You are on page 1of 16

Giallombardo - 1

Nick Giallombardo
AP Calculus
23 March 2015
Mrs. Tallman
Riemann Sums
Integrals are one of the main pillars of calculus. There are many complex
concepts behind the integral that must be understood. However, before those can
comprehended, one must become familiar with Riemann sums, trapezoid rule, and
Simpsons rule, which are simpler versions of the integral that may at times be easier to
visualize.
The Riemann sum is essentially the approximation of a definite integral, or area
under a given graph, using rectangles. These rectangles, since they are being placed
under a curve, can overestimate or underestimate this definite integral. Lets take for
example, the graph in Figure 1.

Figure 1. Example Riemann Sum Graph


Figure 1 displays the approximation of a definite integral for f(x) using Riemann
sums. Each of the sample points c1, c2, etc., determines what height f(c1), f(c2), each

Giallombardo - 2
rectangle will be. The range of the integral is set by a, the lower bound, and b, the upper
bound. Each of these rectangles has a certain width measured by x, which happens to
vary in this case. They are said to be separated by these strips. Formally, these strips
are said to partition the interval [a, b] into n increments. Thus, the area of each
rectangle is f(x) multiplied by the change in x for each particular rectangle. The area
under the curve is found by adding up the areas of each individual rectangle. The
general formula is shown below.
Rn= x ( f ( x 1 ) ) + x ( f ( x 2 ) ) + .
With a greater amount of rectangles, the area under the curve becomes more accurate
since there is a smaller amount of rectangle that lies above or below the curve. On the
flipside, when there are a lower number of rectangles, the accuracy is compromised
since the rectangles are not as capable to account for the area under the curve since it
does not give off the more rounded look. There are 5 types of Riemann sums that
decide where the sample point lands on the graph: upper (sample point is at highest
point for f(x), lower (sample point is at lowest point for f(x), right (sample point is farthest
positive on x), left (sample point is least farthest positive on x), and midpoint (sample
point is in the middle of the subinterval).
The next approximation of definite integrals is the trapezoid rule. They are similar
to Riemann sums in that they are separated into several subintervals, however they
come in the form of trapezoids rather than rectangles and these bases are the
subintervals and its sides connect the sample points. The height, in turn, is x.

Giallombardo - 3

Figure 2. Trapezoid Rule Graph


Figure 2 displays an example of the trapezoid rule for a given graph f(x). It is
seen that when the graph f(x) is concave up, the trapezoid rule overestimates the
definite integral. When the graph f(x) is concave down, the trapezoid rule
underestimates the definite integral. The areas of each trapezoid are found by
averaging the f(x) value of each side of the subinterval and multiplying it by x. The
areas of all of the trapezoids are added up to find the estimation of the definite integral.
Again, as more trapezoids are added and there are smaller subintervals, the
approximation becomes more accurate and vice versa. The general formula appears as
follows.
T = x

f x 1 ) +f ( x 2 )
f ( x 2 ) + f ( x3 )
f ( x 3 ) +f ( x 4 )
+x
+ x
+
2
2
2

((

) (

) (

Simpsons rule is the third and final estimation used for definite integral. Instead
of using definite shapes, this rule utilizes small parabolic regions. Thus, it is able to take
into account the curvature of f(x) more than the other approximations.

Giallombardo - 4

Figure 3. Simpsons Rule Example Graph


Figure 3 illustrates how Simpsons rule estimates a definite integral for a given
graph f(x). The rule underestimates for intervals that are concave down and
overestimates when the interval is concave up. This rule is the only one of the three that
requires a specific number of data points, and that is there must be an even value of
sample points for the formula to work. The first and last f(x) terms (at a and b of the
definite integral) are only multiplied by one, while the second f(x) value is multiplied by
two, then the next is multiplied by four, and this alternation of multiplying by two and four
is repeated until the last term, xn. The general equation appears as follows.
S =

x
(f ( x 0 ) +4 f ( x 1 ) +2 f ( x 2) + 4 f ( x3 ) + +2 f ( x n2 ) +4 f ( x n1 )+ f ( x n ) )
3
All 3 of these methods are some of the most accurate methods to approximate a

definite integral. Each use a given amount of subintervals that are comprised of varying
shapes. The areas of these shapes are calculated and then added together in order to
estimate the curves area in a more simplified manner. Riemann sums using rectangles,
trapezoid rules using trapezoids, and Simpsons rule using parabolic regions, all slightly

Giallombardo - 5
differ in how the subintervals areas are calculated. Riemann and trapezoid are more
diverse in that they can use any number of subintervals, while Simpsons rule is limited
to instances where there is an even number of subintervals. The accuracy of these
methods also differs between each other. Riemann sums are considered to be least
accurate, followed by trapezoid rule, and Simpsons rule is normally the best estimation
for definite integrals. This can be seen in the graphs used for each method. When using
Riemann sums, there is often a lot of area left above or under the graph that throws off
the accuracy of the formula when f(x) varies inside of the subinterval. Trapezoid rule
creates a straight line from each sample point, which decreases this margin of error
between subintervals, but curvature is still not as accounted for. Since Simpsons rule
accounts for this curvature, the fluctuation of f(x) values is more accounted for and the
margin of error is minimized as much as possible. Through it all, however, the accuracy
is affected the most by the number of subintervals used. When this number is
increased, curvature can be accounted for in a more efficient manner and there is a
smaller margin of error. When there is a small amount of subintervals, the variation in
f(x) values cannot be accounted for.
Let us apply this concept of Riemann sums using the function
f(x) = (x-3)4 + 2(x-3)3 4(x-3) + 5. We can estimate the definite integral between [1, 5}
using all five methods (upper, lower, midpoint, left, right). Since there will be two
subintervals for each example, there will be a x of 2 because the interval is 4 units long
and divided by 2 would equal 2. The f(x) values for these approximations were either
found with a table of values or a max and min function on the TI-Nspire graphing
calculator.

Giallombardo - 6
The first calculation is for the upper Riemann sum.

Figure 4. Upper Riemann Sum


Figure 4 displays the graph of the upper Riemann sum for f(x). The graph
overestimates this definite integral since the graph is concave up. The change in x is
again multiplied by the f(x) value for the corresponding type of Riemann sum. In this
case, the upper sum indicates that the highest f(x) value in each subinterval would be
used, f(1) and f(5). Those values are plugged in to the following formula in order to
estimate the definite integral.
R= xf ( 1 ) + xf ( 5 )

R=213+229

R=84 sq . un .

In contrast, the lower Riemann sum is calculated next. The graph grossly
underestimates the definite integral.

Giallombardo - 7

Figure 5. Lower Riemann Sum


Figure 5 displays the graph for the lower Riemann sum of f(x). The f(x) values
used in this case are the lowest f(x) values in each subinterval. In this case, f(3) and
f(3.6777) qualify as those values. These values are plugged in to the following formula
to approximate the definite integral.
R= xf ( 3 )+ xf (3.6777 )

R=25+23.1226

R 16.25 sq . un.

The next Riemann sum that was solved for was the left Riemann sum.

Giallombardo - 8

Figure 6. Left Riemann Sum


Figure 6 displays the graph of the left Riemann sum. The f(x) values used are the
farthest left in each subinterval, f(1) and f(3) in this case. The approximation is
calculated using the following steps.
R= xf ( 1 ) + xf ( 3 )

R=213+25

R=26+ 10

R=36 sq . un .

In contrast, the next Riemann sum solved for is the right Riemann sum.

Giallombardo - 9

Figure 7. Right Riemann Sum


Figure 7 displays the graph for the right Riemann sum. The f(x) value used for
this version of the approximation is the value at the farthest right of each subinterval. In
this case, f(3) and f(5). The calculation runs as follows.
R= xf ( 3 )+ xf (5 )

R=25+229

R=68 sq . un.

The fifth and final Riemann sum used is the midpoint Riemann sum.

Giallombardo - 10

Figure 8. Midpoint Riemann Sum


Figure 8 is the graph of the midpoint Riemann sum for f(x). The f(x) values used
to approximate the definite integral in this case are the midpoints of each subinterval,
f(2) and f(4) in this case. Those values are plugged in to the formula that follows in order
to estimate the definite integral.
R= xf ( 2 ) + xf ( 4 )

R=28+24

R=24 sq . un .

Now, lets try finding the estimated definite integral of the same function f(x)
within the same interval [1, 5], but now utilizing the trapezoid rule. Now, we will use 4
subintervals which will give us a change in x of one per subinterval.

Giallombardo - 11

Figure 9. Trapezoid Rule for f(x)


Figure 9 displays the trapezoid rule approximation for f(x). The sample points
used are at f(1), f(2), f(3), f(4), and f(5). The trapezoids overestimate the definite integral
since the graph is concave up, except from x=2 to x=3 where the graph is concave
down and the graph is underestimated. Once the values at each f(x) are found, they are
plugged in and the calculations run as follows.

T = x

T =1

( f ( 1) +f2 ( 2) )+ x ( f ( 2) +2 f ( 3) )+ x ( f ( 3) +2 f ( 4 ) )+ x ( f ( 4 ) +f2 ( 5) )

( 13+2 8 )+1( 8+52 )+ 1( 5+24 )+1( 4+229 )

T =1

( 212 )+ 1( 132 )+1( 92 )+1( 332 )

T =38 sq . un .
The trapezoid rule produces a more accurate estimation of the definite integral as
a result of using a larger amount of subintervals and the nature of the trapezoid rule
being more accurate than that of Riemann sums.

Giallombardo - 12
The mean value theorem for integrals formally states that if a function, f , is
continuous in the interval [a, b], then there exists a number, c, in the closed interval [a,
b] such that:

f (c)

1 b
a f (x)dx
b a

This essentially means that the area found by this definite integral is equal to the area of


a rectangle created by multiplying
f(c) and the change in

x , the difference the upper

and lower bounds of integration. We can solve for the area under the curve of the
original function f(x) within the same interval [1, 5] using two subintervals [1, 3] and [3,
5] by using MVT for integrals. Referring to the following calculations, c 1 is considered
the rectangle in the first interval [1, 3] while c 2 is the rectangle in the second interval [3,
5]. C represents the x-value needed to find the appropriate height or f(x) value. Since
the integral is being solved for, the equation is evaluated at the upper bound and lower
bound of each subinterval.
3

1 ( x3 ) ( x3 )
f ( c1 ) =
+
2 ( x3 )2+ 5 x
2
5
2

f ( c1 ) =

1
( ( x3 )4 +2 ( x3 )34 ( x3 ) +5 ) dx

31 1

f ( c1 ) =

1 ( 33 ) ( 33 )
1 ( 13 ) (13 )
2
2
+
2 ( 33 ) +53
+
2 ( 13 ) +51
2
5
2
2
5
2

1
f ( c1 ) = (( 15 )(1.4))
2

) (

f ( c1 ) =8.2

The equations for the height of the second rectangle run as follows. Keep in mind that
after the equation is integrated, they are evaluated at the boundaries of the integral and
their difference is found.

Giallombardo - 13
5

5
4
1 ( x3 ) ( x3 )
+
2 ( x 3 )2+ 5 x
2
5
2

f ( c2 ) =

1
( ( x3 )4 +2 ( x3 )34 ( x3 ) +5 ) dx
31 3

f ( c2 ) =

5
4
5
4
1 ( 53 ) ( 53 )
1 ( 33 ) ( 33 )
+
2 ( 53 )2 +55
+
2 ( 33 )2 +53
2
5
2
2
5
2

1
f ( c2 ) = ( ( 31.4 ) (15 ) )
2

) (

f ( c2 ) =

f ( c2 ) =8.2

Figure 10. MVT Graph Example


The graph in Figure 10 displays the function f(x) along with the two rectangles
that can be used to solve for the definite integral on this interval. Although this may look
like it is over or underestimating the definite integral, it is actually equal. This is not an
approximation for the definite integral even though it is using rectangles, for MVT will
solve for the exact definite integral. The formula for this rectangle is fairly simple and
similar to Riemann sums, for it is just the change in x along the subinterval multiplied by
the sum of the heights that was solved for previously. If it were to be distributed, it would

Giallombardo - 14
appear as the addition of two sums of rectangles. The values plugged in to the formula
appear as follows.
A rea= x( f ( c1 ) + f ( c 2 ) )

A rea=2( 8.2+8.2 )

A rea=32.8 sq .un .

According to these equations, the area under f(x) on the [1, 5] interval is 32.8
square units. The left Riemann sum proved to be the most accurate approximation in
this case, for its difference in over and under approximation cancelled each other out
enough to where the margin of error was decreased greatly.
Lets apply these concepts to a real world situation. The radius of a spherical
balloon, in feet, is modeled by a twice-differentiable function r of time t, where t is
measured in minutes. For 0 < t < 12 the graph is concave down. The radius of the
balloon is 32 feet when t = 7.
Let us estimate the radius of the balloon when t = 7.2. We will do so using value
of r(7) and estimating how much the radius would have changed based off of the
provided r(t) value in a given table. The equations to do so appear as follows.
r ( 7.2 ) r ( 7 ) +r ' ( 7 ) t

r ( 7.2 ) r ( 7 ) +r ' ( 7 ) (7.27 )

r ( 7.2 ) 32+1.40.2

r ( 7.2 ) r ( 7 ) +r ' ( 7 )0.2

r ( 7.2 ) 32.28 feet

The derivative is multiplied by the difference in time and added to the original radius.
The graph is concave down in this case, so we can assume that the graph is
overestimating the new radius that was solved for.
Using this knowledge, we can also find the rate of change of the volume of the
balloon with respect to time at t = 7 using the following equations, assuming that we
know the volume of a sphere.

Giallombardo - 15
4
V = r2
3

dV 4
dr
= 3 r 2
dt 3
dt

dV
dr
=4 r 2
dt
dt

dV
=4 3221.4
dt

dV
=18015.15 ft cubed / min The equation was derived in order for the previous and
dt
assumed values to be plugged in and calculated in order to yield the rate of the change
in the volume of the balloon which happens to be 18015.15 cubic feet per minute as per
the formula.
Let us now use a right Riemann Sum with 5 subintervals in order to approximate

12

the integral

r ' ( t ) dt

. Recall that Riemann sums use rectangles and their areas to

approximate area. The change in x is multiplied by the r(t) value for each interval.
R= x r ' ( t 1 ) + x r ' ( t 2 ) + x r ' ( t 3 ) + x r ' ( t 4 )+ x r ' ( t 5 )
R=14 +32+31.4+ 40.5+10.4

R5=16.6 feet The 5 subintervals were taken

from the 5 different sets of points following each other on the table, which was actually
composed of six sets of points. A right Riemann sum was being used, so the change in
x for each subinterval was found by subtracting the farther left or lower t-values point
from the right or higher t-values point. Again, the change in

was multiplied by its

appropriate r(t) value, or the one with t-value to the farthest right. This approximation
produces the change of the radius in feet from the beginning at t=0 to the end of the
table at t=12, which was 16.6 feet.

Giallombardo - 16
The approximation is less than the actual value of the definite integral of

12

r ' ( t ) dt

. This is because the graph is concave down, and the interval given

produces a negatively sloping graph, so all of the rectangles actually end up under the
curve which ends up in an underestimation.

Works Cited
"Calculus II - Approximating Definite Integrals." Pauls Online Notes. Paul Dawkins, n.d.
Web. 21 Mar. 2015.
Foerster, Paul A. Calculus: Concepts and Applications. Berkeley, CA: Key Curriculum,
1998. Print.
"Integration." Computing for Engineers @ JCU. PBWorks, n.d. Web. 21 Mar. 2015.
"Mean Value Theorem for Integrals." Desmos Graphing Calculator. N.p., n.d. Web. 22
Mar. 2015.
"Riemann Sum." Wolfram MathWorld. N.p., n.d. Web. 20 Mar. 2015.

You might also like