Differential Calculus: Techniques of Differentiation ... by Brandon Williams
Main Index...

This tutorial is going to show you the *easy* side of differentiating. In the examples from the article before we had to do quite a bit of work. Believe it or not those examples are pretty trivial. We are going to use that very general method of differentiating to develop some rules that we can use without thinking twice about it.

We will start off easy with something that is actually common sense: the derivative of a constant function. An example of a constant function: f (x) = 4 . The graph of f is only a horizontal line that intersects the y-axis at four. So, the derivative of this function at any point is going to be zero. We have just developed our first rule:
      If f(x) = a , where a is any constant then, f'(x) = 0

          // The derivative of a constant function is always zero

Since we have learned that a derivative is merely a limit many of the same rules apply. One of the best rules we had for limits was that when you are trying to compute the limit of the sum of two functions you could simply evaluate each function separately and then add the two limits. You can do the same with derivatives:

      d                    d         d
      -- [ f(x) + g(x) ] = -- f(x) + -- g(x)
      dx                   dx        dx

          // and of course, if you can do it with addition you can do it with subtraction

      d                    d         d
      -- [ f(x) - g(x) ] = -- f(x) - -- g(x)
      dx                   dx        dx

Next, if you have function which is multiplied by a constant you can simply differentiate the function and then multiply the derivative by the constant. For example, in the previous article we found the derivative of (x2) to be (2x). But what if we had to differentiate (4x2). This new rule tells us that you only need to differentiate (x2) and then multiply the derivative by four. So the derivative of (4x2) would be (8x) because (2x * 4 = 8x). In math terms the rule states:

      d                   d
      -- [ a*f(x) ] = a * -- f(x)  , where a is any constant
      dx                  dx

This is about where the limit rules no longer apply. Say we were give: f(x) = xn , where n is a constant. The limit rules told us that we could just evaluate the limit of x and then take it to the nth power. For differentiating we have a different rule called the power rule. It states:

      d
      -- xn = n*x(n-1)  , where n is a constant
      dx

So, to evaluate the derivative of a variable which has been taken to a constant power simply multiply the term by that power and subtract one from the power. A few examples:

      1.)  f(x) = x3    --so--   f'(x) = 3x2

      2.)  f(x) = x5    --so--   f'(x) = 5x4

      3.)  f(x) = x100  --so--   f'(x) = 100x99

      4.)  f(x) = 3x6   --so--   f'(x) = 3 * (6x5) = 18x5

Note that in example number four I used two rules: the power rule and the rule that told us we could take out the constant, three, evaluate the derivative of (x6), and then multiply the derivative by the constant. See how easy things are becoming already? That last example would have taken (most likely) a good twenty minutes or so to evaluate our old way. I'm probably over estimating there but I would not want to be the person who had to do it. And for those who cannot stand being given information without know why it works here is the derivation of the power rule:

      Differentiate:  f(x) = xn  , where n is a constant

         // I am going to use the non-limit method for differentiating this...mainly
            because it is easier (and shorter) to write out

      y + dy = (x + dx)n

        // use Binomial Theorem to expand the above

                               n*(n-1)
            = xn + nx(n-1)*dx + -------*x(n-2)*dx2 + .... + nx*dx(n-1) + dxn
                                  2

        // subtract our original equation from the above:

         /                          n*(n-1)                                  \
        | y + dy = xn + nx(n-1)*dx + -------*x(n-2)*dx2 + .... + nx*dx(n-1) + dxn |
         \                             2                                     /
       -  [    y = xn  ]
      -----------------------------------------------------------------------------------
                             n*(n-1)
            dy = nx(n-1)*dx + -------*x(n-2)*dx2 + .... + nx*dx(n-1) + dxn
                                2

        // all terms with dx to a power greater than one will cancel out and
           we are left with:

      dy = nx(n-1) * dx

        // divide both sides by dx

      dy
      -- = nx(n-1)   // power rule
      dx
Before we go on to some of the harder rules let us look at an example incorporating everything we have learned so far:
      Differentiate:  f(x) = 3x4 - 2x3 + x2 - 8x + 2


         // from the second rule we know that we can evaluate all the terms
            separately and then add (or subtract) them all together:

      d    d          d          d         d         d
      -- = -- (3x4) - -- (2x3) + -- (x2) - -- (8x) + -- (2)
      dx   dx         dx         dx        dx        dx


         // from the third rule we know that we can take out the coefficients
            of each term and multiply them by the derivative of x-part of the term.  Also
            the first rule we learned tells us that the derivative of the last term in the 
above polynomial is a zero so we do not have to worry about it d d d d d -- = 3 * -- (x4) - 2 * -- (x3) + -- (x2) - 8 * -- (x) dx dx dx dx dx // now that it is all broken up into small parts we can start the actual differentiating d -- = 3*(4x3) - 2*(3x2) + (2x) - 8*(1) dx // And simplify d -- = 12x3 - 6x2 + 2x - 8 dx

How amazingly easy! These new rules helped us evaluate something which would have been a pain to do with our old way. But, in general, the above examples are very easy. Polynomials are the easiest functions to deal with in calculus. The next set of rules will help us tackle complex functions. What if we were to encounter two functions which were being multiplied together. We could not simply differentiate each function and the multiply the derivatives because we would be leaving out some middle terms. Therefore, let us deal with a general case and try to come up with some rule to use. Something like this:

      Differentiate:  y = u * w  , where u and w are both functions of x.  For example,

              y = (2x2 + 3x)(4x3-x+1)  , thus [u = 2x2+3x] and [w = 4x3-x+1]

          // Since we are dealing with a general case here it does not matter what u
             and w are because it should work with anything you substitute for them. To
             come up with a new rule I am going to be using the non-limit way of
             differentiating.

          // All our x's and y's becomes (x+dx)'s and (y+dy)'s thus our u's and w's become
             (u+du)'s and (w+dw)'s :

      y + dy = (u + du)(w + dw)

          // Expand the binomial (u+du)(w+dw)

      y + dy = u*w + u*dw + w*du + du*dw

          // Since du and dw are already indefinitely small when you multiply them
             together they become even small therefore you do not need to worry about
             that term:

      y + dy = u*w + u*dw + w*du

          // Now subtract our original equation (y=u*v) from the one above and you are
             left with:

      dy = u*dw + w*du

          // finally divide everything by dx

      dy     dw     du
      -- = u*-- + w*--
      dx     dx     dx

The above rule is called the power rule. In English: to differentiate the product of two functions multiply the derivative of one function by the other function plus the derivative of the other function multiplied by the first function. This rule can be a lot to take in at once so an example is probably best right now:

      Differentiate:  y = (x3 - x2 + 1)(x2 + 2x)

          // From the above equation we know that:

              u = x3-x2+1  and  w = x2+2x

          // And from the product rule we know that:

              dy     dw     du
              -- = u*-- + w*--
              dx     dx     dx

          // So our strategy is to first differentiate functions u and w, then multiply
             the derivatives by the function of the other term, and finally sum the two
             together...like this:

      dy               d                     d
      -- = (x3-x2+1) * -- (x2+2x) + (x2+2x) * -- (x3-x2+1)
      dx               dx                    dx

          // The derivatives of u and w are pretty trivial so I will not go much into that:

      dy
      -- = (x3-x2+1)*(2x+2) + (x2+2x)*(3x2-2x)
      dx

          // Multiply all terms out:

      dy
      -- = (2x4+2x3-2x3-2x2+2x+2) + (3x4-2x3+6x3-4x2)
      dx

          // simplify

      dy
      -- = 5x4 + 4x3 - 6x2 + 2x + 2
      dx

And that is our final answer. But, instead of doing all of that above work, what if we had just started out by expanding the [(x3-x2+1)(x2+2x)] term into on big polynomial and then proceeding with all the trivial differentiating tricks? Yes, it would have been much easier to do that first. However, soon we will not be able to do that. I cannot show you any examples of when you will not be able to do that just yet because we still have more to learn.

The next thing we are going to attempt are rational functions. What if we had a polynomial divided by another polynomial? Something like this:

      Differentiate:        x4 - 3x2 + 8x + 1
                       y = -------------------
                            2x3 - x2 + 2x - 5

We cannot simply differentiate the top and bottom separately and then divide the two because we would be leaving out a lot of middle terms. We need to develop a general rule for solving this problem. First we will use a function, u, for the numerator and a function, w, for the denominator. Therefore we have:

           u
      y = ---
           w

I am going to use the non-limit way to differentiate this so when the y's and x's become (y+dy)'s and (x+dx)'s the u's and w's will become (u+du)'s and (w+dw)'s and we have:

               u + du
      y + dy = ------
               w + dw

Now a few things to say here. Dividing two polynomials is a pretty difficult task...actually more tedious than anything. It is probably one of the hardest things a person can do in a high school Algebra class. That is what I am about to do here. I'm not going to be able to comment my steps because the way the work is going to be laid out but just try your best to follow along. It is not necessary to get this so do not worry if you are having trouble:

               u    du    u*dw
              --- + --- - ----
               w     w     w2
             |----------------
      w + dw | u + du
              /    u*dw \
             -|u + ---- |
              \     w   /
             -------------
                       u*dw
                  du - ----
                        w
                 /     du*dw \
                -|du + ----- |
                 \       w   /
                --------------
                     -u*dw   du*dw
                     ----- - -----
                       w       w
                    /-u*dw   u*dw2 \
                   -|----- - ----- |
                    \  w       w2  /
                  --------------------
                           -du*dw   u*dw2   // note:  we are now dealing with really small
                           ------ + -----             numbers so we can quit
                              w       w2

Our remainders started getting really small so we could just stop there. But, we are still not done. Take the quotient from above and combine the second and third term. We want to leave the first term as is so that when we subtract our original equation from the above everything will work out nicely. Here is the rest of the work:

      Start with this:

                u    du    u*dw
      y + dy = --- + --- - ----
                w     w     w2

           // get the second term to have the same base as the third (so you can combine
              the two) by multiplying it by (w/w)

                u    du*w    u*dw      u    du*w - u*dw
      y + dy = --- + ----- - ---- =   --- + -----------
                w      w2      w2       w        w2

           // Subtract the original equation [y=u/w] from the above:

           w*du - u*dw
      dy = -----------
               w2

           // divide everything by dx

             du     dw
           w*-- - u*--
      dy     dx     dx     <-- The quotient rule
      -- = ------------
      dx        w2

That right there is the derivation of the quotient rule. You use this when trying to differentiate two functions that are divided by each other. In English: the derivative of the quotient is the denominator times the derivative of the numerator minus the numerator times the derivative of the denominator, all over the denominator squared. This is probably the hardest rule there is to learn so let's look at an example:

      Differentiate:            x2
                       f(x) = ------
                              x2 + 1

          // Let [u=x2] and [w=x2+1] therefore the quotient rules tells us that:

             du     dw
           w*-- - u*--
      dy     dx     dx
      -- = ------------
      dx        w2

          // Plugging the functions we have in the above gives us:

                  d            d
           (x2+1)*--(x2) - (x2)*--(x2+1)
      dy          dx           dx
      -- = -----------------------------
      dx            (x2 + 1)2

          // The derivatives of [x2] and [x2+1] are pretty easy to evaluate so I'll do
             that first:

      dy   (x2+1)(2x) - (x2)(2x)
      -- = ---------------------
      dx        (x2 + 1)2

          // Next I'm going to simplify the numerator by solving the parentheses:

      dy   2x3 + 2x - 2x3
      -- = --------------
      dx     (x2 + 1)2

          // Simplify the numerator more by canceling out the positive and negative
             [2x3] term:

      dy      2x
      -- = ---------   <-- Final answer
      dx    (x2+1)2

And you can stop right about there. You could have expanded the binomial in the denominator but there really is no point since you cannot simplify it anymore. That was probably more difficult than our other examples but still pretty easy.

The next rule we are going to learn is the most useful one. It allows you to break up some of the most complex functions into small steps. For example, what if you were given something like this:

      Differentiate:   f(x) = (x3 - 1)3/4

That is: x to the third minus x all to the three-fourths power. None of the rules we have learned so far would help us. Even using the general rule we learned in the last tutorial would be extremely difficult (if not impossible). The new rule which will help us evaluate the above is called the chain rule. The chain rule is used for finding the derivatives of composite functions. If you did not see it then the above is a composite function. You could think of it like this:

      f(x) = g(h(x))  , where [g(x) = x3/4] and [h(x) = x3-1]

Do you see? You should if you read my tutorial on functions. If you did not, do so now. In a general form, the chain rules says:

     f(x) = g(h(x))   -->   f'(x) = g'(h(x)) * g'(x)

          // here is an easier notation to read:

     y = f(u) , where [u = g(x)], then:

     dy   dy   du
     -- = -- * --
     dx   du   dx

This rule is probably not making too much sense right now. An example will most likely help you out though:

      y = (x3 - 1)3/4

          // From the above we are going to make two other equations:

      u = x3 - 1

      y = u3/4

          // Do you agree that the above is saying pretty much the same as the first
             equation?  The chain rule says all we have to do is differentiate these
             two equations separately and then multiply their derivatives
             together...how easy!...

          // First differentiate u with respect to x

      du
      -- = 3x2
      dx

          // Next differentiate y with respect to u

      dy
      -- = (3/4)*u-1/4
      du

          // Then multiply the two derivatives together for differentiation of y with
             respect to x

      dy   dy   du
      -- = -- * -- = (3/4)*u-1/4 * 3x2
      dx   du   dx

          // Simplify the above a little:

      dy    9x2
      -- =  --- * u-1/4
      dx     4

          // Now that would be the final answer but remember that there was no u in the
             original equation so all we do is put [x3-1] back in for u

      dy   9x2
      -- = ---*(x3-1)-1/4
      dx    4

The above equation is a pretty nasty looking one but it is correct (I think...I double checked my work but I could have gone wrong somewhere).

The chain rule just happens to be our last rule to learn. There is still one more rule (implicit differentiation) and a few good techniques to learn but this is enough for one tutorial. I have a few more things I am going to write about differential calculus and then I plan on making a big list of sample problems as well as the answers (and work). Things to come are derivatives of the trigonometric functions, higher derivatives, real world applications, and maybe more. For those who have gotten this far...you should be proud of yourselves (seriously). When we are done with this calculus stuff we can look at physics topics how they are supposed to be viewed: with knowledge of calculus. Most people go through their first physics class without having calculus which is almost a waste of time. Once you know calculus not only does physics become interesting but it becomes easier. Well, that's enough for now.

Good luck.

- Brandon Williams