16 4
6
Lecture 14 :Linear Approximations and Differentials
2
Consider a point on a smooth curve y = f (x), say P = (a, f (a)), If we draw a tangent line to the curve
at the point P14
6, we can see from the pictures below that as we zoom in towards the point P , the path
10
of the curve is very close to that of the tangent line. If we zoom in far enough the curve looks almost
linear near the point P . (See below the tangent to the curve y = x2 at the point (1, 1) and the tangent
3
9
to the curve y = sin x at the point ( π2 , 1). )
4
12
2
8
4
2
6
10
5
5) = sin(x)
4
2 5 10 15
)=1 3
8
–2
2
– 46
2
5! – 2! 3! –! ! ! ! 3! 2! 5! 3! 7! 4! 9!
– –
1
2 2 2 2 2 2 2 2
–1
–6
–2
4
On the other hand, if we pick a sharp point on a curve, the curve does not look linear near the point as
–3
we zoom in. See the graph of y = |x| below at the point (0, 0).
–4
–8
2
–2
–5
–6
– 10
! –7 ! ! 3!
– –8
2 – 12 25 10 2
The difference between the nature of sharp points and points where the curve is smooth is differentiabil-
– 14
ity. If a curve –1
–is2differentiable on an interval containing the point x = a (smooth at x = a), the points
on the curve, with x values close to a, are very close to the points on the tangent line to the curve at
P . The equation of the tangent line to the curve y = f (x) at x = a is given by,
–4
–2 –2
– 16 L(x) = f 0 (a)(x − a) + f (a).
–4
– 18
1
–6
–2
We can conclude that if f is differentiable in an interval containing a, then
f (x) ≈ L(x) = f (a) + f 0 (a)(x − a).
This is called the linear approximation or Tangent Line Approximation to f (x) at x = a.
The linear function, whose graph is the tangent line to the curve y = f (x) at x = a is called the
Linearization of f at a.
√
Example (a) Find the linearization of the function f (x) = 3 x at a = 27.
f (x) f 0 (x) a f (a) f 0 (a)
L(x) = f (a) + f 0 (a)(x − a) =
√
3
√
(b) Use the linearization above to approximate the numbers 27.01 and 3 26.99.
√
(c) We can get an approximation for 3 x from the linearization of the function, L(x), above, for any
x in the interval 26 ≤ x ≤ 28. We can see, from the table
√ below, that the closer the value of x gets to
27, the better the approximation to the actual value of x.
3
√
√f (x) x From L(x) Actual Value 3
x
3
26.5 26.5 2.9814815 2.9813650
√
3
26.9 26.9 2.996296 2.996292
√
3
26.99 26.99 2.9996296 2.9996296
√
3
27 27 3 3
√
3
27.01 27.01 3.0003704 3.0003703
√
3
27.1 27.1 3.0037037 3.0036991
√
3
27.5 27.5 3.0185185 3.0184054
√
Error of approximation In fact by zooming in on the graph of f (x) = 3 x, you will see that
√ √
| 3 x − L(x)| < 0.001 or − 0.001 < 3 x − L(x) < 0.001
2
when x is in the interval 26.5 ≤ x ≤ 27.5.
Such bounds on the error are useful when using approximations. We will be able to derive such
estimates later when we study Newton’s method.
√
Example (a) Find the linearization of the function f (x) = x + 9 at a = 7.
f (x) f 0 (x) a f (a) f 0 (a)
L(x) =
√ √
(b) Use the linearization above to approximate the numbers 16.03 and 15.98.
Example (commonly used linearizations) (a) Find the Linearization of the functions sin θ and
cos θ at θ = 0.
f (θ) f 0 (θ) a f (a) f 0 (a)
cos(θ) ≈
sin(θ) ≈
π π π
(b) Estimate the value of sin( 100 ), cos 95 and sin 2o = sin 90 .
3
New Notation : ∆y.
Suppose f is differentiable
6.5
on an interval containing the point a. Linear approximation says that the
function f can be approximated by
6
f (x) ≈ f (a) + f 0 (a)(x − a)
where a is fixed and x is a5.5point (in the interval) nearby. This can gives us the following approximation
for the change in function values, when we have a small change in the value of x:
5
f (x) − f (a) ≈ f 0 (a)(x − a).
4.5
As before, we use ∆x to denote
4 a small change in x values. In this case ∆x = x−a. and ∆y = f (x) − f (a)
to denote the corresponding change in the values of y or f (x). This gives us:
3.5
∆y ≈ f 0 (a)∆x
3
where ∆y denotes the change in the value of f between two points a and a + ∆x.
2.5
g (x ) = x L(x)
2
h (x) = 1 + 0.5· (x – 1) f(x)
1.5
f ' (a) x
y
(a, f(a))
1
x
0.5
Example Approximate the change in the surface area of a spherical hot air balloon when the
radius changes from 4 to 3.9 meters. (The surface area of a sphere of radius r is given by S = 4πr2 . )
–3 –2 –1 1 2 3 4 5
0
– 0.5 ∆S ≈ S (a)∆r
–1
S(r) = 4πr2 , S 0 (r) = 8πr.
– 1.5
∆S ≈ S 0 (4)∆r = 32π(−0.1) = −3.2π.
–2
Differentials, dy
We also use the notation– 2.5of differentials to denote changes in L(x), the linear approximation to f (x)
near a.
–3
• The change in the function values on the curve y = f (x) as x changes from a to a + ∆x is denoted
by ∆y as before. (∆y
– 3.5 = f (a + ∆x) − f (a)).
• The differential dx –is4 defined as the change in x, (dx = ∆x.)
• The differential dy is defined as the change in the values of the linear approximation L(x) as x
– 4.5
changes from a to a + ∆x; dy = ∆L = L(a + ∆x) − L(a) = f (a) + f 0 (a)∆x − f (a) = f 0 (a)∆x.
Therefore –5
dy = f 0 (a)∆x in terms of dx: dy = f 0 (a)dx.
4
4
3.5
The differential dy is a dependent variable, depending on the independent variable dx. Check out the
difference between dy and3 ∆y on the graph below.
2.5
g (x ) = x L(x)
2
h (x) = 1 + 0.5· (x – 1) f(x)
1.5
f ' (a) x = L = dy
y
(a, f(a))
1
x = dx
0.5
–3
We can
–2
rephrase– 1our results on linear 1approximation
2
as 3 4 5
– 0.5 dy ≈ ∆y
Example Compare the– 1values of ∆y and dy if y = 3x4 + 2x + 1 and x changes from 2 to 2.04.
a=2 – 1.5
∆y = f (a + ∆a) − f (a) = f (2.04) − f (2) =
57.0367 − 53 = 4.0367 = ∆y
–2
∆x = .04
– 2.5 dy = f 0 (a)dx = (12(4) + 2)(.04) = 98(.04)
f (x) = 3x4 + 2x + 1 f 0 (x) = 12x3 + 2 = 3.92 = dy
–3
Example (a) Find the– 3.5differential for the function y = 3 cos2 x.
–4
dy = f 0 (x)dx = (6 cos x)(− sin x)dx = −6 cos x sin xdx
– 4.5
(b) Use the differential to approximate the change in the values of the function f (x) = 3 cos2 x when
we have a small change in
–5
the value of x, dx at x = π4 .
π π π 1 1
f (x + ∆x) − f (x) ≈ dy = f 0 ( )dx = −6 cos( ) sin( )dx = −6 √ √ dx = −3dx.
4 4 4 2 2
(c) Use differentials to estimate 3 cos2 (44o ) = 3 cos2 ( π4 − π
180
).
π π π π 3 π
3 cos2 (44o ) = 3 cos2 − ≈ 3 cos2 −3 − = + .
4 180 4 180 2 60
Estimating Error
Note that ∆y measures the resulting error in our value for the variable y if we make a mistake in our
calculations of the variable x of size ∆x, ∆y = f (x + ∆x) − f (x). We saw above that dy ≈ ∆y and we
can use differentials to approximate the maximum error in our calculations for y when we have some
bound on our error for the variable x.
Example The radius of a spherical hot air balloon was estimated to be 4 meters with a possible error
of at most 0.5 meters. What is the maximum error you can make in calculating the surface area of the
balloon using the estimate of 4 meters?
5
S = 4πr2 m2
We need bounds for ∆S here, but we will instead use the linear approximation dS ≈ ∆S to approximate
the error.
dS = 8πrdr
When r = 4,
dS = 32πdr
If −0.5 ≤ dr ≤ 0.5, then
−0.5(32π) ≤ dS ≤ 0.5(32π)
or
−50.26 ≤ dS ≤ 50.26.
We can interpret this result as saying that if our estimate of 4 meters for the radius of the balloon is
off by at most 0.5 meters, then our estimate of the surface area of the balloon is off by (approximately)
at most 50.26 meters squared in absolute value.
∆S ∆S
We can also find bounds for the Relative Error = S
and the Percentage Error = S
· 100%.
∆S dS
relative error = ≈
S S
50.26
When r = 4, S = 4π(16) = 201.06. From our calculations above the relative error is at most 201.06
= .25.
The ma percentage error is at most 25%.
6
Commonly Used Linear Approximations
Note that if x ≈ 0, we get the following approximations for some commonly used functions using
Linear approximation:
1. sin x ≈ x if x ≈ 0
2. cos x ≈ 1 if x ≈ 0
√
3. 1 + x ≈ 1 + 12 x if x ≈ 0
4. (1 + x)r ≈ 1 + rx if x ≈ 0.
Recall for x ≈ 0, f (x) ≈ f (0) + f 0 (0)x.
The above results come from the following table which you should verify:
f (x) f (0) f 0 (x) f 0 (0)
sin x 0 cos x 1
cos x 1 − sin x 0
(1 + x)r 1 r(1 + x)r−1 r
