The Setting
For natural numbers , suppose charts on Euclidean space with Euclidean norm , each cut by families of parallel hyperplanes at spacing , carry a map into for a finite coordinate set . Extending to piecewise-constant chart step functions on for and creates jump discontinuities across the threshold hyperplanes, so convolving each component with a product mollifier of width yields smooth functions on . The central question is how closely approximates and how the total gradient at is controlled by cell-interface jumps and cross-chart mismatches when direction vectors are small against dual vectors on at least half of the charts.
Chart Step Functions
Equip with the standard inner product , writing coordinates as . For , the cube of radius in is .✦ Given a family of vectors in and a scale , a point is -regular for if for all and ✦, and its integer label is with entries .✦ Given families in , write and ; the rounded chart vertex set consists of all pairs with and for which some -regular point for satisfies ✦, and each such defines a chart vertex .✦
Let and identify with via the linear homeomorphism , where .✦ For any vector , the non-pivot linear functional on is defined by ✦, and when , the hyperplane slope functional on is defined by .✦ For a finite index set and vectors in with nonzero pivot coordinates , the slope family of is the family of linear functionals on defined by for each .✦ On , each slope functional defines the slice linear form by ✦, and the pivoted slice functionals are defined by , which satisfy .✦
The split cube domain is the image subset .✦ For a family of continuous linear functionals on and a subset , a point is -regular if for all and ✦, with integer label ✦; the label vertex set consists of all integer labels realized by -regular points in , each associated with its label vertex.✦ For the pivoted slice functionals of with , maps onto and identifies with the chart- vertices of .
Given a finite threshold set , a scale , continuous linear functionals on , and nonzero pivots , the sign pattern of a point is the subset of defined by .✦ A pattern extension assigns to each map a map defined on all subsets and valued in ✦; specializing to the pivoted slice functionals and pivots of gives the cell pattern values.✦ Given slopes , offsets , and a real-valued function on subsets , the slice cell function sends to ✦, and setting and composing with yields the Euclidean chart function for , , , and .✦ The chart label range is defined by (1): (1)✦ For any collection with and vertex map , let and let be the cell pattern extension of from the chart- slice label vertices over to subsets of ; then the concrete chart step function for and is defined by (2): (2)✦
Mollifiers and Constants
Let be the smooth () bump function centered at with inner radius and outer radius , equal to on and supported in .✦ Normalizing to unit Lebesgue integral produces the standard mollifier , which is smooth, nonnegative, supported in , and of integral .✦
For a width , the scaled mollifier density is the function defined by .✦ Given any , the product kernel on is defined by ✦, and taking yields the product mollifier kernel in (3): (3)✦ For , the mollifier Lipschitz constant is given by (4): (4)✦
For a kernel and a function , the convolution smoothing with respect to Lebesgue measure on is defined by .✦ For each chart and coordinate , the smoothed chart function is the convolution defined by (5): (5)
The profile function is given by ✦, the profile constant is the real number ✦, and the source metric constant is the real number .✦
The Claim
Definition 1. The mollifier derivative integral is defined by (6): (6)✦ In terms of , , and , the contraction constant is given by (7): (7)✦ The contraction weight function maps each natural number to the real number (8): (8)✦
Theorem 1. Let and let be a finite set. Suppose vectors in satisfy and for all and , let and , and let satisfy for some and all , . Suppose further that there is such that whenever and is -regular for both and . Then for all and every , the gradients of the smoothed chart functions satisfy (9): (9)✦
Theorem 2. Given , a finite set , vectors in with and , scales and , and a map satisfying for some , suppose and are such that for every and all that are -regular for . Then for any and any that is -regular for , the smoothing error satisfies (10): (10)✦
Theorem 3. Take , a finite set , scales and , parameters , , and , and vectors in satisfying , , and whenever and . Let satisfy for some and obey the cross-chart bound for all and that is -regular for and , together with the interface jump bound whenever and in share a chart and satisfy at one index and for all . Assume also that for all and , and that for every continuous linear functional on there exists a subset with such that for all and . Then for every chart and every , the smoothed gradient satisfies (11): (11)✦
Theorem 4. In any real inner-product space , let and be finite index sets, let and for and , and let . Suppose for every , and suppose a subset with satisfies for all and . Then (12) holds: (12)✦
Theorem 5. Suppose , , and are finite index sets with , is a real inner-product space, and , . Let and for , , and satisfy for all . Assume that for every there exists a subset with such that for all and , and that for every the budgets and hold. Then the total norm satisfies (13): (13)✦
What Is Checked
The definitions (6), (7), and (8) and the bounds (9), (10), and (11) are stated over the standard mollifier , the constants and , and the smoothed chart step functions on for bounded vertex maps and direction families with nonzero zeroth coordinates and at scales and , whereas (12) and (13) are stated over finite vector decompositions in an arbitrary real inner-product space.