Each result cited below is universally quantified over the data in its own statement.
Let c c c , r r r and Ξ· \eta Ξ· be as in A Continuously Differentiable Map Whose Jacobian Matrix is Close to the Identity Maps a Cube into a Slightly Larger Cube Β§cube , fix x β Q ( c , r ) x\in Q(c,r) x β Q ( c , r ) , and put h = x β c h=x-c h = x β c , so that h k = x k β c k h_{k}=x_{k}-c_{k} h k β = x k β β c k β and β£ h k β£ β€ r |h_{k}|\le r β£ h k β β£ β€ r for every k β [ d ] k\in[d] k β [ d ] . Fix i β [ d ] i\in[d] i β [ d ] . The i i i th coordinate of G ( c ) G(c) G ( c ) is G i ( c ) G_{i}(c) G i β ( c ) , so it suffices to show β£ G i ( x ) β G i ( c ) β£ β€ ( 1 + Ξ· ) r |G_{i}(x)-G_{i}(c)|\le(1+\eta)r β£ G i β ( x ) β G i β ( c ) β£ β€ ( 1 + Ξ· ) r ; note that ( 1 + Ξ· ) r (1+\eta)r ( 1 + Ξ· ) r is nonnegative.
Step 1 (Derivative along the segment). The function G i G_{i} G i β is of class C 1 C^{1} C 1 on the open set R d \mathbb{R}^{d} R d (read through clause 3 of C^k Maps on a Euclidean Open Set ), so by A Real-Valued C^1 Function is Differentiable at Every Point it is differentiable at every point of R d \mathbb{R}^{d} R d . Let J = ( β 1 , 2 ) = { Ο β R : β 1 < Ο < 2 } J=(-1,2)=\{\tau\in\mathbb{R}:-1<\tau<2\} J = ( β 1 , 2 ) = { Ο β R : β 1 < Ο < 2 } . It is an interval : if u , w β J u,w\in J u , w β J and u β€ v β€ w u\le v\le w u β€ v β€ w , then β 1 < v < 2 -1<v<2 β 1 < v < 2 by claim 2 of Elementary Order Arithmetic in an Ordered Field . Every Ο β J \tau\in J Ο β J is an interior point of J J J , the midpoints Ο β 1 2 \frac{\tau-1}{2} 2 Ο β 1 β and Ο + 2 2 \frac{\tau+2}{2} 2 Ο + 2 β lying in J J J and satisfying Ο β 1 2 < Ο < Ο + 2 2 \frac{\tau-1}{2}<\tau<\frac{\tau+2}{2} 2 Ο β 1 β < Ο < 2 Ο + 2 β . Define F : J β R F:J\to\mathbb{R} F : J β R by F ( Ο ) = G i ( c + Ο h ) F(\tau)=G_{i}(c+\tau h) F ( Ο ) = G i β ( c + Ο h ) . For every Ο β J \tau\in J Ο β J , Chain Rule Along an Affine Path (with U = R d U=\mathbb{R}^{d} U = R d , open by claim 1 of Euclidean Space is Open in Itself, and C k C^k C k Maps are Continuous , base point c c c , direction h h h and Ο 0 = Ο \tau_{0}=\tau Ο 0 β = Ο ) shows that F F F is differentiable at Ο \tau Ο with
F β² ( Ο ) = β j = 1 d β j G i ( c + Ο h ) β h j . F'(\tau)=\sum_{j=1}^{d}\partial_{j}G_{i}(c+\tau h)\,h_{j}. F β² ( Ο ) = j = 1 β d β β j β G i β ( c + Ο h ) h j β .
Step 2 (Mean value). By Mean Value Theorem on an Open Interval , applied to F F F on the open interval ( β 1 , 2 ) (-1,2) ( β 1 , 2 ) with the points 0 < 1 0<1 0 < 1 of that interval, there is ΞΎ β ( 0 , 1 ) \xi\in(0,1) ΞΎ β ( 0 , 1 ) with F ( 1 ) β F ( 0 ) = F β² ( ΞΎ ) ( 1 β 0 ) F(1)-F(0)=F'(\xi)(1-0) F ( 1 ) β F ( 0 ) = F β² ( ΞΎ ) ( 1 β 0 ) . Since F ( 1 ) = G i ( x ) F(1)=G_{i}(x) F ( 1 ) = G i β ( x ) and F ( 0 ) = G i ( c ) F(0)=G_{i}(c) F ( 0 ) = G i β ( c ) , with z = c + ΞΎ h z=c+\xi h z = c + ΞΎ h we get
G i ( x ) β G i ( c ) = β j = 1 d β j G i ( z ) β h j . G_{i}(x)-G_{i}(c)=\sum_{j=1}^{d}\partial_{j}G_{i}(z)\,h_{j}. G i β ( x ) β G i β ( c ) = j = 1 β d β β j β G i β ( z ) h j β .
Step 3 (The intermediate point lies in the cube). For k β [ d ] k\in[d] k β [ d ] we have z k β c k = ΞΎ h k z_{k}-c_{k}=\xi h_{k} z k β β c k β = ΞΎ h k β , and β£ ΞΎ h k β£ = β£ ΞΎ β£ β β£ h k β£ = ΞΎ β£ h k β£ |\xi h_{k}|=|\xi|\,|h_{k}|=\xi|h_{k}| β£ ΞΎ h k β β£ = β£ ΞΎ β£ β£ h k β β£ = ΞΎ β£ h k β β£ by claim 4 of Properties of the Absolute Value in an Ordered Field and Absolute Value in an Ordered Field , as 0 < ΞΎ 0<\xi 0 < ΞΎ . Since ΞΎ β€ 1 \xi\le1 ΞΎ β€ 1 and 0 β€ β£ h k β£ 0\le|h_{k}| 0 β€ β£ h k β β£ (claim 1 of Properties of the Absolute Value in an Ordered Field ), claim 5 of Elementary Arithmetic in an Ordered Field gives ΞΎ β£ h k β£ β€ β£ h k β£ β€ r \xi|h_{k}|\le|h_{k}|\le r ΞΎ β£ h k β β£ β€ β£ h k β β£ β€ r . Hence z β Q ( c , r ) z\in Q(c,r) z β Q ( c , r ) , and the hypothesis gives β£ β j G i ( z ) β Ξ΄ i j β£ β€ Ξ· d |\partial_{j}G_{i}(z)-\delta_{ij}|\le\frac{\eta}{d} β£ β j β G i β ( z ) β Ξ΄ ij β β£ β€ d Ξ· β for every j β [ d ] j\in[d] j β [ d ] .
Step 4 (Estimate). By claim 7 of Properties of Finite Sums , β j = 1 d Ξ΄ i j h j = h i \sum_{j=1}^{d}\delta_{ij}h_{j}=h_{i} β j = 1 d β Ξ΄ ij β h j β = h i β , so by Step 2 and claims 2 and 3 of Properties of Finite Sums ,
G i ( x ) β G i ( c ) β h i = β j = 1 d a j , a j = ( β j G i ( z ) β Ξ΄ i j ) h j . G_{i}(x)-G_{i}(c)-h_{i}=\sum_{j=1}^{d}a_{j},\qquad a_{j}=\bigl(\partial_{j}G_{i}(z)-\delta_{ij}\bigr)h_{j}. G i β ( x ) β G i β ( c ) β h i β = j = 1 β d β a j β , a j β = ( β j β G i β ( z ) β Ξ΄ ij β ) h j β .
By claim 4 of Properties of the Absolute Value in an Ordered Field , Step 3, β£ h j β£ β€ r |h_{j}|\le r β£ h j β β£ β€ r and claim 5 of Elementary Arithmetic in an Ordered Field (all factors being nonnegative), β£ a j β£ β€ Ξ· d β r |a_{j}|\le\frac{\eta}{d}\,r β£ a j β β£ β€ d Ξ· β r , hence β Ξ· r d β€ a j β€ Ξ· r d -\frac{\eta r}{d}\le a_{j}\le\frac{\eta r}{d} β d Ξ·r β β€ a j β β€ d Ξ·r β by claim 6 of Properties of the Absolute Value in an Ordered Field . Applying claim 5 of Properties of Finite Sums to the nonnegative summands Ξ· r d β a j \frac{\eta r}{d}-a_{j} d Ξ·r β β a j β and Ξ· r d + a j \frac{\eta r}{d}+a_{j} d Ξ·r β + a j β , and using claims 2 and 3 there together with β j = 1 d Ξ· r d = d β
Ξ· r d = Ξ· r \sum_{j=1}^{d}\frac{\eta r}{d}=d\cdot\frac{\eta r}{d}=\eta r β j = 1 d β d Ξ·r β = d β
d Ξ·r β = Ξ·r , we get β Ξ· r β€ β j a j β€ Ξ· r -\eta r\le\sum_{j}a_{j}\le\eta r β Ξ·r β€ β j β a j β β€ Ξ·r , so
β£ G i ( x ) β G i ( c ) β ( x i β c i ) β£ β€ Ξ· r \bigl|G_{i}(x)-G_{i}(c)-(x_{i}-c_{i})\bigr|\le\eta r β G i β ( x ) β G i β ( c ) β ( x i β β c i β ) β β€ Ξ·r
by claim 6 of Properties of the Absolute Value in an Ordered Field . Finally, by claim 5 of Properties of the Absolute Value in an Ordered Field ,
β£ G i ( x ) β G i ( c ) β£ β€ β£ x i β c i β£ + β£ G i ( x ) β G i ( c ) β ( x i β c i ) β£ β€ r + Ξ· r = ( 1 + Ξ· ) r . |G_{i}(x)-G_{i}(c)|\le|x_{i}-c_{i}|+\bigl|G_{i}(x)-G_{i}(c)-(x_{i}-c_{i})\bigr|\le r+\eta r=(1+\eta)r . β£ G i β ( x ) β G i β ( c ) β£ β€ β£ x i β β c i β β£ + β G i β ( x ) β G i β ( c ) β ( x i β β c i β ) β β€ r + Ξ·r = ( 1 + Ξ· ) r .
As i β [ d ] i\in[d] i β [ d ] was arbitrary, G ( x ) β Q ( G ( c ) , ( 1 + Ξ· ) r ) G(x)\in Q\bigl(G(c),(1+\eta)r\bigr) G ( x ) β Q ( G ( c ) , ( 1 + Ξ· ) r ) .