begin{equation}\label{eq:bKoAZNqIo5r7chSldqk} \lim_{\Delta{x}\to0} \Big(\frac{g(x+\Delta{x})-g(x)}{\Delta{x}} \Big)=g'(x) \end{equation}$$

Let's now examine the left limit in \eqref{eq:KFof5YOVEjfm0e8U5LZ}:

$$\begin{equation}\label{eq:aApjqRePT9fJreejxoV} \lim_{\Delta{x}\to0} \Big( \frac{f(g(x+\Delta{x}))-f(g(x))}{g(x+\Delta{x})-g(x)} \Big) \end{equation}$$

Let's define a new variable $k$ to represent the denominator, that is:

$$\begin{equation}\label{eq:FbkqJwantXmoE193iYc} k=g(x+\Delta{x})-g(x) \end{equation}$$

We can see that as $\Delta{x}$ tends to zero, $k$ would also tend to zero since $g(x+\Delta{x})\approx{g(x)}$. Next, let's rearrange \eqref{eq:FbkqJwantXmoE193iYc} like so:

$$\begin{equation}\label{eq:dPXTyGR83FGoV9YIfbl} g(x+\Delta{x})=k+g(x) \end{equation}$$

We can now express the limit \eqref{eq:aApjqRePT9fJreejxoV} using $k$ instead of $\Delta{x}$ like so:

$$\begin{equation}\label{eq:ZxLSyYqUzOjZ8cM9B0U} \lim_{\Delta{x}\to0} \Big( \frac{f(g(x+\Delta{x}))-f(g(x))}{g(x+\Delta{x})-g(x)} \Big)= \lim_{k\to0} \Big( \frac{f(k+g(x))-f(g(x))}{k} \Big) \end{equation}$$

The key here is to notice that \eqref{eq:ZxLSyYqUzOjZ8cM9B0U} is the definition of the derivative of $f(g(x))$. To make this clear, let's define a new variable $u$ such that $u=g(x)$. Therefore, \eqref{eq:ZxLSyYqUzOjZ8cM9B0U} becomes:

$$\begin{equation}\label{eq:RcHty25PuyXEuylAV6j} \lim_{k\to0} \Big( \frac{f(k+u)-f(u)}{k} \Big) = f'(u) = f'(g(x)) \end{equation}$$

Now, substituting the expression for the two limits in \eqref{eq:KFof5YOVEjfm0e8U5LZ} gives us the chain rule:

$$\begin{equation}\label{eq:kSDVuhwt1kNE7mY5ZaQ} \frac{dy}{dx} = f'(g(x))\cdot{g'(x)} \end{equation}$$

Finally, let's also express the chain rule using Leibniz's notation. Recall that $y=f(g(x))$ and suppose we let $u=g(x)$ again. Let's take the derivative of $y=f(g(x))=f(u)$ with respect to $u$ to get:

$$\begin{equation}\label{eq:nw5ccQkWTBhasxKCEZV} \frac{dy}{du}=f'(u)=f'(g(x)) \end{equation}$$

Next, we take the derivative of both sides of $u=g(x)$ with respect to $x$ to get:

$$\begin{equation}\label{eq:cXz2kXT5P4Qdc5KuDR2} \frac{du}{dx}=g'(x) \end{equation}$$

Substituting \eqref{eq:nw5ccQkWTBhasxKCEZV} and \eqref{eq:cXz2kXT5P4Qdc5KuDR2} into \eqref{eq:kSDVuhwt1kNE7mY5ZaQ} gives us the chain rule in Leibniz's notation:

$$\frac{dy}{dx} = \frac{dy}{du} \cdot \frac{du}{dx}$$

This completes the proof of the chain rule.

Theorem.

Derivative of natural log

The derivative of natural log is its reciprocal:

$$\frac{d}{dx}\ln(x)=\frac{1}{x}$$

Solution. We begin with the formal definition of derivatives and use basic properties of logarithm to simplify:

$$\begin{equation}\label{eq:fTAFYivb5a5WjPszK3z} \begin{aligned}[b] \frac{d}{dx}\ln(x)&= \lim_{\Delta{x}\to0}\frac{\ln(x+\Delta{x})-\ln(x)}{\Delta{x}}\\ &= \lim_{\Delta{x}\to0} \Big[\ln\Big(\frac{x+\Delta{x}}{x}\Big)\cdot\frac{1}{\Delta{x}}\Big]\\ &= \lim_{\Delta{x}\to0} \Big[\ln\Big(1+\frac{\Delta{x}}{x}\Big)\cdot\frac{1}{\Delta{x}}\Big]\\ &= \lim_{\Delta{x}\to0} \Big[\ln\Big[\Big(1+\frac{\Delta{x}}{x}\Big)^{1/\Delta{x}}\Big]\Big] \end{aligned} \end{equation}$$

Suppose we define a new variable $n$ like so:

$$\begin{equation}\label{eq:OFlT1tIhdOqqD5E0c1F} n=\frac{\Delta{x}}{x} \end{equation}$$

From \eqref{eq:OFlT1tIhdOqqD5E0c1F}, we know that as $\Delta{x}$ approaches zero, $n$ tends to zero as well. Let's rearrange \eqref{eq:OFlT1tIhdOqqD5E0c1F} to the following form:

$$\begin{equation}\label{eq:lMPx5HbaDo54Eh3drQN} \frac{1}{\Delta{x}}=\frac{1}{nx} \end{equation}$$

Now, substituting \eqref{eq:OFlT1tIhdOqqD5E0c1F} and \eqref{eq:lMPx5HbaDo54Eh3drQN} into \eqref{eq:fTAFYivb5a5WjPszK3z} gives:

$$\begin{align*} \frac{d}{dx}\ln(x)&=\lim_{n\to0} \Big[\ln\Big[\Big(1+n\Big)^{1/(nx)}\Big]\Big]\\ &=\lim_{n\to0} \Big[\frac{1}{x}\cdot\ln\Big[\Big(1+n\Big)^{1/n}\Big]\Big]\\ &=\frac{1}{x}\cdot\lim_{n\to0} \Big[\ln\Big[\Big(1+n\Big)^{1/n}\Big]\Big] \end{align*}$$

Now, we use the property of limit of composite functionslink to swap the order of the limit and natural log:

$$\begin{align*} \frac{d}{dx}\ln(x) &=\frac{1}{x}\cdot\ln\Big[\lim_{n\to0} \Big[\Big(1+n\Big)^{1/n}\Big]\Big] \end{align*}$$

The term inside the natural log is the formal definitionlink of $e$ and thus we get:

$$\begin{align*} \frac{d}{dx}\ln(x) &=\frac{1}{x}\cdot\ln(e)\\ &=\frac{1}{x}\cdot1\\ &=\frac{1}{x} \end{align*}$$

This completes the proof.

Theorem.

Power rule

The derivative of $y=x^n$ where $n\in\mathbb{R}$ with respect to $x$ is:

$$\frac{dy}{dx}=nx^{n-1}$$

Solution. Suppose we have the following equation:

$$\begin{equation}\label{eq:hsZr1Fc8hHHMZf4UT3X} y=x^n \end{equation}$$

Where $n\in\mathbb{R}$. Taking the natural logarithm on both sides gives:

$$\ln(y)=n\cdot\ln(x)$$

Now, we are going to perform implicit differentiation, which involves taking the derivative of both sides with respect to $x$ like so:

$$\begin{equation}\label{eq:CZGjfAEGwVoamin4gJQ} \begin{aligned}[b] \frac{d}{dx}\Big(\ln(y)\Big)&=\frac{d}{dx}\Big(n\cdot\ln(x)\Big)\\ \frac{1}{y}\cdot{y'}&=n\cdot\frac{1}{x}\\ y'&=\frac{ny}{x} \end{aligned} \end{equation}$$

Now, we substitute \eqref{eq:hsZr1Fc8hHHMZf4UT3X} into \eqref{eq:CZGjfAEGwVoamin4gJQ} to get:

$$\begin{align*} y'&=\frac{nx^n}{x}\\ &=nx^{n-1} \end{align*}$$

This completes the proof.

Theorem.

Product rule

If $f(x)$ and $g(x)$ are differentiable functions, then their product $P(x)=f(x)\cdot{g(x)}$ is also a differentiable function whose derivative is:

$$P'(x)= f(x)\cdot{g'(x)}+ f'(x)\cdot{g(x)}$$

Proof. We are given that:

$$\begin{equation}\label{eq:JKNwEXxB1iFGkN5wvSO} P(x)=f(x)\cdot{g(x)} \end{equation}$$

Taking the natural log of both sides:

$$\begin{align*} \ln(P(x))&=\ln[f(x)\cdot{g(x)}]\\ &=\ln[f(x)]+\ln[g(x)] \end{align*}$$

Now, let's perform implicit differentiation by taking the derivative of both sides with respect to $x$ like so:

$$\begin{align*} \frac{d}{dx}\ln(P(x))&=\frac{d}{dx}\Big(\ln[f(x)]+\ln[g(x)]\Big)\\ \frac{1}{P(x)}\cdot{P'(x)}&=\frac{d}{dx}\ln[f(x)]+\frac{d}{dx}\ln[g(x)]\\ \frac{1}{P(x)}\cdot{P'(x)}&=\frac{1}{f(x)}\cdot{f'(x)}+\frac{1}{g(x)}\cdot{g'(x)}\\ P'(x)&=P(x)\cdot\Big(\frac{f'(x)}{f(x)}+\frac{g'(x)}{g(x)}\Big)\\ \end{align*}$$

Here, for the second step, we relied on the derivative of natural logarithmlink, chain rulelink and sum rule of differentiationlink. Substituting \eqref{eq:JKNwEXxB1iFGkN5wvSO} into the above gives us the product rule:

$$\begin{align*} P'(x)&=(f(x)\cdot{g(x)})\cdot\Big(\frac{f'(x)}{f(x)}+\frac{g'(x)}{g(x)}\Big)\\ &=g(x)\cdot{f'(x)}+f(x)\cdot{g'(x)} \end{align*}$$

This completes the proof.

Theorem.

Differentiability implies continuity

If $f(x)$ is differentiable at point $x=c$, then $f(x)$ is continuouslink at this point.

Proof. For $f(x)$ to be continuous at point $x=c$, the following conditions of continuitylink must be satisfied:

  • $f(c)$ exists.

  • $\lim_{x\to{c}}f(x)=f(c)$.

We know that if $f(x)$ is differentiable at point $x=c$, then the following limit exists:

$$\begin{equation}\label{eq:g2IZUNvydNgUGjRU41s} \begin{aligned} {\color{#ADF0A6}f'(c)}&= {\color{#ADF0A6}\lim_{x\to{c}}\frac{f(x)-f(c)}{x-c}} \end{aligned} \end{equation}$$

The fact that this limit exists implies $f(c)$ exists, which is the first criteria of continuity.

Let's now forget about \eqref{eq:g2IZUNvydNgUGjRU41s} for a moment and focus on the following limit:

$$\lim_{x\to{c}}\Big(f(x)-f(c)\Big)$$

From the product rule of limitslink, we have that:

$$\begin{align*} \lim_{x\to{c}}\Big(f(x)-f(c)\Big) &=\lim_{x\to{c}}\Big(\frac{f(x)-f(c)}{x-c}(x-c)\Big)\\ &={\color{#ADF0A6}\lim_{x\to{c}} \Big(\frac{f(x)-f(c)}{x-c}\Big)}\cdot\lim_{x\to{c}}(x-c)\\ &={\color{#ADF0A6}f'(c)}\cdot{(0)}\\ &=0 \end{align*}$$

Here, the green limit exists because we are given that $f(x)$ is differentiable at point $x=c$.

To summarize what we've shown:

$$\lim_{x\to{c}}\Big(f(x)-f(c)\Big)=0$$

By the summation rule of limitslink, we have that:

$$\lim_{x\to{c}}f(x)-\lim_{x\to{c}}f(c)=0$$

Since $f(c)$ is some fixed value, the limit as $x$ approaches $c$ is still equal to $f(c)$ and thus:

$$\lim_{x\to{c}}f(x)-f(c)=0$$

Finally, placing $f(c)$ to the right-hand side gives:

$$\lim_{x\to{c}}f(x)=f(c)$$

This is definition of continuity! Therefore, we have just shown that differentiability implies continuity!

Theorem.

Quotient rule

If $f(x)$ and $g(x)$ are differentiable functions, then:

$$\frac{d}{dx} \frac{f(x)}{g(x)}= \frac{g(x)\cdot{f'(x)}-f(x)\cdot{g'(x)}} {\big(g(x)\big)^2}$$

Proof. Let's define a new function $h(x)$ like so:

$$\begin{equation}\label{eq:qRXsic0qJKNYM0zRcnb} h(x)= \frac{f(x)}{g(x)} \end{equation}$$

Multiplying both sides by $g(x)$ gives:

$$f(x)= h(x)\cdot{g(x)}$$

Let's now take the derivative of both sides with respect to $x$ like so:

$$\begin{equation}\label{eq:OhQvSpBcsoMLR2Aej6n} f'(x)= h'(x)\cdot{g(x)}+ g'(x)\cdot{h(x)} \end{equation}$$

Here, we have used the product rule of differentiation. The reason we did this is so that we get $h'(x)$, which is what we wish to find the expression of!

Let's now rearrange \eqref{eq:OhQvSpBcsoMLR2Aej6n} to make $h'(x)$ the subject:

$$\begin{equation}\label{eq:dJNwqlRrJHVCGFUnFSy} h'(x)= \frac{f'(x)-g'(x)\cdot{h(x)}} {g(x)} \end{equation}$$

Now, let's substitute the definition of $h(x)$, that is \eqref{eq:qRXsic0qJKNYM0zRcnb}, into \eqref{eq:dJNwqlRrJHVCGFUnFSy} to get:

$$\begin{align*} h'(x)&= \frac{f'(x)-g'(x)\cdot\frac{f(x)}{g(x)}} {g(x)}\\ &= \frac{g(x)\cdot{f'(x)}-g'(x)\cdot{f(x)}} {\big(g(x)\big)^2}\\ \end{align*}$$

This completes the proof.

robocat
Published by Isshin Inada
Edited by 0 others
Did you find this page useful?
thumb_up
thumb_down
Comment
Citation
Ask a question or leave a feedback...
thumb_up
0
thumb_down
0
chat_bubble_outline
0
settings
Enjoy our search
Hit / to insta-search docs and recipes!