Maths

Various notes (crib sheet) from trying to understand various things... really just stuff you'd find in any old math text book. Oh and this is a useful MathJax reference on StackExchange. And Detxify can be used to draw the symbol you're looking for! For an interactive view of your mathjax try here. For graphing functions try Desmos.

Page Contents

References and Resources

The following are absolutely amazing, completely free, well taught resources that just put things in plain English and make concepts that much easier to understand! Definitely worth a look!

  1. The amazing Khan Achademy.
  2. The amazing CK12 Foundation.

For a more in-depth look at calculus the following book is AMAZING! Written really well. I wish I'd had this book when I was at school!

  1. The Calculus Lifesaver, Adrian Banner.

Also, not free, but still looks good and encourages learning by problem solving, is the excellent Brialliant.org.

Oooh and this guy's animations look pretty cool too.

Todo / To Read

https://math.stackexchange.com/questions/706282/how-are-the-taylor-series-derived
Laplace Transform Explained and Visualized Intuitively - https://www.youtube.com/watch?v=6MXMDrs6ZmA
Derivatives of exponentials | Chapter 5, Essence of calculus - https://www.youtube.com/watch?v=m2MIpDrF7Es
The Laplace Transform - A Graphical Approach - https://www.youtube.com/watch?v=ZGPtPkTft8g

Function Scaling


<View source>.

Partial fractions

The point of partial fractions is to do the reverse of: 5x4+3x+1=8x7x23x4

In other words, given the following, 8x7x23x4 Partial fractions let us go back to, 5x4+3x+1 We would start this by first factoring the above to get, 8x7x23x4=8x7(x4)(x+1) We know that we want to get to something like, 8x7(x4)(x+1)=A(x4)B(x+1) We can remove the denominators by multiplying through by (x4)(x+1) to give, 8x7=A(x+1)B(x4)=(AB)x(A4B) Matching terms between 8x7 and (AB)x+(A+4B), we get, AB=8 and, A+4B=7 Solving these simultaneous linear equations, we get, A=745, and B=15 And so we can get back to 8x7(x4)(x+1)=745(x4)+15(x+1) Yucky example, sorry!

Summary Of Partial Fraction Rules

For example, 3s24s+11(s+3)(s2)2 Has partial fractions of the form... A(s+3)+B(s2)+C(s2)2

Difference Of Two Cubes

a3b3=(ab)(a2+ab+b2) So for example... x327=x333=(x3)(x2+3x+9)

Trigonometry

Wikipedia has a nice list of proofs for these :)

The "cut the knot" website lists some excellent proofs of the fundamental Pythagorean Theorem

Basics

sin(θ)=oppositehypotenuse, cos(θ)=adjacenthypotenuse, tan(θ)=oppositeadjacent

tan(θ)=sinθcosθ csc(θ)=1sin(θ), sec(θ)=1cos(θ), cot(θ)=1tan(θ)

cot(θ)=cosθsinθ

Some trig functions are "complementary" in the sense that a in pair of such functions, one equals the other shifted by 90 degrees (π/2 radians). So, "cos" is "complementary sine", for example, and "cot" is "complementary tan". The one that doesn't quite fit this pattern is the "csc" and "sec" complements :'(... sin(θ)=cos(π2θ), cos(θ)=sin(π2θ) tan(θ)=cot(π2θ), cot(θ)=tan(π2θ) sec(θ)=csc(π2θ), csc(θ)=sec(π2θ)

Radians

Here is a fantastic page about the why of radians. In summary radians define a circle such that the angle measured in radians is also the arc length around the unit circle for that arc. angle in radians =π180× angle in degrees


[Image taken from Ask A Mathamatition.]

Common Values For Sine and Cosine

0 π6 π4 π3 π2
sin(θ) 0 12 12 32 1
cos(θ) 1 32 12 12 0
tan(θ) 0 13 1 3

Reference Angles and the ATSC Method

Sum and product formulas

sin(a)cos(b)=12[sin(a+b)+sin(ab)] cos(a)sin(b)=12[sin(a+b)sin(ab)] sin(a)sin(b)=12[cos(ab)cos(a+b)] cos(a)cos(b)=12[cos(a+b)+cos(ab)] sin(a)+sin(b)=2 sin(a+b2) cos(ab2) sin(a)sin(b)=2 cos(a+b2) sin(ab2) cos(a)+cos(b)=2 cos(a+b2) cos(ab2) cos(a)cos(b)=2 sin(a+b2) sin(ab2)

Reduction Formulas

sin(θ)=sin(θ) sin(θ)=sin(θπ) sin(θ)=cos(θ±π2) cos(θ)=cos(θ) cos(θ)=cos(θπ) ±cos(θ)=sin(θ±π2) tan(θ)=tan(θ) tan(θ)=tan(θπ)

Pythagorean Identities

sin(a)cos(a)=tan(a) sin2(θ)+cos2(θ)=1 tan2(θ)+1=sec2(θ) cot2(θ)+1=csc2(θ)

Sum or Difference of Two Angles

sin(a±b)=sin(a)cos(b)±cos(a)sin(b) cos(a±b)=cos(a)cos(b)sin(a)sin(b) tan(a±b)=tan(a)±tan(b)1tan(a)tan(b)

Double Angle Formulas

sin(2θ)=2sin(θ)cos(θ) cos(2θ)=12sin2(θ) cos(2θ)=2cos2(θ)1 cos(2θ)=cos2(θ)sin2(θ) tan(2θ)=2tan(θ)1tan2(θ)

A Nice GIF Relating Sin And Cos

Animation of trig functions

Found the above here.

Logarithms

loga(au)=u aloga(u)=u loga(uv)=loga(u)+loga(v) loga(uv)=loga(u)loga(v) loga(un)=nloga(u) loga(u)=logb(u)logb(a)

Series

Basic Series

Consider summing the numbers in the series 1,2,3,4. For each number in the series, n imagine stacking n blocks to the left of the last stack.... you'd build up the triangle shown to the left. It's basically one half of a square. Trouble is if you divide the square by 2 you would chop off the top half of each block at the top of its stack. Therefore for each number in the series you have "lost" half a block.

Therefore the total number of blocks is... number_of_blocks=n22+n2=n(n+1)2 Where n2/2 is half of the square and n/2 is the total of all the halves that we "cut" off (but didn't mean to) when we took the half.

Therefore, we can say... 1nn=n(n+1)2 And... 1nn2=n(n+1)(2n+1)6 1nn3=(n(n+1)2)2

Arithmetic Progression

x(n)=a1,a1+d,a1+2d, Sn=na+(n1)d2

Geometric Progression

x(n)=a,ar,ar2,ar3,,arn, Sn=a(1rn)1r,n>0 If |r|<1, then as n, rn0, S=a1r Or, if n<0 (useful for the z-transform), then S=arr1

Binomial Theorem

The binomial theorem is summarised as... (1+x)n=1+nx+n(n1)2!x2+n(n1)(n2)3!x3++n(n1)(n2)1(n1)!xn Where the nth term is given by the following equation. ua=xn!a!(na)!

Exponential Series

The exponential series is defined as... e=(1+1x)n, n This is why e=2.71828

The constant e can be expanded as follows... e=1+1+12!+13!+14!+ And powers of e as ... ex=1+x+x22!+x33!+ ekx=1+kx+(kx)22!+(kx)33!+

Malclauren's Theorem

Attempts to express a function as a polynomial. f(x)=a0+a1x+a2x2+a3x3++anxn Note that the series must be shown to converge. Need to figure out the ai coefficients. Notice the following. f(x)=a0+a1x+a2x2+a3x3+a4x4+a5x5+a6x6+ f(x)=a1+2a2x+3a3x2+4a4x3+5a5x4+6a6x5+ f(x)=2a2+32a3x+43a4x2+54a5x3+65a6x4+ f(x)=32a3+432a4x+543a5x2+654a6x3+ Notice then... f(0)=a0 f(0)=a1 f(0)=2!a2 f(0)=3!a3 And so on...
It thus looks like, and indeed is the case, that: an=dnfdxn/n!

The above definition can then be used to derive the expansion of ex, which is why we were able to say: ex=1+x+x22!+x33!+ Remember, the series must be shown to converge! This is easily seen because the denominator is growing at a faster rate than the numerator.

Using this we can derive Euler's Formula. cos(x)=1x22!+x44!x66!+x88! sin(x)=xx33!+x55!x77!+x99!

Summing the above two expansions we get... cos(x)+sin(x)=1+xx22!x33!+x44!+x55!x66!x77!+ Doing a similar expansion for ejx we get the following and can then see how we get Euler's formula... ejx=1+jx+(jx)22!+(jx)33!+(jx)44!+=1+jx+x22!+jx33!+x44!++jx55!+=cos(x)+jsin(x)

Imaginary & Complex Numbers

Intro

The definition of a imaginary number, i, is i2=1, which is also often seen written as i=1. It's called "imaginary" because the square root of -1 doesn't exist in real terms: there is no real number that when multiplied by itself is negative. Thus, the set of real numbers has been "extended". Sometimes j is used instead of i to denote an imaginary number.

A complex number is one that has a real and imaginary part, even if the real part is zero. Thus real numbers are a subset of complex numbers.

Rectangular Form

In rectangular form we write complex numbers like this: z=a+bi, where a and b are real numbers. If z=a+bi then it we can write Re(z)=a and Im(z)=b.

The representation of a complex number as z=a+bi is called rectangular form.

A complex number can be viewed graphically on an argand diagram, which is a little like a Cartesian diagram, except the horizontal axis is the real component and the vertical axis is the imaginary component of the complex number:

argand diagram of complex numnbers

Add z1+z2=(a1+b1i)+(a2+b2i)=(a1+a2)+(b1+b2)i
Subtract z1z2=(a1+b1i)(a2+b2i)=(a1a2)+(b1b2)i
Multiply z1z2=(a1+b1i)(a2+b2i)=a1a2+a1b2i+a2b1ib1b2
Divide Need complex conjugates for this. See below... z1÷z2=(a1+b1i)(a2+b2i)=(a1+b1i)(a2+b2i)(a2b2i)(a2b2i)=(a1+b1i)(a2b2i)a22+b22=a1a2+a1b2i+a2b1ib1b2|z2|2

Complex conjugates. The complex conjugate of z=a+bi is defined as z¯=z=abi (where z¯ and z are just two different notations that mean the same thing). The can be visualised as shown below:

complex conjugate argand diagram

The next thing to talk about is the magnitude of a complex number. In the above diagrams we can see that the vector forms a right-angled triangle with the horizontal axis. The magnitude is the length of the vector and this is defined as follows: |z|=Re(z)2+Im(z)2

Multiplying By j Rotates 90 Degrees Anticlockwise

Multiplying a real or complex number by j rotates it, in the complex plan, by 90 degrees anticlockwise, whilst dividing by j rotates in 90 degrees clockwise.

Rotating by j rotates AC in complex plane

Polar Form

Complex numbers can also be written in polar form by specifying magnitude (distance from the origin) and angle...

diagram of complex number in polar form

Arg(z)=θ|z|=r=a2+b2

How do we go from rectangular form (z=a+bi) to polar form ((r,θ))? Recall tan(θ)=ba and that therefore θ=tan1(ba). We already know row to calculate r.

Lets do a quick example. Let z=2+3i: tanθ=32 θ=0.983 radians, rounded to 3dp |z|=r=13 Thus, in polar form the complex number z=2+3i is expressed as (13,0.983).

Exponential Form

From polar form we go to exponential form. cos(θ)=arrcos(θ)=asin(θ)=brrsin(θ)=b This means that we can re-write z=a+bi as z=rcos(θ)+rsin(θ)i=r[cos(θ)+isin(θ)] And using Euler's amazing formula eiθ=cos(θ)+isin(θ), we can write... z=reiθ This is the exponential form. It is especially useful if you want to do division of complex numbers.

We can go back from exponential form to rectangular form using the same two formulas we saw above: rcos(θ)=a and rsin(θ)=b.

Taking the example from the previous section where we saw that we could represent z=2+3i as (13,0.983), we can now see this is easily converted to exponential form: z=13e0.983i.

And, if we wish to go back we can see that a13cos(0.983)2 (approximately because the 0.983 value is rounded).

The exponential form is particularly useful in subjects like DSP for doing things like changing a signal's phase, which ends up just being a multiplication like this: z=zejα The number z is z rotated anticlockwise by α radians!

To explain a little more clearly, a discrete complex exponential is generally represented as: A[cos(ωn+θ)+jsin(ωn+θ)]=Aejωn+θ Where ω is the angular frequency is radians per sample period, A is the amplitude and θ is the initial phase of the signal in radians. If we had to do a phase shift of the signal we'd either have to remember a whole load of trig identities or use the complex exponetial form, which makes life easier...

We can see this very simply by noting the following: cos(ωn+θ)=R{ej(ωn+θ)}=R{ejωnejθ} The phase shift has become just multiplication.

Combinations & Permutations

Selecting r samples from a set of n samples at random. If the order of the elements matters then it is a permutation, otherwise it is a combination. Permutations=nPr=n!(nr)! If I have a set of n objects and I select r=2 objects, on the first selections I can choose from n objects. For the second selection I can choose from n1 objects. This looks like the beginning of a factorial, but the factorial would look like this. n!=n(n1)(n2)(n3)...1 We only want the first 2 terms. To get rid of the remaining terms we need to divide by (n2)(n3)...1, in other words, (nr)!. Hence the formula above.

With combinations the order is not important. This time take r=3... for every choice I make there will be 3×2 permutations with the same elements, which is one combination (as order is now not important). This is the case because the 3 elements can be ordered any how and count as the same object. (a,b,c) and (b,a,c) are, for example, no different. Out of all sets containing only these elements there must be 3! permutations. Therefore need to get rid of this count from nPr by divifing by r! in the general case... Combinations=nCr=nPrr!=n!r!(nr)!

For a fairly explicit example of combinations v.s. permutations and their application in statistics see this little example.

Square Roots

Reference: Mathwords.

Limits

Definition

When we talking about the limit of a function, we are asking what value the function will give as the input gets closer and closer to some value, but without reaching that value. The animation below tries to explain this:


<View source>

What you should see in the animation is that we can always get closer to the point x=3 without ever reaching x=3 by adding smaller and smaller amounts. In other words, you can get f(x) as close to the limit (in the above case 9) by getting x sufficiently close to some input (in the above case 3)

Intuitively, limxcf(x)=L If f is defined near, but not necessarily at, c, then f(x) will approach L as x approaches c. More rigorously, let f be defined at all x in an open interval containing c, except possibly at c itself. Then, limxcf(x)=L If and only if for each e>0, there exists d>0 such that if 0<Z|xc|<d then |f(x)L|<e.

I.e., no matter how close to L we get, there is always a value of x arbitrarily close to the limit value, c, but which is not c, that yields this value close to L.

Ie. We can pick a window around the y-axis value L and if we keep shrinking this window, we will always find an x value, either side of c that will yield a y-value in this error window. So, the error can be arbitrarily small and we will always find an x either side of c to satisfy it :) This is the two sided limit.

Limit Properties

Most of these are only true when the limits are finite...

Uniquness: If limf(x)=L1 and limf(x)=L2 then L1=L2
Addition: lim[f(x)+g(x)]=limf(x)+limg(x)
Scalar multiplication: lim[αf(x)]=αlimf(x)
Multiplication: lim[f(x)g(x)]=limf(x)limg(x)
Division: lim[f(x)g(x)]=limf(x)limg(x), g(x)0
Powers: lim[f(x)n]=[limf(x)]n, n>0

Ways To "Solve" Limits

If the limit is not of the form x, i.e., x tends to some known number, first try pluggin that number into the equation. If you get a determinate answer thats great.

Generally though we get an indeterminate answer because we get a divide by zero...

If you can transform the function you're taking the limit of to something where there is no longer a divide by zero, note that original function and the transform are not exactly identical. Let's say the original function is f(x) and you have transformed it to g(x).

If xn and at n f(n) is indeterminate, our transform to g(n) is not. This means f(x) and g(x) are not the same function. But because limits are only concerned with values near n and not at n, this is okay as the functions are identical everywhere else.

When numerator non-zero but denominator zero, there is at least one vertical asymptote in your function. Either the limit will exist here or you will only have a left or right sided limit if on one side the y-axis tends in the opposite direction to the other side of the asymptote.

Factor Everything

If f(x)=2x2+12x+102x+2 then factor to f(x)=(2x+2)(x+5)(2x+2) and cancel out the denominator. You can also use the difference of two cubes to help with more complex functions: a3b3=(ab)(a2+ab+b2).

Get Rid Of Square Roots

For problems where xa, where a is not you get rid of square roots by multiplying by the conjugate. If you have expr+C you multiply by exprC. If you have exprC you multiply by expr+C: limxa(exprCxD×expr+Cexpr+C)=limxaexprc2(xD)(expr+C) Now when x=a, hopefully you won't have an indeterminate fraction!

Rational Functions

When x, the leading term dominates. If p(x)=a0xn+a1xn1+an then the leading term is a1xn. Putting pL(x)=a1xn, we say that... limxp(x)pL(x)=1 This does not mean that p(x) ever equals pL(x), just that the ratio of the two tends to one as x tends to infinity!

If you are taking the limit of a rational function like the one below... limx(n(x)d(x)) ... you can't just substitute infinity for x because you get which doesn't make a lot of sense...

So do this... limx(n(x)×nL(x)nL(x)d(x)×dL(x)dL(x))=limx(n(x)nL(x)×nL(x)d(x)dL(x)×dL(x))=limx(n(x)nL(x)d(x)dL(x)×nL(x)dL(x)) We know that because the leading term dominates... As x the fraction n(x)nL(x) tends to 1 as does d(x)dL(x) Which means that we will really be looking at the limit of... nL(x)dL(x) ... as x tends to infinity!

A numerical example. Solve limx(x+5x54x8+7x4+x+1234) So we do... limx(x+5x55x54x8+7x4+x+12344x8×5x54x8) Everything tends to one except 5x54x8, which is equal to 54x3, so we know that this tends to zero!

In general, for our polynomials n(x) and d(x):

  1. If degree of n == degree of d, then limit is finite and nonzero as x±
  2. If degree of n > degree of d, then limit is or as as x±
  3. If degree of n < degree of d, then limit is 0 as x+

N-th Roots...

When there are square roots, or indeed n-th roots in the equation the leading term idea still works. All that happens is that when you divide the n-th root by the leading term you bring it back under the square root.

Be careful when x because x2x when x is negative! It equals x. Use the following rule. If you write... nxsome power=xm You need a minus in front of xm when n is even and m is odd.

The Sandwich Principle / Squeeze Theorem

If g(x)f(x)h(x) for all x near a, and limxag(x)=limxah(x)=L, then limxaf(x)=L too.

Sal on Khan Achademy gives a good example of using the squeeze theorem to solve the following limit. limθ0(sinθθ)=1 The limit cannot be defined at θ=0, so the function is not continuous at that point, but it can still have a limit as θ tends to 0, as it does not have to equal 0... we only have to be able to get arbitrarily close.

L'Hopital's Rule

Can be summarised as follows... limxa{f(x)g(x)}=limxa{f(x)g(x)} L'Hopital's rule can be used when the normal limit is indeterminate. For example... limx0cosh(x)exx Substituting in 0 gives (11)/0, which is indeterminate. So, apply L'Hopital's rule by differentiating numerator and denominator separately... limx0cosh(x)exx=limx0sinh(x)ex1 Substitute in 0 for x gives 1. Therefore... limx0cosh(x)exx=1

Trig functions

For small values of x, y=sin(x) and y=x are approximately the same. For really small values, I'm not even sure my computer has enough precision to be able to accurately show the difference. But, using a small python script we can get the idea:

y=sin(x) vs y=x for very small values of x

The above image was generated using the following script:

import matplotlib.pyplot as pl
import numpy as np

x = np.arange(-0.0001,0.0001, 0.000001)
y_lin = x
y_sin = np.sin(x)

pl.plot(x,y_sin, color="green")
pl.plot(x,y_lin, color="red")
pl.grid()
pl.legend(['y=sin(x)', 'y=x'])
pl.show()

In fact if your print(y_sin/y_lin) you just get an array of 1's because the computer does not have the precision to do any better. I think even at small values of x it is not the case that sin(x)=x, but it is very close. So in reality, even for tiny x: sin(x)x1 But, the following does hold: limx0sin(x)x=1

The following also hold: limx0cos(x)=1 limx0tan(x)x=1

"Continuous" In Terms Of Limits

Continuous At A Point

A function is continuous at a point if it can "be drawn without taking the pen off the paper" [Ref]. This means, more formally, that a function f(x) is continuous at x=a if limxaf(x)=f(a).

As Adrian Banner says in his book "The Calculus Lifesaver", it is continuity that connects the "near" with the "at" in terms of limits. It is what allows us to find limits by direct substitution.

Continuity Over An Interval

If a function is continuous over the interval [a, b]:

  1. The function is continuous at over point in (a,b).
  2. The function is right-continuous at x = a: f(x)=a and limxa+f(x) exists.
  3. The function is left-continuous at x = b: f(x)=b and limxaf(x) exists.

Intermediate Value Theorem

If f is a function continuous at every point of the interval [a,b], then:

Min/Max Theorem

If f is continuous on [a,b], then f has at least one maximum and one minimum on [a,b].

Differentiation

Definition

Recall that a function is continuous at a point if it can "be drawn without taking the pen off the paper" [Ref]. This means, more formally, that a function f(x) is continuous at x=a if limxaf(x)=f(a). For a function to be differentiable at a point, it must be continuous at that point....

The derivative of a function f(x) with respect to it's variable x is defined by the following limit, assuming that limit exists. f(x)=limh0f(x+h)f(x)h If the limit does not exist the function does not have a derivative at that point.

This is why, for example the function f(x)=|x| does not have a defined derivative at x=0, because the limit (of the differentiation) does not exist at that point. The function is continuous at that point because the left and right limits are the same, but the derivative function does not exist.

A function is continuous over an interval [a,b] if:

  1. it is continuous at every point in (a,b),
  2. it is right-continuous at x=a, i.e., limxa+f(x)=f(a)
  3. it is left-continuous at x=b, i.e., limxbf(x)=f(b)

If a function is differentiable then it must also be continuous.

If h is replaced with Δx we would write f(x)=limΔx0f(x+Δx)f(x)Δx Where Δ just means a "small change in". This small change in x leads to a small change in y, which is given by f(x+Δx)f(x). Therefore we can write: f(x)=limΔx0ΔyΔx=dydx I.e, if y=f(x),f(x) can be written as dydx.

Now at school, as we'll see in the integration by parts section, sometimes dydx was treated as a fraction although my teacher always said that it wasn't a fraction. The above, which I didn't know at the time, explains why. And, finally, in the awesome book "The Calculus Lifesaver" by Adrian Banner [Ref] I know why :)

Unfortunately neither dy or dx means anything by itself ...

... the quantity dydx is not actually a fraction at all - it's the limit of the fraction dydx as Δx0.

Example

We can use the definition above to take the derivative of y=4x3: f(x)=limh04(x+h)34x3h=4limh0(x+h)3x3h=4limh03hx2+3h2x+h3h=4limh03x2+3hx+h2 We know that this limit exists because the function f is continuous by virtue of it being a polynomial. We also know that h is tending to zero, so those terms just dissapear, which leaves us with... f(x)=4limh03x2+3hx+h2=12x2 Thus, ddx(4x3)=f(x)=12x2

Basic Differential Coefficients

ddx(xn)=nxn1ddx(lnx)=1xddx(ekx)=kekxddx(ax)=axlnaddx(cosx)=sinxddx(sinx)=cosxddx(tanx)=sec2x
ddx(coshx)=sinhxddx(sin1x)=11x2ddx(cos1x)=11x2ddx[ln(f(x))]=f(x)f(x)ddx(cax)=caxln(ca)ddx(xx)=xx(1+lnx)
Differentiate sine to cosine and vice versa

Chain Rule

Using this when you need to differentiate a function that is made up of the result of one function passed into the next and so on. I.e. if h(x)=f(g(x)).

dydx=dydu×dudx

Quotient Rule

When you have a function h(x) that can be expressed as a fraction, f(x)g(x) use the quotient rule to allow you do differentiate f(x) and g(x) independently (easier!) and the combine the results using this rule. f(x) becomes u and g(x) becomes v:

dydx=vdudxudvdxv2

Example

Solve the following: y=3x4+2x3+149x To do this we note that it is of the form uv where: u=3x4+2x3+1, and v=49x So we can calculate: dudx=12x3+6x2, and dvdx=9 Plugging these into the quotient rule formula we get, dydx=vdudxudvdxv2=v(12x3+6x2)u(9)v2=(49x)(12x3+6x2)(3x4+2x3)(9)(49x)2=48x3+24x2108x454x3+18x4+18x3(49x)2=90x4+12x3+24x2(49x)2

Product Rule

Use the product rule when trying to differentiate a quantity that is the multiplication of two function. I.e when trying to find the derivative of h(x) when h(x)=f(x)g(x). One of these functions becomes u, the other v. Use this so you can differentiate the simpler functions f(x) and g(x) independently and then combine the results.

For two functions...

dydx=udvdx+vdudx

Or for three functions... dydx=dudxvw+udvdxw+uvdwdx

Or, even, for any number of funtions...

... add up the group abc...uvw n times and put a d/dx in front of a different variable in each term ...

The following is a really cool visual tutorial by Eugene Khutoryansky...

Example

A fairly simple example - find the derivative of y=(2x+1)3(x1)4. Okay, so we could just expand this out but it would be pretty tedious. The product rule help make it easier!

In this case if we let f(x)=(2x+1)3 and g(x)=3(x1)4, then we can write h(x)=f(x)g(x)=(2x+1)3(x1)4, which is in the exact pattern we need for the product rule where we will label f(x) as u and g(x) as v: u=(2x+1)3v=(x1)4 We can find their derivatives as follows. Let a=(2x+1). u=a3duda=3a2 And ... dadx=2 Therefore... dudx=dadxduda=23a2=6a2=6(2x+1)2 Let b=(x1). v=b4dvdb=4b3 And... dbdx=1 Therefore... dvdx=dbdxdvdb=14b3=4(x1)3 Now that we have the above we can use the product rule: dydx=udvdx+vdudx dydx=u(4(x1)3)+v(6(2x+1)2)=(2x+1)34(x1)3+(x1)46(2x+1)2=2(2x+1)2(x1)3[2(2x+1)+3(x1)]=2(2x+1)2(x1)3(7x1)

Integration

Basic Integrals

xndx=xn+1n+1+C1xdx=lnx+Cekxdx=ekxk+Caxdx=axlna+Csinxdx=cosx+Ccosxdx=sinx+C
sec2xdx=tanx+Csinhxdx=coshx+C11x2dx=sin1x+C11x2dx=cos1x+Cf(x)f(x)dx=ln[f(x)]+C

Integration Of Linear Factors

Use variable substitution, for example, if the integral is y=(ax+b)ndx Then put u=ax+b, giving... y=undx But now we need to integrate with respect to u, not x!. In other words we need to go from the above integral to something like, y=???du We can use the chain rule as follows... dydu=dydxdxdu By integrating both sides of the equation with respect to u we get.... y=dydxdxdudu We now have an integral of some function with respect to u. We can find dy/dx from the definition of the original integral (just differentiate both sides!). We can find dx/du by firstly rearranging u=ax+b to... x=uba Then take the derivative with respect to u to get dxdu=1a Notice how, that because the "thing" (factor) that is raised to a power in the integral is a linear function, the x terms will always dissapear, leaving only a constant, which can be differentiated w.r.t u. If x terms remained after differentiating the "thing" (factor), it would stop us doing our desired integration w.r.t u! Now we can see that... dydxdxdu=(ax+b)n1a=un1a Substitute this into our integral and we have... y=un1adu This we know how to integrate the above using the list of standard integrals. y=un+1a(n+1)+C Now we can substitute back in for u to obtain the answer y=(ax+b)n+1a(n+1)+C

One point to note is as follows. I always remember being taught to re-arrange the substitution that was made... dxdu=1a To... dx=dua ..And then subsitute for the dx term in the integral. However, this is not strictly correct as far as I understand because a differential coefficient is not a fraction... it is a limit: dydx=limaf(x+a)f(x)a So, as we can see dy/dx is not really a fraction... hence the above method and explanation, even if the "trick" I was taught works.

Integration By Parts

Phasors

This awesome GIF is produced by RadarTutorial.eu, although I couldn't find it on their site. I originally found the image on this forum thread and the watermark bears RadarTutorial's site address (it's not too visible on the white background of this page).

Fourier

References

This is an awesome YouTube tutorial explaining how sine waves can be combined to produce any imaginable waveform: