Physlib

PhyslibAlpha.ProbabilisticTheory.HilbertSpace.Unbounded.SpectralPointMass

Spectral measures taking values `0` and `1`

i. Overview

A measure on `ℝ` with bounded support that only takes the values `0` and `1` is a point mass. Bisecting the support and keeping the half of measure `1` gives nested intervals shrinking to the point.

ii. Key results

- `WOTSpectralMeasure.exists_forall_notMem_measure_eq_zero` : such a measure vanishes on every set avoiding some point `r`.

30 declarations

definition

One-step bisection of an interval [a,b][a, b] via measure μ\mu

Given a measure μ\mu and a pair of endpoints p=(a,b)∈R2p = (a, b) \in \mathbb{R}^2 representing the closed interval [a,b][a, b], this function computes the midpoint m=a+b2m = \frac{a + b}{2}. It returns the pair (a,m)(a, m) if the measure of the left sub-interval μ([a,m])=1\mu([a, m]) = 1; otherwise, it returns the pair (m,b)(m, b).

definition

The nn-th bisection interval (an,bn)(a_n, b_n) starting from (a,b)(a, b)

Given a measure μ\mu and initial endpoints a,b∈Ra, b \in \mathbb{R}, this function defines a sequence of pairs (an,bn)∈R×R(a_n, b_n) \in \mathbb{R} \times \mathbb{R} representing the endpoints of nested intervals. The sequence is defined recursively by (a0,b0)=(a,b)(a_0, b_0) = (a, b), and the (n+1)(n+1)-th interval is obtained by applying a bisection step `bisectStep` to the nn-th interval. At each step nn, the width of the interval is halved, such that bn−an=b−a2nb_n - a_n = \frac{b-a}{2^n}.

theorem

Bisection Step Preserves the Order of Endpoints a≤ba \le b

Let μ\mu be a measure on R\mathbb{R}. For any pair of real numbers p=(a,b)p = (a, b) representing an interval such that a≤ba \le b, the pair (a′,b′)=bisectStep(μ,p)(a', b') = \text{bisectStep}(\mu, p) resulting from the bisection step also satisfies a′≤b′a' \le b'.

theorem

Bisection Step Halves the Interval Width

For any pair of real numbers p=(a,b)p = (a, b) representing an interval, the width of the interval resulting from one bisection step bisectStep(μ,p)\text{bisectStep}(\mu, p) is equal to half the width of the original interval. That is, if (a′,b′)=bisectStep(μ,p)(a', b') = \text{bisectStep}(\mu, p), then b′−a′=b−a2b' - a' = \frac{b - a}{2}.

theorem

The left endpoint is non-decreasing under the bisection step.

Let μ\mu be a measure on R\mathbb{R}. For any pair p=(a,b)∈R2p = (a, b) \in \mathbb{R}^2 representing an interval [a,b][a, b] such that a≤ba \le b, the first component (left endpoint) of the pair produced by the bisection step bisectStep(μ,p)\text{bisectStep}(\mu, p) is greater than or equal to aa.

theorem

The right endpoint is non-increasing under the bisection step

Let μ\mu be a measure on R\mathbb{R}. For any pair of real numbers p=(a,b)p = (a, b) representing the closed interval [a,b][a, b] such that a≤ba \le b, the second component (right endpoint) of the pair produced by the bisection step bisectStep(μ,p)\text{bisectStep}(\mu, p) is less than or equal to the original right endpoint bb.

theorem

The interval resulting from a bisection step is a subset of the original interval

Let p=(a,b)p = (a, b) be a pair of real numbers representing the closed interval [a,b][a, b] such that a≤ba \le b. Let (a′,b′)=bisectStep(μ,p)(a', b') = \text{bisectStep}(\mu, p) be the pair of endpoints resulting from one step of the bisection process with respect to the measure μ\mu. Then the closed interval [a′,b′][a', b'] is a subset of the original closed interval [a,b][a, b], i.e., [a′,b′]⊆[a,b][a', b'] \subseteq [a, b].

theorem

The nn-th bisection interval endpoints satisfy an≤bna_n \le b_n

Let a,b∈Ra, b \in \mathbb{R} be real numbers such that a≤ba \le b. For any natural number nn, let (an,bn)(a_n, b_n) be the nn-th bisection interval starting from (a,b)(a, b) with respect to a measure μ\mu. Then, the left endpoint is less than or equal to the right endpoint, i.e., an≤bna_n \le b_n.

theorem

The width of the nn-th bisection interval is b−a2n\frac{b - a}{2^n}

Let (an,bn)(a_n, b_n) be the nn-th interval in the bisection sequence starting from (a,b)(a, b) relative to a spectral measure μ\mu. For any a,b∈Ra, b \in \mathbb{R} and any n∈Nn \in \mathbb{N}, the width of the nn-th interval satisfies bn−an=b−a2nb_n - a_n = \frac{b - a}{2^n}.

theorem

The sequence of left endpoints ana_n in the spectral measure bisection is non-decreasing.

Let μ\mu be a spectral measure on R\mathbb{R} and let [a,b][a, b] be an initial interval with a≤ba \le b. Let (an,bn)(a_n, b_n) denote the endpoints of the nn-th interval produced by the bisection process (which recursively selects the half-interval where the measure μ\mu is 1). Then, for any natural number nn, the left endpoint of the nn-th interval is less than or equal to the left endpoint of the (n+1)(n+1)-th interval, i.e., an≤an+1a_n \le a_{n+1}.

theorem

The right endpoints of bisection intervals are non-increasing (bn+1≤bnb_{n+1} \le b_n)

Let μ\mu be a measure on R\mathbb{R}. For any real numbers aa and bb such that a≤ba \le b, let (an,bn)(a_n, b_n) denote the nn-th interval produced by the bisection process starting from initial endpoints aa and bb. Then for any natural number nn, the right endpoint of the (n+1)(n+1)-th interval is less than or equal to the right endpoint of the nn-th interval, i.e., bn+1≤bnb_{n+1} \le b_n.

theorem

Successive bisection intervals are nested: [an+1,bn+1]⊆[an,bn][a_{n+1}, b_{n+1}] \subseteq [a_n, b_n]

Let μ\mu be a measure on R\mathbb{R} and a,b∈Ra, b \in \mathbb{R} such that a≤ba \leq b. Let (an,bn)(a_n, b_n) be the nn-th bisection interval obtained from the initial interval [a,b][a, b] with respect to μ\mu. Then for any n∈Nn \in \mathbb{N}, the (n+1)(n+1)-th interval is contained within the nn-th interval, i.e., [an+1,bn+1]⊆[an,bn][a_{n+1}, b_{n+1}] \subseteq [a_n, b_n].

theorem

Single Bisection Step Preserves Measure 1 for 00-11 Valued Measures

Let μ\mu be a measure on R\mathbb{R} such that for every measurable set E⊆RE \subseteq \mathbb{R}, μ(E)\mu(E) is either 00 or 11. Let p=(a,b)p = (a, b) be a pair of real numbers with a≤ba \le b such that the measure of the closed interval is μ([a,b])=1\mu([a, b]) = 1. Then the interval resulting from a single bisection step, [a′,b′]=bisectStep(μ,p)[a', b'] = \text{bisectStep}(\mu, p), also has measure μ([a′,b′])=1\mu([a', b']) = 1. (Note: The bisection step bisectStep(μ,p)\text{bisectStep}(\mu, p) returns the left sub-interval [a,a+b2][a, \frac{a+b}{2}] if its measure is 1, and the right sub-interval [a+b2,b][\frac{a+b}{2}, b] otherwise.)

theorem

μ([an,bn])=1\mu([a_n, b_n]) = 1 for Bisection Intervals of 0-1 Valued Measures

Let μ\mu be a measure on R\mathbb{R} that only takes the values 00 and 11. For any real numbers a≤ba \le b such that the measure of the closed interval [a,b][a, b] is μ([a,b])=1\mu([a, b]) = 1, the sequence of intervals [an,bn][a_n, b_n] produced by the bisection process starting from [a,b][a, b] satisfies μ([an,bn])=1\mu([a_n, b_n]) = 1 for all n∈Nn \in \mathbb{N}.

theorem

μ(In∖In+1)=0\mu(I_n \setminus I_{n+1}) = 0 for 00-11 spectral measure bisection intervals

Let μ\mu be a measure on R\mathbb{R} such that for every measurable set EE, μ(E)\mu(E) is either 00 or 11. Let [a,b][a, b] be an interval with a≤ba \le b such that μ([a,b])=1\mu([a, b]) = 1. If In=[an,bn]I_n = [a_n, b_n] denotes the nn-th interval in the bisection sequence starting from [a,b][a, b], then for any n∈Nn \in \mathbb{N}, the measure of the set difference between the nn-th and (n+1)(n+1)-th intervals is zero, i.e., μ(In∖In+1)=0\mu(I_n \setminus I_{n+1}) = 0.

theorem

Monotonicity of the left endpoints ana_n in the bisection process

Let μ\mu be a measure on R\mathbb{R} and let a,b∈Ra, b \in \mathbb{R} be such that a≤ba \leq b. Let (an,bn)(a_n, b_n) be the sequence of intervals defined by the bisection process starting from [a,b][a, b] with respect to μ\mu. Then the sequence of left endpoints n↦ann \mapsto a_n is monotone (non-decreasing).

theorem

Right endpoints of bisection intervals are antitone

Let μ\mu be a measure on R\mathbb{R} and let a,ba, b be real numbers such that a≤ba \leq b. Let (an,bn)(a_n, b_n) denote the endpoints of the nn-th interval produced by the bisection process bisect(μ,a,b,n)\text{bisect}(\mu, a, b, n). The sequence of right endpoints bnb_n is antitone; that is, for any n,m∈Nn, m \in \mathbb{N}, if n≤mn \leq m, then bm≤bnb_m \leq b_n.

theorem

The right endpoint bnb_n of the nn-th bisection interval is bounded by bb

Let μ\mu be a measure and a,b∈Ra, b \in \mathbb{R} such that a≤ba \leq b. Let (an,bn)(a_n, b_n) be the nn-th bisection interval starting from (a,b)(a, b). For any natural number nn, the right endpoint bnb_n of the nn-th interval is less than or equal to the initial right endpoint bb.

theorem

The left endpoint ana_n is bounded below by the initial value aa in the bisection process

Let μ\mu be a spectral measure and let a,b∈Ra, b \in \mathbb{R} be such that a≤ba \le b. If (an,bn)(a_n, b_n) denotes the endpoints of the nn-th bisection interval starting from [a,b][a, b], then for any natural number nn, the left endpoint ana_n satisfies a≤ana \le a_n.

theorem

The sequence of left endpoints ana_n is bounded above

Let μ\mu be a spectral measure and let aa and bb be real numbers such that a≤ba \leq b. Consider the sequence of intervals [an,bn][a_n, b_n] generated by the bisection process starting from [a,b][a, b] (defined by `bisect μ a b n`). Then the set of left endpoints {an:n∈N}\{a_n : n \in \mathbb{N}\} is bounded above.

theorem

The set of right endpoints bnb_n of the bisection intervals is bounded below

Let a,b∈Ra, b \in \mathbb{R} be such that a≤ba \leq b. For a measure μ\mu, let (an,bn)(a_n, b_n) denote the nn-th interval produced by the bisection process starting from [a,b][a, b]. Then the set of right endpoints {bn:n∈N}\{b_n : n \in \mathbb{N}\} is bounded from below.

definition

Limit point of bisection intervals sup⁡nan\sup_n a_n

Given a measure μ\mu and an initial interval [a,b][a, b] with a≤ba \leq b, let (an,bn)(a_n, b_n) be the sequence of nested intervals produced by the bisection procedure (where each step selects the sub-interval of measure 1). This value is defined as the supremum of the left endpoints, sup⁡n∈Nan\sup_{n \in \mathbb{N}} a_n, which represents the unique limit point of the nested intervals.

theorem

The left endpoints ana_n converge to the limit point rr of the bisection intervals

Let a,b∈Ra, b \in \mathbb{R} be such that a≤ba \leq b. For a measure μ\mu, let (an,bn)(a_n, b_n) denote the nn-th interval generated by the bisection process starting from [a,b][a, b], where at each step the sub-interval of measure 11 is selected. Let rr be the limit point of these intervals, defined as r=sup⁡n∈Nanr = \sup_{n \in \mathbb{N}} a_n. Then the sequence of left endpoints ana_n converges to rr as n→∞n \to \infty.

theorem

lim⁡n→∞(bn−an)=0\lim_{n \to \infty} (b_n - a_n) = 0 for bisection intervals

For any real numbers aa and bb, let (an,bn)(a_n, b_n) denote the nn-th interval generated by the bisection process starting from initial endpoints aa and bb. Then the width of these intervals, given by the difference bn−anb_n - a_n, converges to 00 as n→∞n \to \infty.

theorem

The sequence of right endpoints bnb_n converges to the bisection limit point rr

Given a spectral measure μ\mu and initial endpoints a,b∈Ra, b \in \mathbb{R} with a≤ba \leq b, let (an,bn)(a_n, b_n) be the sequence of nested intervals produced by the bisection process. Then the sequence of right endpoints bnb_n converges to the bisection limit point r=sup⁡n∈Nanr = \sup_{n \in \mathbb{N}} a_n as n→∞n \to \infty.

theorem

an≤x∗a_n \leq x^* for bisection intervals

Let μ\mu be a spectral measure taking values in {0,1}\{0, 1\}. For an initial interval [a,b][a, b] with a≤ba \leq b, let (an,bn)(a_n, b_n) be the sequence of nested intervals produced by recursively bisecting the interval and selecting the sub-interval with measure 11. Let x∗x^* be the limit point of these intervals, defined as the supremum of the left endpoints, x∗=sup⁡k∈Nakx^* = \sup_{k \in \mathbb{N}} a_k. Then for any n∈Nn \in \mathbb{N}, the left endpoint of the nn-th interval satisfies an≤x∗a_n \leq x^*.

theorem

x∗≤bnx^* \le b_n for bisection intervals

Let μ\mu be a spectral measure on R\mathbb{R} that takes values in {0,1}\{0, 1\}. For an initial interval [a,b][a, b] with a≤ba \le b, let (an,bn)(a_n, b_n) be the sequence of nested intervals produced by recursively bisecting the interval and selecting the sub-interval with measure 11. Let x∗x^* be the limit point of these intervals, defined as the supremum of the left endpoints, x∗=sup⁡k∈Nakx^* = \sup_{k \in \mathbb{N}} a_k. Then for any n∈Nn \in \mathbb{N}, the bisection point x∗x^* satisfies x∗≤bnx^* \le b_n.

theorem

The intersection of bisection intervals ⋂n[an,bn]\bigcap_n [a_n, b_n] is the limit point {r}\{r\}

Let a,b∈Ra, b \in \mathbb{R} be real numbers such that a≤ba \leq b. Let (an,bn)(a_n, b_n) denote the sequence of endpoints of the nested intervals produced by the bisection procedure starting from the interval [a,b][a, b]. Then the intersection of all the closed intervals [an,bn][a_n, b_n] for n∈Nn \in \mathbb{N} is the singleton set {r}\{r\}, where rr is the bisection limit point defined as r=sup⁡n∈Nanr = \sup_{n \in \mathbb{N}} a_n. That is, ⋂n∈N[an,bn]={r}.\bigcap_{n \in \mathbb{N}} [a_n, b_n] = \{r\}.

theorem

The bisection point of a {0,1}\{0, 1\}-valued measure has measure 11

Let μ\mu be a measure on R\mathbb{R} such that for every measurable set E⊆RE \subseteq \mathbb{R}, μ(E)∈{0,1}\mu(E) \in \{0, 1\}. If a≤ba \le b and the measure of the closed interval [a,b][a, b] is μ([a,b])=1\mu([a, b]) = 1, then μ({x∗})=1\mu(\{x^*\}) = 1, where x∗x^* is the bisection point (the unique limit point of the sequence of nested sub-intervals of [a,b][a, b] that each have measure 11).

theorem

A {0,1}\{0, 1\}-valued measure on R\mathbb{R} with bounded support is a point mass

Let μ\mu be a measure on the real numbers R\mathbb{R} such that for every measurable set E⊆RE \subseteq \mathbb{R}, the measure μ(E)\mu(E) is either 00 or 11. Suppose there exists a closed interval [a,b][a, b] with a≤ba \le b such that μ([a,b])=1\mu([a, b]) = 1. Then there exists a point r∈Rr \in \mathbb{R} such that μ({r})=1\mu(\{r\}) = 1, and for any measurable set EE, if r∉Er \notin E, then μ(E)=0\mu(E) = 0.