Throughout, ∂ i φ \partial_i\varphi ∂ i φ is the partial derivative of φ \varphi φ with respect to the i i i -th coordinate and ∂ i ∂ j φ \partial_i\partial_j\varphi ∂ i ∂ j φ is the iterated partial derivative of clause 4 of C^k Maps on a Euclidean Open Set , defined on all of U U U by clause 2 there. Sums are finite sums of real numbers, t 2 t^{2} t 2 abbreviates t ⋅ t t\cdot t t ⋅ t , and ∣ t ∣ |t| ∣ t ∣ is the absolute value . Write 2 = 1 + 1 2=1+1 2 = 1 + 1 and 4 = 2 ⋅ 2 4=2\cdot 2 4 = 2 ⋅ 2 ; then 0 < 2 0<2 0 < 2 and 0 < 4 0<4 0 < 4 by claims 8 and 5 of Elementary Order Arithmetic in an Ordered Field , so 2 − 1 2^{-1} 2 − 1 and 4 − 1 4^{-1} 4 − 1 exist, and we abbreviate a ⋅ 2 − 1 a\cdot 2^{-1} a ⋅ 2 − 1 by a 2 \tfrac{a}{2} 2 a .
Convention on absolute values. If a ∈ R a\in\mathbb{R} a ∈ R satisfies 0 ≤ a 0\le a 0 ≤ a then ∣ a ∣ = a |a|=a ∣ a ∣ = a . Indeed ∣ a ∣ |a| ∣ a ∣ equals a a a or − a -a − a by claim 1 of Properties of the Absolute Value in an Ordered Field ; in the second case 0 ≤ ∣ a ∣ = − a 0\le|a|=-a 0 ≤ ∣ a ∣ = − a gives a ≤ 0 a\le 0 a ≤ 0 by claim 4 of Elementary Order Arithmetic in an Ordered Field , so a = 0 a=0 a = 0 by antisymmetry of the total order and ∣ a ∣ = − 0 = 0 = a |a|=-0=0=a ∣ a ∣ = − 0 = 0 = a . In particular ∣ 2 ∣ = 2 |2|=2 ∣2∣ = 2 , and ∣ − 1 ∣ = ∣ 1 ∣ = 1 |-1|=|1|=1 ∣ − 1∣ = ∣1∣ = 1 by claim 2 of Properties of the Absolute Value in an Ordered Field together with claim 6 of Elementary Order Arithmetic in an Ordered Field .
Symmetric points. For h ∈ R n h\in\mathbb{R}^{n} h ∈ R n we have x 0 − h = x 0 + ( − 1 ) h x_{0}-h=x_{0}+(-1)h x 0 − h = x 0 + ( − 1 ) h by claims 3 and 2 of Euclidean Space R n \mathbb{R}^n R n is a Real Vector Space , and therefore, by claim 5 of Elementary Properties of the Euclidean Norm on R n \mathbb{R}^n R n ,
∥ − h ∥ = ∥ ( − 1 ) h ∥ = ∣ − 1 ∣ ∥ h ∥ = ∥ h ∥ . \lVert -h\rVert=\lVert(-1)h\rVert=|-1|\,\lVert h\rVert=\lVert h\rVert . ∥ − h ∥ = ∥( − 1 ) h ∥ = ∣ − 1∣ ∥ h ∥ = ∥ h ∥ .
By claim 2 of Elementary Properties of the Euclidean Norm on R n \mathbb{R}^n R n , d ( x 0 , x 0 + h ) = ∥ h ∥ d(x_{0},x_{0}+h)=\lVert h\rVert d ( x 0 , x 0 + h ) = ∥ h ∥ and d ( x 0 , x 0 − h ) = ∥ − h ∥ = ∥ h ∥ d(x_{0},x_{0}-h)=\lVert -h\rVert=\lVert h\rVert d ( x 0 , x 0 − h ) = ∥ − h ∥ = ∥ h ∥ .
Step 1: a doubling inequality for φ \varphi φ .
Because the function x ↦ w ( x ) − φ ( x ) x\mapsto w(x)-\varphi(x) x ↦ w ( x ) − φ ( x ) has a local maximum at x 0 x_{0} x 0 relative to U U U , there is δ 1 ∈ R \delta_{1}\in\mathbb{R} δ 1 ∈ R with 0 < δ 1 0<\delta_{1} 0 < δ 1 such that every y ∈ U y\in U y ∈ U with d ( x 0 , y ) < δ 1 d(x_{0},y)<\delta_{1} d ( x 0 , y ) < δ 1 satisfies w ( y ) − φ ( y ) ≤ w ( x 0 ) − φ ( x 0 ) w(y)-\varphi(y)\le w(x_{0})-\varphi(x_{0}) w ( y ) − φ ( y ) ≤ w ( x 0 ) − φ ( x 0 ) . Because U U U is open in ( R n , d ) (\mathbb{R}^{n},d) ( R n , d ) and x 0 ∈ U x_{0}\in U x 0 ∈ U , Open Subset of a Metric Space supplies δ 2 ∈ R \delta_{2}\in\mathbb{R} δ 2 ∈ R with 0 < δ 2 0<\delta_{2} 0 < δ 2 such that the open ball of centre x 0 x_{0} x 0 and radius δ 2 \delta_{2} δ 2 , that is the set of y ∈ R n y\in\mathbb{R}^{n} y ∈ R n with d ( x 0 , y ) < δ 2 d(x_{0},y)<\delta_{2} d ( x 0 , y ) < δ 2 , is contained in U U U . By claim 9 of Elementary Order Arithmetic in an Ordered Field there is r ∈ R r\in\mathbb{R} r ∈ R with r ≤ δ 1 r\le\delta_{1} r ≤ δ 1 , r ≤ δ 2 r\le\delta_{2} r ≤ δ 2 , and r r r equal to δ 1 \delta_{1} δ 1 or to δ 2 \delta_{2} δ 2 ; in either case 0 < r 0<r 0 < r .
We claim that every h ∈ R n h\in\mathbb{R}^{n} h ∈ R n with ∥ h ∥ < r \lVert h\rVert<r ∥ h ∥ < r satisfies x 0 + h ∈ U x_{0}+h\in U x 0 + h ∈ U , x 0 − h ∈ U x_{0}-h\in U x 0 − h ∈ U and
( − μ ) ∥ h ∥ 2 ≤ φ ( x 0 + h ) + φ ( x 0 − h ) − 2 φ ( x 0 ) . (-\mu)\lVert h\rVert^{2}\;\le\;\varphi(x_{0}+h)+\varphi(x_{0}-h)-2\varphi(x_{0}). ( − μ ) ∥ h ∥ 2 ≤ φ ( x 0 + h ) + φ ( x 0 − h ) − 2 φ ( x 0 ) .
We call this the doubling inequality .
To prove it, fix such an h h h . By the computation of the distances above and mixed transitivity (claim 2 of Elementary Order Arithmetic in an Ordered Field ) we have d ( x 0 , x 0 ± h ) < δ 2 d(x_{0},x_{0}\pm h)<\delta_{2} d ( x 0 , x 0 ± h ) < δ 2 , so both points lie in U U U , and d ( x 0 , x 0 ± h ) < δ 1 d(x_{0},x_{0}\pm h)<\delta_{1} d ( x 0 , x 0 ± h ) < δ 1 , so the local maximum property gives
w ( x 0 + h ) − φ ( x 0 + h ) ≤ w ( x 0 ) − φ ( x 0 ) , w ( x 0 − h ) − φ ( x 0 − h ) ≤ w ( x 0 ) − φ ( x 0 ) . w(x_{0}+h)-\varphi(x_{0}+h)\le w(x_{0})-\varphi(x_{0}),\qquad
w(x_{0}-h)-\varphi(x_{0}-h)\le w(x_{0})-\varphi(x_{0}). w ( x 0 + h ) − φ ( x 0 + h ) ≤ w ( x 0 ) − φ ( x 0 ) , w ( x 0 − h ) − φ ( x 0 − h ) ≤ w ( x 0 ) − φ ( x 0 ) .
Adding these two inequalities — each is equivalent by claim 3 of Elementary Arithmetic in an Ordered Field to the statement that a certain difference is nonnegative, the two differences have nonnegative sum by claim 2 there, and claim 3 converts back — yields
w ( x 0 + h ) + w ( x 0 − h ) ≤ 2 w ( x 0 ) − 2 φ ( x 0 ) + φ ( x 0 + h ) + φ ( x 0 − h ) . w(x_{0}+h)+w(x_{0}-h)\;\le\;2w(x_{0})-2\varphi(x_{0})+\varphi(x_{0}+h)+\varphi(x_{0}-h). w ( x 0 + h ) + w ( x 0 − h ) ≤ 2 w ( x 0 ) − 2 φ ( x 0 ) + φ ( x 0 + h ) + φ ( x 0 − h ) .
Next, x 0 + h x_{0}+h x 0 + h and x 0 − h x_{0}-h x 0 − h lie in U U U , hence in Ω \Omega Ω . Computing coordinatewise with Euclidean Space R n \mathbb{R}^n R n , Sum of Points of R n \mathbb{R}^n R n and Scalar Multiple of a Point of R n \mathbb{R}^n R n ,
1 2 ( x 0 + h ) + ( 1 − 1 2 ) ( x 0 − h ) = x 0 , ( x 0 + h ) − ( x 0 − h ) = 2 h , \tfrac{1}{2}(x_{0}+h)+\bigl(1-\tfrac{1}{2}\bigr)(x_{0}-h)=x_{0},\qquad
(x_{0}+h)-(x_{0}-h)=2h, 2 1 ( x 0 + h ) + ( 1 − 2 1 ) ( x 0 − h ) = x 0 , ( x 0 + h ) − ( x 0 − h ) = 2 h ,
the first because 1 − 1 2 = 1 2 1-\tfrac12=\tfrac12 1 − 2 1 = 2 1 and 1 2 ( x 0 i + h i ) + 1 2 ( x 0 i − h i ) = x 0 i \tfrac12(x_{0i}+h_i)+\tfrac12(x_{0i}-h_i)=x_{0i} 2 1 ( x 0 i + h i ) + 2 1 ( x 0 i − h i ) = x 0 i , the second because ( x 0 i + h i ) − ( x 0 i − h i ) = h i + h i = 2 h i (x_{0i}+h_i)-(x_{0i}-h_i)=h_i+h_i=2h_i ( x 0 i + h i ) − ( x 0 i − h i ) = h i + h i = 2 h i . Moreover ∥ 2 h ∥ = ∣ 2 ∣ ∥ h ∥ = 2 ∥ h ∥ \lVert 2h\rVert=|2|\,\lVert h\rVert=2\lVert h\rVert ∥ 2 h ∥ = ∣2∣ ∥ h ∥ = 2 ∥ h ∥ by claim 5 of Elementary Properties of the Euclidean Norm on R n \mathbb{R}^n R n , so ∥ 2 h ∥ 2 = 4 ∥ h ∥ 2 \lVert 2h\rVert^{2}=4\lVert h\rVert^{2} ∥ 2 h ∥ 2 = 4 ∥ h ∥ 2 . Since 0 ≤ 1 2 0\le\tfrac12 0 ≤ 2 1 and 1 2 ≤ 1 \tfrac12\le1 2 1 ≤ 1 by claim 8 of Elementary Order Arithmetic in an Ordered Field applied with ε = 1 \varepsilon=1 ε = 1 , Quadratic Increment Characterisation of Semiconvexity , applied to w w w on Ω \Omega Ω with x = x 0 + h x=x_{0}+h x = x 0 + h , y = x 0 − h y=x_{0}-h y = x 0 − h and t = 1 2 t=\tfrac12 t = 2 1 , gives
w ( x 0 ) ≤ 1 2 w ( x 0 + h ) + 1 2 w ( x 0 − h ) + μ 2 ⋅ 1 2 ⋅ 1 2 ⋅ 4 ∥ h ∥ 2 . w(x_{0})\;\le\;\tfrac{1}{2}w(x_{0}+h)+\tfrac{1}{2}w(x_{0}-h)+\frac{\mu}{2}\cdot\tfrac{1}{2}\cdot\tfrac{1}{2}\cdot 4\lVert h\rVert^{2}. w ( x 0 ) ≤ 2 1 w ( x 0 + h ) + 2 1 w ( x 0 − h ) + 2 μ ⋅ 2 1 ⋅ 2 1 ⋅ 4 ∥ h ∥ 2 .
Here 1 2 ⋅ 1 2 ⋅ 4 = 2 − 1 2 − 1 ( 2 ⋅ 2 ) = 1 \tfrac12\cdot\tfrac12\cdot4=2^{-1}2^{-1}(2\cdot2)=1 2 1 ⋅ 2 1 ⋅ 4 = 2 − 1 2 − 1 ( 2 ⋅ 2 ) = 1 , so the last term is μ 2 ∥ h ∥ 2 \tfrac{\mu}{2}\lVert h\rVert^{2} 2 μ ∥ h ∥ 2 . Multiplying the whole inequality by 2 2 2 , which is legitimate by claim 5 of Elementary Arithmetic in an Ordered Field because 0 ≤ 2 0\le 2 0 ≤ 2 , and using 2 ⋅ 2 − 1 = 1 2\cdot 2^{-1}=1 2 ⋅ 2 − 1 = 1 , we obtain
2 w ( x 0 ) ≤ w ( x 0 + h ) + w ( x 0 − h ) + μ ∥ h ∥ 2 . 2w(x_{0})\;\le\;w(x_{0}+h)+w(x_{0}-h)+\mu\lVert h\rVert^{2}. 2 w ( x 0 ) ≤ w ( x 0 + h ) + w ( x 0 − h ) + μ ∥ h ∥ 2 .
Adding μ ∥ h ∥ 2 \mu\lVert h\rVert^{2} μ ∥ h ∥ 2 to both sides of the previous displayed inequality for w ( x 0 + h ) + w ( x 0 − h ) w(x_{0}+h)+w(x_{0}-h) w ( x 0 + h ) + w ( x 0 − h ) (again claim 3 of Elementary Arithmetic in an Ordered Field , the two differences being equal) and chaining the two inequalities gives
2 w ( x 0 ) ≤ 2 w ( x 0 ) − 2 φ ( x 0 ) + φ ( x 0 + h ) + φ ( x 0 − h ) + μ ∥ h ∥ 2 . 2w(x_{0})\;\le\;2w(x_{0})-2\varphi(x_{0})+\varphi(x_{0}+h)+\varphi(x_{0}-h)+\mu\lVert h\rVert^{2}. 2 w ( x 0 ) ≤ 2 w ( x 0 ) − 2 φ ( x 0 ) + φ ( x 0 + h ) + φ ( x 0 − h ) + μ ∥ h ∥ 2 .
Cancelling the common summand 2 w ( x 0 ) 2w(x_{0}) 2 w ( x 0 ) , by two applications of claim 3 of Elementary Arithmetic in an Ordered Field , and then using claim 3 once more together with the identity ( − μ ) ∥ h ∥ 2 = − ( μ ∥ h ∥ 2 ) (-\mu)\lVert h\rVert^{2}=-\bigl(\mu\lVert h\rVert^{2}\bigr) ( − μ ) ∥ h ∥ 2 = − ( μ ∥ h ∥ 2 ) of claim 2 of Zero Products and Elementary Identities in a Field , we arrive at the doubling inequality. This proves the claim.
Step 2: reduction to a quadratic form.
Let z = ( z 1 , … , z n ) ∈ R n z=(z_{1},\dots,z_{n})\in\mathbb{R}^{n} z = ( z 1 , … , z n ) ∈ R n . By The Positive Semidefinite Ordering on Symmetric Matrices it suffices to prove that
z ⋅ ( ( − μ ) I n z ) ≤ z ⋅ ( D 2 φ ( x 0 ) z ) . z\cdot\bigl((-\mu)I_{n}z\bigr)\;\le\;z\cdot\bigl(D^{2}\varphi(x_{0})z\bigr). z ⋅ ( ( − μ ) I n z ) ≤ z ⋅ ( D 2 φ ( x 0 ) z ) .
By Identity Matrix and Scalar Multiple of a Real Matrix the entry of ( − μ ) I n (-\mu)I_{n} ( − μ ) I n in row i i i and column j j j is ( − μ ) (-\mu) ( − μ ) when i = j i=j i = j and is ( − μ ) ⋅ 0 = 0 (-\mu)\cdot 0=0 ( − μ ) ⋅ 0 = 0 otherwise, by claim 1 of Zero Products and Elementary Identities in a Field . Hence by Matrix-Vector Product and claim 7 of Properties of Finite Sums , ( ( − μ ) I n z ) i = ( − μ ) z i \bigl((-\mu)I_{n}z\bigr)_{i}=(-\mu)z_{i} ( ( − μ ) I n z ) i = ( − μ ) z i , and therefore, by the definition of the dot product , claim 3 of Properties of Finite Sums and claim 1 of Elementary Properties of the Euclidean Norm on R n \mathbb{R}^n R n ,
z ⋅ ( ( − μ ) I n z ) = ∑ i = 1 n z i ( ( − μ ) z i ) = ( − μ ) ∑ i = 1 n z i 2 = ( − μ ) ∥ z ∥ 2 . z\cdot\bigl((-\mu)I_{n}z\bigr)=\sum_{i=1}^{n}z_{i}\bigl((-\mu)z_{i}\bigr)=(-\mu)\sum_{i=1}^{n}z_{i}^{2}=(-\mu)\lVert z\rVert^{2}. z ⋅ ( ( − μ ) I n z ) = i = 1 ∑ n z i ( ( − μ ) z i ) = ( − μ ) i = 1 ∑ n z i 2 = ( − μ ) ∥ z ∥ 2 .
By Hessian Matrix of a C^2 Function and Matrix-Vector Product , ( D 2 φ ( x 0 ) z ) i = ∑ j = 1 n ∂ i ∂ j φ ( x 0 ) z j \bigl(D^{2}\varphi(x_{0})z\bigr)_{i}=\sum_{j=1}^{n}\partial_i\partial_j\varphi(x_{0})z_{j} ( D 2 φ ( x 0 ) z ) i = ∑ j = 1 n ∂ i ∂ j φ ( x 0 ) z j , so, writing Q Q Q for the number z ⋅ ( D 2 φ ( x 0 ) z ) z\cdot\bigl(D^{2}\varphi(x_{0})z\bigr) z ⋅ ( D 2 φ ( x 0 ) z ) and using claim 3 of Properties of Finite Sums on the inner sum,
Q = ∑ i = 1 n ∑ j = 1 n ∂ i ∂ j φ ( x 0 ) z i z j = ∑ i = 1 n ∑ j = 1 n ∂ j ∂ i φ ( x 0 ) z i z j , Q=\sum_{i=1}^{n}\sum_{j=1}^{n}\partial_i\partial_j\varphi(x_{0})\,z_{i}z_{j}
=\sum_{i=1}^{n}\sum_{j=1}^{n}\partial_j\partial_i\varphi(x_{0})\,z_{i}z_{j}, Q = i = 1 ∑ n j = 1 ∑ n ∂ i ∂ j φ ( x 0 ) z i z j = i = 1 ∑ n j = 1 ∑ n ∂ j ∂ i φ ( x 0 ) z i z j ,
the second equality because ∂ i ∂ j φ ( x 0 ) = ∂ j ∂ i φ ( x 0 ) \partial_i\partial_j\varphi(x_{0})=\partial_j\partial_i\varphi(x_{0}) ∂ i ∂ j φ ( x 0 ) = ∂ j ∂ i φ ( x 0 ) for all i , j i,j i , j by claim 1 of Equality of Mixed Second Partial Derivatives and Symmetry of the Hessian . Write also L = ∑ i = 1 n ∂ i φ ( x 0 ) z i L=\sum_{i=1}^{n}\partial_i\varphi(x_{0})z_{i} L = ∑ i = 1 n ∂ i φ ( x 0 ) z i . We must show ( − μ ) ∥ z ∥ 2 ≤ Q (-\mu)\lVert z\rVert^{2}\le Q ( − μ ) ∥ z ∥ 2 ≤ Q .
Suppose first that z z z is the origin of R n \mathbb{R}^{n} R n . Then ∥ z ∥ = 0 \lVert z\rVert=0 ∥ z ∥ = 0 by claim 3 of Elementary Properties of the Euclidean Norm on R n \mathbb{R}^n R n , so ( − μ ) ∥ z ∥ 2 = 0 (-\mu)\lVert z\rVert^{2}=0 ( − μ ) ∥ z ∥ 2 = 0 by claim 1 of Zero Products and Elementary Identities in a Field ; and every z i = 0 z_{i}=0 z i = 0 , so each summand of the sum defining Q Q Q equals 0 ⋅ a i 0\cdot a_{i} 0 ⋅ a i with a i = ( D 2 φ ( x 0 ) z ) i a_{i}=\bigl(D^{2}\varphi(x_{0})z\bigr)_{i} a i = ( D 2 φ ( x 0 ) z ) i , whence Q = 0 ⋅ ∑ i = 1 n a i = 0 Q=0\cdot\sum_{i=1}^{n}a_{i}=0 Q = 0 ⋅ ∑ i = 1 n a i = 0 by claims 3 and 1 of Properties of Finite Sums and Zero Products and Elementary Identities in a Field respectively. The required inequality reads 0 ≤ 0 0\le 0 0 ≤ 0 and holds.
Step 3: the case z ≠ 0 z\ne 0 z = 0 , by Taylor expansion.
Suppose now that z z z is not the origin. By claims 3 and 1 of Elementary Properties of the Euclidean Norm on R n \mathbb{R}^n R n we have ∥ z ∥ ≠ 0 \lVert z\rVert\ne 0 ∥ z ∥ = 0 and 0 ≤ ∥ z ∥ 0\le\lVert z\rVert 0 ≤ ∥ z ∥ , hence 0 < ∥ z ∥ 0<\lVert z\rVert 0 < ∥ z ∥ and, by claim 5 of Elementary Order Arithmetic in an Ordered Field , 0 < ∥ z ∥ 2 0<\lVert z\rVert^{2} 0 < ∥ z ∥ 2 ; by claim 7 there, ( ∥ z ∥ 2 ) − 1 \bigl(\lVert z\rVert^{2}\bigr)^{-1} ( ∥ z ∥ 2 ) − 1 exists and is positive.
Assume, seeking a contradiction, that the desired inequality fails. Since the order on R \mathbb{R} R is total, this means Q < ( − μ ) ∥ z ∥ 2 Q<(-\mu)\lVert z\rVert^{2} Q < ( − μ ) ∥ z ∥ 2 . Put
c = ( − μ ) ∥ z ∥ 2 − Q , ε = ( 1 2 ⋅ c 2 ) ⋅ ( ∥ z ∥ 2 ) − 1 . c=(-\mu)\lVert z\rVert^{2}-Q,\qquad \varepsilon=\Bigl(\tfrac{1}{2}\cdot\tfrac{c}{2}\Bigr)\cdot\bigl(\lVert z\rVert^{2}\bigr)^{-1}. c = ( − μ ) ∥ z ∥ 2 − Q , ε = ( 2 1 ⋅ 2 c ) ⋅ ( ∥ z ∥ 2 ) − 1 .
By claim 1 of Elementary Order Arithmetic in an Ordered Field we have 0 < c 0<c 0 < c , hence 0 < c 2 0<\tfrac{c}{2} 0 < 2 c and 0 < 1 2 ⋅ c 2 0<\tfrac12\cdot\tfrac{c}{2} 0 < 2 1 ⋅ 2 c by claim 8 there, and so 0 < ε 0<\varepsilon 0 < ε by claim 5 there.
Apply Second-Order Taylor Expansion with Peano Remainder to φ \varphi φ at the point x 0 x_{0} x 0 with this ε \varepsilon ε : there is δ ∈ R \delta\in\mathbb{R} δ ∈ R with 0 < δ 0<\delta 0 < δ such that every k ∈ R n k\in\mathbb{R}^{n} k ∈ R n with ∥ k ∥ < δ \lVert k\rVert<\delta ∥ k ∥ < δ satisfies x 0 + k ∈ U x_{0}+k\in U x 0 + k ∈ U and
∣ φ ( x 0 + k ) − φ ( x 0 ) − ∑ i = 1 n ∂ i φ ( x 0 ) k i − 1 2 ∑ i = 1 n ∑ j = 1 n ∂ j ∂ i φ ( x 0 ) k i k j ∣ ≤ ε ∥ k ∥ 2 . \Bigl|\,\varphi(x_{0}+k)-\varphi(x_{0})-\sum_{i=1}^{n}\partial_i\varphi(x_{0})k_{i}-\frac{1}{2}\sum_{i=1}^{n}\sum_{j=1}^{n}\partial_j\partial_i\varphi(x_{0})k_{i}k_{j}\,\Bigr|\;\le\;\varepsilon\lVert k\rVert^{2}. φ ( x 0 + k ) − φ ( x 0 ) − i = 1 ∑ n ∂ i φ ( x 0 ) k i − 2 1 i = 1 ∑ n j = 1 ∑ n ∂ j ∂ i φ ( x 0 ) k i k j ≤ ε ∥ k ∥ 2 .
By claim 9 of Elementary Order Arithmetic in an Ordered Field choose ρ ∈ R \rho\in\mathbb{R} ρ ∈ R with ρ ≤ r \rho\le r ρ ≤ r , ρ ≤ δ \rho\le\delta ρ ≤ δ and ρ \rho ρ equal to r r r or to δ \delta δ , so that 0 < ρ 0<\rho 0 < ρ . Put
s = ρ 2 ⋅ ∥ z ∥ − 1 , h = s z . s=\tfrac{\rho}{2}\cdot\lVert z\rVert^{-1},\qquad h=s\,z . s = 2 ρ ⋅ ∥ z ∥ − 1 , h = s z .
Then 0 < s 0<s 0 < s by claims 8, 7 and 5 of Elementary Order Arithmetic in an Ordered Field , so ∣ s ∣ = s |s|=s ∣ s ∣ = s by the convention above, and by claim 5 of Elementary Properties of the Euclidean Norm on R n \mathbb{R}^n R n together with ∥ z ∥ − 1 ∥ z ∥ = 1 \lVert z\rVert^{-1}\lVert z\rVert=1 ∥ z ∥ − 1 ∥ z ∥ = 1 ,
∥ h ∥ = ∣ s ∣ ∥ z ∥ = ρ 2 < ρ , \lVert h\rVert=|s|\,\lVert z\rVert=\tfrac{\rho}{2}<\rho , ∥ h ∥ = ∣ s ∣ ∥ z ∥ = 2 ρ < ρ ,
the last step by claim 8 of Elementary Order Arithmetic in an Ordered Field . Consequently ∥ h ∥ < r \lVert h\rVert<r ∥ h ∥ < r and ∥ h ∥ < δ \lVert h\rVert<\delta ∥ h ∥ < δ by mixed transitivity, and likewise ∥ − h ∥ = ∥ h ∥ < δ \lVert -h\rVert=\lVert h\rVert<\delta ∥ − h ∥ = ∥ h ∥ < δ .
By Scalar Multiple of a Point of R n \mathbb{R}^n R n we have h i = s z i h_{i}=s\,z_{i} h i = s z i , and − h = ( − s ) z -h=(-s)z − h = ( − s ) z with ( − h ) i = ( − s ) z i (-h)_{i}=(-s)z_{i} ( − h ) i = ( − s ) z i , by claim 2 of Zero Products and Elementary Identities in a Field applied coordinatewise. Claim 3 of Properties of Finite Sums , applied to the outer and the inner sums, therefore gives
∑ i = 1 n ∂ i φ ( x 0 ) h i = s L , ∑ i = 1 n ∑ j = 1 n ∂ j ∂ i φ ( x 0 ) h i h j = s 2 Q , \sum_{i=1}^{n}\partial_i\varphi(x_{0})h_{i}=sL,\qquad
\sum_{i=1}^{n}\sum_{j=1}^{n}\partial_j\partial_i\varphi(x_{0})h_{i}h_{j}=s^{2}Q, i = 1 ∑ n ∂ i φ ( x 0 ) h i = s L , i = 1 ∑ n j = 1 ∑ n ∂ j ∂ i φ ( x 0 ) h i h j = s 2 Q ,
and, since ( − s ) z i ( − s ) z j = s 2 z i z j (-s)z_{i}(-s)z_{j}=s^{2}z_{i}z_{j} ( − s ) z i ( − s ) z j = s 2 z i z j by claim 2 of Zero Products and Elementary Identities in a Field ,
∑ i = 1 n ∂ i φ ( x 0 ) ( − h ) i = − ( s L ) , ∑ i = 1 n ∑ j = 1 n ∂ j ∂ i φ ( x 0 ) ( − h ) i ( − h ) j = s 2 Q . \sum_{i=1}^{n}\partial_i\varphi(x_{0})(-h)_{i}=-(sL),\qquad
\sum_{i=1}^{n}\sum_{j=1}^{n}\partial_j\partial_i\varphi(x_{0})(-h)_{i}(-h)_{j}=s^{2}Q . i = 1 ∑ n ∂ i φ ( x 0 ) ( − h ) i = − ( s L ) , i = 1 ∑ n j = 1 ∑ n ∂ j ∂ i φ ( x 0 ) ( − h ) i ( − h ) j = s 2 Q .
Also ∥ h ∥ 2 = ( s ∥ z ∥ ) 2 = s 2 ∥ z ∥ 2 \lVert h\rVert^{2}=\bigl(s\lVert z\rVert\bigr)^{2}=s^{2}\lVert z\rVert^{2} ∥ h ∥ 2 = ( s ∥ z ∥ ) 2 = s 2 ∥ z ∥ 2 and ∥ − h ∥ 2 = s 2 ∥ z ∥ 2 \lVert -h\rVert^{2}=s^{2}\lVert z\rVert^{2} ∥ − h ∥ 2 = s 2 ∥ z ∥ 2 .
Write
A + = φ ( x 0 + h ) − φ ( x 0 ) − s L − 1 2 s 2 Q , A − = φ ( x 0 − h ) − φ ( x 0 ) + s L − 1 2 s 2 Q , A_{+}=\varphi(x_{0}+h)-\varphi(x_{0})-sL-\tfrac{1}{2}s^{2}Q,\qquad
A_{-}=\varphi(x_{0}-h)-\varphi(x_{0})+sL-\tfrac{1}{2}s^{2}Q, A + = φ ( x 0 + h ) − φ ( x 0 ) − s L − 2 1 s 2 Q , A − = φ ( x 0 − h ) − φ ( x 0 ) + s L − 2 1 s 2 Q ,
so that the two applications of the Taylor estimate, at k = h k=h k = h and at k = − h k=-h k = − h , read ∣ A + ∣ ≤ ε s 2 ∥ z ∥ 2 |A_{+}|\le\varepsilon s^{2}\lVert z\rVert^{2} ∣ A + ∣ ≤ ε s 2 ∥ z ∥ 2 and ∣ A − ∣ ≤ ε s 2 ∥ z ∥ 2 |A_{-}|\le\varepsilon s^{2}\lVert z\rVert^{2} ∣ A − ∣ ≤ ε s 2 ∥ z ∥ 2 . By claims 3 and 5 of Properties of the Absolute Value in an Ordered Field and the addition of inequalities as in Step 1,
A + + A − ≤ ∣ A + + A − ∣ ≤ ∣ A + ∣ + ∣ A − ∣ ≤ ε s 2 ∥ z ∥ 2 + ε s 2 ∥ z ∥ 2 , A_{+}+A_{-}\;\le\;|A_{+}+A_{-}|\;\le\;|A_{+}|+|A_{-}|\;\le\;\varepsilon s^{2}\lVert z\rVert^{2}+\varepsilon s^{2}\lVert z\rVert^{2}, A + + A − ≤ ∣ A + + A − ∣ ≤ ∣ A + ∣ + ∣ A − ∣ ≤ ε s 2 ∥ z ∥ 2 + ε s 2 ∥ z ∥ 2 ,
while A + + A − = φ ( x 0 + h ) + φ ( x 0 − h ) − 2 φ ( x 0 ) − s 2 Q A_{+}+A_{-}=\varphi(x_{0}+h)+\varphi(x_{0}-h)-2\varphi(x_{0})-s^{2}Q A + + A − = φ ( x 0 + h ) + φ ( x 0 − h ) − 2 φ ( x 0 ) − s 2 Q , the terms ∓ s L \mp sL ∓ s L cancelling and 1 2 s 2 Q + 1 2 s 2 Q = s 2 Q \tfrac12 s^{2}Q+\tfrac12 s^{2}Q=s^{2}Q 2 1 s 2 Q + 2 1 s 2 Q = s 2 Q by claim 8 of Elementary Order Arithmetic in an Ordered Field . Hence, by claim 3 of Elementary Arithmetic in an Ordered Field ,
φ ( x 0 + h ) + φ ( x 0 − h ) − 2 φ ( x 0 ) ≤ s 2 Q + ε s 2 ∥ z ∥ 2 + ε s 2 ∥ z ∥ 2 . \varphi(x_{0}+h)+\varphi(x_{0}-h)-2\varphi(x_{0})\;\le\;s^{2}Q+\varepsilon s^{2}\lVert z\rVert^{2}+\varepsilon s^{2}\lVert z\rVert^{2}. φ ( x 0 + h ) + φ ( x 0 − h ) − 2 φ ( x 0 ) ≤ s 2 Q + ε s 2 ∥ z ∥ 2 + ε s 2 ∥ z ∥ 2 .
Since ∥ h ∥ < r \lVert h\rVert<r ∥ h ∥ < r , the doubling inequality of Step 1 applies to h h h and gives
( − μ ) s 2 ∥ z ∥ 2 = ( − μ ) ∥ h ∥ 2 ≤ φ ( x 0 + h ) + φ ( x 0 − h ) − 2 φ ( x 0 ) . (-\mu)s^{2}\lVert z\rVert^{2}=(-\mu)\lVert h\rVert^{2}\;\le\;\varphi(x_{0}+h)+\varphi(x_{0}-h)-2\varphi(x_{0}). ( − μ ) s 2 ∥ z ∥ 2 = ( − μ ) ∥ h ∥ 2 ≤ φ ( x 0 + h ) + φ ( x 0 − h ) − 2 φ ( x 0 ) .
Chaining the two displays and multiplying by ( s 2 ) − 1 \bigl(s^{2}\bigr)^{-1} ( s 2 ) − 1 , which is positive by claims 5 and 7 of Elementary Order Arithmetic in an Ordered Field and so may be applied by claim 5 of Elementary Arithmetic in an Ordered Field , we obtain
( − μ ) ∥ z ∥ 2 ≤ Q + ε ∥ z ∥ 2 + ε ∥ z ∥ 2 . (-\mu)\lVert z\rVert^{2}\;\le\;Q+\varepsilon\lVert z\rVert^{2}+\varepsilon\lVert z\rVert^{2}. ( − μ ) ∥ z ∥ 2 ≤ Q + ε ∥ z ∥ 2 + ε ∥ z ∥ 2 .
By the choice of ε \varepsilon ε and ( ∥ z ∥ 2 ) − 1 ∥ z ∥ 2 = 1 \bigl(\lVert z\rVert^{2}\bigr)^{-1}\lVert z\rVert^{2}=1 ( ∥ z ∥ 2 ) − 1 ∥ z ∥ 2 = 1 we have ε ∥ z ∥ 2 = 1 2 ⋅ c 2 \varepsilon\lVert z\rVert^{2}=\tfrac12\cdot\tfrac{c}{2} ε ∥ z ∥ 2 = 2 1 ⋅ 2 c , so ε ∥ z ∥ 2 + ε ∥ z ∥ 2 = c 2 \varepsilon\lVert z\rVert^{2}+\varepsilon\lVert z\rVert^{2}=\tfrac{c}{2} ε ∥ z ∥ 2 + ε ∥ z ∥ 2 = 2 c by claim 8 of Elementary Order Arithmetic in an Ordered Field . Since c = c 2 + c 2 c=\tfrac{c}{2}+\tfrac{c}{2} c = 2 c + 2 c and Q + c = ( − μ ) ∥ z ∥ 2 Q+c=(-\mu)\lVert z\rVert^{2} Q + c = ( − μ ) ∥ z ∥ 2 , the right-hand side equals ( − μ ) ∥ z ∥ 2 − c 2 (-\mu)\lVert z\rVert^{2}-\tfrac{c}{2} ( − μ ) ∥ z ∥ 2 − 2 c , so
( − μ ) ∥ z ∥ 2 ≤ ( − μ ) ∥ z ∥ 2 − c 2 . (-\mu)\lVert z\rVert^{2}\;\le\;(-\mu)\lVert z\rVert^{2}-\tfrac{c}{2}. ( − μ ) ∥ z ∥ 2 ≤ ( − μ ) ∥ z ∥ 2 − 2 c .
By claim 3 of Elementary Arithmetic in an Ordered Field this gives 0 ≤ − c 2 0\le-\tfrac{c}{2} 0 ≤ − 2 c , hence c 2 ≤ 0 \tfrac{c}{2}\le 0 2 c ≤ 0 by claim 4 of Elementary Order Arithmetic in an Ordered Field ; combined with 0 < c 2 0<\tfrac{c}{2} 0 < 2 c and mixed transitivity this yields 0 < 0 0<0 0 < 0 , which is false.
Therefore ( − μ ) ∥ z ∥ 2 ≤ Q (-\mu)\lVert z\rVert^{2}\le Q ( − μ ) ∥ z ∥ 2 ≤ Q also when z z z is not the origin. In both cases z ⋅ ( ( − μ ) I n z ) ≤ z ⋅ ( D 2 φ ( x 0 ) z ) z\cdot\bigl((-\mu)I_{n}z\bigr)\le z\cdot\bigl(D^{2}\varphi(x_{0})z\bigr) z ⋅ ( ( − μ ) I n z ) ≤ z ⋅ ( D 2 φ ( x 0 ) z ) , and since z ∈ R n z\in\mathbb{R}^{n} z ∈ R n was arbitrary, ( − μ ) I n ⪯ D 2 φ ( x 0 ) (-\mu)I_{n}\preceq D^{2}\varphi(x_{0}) ( − μ ) I n ⪯ D 2 φ ( x 0 ) .