Linear Approximations

After studying the differentials, we know that if Δy=f(a+Δx)f(a)\Delta{y}={f{{\left({a}+\Delta{x}\right)}}}-{f{{\left({a}\right)}}} and dy=f(x)Δx{d}{y}={f{'}}{\left({x}\right)}\Delta{x}. This means that Δx\Delta{x} becomes very small, i.e. if we let Δx0\Delta{x}\to{0}, we can write that dyΔy{d}{y}\approx\Delta{y}.

This can be rewritten as f(a+Δx)f(a)f(x)Δx{f{{\left({a}+\Delta{x}\right)}}}-{f{{\left({a}\right)}}}\approx{f{'}}{\left({x}\right)}\Delta{x}.

If we now let x=a+Δx{x}={a}+\Delta{x}, we obtain that f(x)f(a)f(x)(xa){f{{\left({x}\right)}}}-{f{{\left({a}\right)}}}\approx{f{'}}{\left({x}\right)}{\left({x}-{a}\right)}.

Definition. The linearization of a function y=f(x){y}={f{{\left({x}\right)}}} at a point a{a} is f(x)f(a)+f(x)(xa){f{{\left({x}\right)}}}\approx{f{{\left({a}\right)}}}+{f{'}}{\left({x}\right)}{\left({x}-{a}\right)}.

We need this linearization because it might be easy to calculate the value of a function f(a){f{{\left({a}\right)}}} but it might be difficult (or even impossible) to compute the nearby values of f{f{}}. So, we settle for the easily computed values of the linear function (tangent line).

Example 1. Find the linear approximation of y=x{y}=\sqrt{{{x}}} at (4,2){\left({4},{2}\right)}. Then approximate 3.99\sqrt{{{3.99}}}, 4.01\sqrt{{{4.01}}}, and 4.05\sqrt{{{4.05}}}.

Since f(x)=12x{f{'}}{\left({x}\right)}=\frac{{1}}{{{2}\sqrt{{{x}}}}}, we have that f(4)=14{f{'}}{\left({4}\right)}=\frac{{1}}{{4}}.

Therefore, linear approximation L(2)=14(x4)+2{L}{\left({2}\right)}=\frac{{1}}{{4}}{\left({x}-{4}\right)}+{2}, or x14x+1\sqrt{{{x}}}\approx\frac{{1}}{{4}}{x}+{1}.

3.99143.99+1=1.9975\sqrt{{{3.99}}}\approx\frac{{1}}{{4}}{3.99}+{1}={1.9975}. The true value is given for comparison: 3.991.997498435544\sqrt{{{3.99}}}\approx{1.997498435544}.

4.01144.01+1=2.0025\sqrt{{{4.01}}}\approx\frac{{1}}{{4}}{4.01}+{1}={2.0025}. The true value is given for comparison: 4.012.00249843945\sqrt{{{4.01}}}\approx{2.00249843945}.

4.05144.05+1=2.0125\sqrt{{{4.05}}}\approx\frac{{1}}{{4}}{4.05}+{1}={2.0125}. The true value is given for comparison: 4.052.01246117974981\sqrt{{{4.05}}}\approx{2.01246117974981}.

As can be seen, the approximate values are overestimated. That's because the tangent line at x=4{x}={4} lies above f(x)=x{f{{\left({x}\right)}}}=\sqrt{{{x}}}.

Let's do another example.

Example 2. Suppose that the temperature of tea is 950C95^{0}C. After one minute, its temperature is 850C85^{0}C. Find the temperature of the tea in two minutes.

If T(t){T}{\left({t}\right)} represents the temperature of the tea after t{t} minutes, T(0)=95{T}{\left({0}\right)}={95}, T(1)=85{T}{\left({1}\right)}={85}.

In order to make linear approximation with a=1{a}={1}, we need to approximate T(1){T}'{\left({1}\right)}.

Since T(1)=limt1T(t)T(1)t1{T}'{\left({1}\right)}=\lim_{{{t}\to{1}}}\frac{{{T}{\left({t}\right)}-{T}{\left({1}\right)}}}{{{t}-{1}}}, we can't find the exact value of the derivative, but we could estimate T(1){T}'{\left({1}\right)} by taking t=0{t}={0}: T(1)T(0)T(1)01=958501=10{T}'{\left({1}\right)}\approx\frac{{{T}{\left({0}\right)}-{T}{\left({1}\right)}}}{{{0}-{1}}}=\frac{{{95}-{85}}}{{{0}-{1}}}=-{10}.

Thus, we approximated the instantaneous rate of change by the average rate of change between t=0{t}={0} and t=1{t}={1}, which is 100C-10^{0}C per minute. Thus, T(2)T(1)(21)+T(1)=10(32)+85=750C{T}{\left({2}\right)}\approx{T}'{\left({1}\right)}{\left({2}-{1}\right)}+{T}{\left({1}\right)}=-{10}{\left({3}-{2}\right)}+{85}={{75}}^{{0}}{C}.