Throughout, β£ β
β£ |\cdot| β£ β
β£ is the absolute value on R \mathbb{R} R , so that d R ( s , t ) = β£ s β t β£ d_{\mathbb{R}}(s,t)=|s-t| d R β ( s , t ) = β£ s β t β£ for all real s , t s,t s , t by The Absolute Value Metric on the Real Line .
Write L = f β² ( x 0 ) L=f'(x_0) L = f β² ( x 0 β ) . Applying the definition of the derivative with Ξ΅ = 1 \varepsilon=1 Ξ΅ = 1 , there is a real Ξ΄ 1 > 0 \delta_1>0 Ξ΄ 1 β > 0 such that every real h h h with 0 < β£ h β£ < Ξ΄ 1 0<|h|<\delta_1 0 < β£ h β£ < Ξ΄ 1 β and x 0 + h β I x_0+h\in I x 0 β + h β I satisfies
β£ f ( x 0 + h ) β f ( x 0 ) h β L β£ < 1. \left|\frac{f(x_0+h)-f(x_0)}{h}-L\right|<1 . β h f ( x 0 β + h ) β f ( x 0 β ) β β L β < 1.
For such h h h , the triangle inequality for the absolute value gives
β£ f ( x 0 + h ) β f ( x 0 ) h β£ < β£ L β£ + 1 , \left|\frac{f(x_0+h)-f(x_0)}{h}\right|<|L|+1 , β h f ( x 0 β + h ) β f ( x 0 β ) β β < β£ L β£ + 1 ,
and multiplying by β£ h β£ > 0 |h|>0 β£ h β£ > 0 , together with β£ f ( x 0 + h ) β f ( x 0 ) β£ = β£ h β£ β
β£ f ( x 0 + h ) β f ( x 0 ) h β£ |f(x_0+h)-f(x_0)|=|h|\cdot\bigl|\frac{f(x_0+h)-f(x_0)}{h}\bigr| β£ f ( x 0 β + h ) β f ( x 0 β ) β£ = β£ h β£ β
β h f ( x 0 β + h ) β f ( x 0 β ) β β ,
β£ f ( x 0 + h ) β f ( x 0 ) β£ β€ ( β£ L β£ + 1 ) β β£ h β£ . ( β ) \bigl|f(x_0+h)-f(x_0)\bigr|\le(|L|+1)\,|h| . \tag{$*$} β f ( x 0 β + h ) β f ( x 0 β ) β β€ ( β£ L β£ + 1 ) β£ h β£. ( β )
Now let Ξ΅ β R \varepsilon\in\mathbb{R} Ξ΅ β R with Ξ΅ > 0 \varepsilon>0 Ξ΅ > 0 be given. Since β£ L β£ β₯ 0 |L|\ge0 β£ L β£ β₯ 0 , the number β£ L β£ + 1 |L|+1 β£ L β£ + 1 is positive, so
Ξ΄ = min β‘ { Ξ΄ 1 , Β Ξ΅ β£ L β£ + 1 } \delta=\min\Bigl\{\delta_1,\ \frac{\varepsilon}{|L|+1}\Bigr\} Ξ΄ = min { Ξ΄ 1 β , Β β£ L β£ + 1 Ξ΅ β }
is a positive real. Let y β I y\in I y β I satisfy d R ( x 0 , y ) < Ξ΄ d_{\mathbb{R}}(x_0,y)<\delta d R β ( x 0 β , y ) < Ξ΄ , that is β£ y β x 0 β£ < Ξ΄ |y-x_0|<\delta β£ y β x 0 β β£ < Ξ΄ , and put h = y β x 0 h=y-x_0 h = y β x 0 β , so that x 0 + h = y β I x_0+h=y\in I x 0 β + h = y β I .
If h = 0 h=0 h = 0 then y = x 0 y=x_0 y = x 0 β and d R ( f ( y ) , f ( x 0 ) ) = 0 < Ξ΅ d_{\mathbb{R}}(f(y),f(x_0))=0<\varepsilon d R β ( f ( y ) , f ( x 0 β )) = 0 < Ξ΅ . If h β 0 h\ne0 h ξ = 0 then 0 < β£ h β£ < Ξ΄ β€ Ξ΄ 1 0<|h|<\delta\le\delta_1 0 < β£ h β£ < Ξ΄ β€ Ξ΄ 1 β , so ( β ) (*) ( β ) applies and
d R ( f ( y ) , f ( x 0 ) ) = β£ f ( y ) β f ( x 0 ) β£ β€ ( β£ L β£ + 1 ) β β£ h β£ < ( β£ L β£ + 1 ) β Ξ΄ β€ ( β£ L β£ + 1 ) β Ξ΅ β£ L β£ + 1 = Ξ΅ . d_{\mathbb{R}}\bigl(f(y),f(x_0)\bigr)=\bigl|f(y)-f(x_0)\bigr|\le(|L|+1)\,|h|<(|L|+1)\,\delta\le(|L|+1)\,\frac{\varepsilon}{|L|+1}=\varepsilon . d R β ( f ( y ) , f ( x 0 β ) ) = β f ( y ) β f ( x 0 β ) β β€ ( β£ L β£ + 1 ) β£ h β£ < ( β£ L β£ + 1 ) Ξ΄ β€ ( β£ L β£ + 1 ) β£ L β£ + 1 Ξ΅ β = Ξ΅ .
In both cases d R ( f ( y ) , f ( x 0 ) ) < Ξ΅ d_{\mathbb{R}}(f(y),f(x_0))<\varepsilon d R β ( f ( y ) , f ( x 0 β )) < Ξ΅ . Since Ξ΅ > 0 \varepsilon>0 Ξ΅ > 0 was arbitrary, f f f is continuous at x 0 x_0 x 0 β relative to I I I . β \blacksquare β