Linear combinations and constants follow from the one-variable sum, constant-multiple and constant rules applied to the sections. For a function on , the Fréchet expansion along the line x+te_k gives the first section derivative <DPhi(x+te_k),e_k>, and the second-derivative expansion of the gradient, tested against , gives ,.
Each result cited is universally quantified over the data in its own statement.
Elementary arithmetic, order and absolute values of real numbers are used freely under The Real Numbers: Standing Notation and Background §background. Differentiability of a function at a point , and its derivative , are those of Single-Variable Calculus on an Interval §derivative with , as fixed in Derivatives of a Function on the Sobolev Space of Order -1 along the Fourier Modes; every point of is an interior point of , and the condition "" in that definition is automatic. For , and we write for the section of Derivatives of a Function on the Sobolev Space of Order -1 along the Fourier Modes.
Step 1 (Clause linear: combinations). Let , be as in clause 1, put , and fix and . Since functions are combined pointwise, for every , that is, as functions on . By Derivatives of a Function on the Sobolev Space of Order -1 along the Fourier Modes §twice, and are differentiable at every , and their derivatives are differentiable at . By claim 2 of Sum, Constant Multiple, and Product Rules for One-Dimensional Derivatives (the sum and constant multiple rules, applied on the interval , first to the two multiples and then to their sum), is differentiable at every with
so that as functions on ; and the same claim, applied at the point to these derivative functions, shows that is differentiable at with
As and were arbitrary, is twice differentiable along the modes (Derivatives of a Function on the Sobolev Space of Order -1 along the Fourier Modes §twice), and evaluating the first display at and reading both displays through Derivatives of a Function on the Sobolev Space of Order -1 along the Fourier Modes §derivatives gives the two asserted identities.
Step 2 (Clause linear: constants). Let have constant value , and fix and . Then is the constant function with value on , which by claim 1 of Sum, Constant Multiple, and Product Rules for One-Dimensional Derivatives is differentiable at every with derivative . So is the constant function with value , which by the same claim is differentiable at with derivative . Hence is twice differentiable along the modes, and and by Derivatives of a Function on the Sobolev Space of Order -1 along the Fourier Modes §derivatives.
Step 3 (Clause sobolev: setup). Let and be as in clause 2, and write and for the inner product and norm of . Both and are linear subspaces of under the pointwise operations (The Negative-Order Sobolev Spaces of the Torus §space), and as recorded in the statement; so for and the vector is the same family whether formed in or in , and the section of at along is . The set is open in itself (Real Hilbert Spaces: Standing Notation and Background §topology). Since is of class on , it is of class there, hence differentiable at every point of , and has a second derivative at every point of (The Classes and on an Open Subset of a Real Inner Product Space §c2, The Classes and on an Open Subset of a Real Inner Product Space §c1). Fix and , and put , a nonnegative real number; then and .
Step 4 (Clause sobolev: the first derivative). Let and , and let , which by Fréchet Differentiability and the Gradient on an Open Subset of a Real Inner Product Space §gradient has the property of Fréchet Differentiability and the Gradient on an Open Subset of a Real Inner Product Space §differentiable at . We show that is differentiable at with derivative . Let be given; the choices are made in the order , then , then . Put , which is positive, and let be the radius that Fréchet Differentiability and the Gradient on an Open Subset of a Real Inner Product Space §differentiable provides for at with gradient and in place of ; put . Let with , and . By Elementary Identities in a Real Inner Product Space §homogeneity, , so
Here by the vector space axioms, so and , and by Elementary Identities in a Real Inner Product Space §bilinear. Dividing by ,
Hence is differentiable at every , and
in particular by Derivatives of a Function on the Sobolev Space of Order -1 along the Fourier Modes §derivatives.
Step 5 (Clause sobolev: the second derivative). Since has a second derivative at , The Second Derivative and the Hessian on an Open Subset of a Real Inner Product Space §expansion provides a form and a positive such that is a second derivative of at (with and differentiable on that ball); by The Second Derivative and the Hessian on an Open Subset of a Real Inner Product Space §hessian, . We show that is differentiable at with derivative . Let be given; the choices are made in the order , then , then . Put , let (positive, with ) be the radius that The Second Derivative and the Hessian on an Open Subset of a Real Inner Product Space §expansion provides for with in place of , and put . Let with , and ; as in Step 4, . Applying the expansion inequality with this and with in the second slot,
By Step 4 and Elementary Identities in a Real Inner Product Space §bilinear, ; and by the homogeneity condition of Bounded Symmetric Bilinear Forms on a Real Inner Product Space: Norm, Order, Identity Form and Restriction §form. Dividing by ,
So is differentiable at with .
Step 6 (Conclusion of clause sobolev). Steps 4 and 5 hold for every and , so is twice differentiable along the modes (Derivatives of a Function on the Sobolev Space of Order -1 along the Fourier Modes §twice), and by Derivatives of a Function on the Sobolev Space of Order -1 along the Fourier Modes §derivatives
Loading…