Separation Theorems and Supporting Functionals in Normed Spaces: Structure, Duality, and Applications

Engineering and Technology Journal e-ISSN: 2456-3358 Volume 10 Issue 10 October-2025, Page No.-7637-7647 DOI: 10.47191/etj/v10i10.38, I.F. – 8.482 © 2025, ETJ 7637 ETJ Volume 10 Issue 10 October 2025, Santosh Kumar Separation Theorems and Supporting Functionals in Normed Spaces: Structure, Duality, and Applications Santosh Kumar Department of Mathematics, Shree Radha Krishna Goenka College, Sitamarhi, B.R.A. Bihar University, Muzaffarpur, Bihar, India. ABSTRACT: This paper presents a rigorous, application-oriented survey of separation theorems and supporting functionals for convex sets in normed and Banach spaces. Emphasizing geometric version of the Hahn-Banach theorem, this work develops clean conditions for strict and non-strict separation, existence of supporting hyperplanes, and links to dual cones and polar sets. We highlight the role of weak and weak-star topologies in separation and illustrate how these results undergird feasibility, sensitivity, and duality principles in convex optimization. Short, self-contained proofs and examples are provided to keep the exposition accessible while maintaining mathematical precision. The paper thereby complements classical treatments of linear functional extension by focusing on geometric separation mechanisms and their applied consequences, especially in linear programming, convex feasibility, and basic duality frameworks. This comprehensive investigation into separation theorems reveals fundamental geometric structures that underpin both theoretical functional analysis and practical optimization applications, providing a unified framework for understanding convex separation phenomena across diverse mathematical contexts. KEYWORDS: separation theorems, supporting functionals, convex sets, normed spaces, duality theory, geometric Hahn-Banach, hyperplane separation, polar cones (I} INTRODUCTION Functional analysis has witnessed a transformative shift with the rise of geometric perspectives, particularly those rooted in the Hahn-Banach theorem. Moving beyond the classical focus on extending linear functionals, these geometric insights have revealed powerful tools for unveiling the structure of convex sets and the logic of their separation in infinite-dimensional environments. This change in viewpoint deepens our understanding, offering not only theoretical advances but also new possibilities for practical algorithms in optimization and convex analysis. In contemporary mathematics, the ability to separate and support convex sets shapes duality theory, informs feasibility analysis, and forms the backbone of algorithmic strategies in convex programming. The influence of these concepts extends into areas such as machine learning, economics, and engineering, highlighting the continued impact and relevance of separation principles well beyond their original analytic context.[1-4] This transition from extension theorems to separation results marks a fundamental shift in perspective, moving from analytic considerations of functional extension to geometric insights about convex sets and hyperplanes. This geometric viewpoint has profound implications for optimization theory, where separation principles form the foundation of duality theory, feasibility analysis, and algorithmic approaches to convex programming.[5-8] Modern applications in machine learning, economic theory, and engineering optimization demonstrate the continuing relevance of these classical results. Support vector machines explicitly utilize optimal separating hyperplanes, while convex optimization algorithms rely fundamentally on separation-based feasibility certificates and duality relationships. The mathematical rigor of separation theorems provides the theoretical foundation that ensures convergence and optimality in these applied contexts. [9-11] This work of ours aims to provide a thorough yet accessible treatment of the separation theorems and its supporting functionals, emphasizing their geometric interpretation and practical applications. We focus particularly on the interplay between topological properties of convex sets and the existence of separating hyperplanes, developing the theory with sufficient generality to cover infinite-dimensional applications while maintaining clarity through concrete examples and illustrations. (II) PRELIMINARIES (A) Normed Spaces and Convex Sets Assume a real normed space 𝔼 having norm ‖⋅‖. The fundamental objects of our study are convex sets/subsets, as they preserve the essential geometric structure necessary for separation results[3, 4]. Definition 2.1 Following the classical formulation by Rudin,(1991)[3] and Rockafellar (1970)[9], Let 𝔼 be a “Separation Theorems and Supporting Functionals in Normed Spaces: Structure, Duality, and Applications” 7638 ETJ Volume 10 Issue 10 October 2025, Santosh Kumar real normed space. A subset ℂ⊆𝔼 is called convex if, given any two points x,y∈ℂ , every intermediate point 𝑧=𝑡𝑥+(1−𝑡)𝑦 Where Scalar 𝑡∈[0,1] also lies in ℂ. The geometric significance of convexity lies in its preservation under various operations and its compatibility with linear functionals. Classical examples from functional analysis literature [3-5] demonstrate this below structures: • Unit balls in normed spaces: 𝐵={x∈𝔼∶‖x‖≤ 1}, which serve as fundamental convex prototypes • Hyperplanes in a real normed space 𝔼 is any set of the form 𝐻𝑓,𝛼 ={ 𝑥∈𝔼:𝑓(𝑥)=𝛼}, where 𝑓:𝔼→ℝ is a continuous linear functional and 𝛼∈ℝ. • Half-Space : a closed half-space is determined by 𝑓 and 𝛼 is 𝐻𝑓,𝛼 +={ 𝑥∈𝔼:𝑓(𝑥)≥𝛼}, and the corresponding open half-space is 𝐻𝑓,𝛼 −={ 𝑥∈𝔼:𝑓(𝑥)<𝛼}. Definition 2.2 (Minkowski Functional, also called gauge function): Following the classical formulation in Rudin [3] and Rockafellar [9]; Let 𝔼 be a real normed space and let ℂ⊆𝔼 be a convex set containing the origin. The Minkowski functional associated with ℂ is the map 𝑝ℂ:𝔼 ⟶ [0,∞],𝑝ℂ(𝑥) = inf{ 𝑡>0:𝑥∈𝑡 ℂ }, where 𝑡ℂ={ 𝑡 𝑦:𝑦∈ℂ}. By construction, 𝑝ℂ is: • positively homogeneous: 𝑝ℂ(𝜆 𝑥)=𝜆 𝑝ℂ(𝑥) for all 𝜆≥0, • subadditive: 𝑝ℂ(𝑥+𝑦)≤𝑝ℂ(𝑥)+𝑝ℂ(𝑦). Properties and Significance [3, 8, 9]: The Minkowski functional establishes a fundamental connection between the geometric structure of convex sets and their analytic characterization through sublinear functions. This functional possesses essential sublinearity properties—specifically, positive homogeneity and subadditivity—that enable the application of the analytic Hahn-Banach extension principle [3, 8]. Consequently, it creates a direct correspondence between geometric convexity assumptions and the theory of linear functional extension, making it an indispensable tool in the development of separation theorems. Table 1: Topological Properties of Convex Sets by Structural Type Topological Property Open Convex Sets Closed Convex Sets Compact Convex Sets Boundary Structure Excludes boundary points Incorporates boundary Complete boundary inclusion Interior Characterization Inherently non-empty core Potentially empty interior Guaranteed interior (finite dimensions) Separation Behavior Mazur-type separation Basic separation results Strict separation achievable Compactness Properties Non-compact in general Conditional compactness Compact by construction (B) Hyperplanes and Half-Spaces The geometric foundation of separation theory rests on the proper understanding of hyperplanes and their relationship to linear functionals.[11, 3] Definition 2.3 (Affine Hyperplanes and HalfSpace Decomposition): In the framework established by Rudin [3] and Holmes [7] and building upon the concepts introduced in Definition 2.1, where hyperplanes and half-spaces were defined, we consider the structure of linear separation in a real normed space 𝔼. Let 𝑓:𝔼→ℝ be a nonzero continuous linear functional and 𝛼∈ ℝ a scalar parameter. The hyperplane associated with (𝑓,𝛼) is the affine subset 𝐻𝑓,𝛼 ={𝑥∈𝔼:𝑓(𝑥)=𝛼}. This hyperplane divides the space into two complementary subsets: • The closed half-space 𝐻𝑓,𝛼 +={𝑥∈𝔼:𝑓(𝑥)≥𝛼}, • The open half-space 𝐻𝑓,𝛼 −={𝑥∈𝔼:𝑓(𝑥)<𝛼}. These sets play a central role in the geometric theory of linear separation. Proposition 2.4 (Topological Closedness of Hyperplanes). For a linear functional 𝑓:𝔼→ℝ and scalar 𝛼∈ℝ, the hyperplane 𝐻𝑓,𝛼 is closed in the norm topology of 𝔼 if and only if 𝑓 is continuous. Proof. The forward direction follows from observing that 𝐻𝑓,𝛼 =𝑓−1({𝛼}); if 𝑓 is continuous, the preimage of the closed singleton {𝛼}⊆ℝ must be closed. Conversely, if 𝐻𝑓,𝛼 is closed and 𝑓 is discontinuous, one can construct a sequence converging to the hyperplane while 𝑓 oscillates unboundedly, yielding a contradiction. This correspondence [3, 4, 7] establishes that normclosed separation necessitates continuous dual functionals, thereby linking geometric separation with dualspace structure. (C ) Topological Concepts The effectiveness of separation theorems depends critically on the topological properties of the underlying convex sets. In infinite-dimensional spaces, we must carefully distinguish between different topological structures.[2, 13] Definition 2.5 (Topological Framework). Following the systematic development in Rudin [3] and Phelps [13], the geometric theory of separation requires careful attention to “Separation Theorems and Supporting Functionals in Normed Spaces: Structure, Duality, and Applications” 7639 ETJ Volume 10 Issue 10 October 2025, Santosh Kumar distinct topological structures on normed spaces and their duals. For a normed space 𝔼 with continuous dual 𝔼∗, we distinguish three fundamental topological regimes [2,13]: • Norm-induced Structure: This is the topology on 𝔼 induced by the metric 𝑑(𝑥,𝑦)=‖𝑥−𝑦‖, where ‖⋅‖ denotes the given norm. It provides the strongest framework for geometric separation. • Weak Functional Topology: The coarsest topology 𝜏𝑤(𝔼,𝔼∗)on 𝔼 ensuring continuity of all elements 𝑓 ∈ 𝔼∗, creating the minimal topological structure compatible with dual space pairing [3,13]. • Dual Weak-* Structure: Denoted as 𝜏𝑤∗(𝔼∗,𝔼), this is the minimal topology on 𝔼∗ ensuring that each evaluation map 𝜑𝑥:𝔼∗→ℝ,𝜑𝑥(𝑓)=𝑓(𝑥) is continuous for all 𝑥∈𝔼. This establishes the fundamental biduality framework [2, 3]. These topological distinctions [13,4] prove essential for separation analysis, particularly when strong compactness arguments fail and weak compactness provides the necessary substitute for geometric separation theorems. Example 2.6 (Topological Distinctions in ℓ1). Consider the sequence space ℓ1 with the standard coordinate basis vectors 𝑢𝑛 (where 𝑢𝑛 has 1 in position n and 0 elsewhere). Examine the collection: 𝑆 = {𝑢𝑛: 𝑛 ∈ℕ} This configuration exhibits remarkable topological behavior [2,13]: • Norm Topology Analysis: Under the ℓ1-norm ||· ||₁, the set 𝑆 lacks sequential compactness since ||𝑢𝑛− 𝑢𝑚||1= 2 for all 𝑛 ≠ 𝑚, preventing the extraction of convergent subsequences. • Weak Topology Analysis: The weak closure 𝑆𝜎(ℓ1,ℓ∞) incorporates additional limit points not present in the norm closure, fundamentally altering separation characteristics. This dichotomy [2,13] demonstrates how topological framework selection critically impacts separation theorem applicability in infinite-dimensional functional analysis, where weak topological properties may enable separation results impossible under norm topology. (III) SEPARATION THEOREMS (A) Basic Separation Results Geometric functional analysis finds its foundation in separation principles that ensure hyperplane boundaries between non-intersecting convex regions when suitable topological conditions hold [3,5]. Theorem 3.1 (Separation via Continuous Functionals). Consider a real normed space 𝔼 and two nonempty convex subsets 𝐴,𝐵⊆𝔼 with 𝐴∩𝐵=∅. If 𝐴 possesses the additional property of being open in the norm topology, then one can construct a nonzero continuous linear functional 𝑓∈ 𝔼∗ together with a threshold value 𝛼∈ℝ satisfying 𝑓(𝑥)<𝛼≤𝑓(𝑦) for all 𝑥∈𝐴 and 𝑦∈𝐵. This provides a hyperplane {𝑧:𝑓(𝑧)=𝛼} that strictly separates 𝐴 from 𝐵 [3, 1]. Proof. Step 1: Establishing a reference framework. Fix an arbitrary element 𝑎0∈𝐴. The openness and convexity of 𝐴 ensure that the translated collection 𝐴′=𝐴−𝑎0={𝑥−𝑎0:𝑥∈𝐴} forms an open convex neighborhood of the origin in 𝔼. Step 2: Constructing the gauge. Define the gauge functional 𝑝:𝔼→ℝ associated with 𝐴′ by 𝑝(𝑥)=inf{𝑡>0:𝑥∈𝑡 𝐴′}. This function is sublinear (satisfying 𝑝(𝜆𝑥)=𝜆𝑝(𝑥) for 𝜆≥ 0 and 𝑝(𝑥+𝑦)≤𝑝(𝑥)+𝑝(𝑦)). Moreover, for any 𝑥∈𝐴′, one has 𝑝(𝑥)<1, while for every element 𝑏∈𝐵′:=𝐵−𝑎0, the condition 𝑝(𝑏)≥1 holds due to disjointness. Step 3: Functional extension. By the analytic Hahn-Banach theorem, there exists a linear map 𝑓:𝔼→ℝ satisfying 𝑓(𝑥)≤𝑝(𝑥) for all 𝑥∈𝔼. The boundedness of 𝑝 on the unit sphere implies continuity of 𝑓, hence 𝑓∈𝔼∗ with ‖𝑓‖≤1 [1]. Step 4: Verifying the separation inequalities. For each 𝑥∈ 𝐴, write 𝑥=𝑎0+𝑥′ where 𝑥′∈𝐴′. Then 𝑓(𝑥)=𝑓(𝑎0)+𝑓(𝑥′)<𝑓(𝑎0)+1. Similarly, for any 𝑦∈𝐵, express 𝑦=𝑎0+𝑦′ with 𝑦′∈𝐵′, yielding 𝑓(𝑦)=𝑓(𝑎0)+𝑓(𝑦′)≥𝑓(𝑎0)+1. Defining 𝛼:=𝑓(𝑎0)+1 establishes the desired strict inequality: 𝑓(𝑥)<𝛼≤𝑓(𝑦) for all 𝑥∈𝐴 and 𝑦∈𝐵. Step 5: Conclusion. The constructed functional 𝑓∈𝔼∗ and scalar 𝛼 provide the required separation, confirming that openness of one set suffices for strict hyperplane separation in normed spaces [4, 1]. Example 3.2 (Separation Failure in Sequence Spaces). Consider the Hilbert space ℓ2 and define the sets 𝐴={𝑒𝑛:𝑛∈ℕ} and 𝐵={0}, where 𝑒𝑛 denotes the standard basis vector with 1 in position 𝑛 and 0 elsewhere. Although 𝐴 and 𝐵 are convex and disjoint, no closed affine subspace of codimension one can achieve their separation [2, 3]. To verify this claim, suppose a continuous linear functional 𝑓∈(ℓ2)∗ and scalar 𝛼∈ℝ satisfy 𝑓(𝑒𝑛)<𝛼≤𝑓(0)=0 for all 𝑛∈ℕ. Then 𝑓(𝑒𝑛)<0 for each 𝑛, yet by the Riesz representation theorem, 𝑓 corresponds to some sequence 𝑦=(𝑦𝑛)∈ℓ2, so 𝑓(𝑒𝑛)=𝑦𝑛. The requirement ∑ ∞ 𝑛=1 |𝑦𝑛|2<∞ forces 𝑦𝑛→ 0, contradicting 𝑦𝑛<0 uniformly. This demonstrates that interior point assumptions are indispensable for separation results in infinitedimensional normed spaces [3, 2]. “Separation Theorems and Supporting Functionals in Normed Spaces: Structure, Duality, and Applications” 7640 ETJ Volume 10 Issue 10 October 2025, Santosh Kumar (B) Strict Separation Theorems When additional compactness assumptions are satisfied, strict separation becomes achievable, ensuring a quantifiable gap separating the disjoint convex regions [11,3]. Theorem 3.3 (Strict Hyperplane Separation Under Compactness). Considering as previously done, take 𝔼 to be a real normed space containing two nonempty convex subsets 𝐴 and 𝐵 satisfying 𝐴∩𝐵=∅. Suppose further that 𝐴 is closed in the norm topology and 𝐵 is compact. Then there exists a nonzero continuous linear functional 𝑓∈𝔼∗ and real scalars 𝛽1<𝛽2 such that 𝑓(𝑎)≤𝛽1<𝛽2≤𝑓(𝑏) for all 𝑎∈𝐴 and 𝑏∈𝐵. This establishes strict separation with a positive gap 𝛽2− 𝛽1>0 between the two sets [2, 1]. Proof. Stage I: Normalizing the configuration. Select an arbitrary base element 𝑎∗∈𝐴 and define the translated collections 𝐴˜=𝐴−𝑎∗={𝑥−𝑎∗:𝑥∈𝐴},𝐵˜=𝐵−𝑎∗={𝑦−𝑎∗:𝑦 ∈𝐵}. Both 𝐴˜ and 𝐵˜ inherit convexity and disjointness from 𝐴 and 𝐵. Moreover, 𝐴˜ remains closed and 𝐵˜ remains compact under this translation. Without loss of generality, we assume 𝑎∗= 0, so that 0∈𝐴 [1]. Stage II: Quantifying the separation distance. Since 𝐴 is closed and 𝐵 is compact with 𝐴∩𝐵=∅, the quantity 𝛿:=inf{‖𝑎−𝑏‖:𝑎∈𝐴,𝑏∈𝐵} is strictly positive. By compactness of 𝐵 and closedness of 𝐴, this infimum is attained: there exist 𝑎0∈𝐴 and 𝑏0∈𝐵 such that ‖𝑎0−𝑏0‖=𝛿>0 [2]. Stage III: Constructing the direction of separation. Define the unit vector 𝜉:= 𝑏0−𝑎0 ‖𝑏0−𝑎0‖. This vector points from 𝐴 toward 𝐵 and will determine the orientation of the separating hyperplane. Consider the linear functional 𝜑:span{𝜉}→ℝ given by 𝜑(𝑡𝜉)=𝑡 for 𝑡∈ℝ [1]. Stage IV: Extension via Hahn-Banach [1]. By the HahnBanach extension theorem (analytic form), there exists a continuous linear functional 𝑓:𝐸→ℝ extending 𝜑 with ‖𝑓‖≤1. In particular, 𝑓(𝜉)=1 [1]. Stage V: Establishing the strict separation inequality. For any 𝑎∈𝐴, we have 𝑓(𝑏0)−𝑓(𝑎)=𝑓(𝑏0−𝑎)=𝑓((𝑏0−𝑎0)+(𝑎0−𝑎)). Using the fact that ‖𝑏0−𝑎0‖=𝛿 and 𝑓(𝜉)=1, we obtain 𝑓(𝑏0−𝑎0)=𝑓(𝛿⋅𝜉)=𝛿. Since ‖𝑓‖≤1 and ‖𝑎0−𝑎‖≥0, the functional 𝑓 satisfies 𝑓(𝑏0)≥𝑓(𝑎)+ 𝛿 for all 𝑎∈𝐴. By a symmetric argument applied to all 𝑏∈𝐵 using compactness, we deduce 𝑓(𝑏)≥𝑓(𝑎)+𝛿 2 for all 𝑎∈𝐴,𝑏∈𝐵. Stage VI: Defining separation constants. Set 𝛽1:=sup 𝑎∈𝐴 𝑓(𝑎),𝛽2:=inf 𝑏∈𝐵 𝑓(𝑏). By the preceding inequality, 𝛽2−𝛽1≥𝛿/2>0, confirming strict separation [2]. Stage VII: Translating back. If the original sets were not normalized (i.e., 𝑎∗≠0), define 𝑓˜(𝑥)=𝑓(𝑥−𝑎∗) to obtain the separating functional for the original configuration. This completes the proof. Corollary 3.4 (Separation for Closed-Open Convex Pairs) [7, 8]. Consider a real normed space 𝔼 containing two nonempty, disjoint convex subsets with complementary topological properties: let 𝐴⊆𝔼 be closed in the norm topology, and let 𝐵⊆𝔼 possess the property of openness. Under these conditions, one can construct a continuous nonzero linear functional 𝑓∈𝔼∗ together with threshold values 𝛾1,𝛾2∈ℝ satisfying 𝛾1<𝛾2 such that 𝑓(𝑎)≤𝛾1<𝛾2≤𝑓(𝑏) for all 𝑎∈𝐴 and 𝑏∈𝐵. This result, attributed to Eidelheit, demonstrates that the topological contrast between closedness and openness suffices to guarantee strict separation with a positive gap[1, 2]. Proof. The argument proceeds by reducing the closed-open configuration to the closed-compact framework of Theorem 3.3. Step 1: Local compactness via openness. Since 𝐵 is open and nonempty, select any point 𝑏0∈𝐵. The openness ensures the existence of 𝜀>0 such that the closed ball 𝐾:=𝐵𝜀(𝑏0)={𝑥∈𝔼:‖𝑥−𝑏0‖≤𝜀} lies entirely within 𝐵. This closed ball 𝐾 is a compact subset (in finite dimensions) or can be replaced by a weakly compact subset in infinite-dimensional Banach spaces. Step 2: Applying strict separation. By Theorem 3.3, the disjoint pair (𝐴,𝐾) admits strict separation: there exists 𝑓1∈ 𝔼∗ and constants 𝛼1<𝛽1 with 𝑓1(𝑎)≤𝛼1<𝛽1≤𝑓1(𝑘) for all 𝑎∈𝐴,𝑘∈𝐾. Step 3: Extension to the entire open set. Since 𝐾⊆𝐵 and 𝑓1 is continuous, the inequality extends by continuity to a neighborhood of 𝐾. For any 𝑏∈𝐵, the openness of 𝐵 allows us to find a compact subset containing 𝑏 and apply the same separation argument. By considering the infimum and supremum of 𝑓1 over 𝐴 and 𝐵 respectively, we obtain 𝛾1:=sup 𝑎∈𝐴 𝑓1(𝑎),𝛾2:=inf 𝑏∈𝐵 𝑓1(𝑏), with 𝛾1<𝛾2 by construction. Step 4: Conclusion. Setting 𝑓:=𝑓1 yields the required strict separation of 𝐴 and 𝐵 with the desired gap 𝛾2− 𝛾1>0 [2, 1]. (C ) Mazur Separation Theorem The Mazur separation theorem provides a sophisticated treatment of separation involving open convex sets, with important connections to the weak topology[2, 13]. Theorem 3.5 (Mazur Separation via Relative Interior Points). Consider a real normed space 𝔼 and two nonempty, disjoint convex subsets 𝐴,𝐵⊆𝔼. Suppose that 𝐴 possesses at least “Separation Theorems and Supporting Functionals in Normed Spaces: Structure, Duality, and Applications” 7641 ETJ Volume 10 Issue 10 October 2025, Santosh Kumar one point lying in its relative interior—that is, a point 𝑎0∈ 𝐴 that is interior to 𝐴 when viewed within the affine subspace aff(𝐴) generated by 𝐴. Under this geometric condition, one can construct a nonzero continuous linear functional 𝑓∈𝔼∗ and a threshold 𝛼∈ℝ satisfying 𝑓(𝑎)≤𝛼≤𝑓(𝑏) for all 𝑎∈𝐴 and 𝑏∈𝐵. This result, due to Mazur, establishes that relative interior points enable non-strict separation even when neither set is open in the ambient space topology[2]. Proof. Phase I: Coordinate normalization via affine translation. Since 𝑎0 lies in the relative interior of 𝐴 with respect to aff(𝐴), we may translate the configuration so that 𝑎0 becomes the origin. Define 𝐴˜=𝐴−𝑎0,𝐵˜=𝐵−𝑎0. Both sets inherit convexity and disjointness from 𝐴 and 𝐵. By construction, 0∈𝐴˜ and 0 is a relative interior point of 𝐴˜ within aff(𝐴˜). Without loss of generality, we work with 𝐴˜ and 𝐵˜, and assume 𝑎0=0 [2]. Phase II: Exploiting the relative interior property. The hypothesis that 0 is a relative interior point of 𝐴 within aff(𝐴) implies the existence of 𝑟>0 such that the intersection 𝑟⋅𝑈∩aff(𝐴)⊆𝐴, where 𝑈={𝑥∈𝔼:‖𝑥‖≤1} denotes the closed unit ball. This provides a neighborhood base for constructing the gauge functional. Phase III: Constructing the Minkowski gauge. Define the functional 𝑝:𝔼→[0,∞] by 𝑝(𝑥)=inf{𝑡>0:𝑥∈𝑡⋅𝐴}. This gauge is sublinear: it satisfies 𝑝(𝜆𝑥)=𝜆𝑝(𝑥) for 𝜆≥0 and 𝑝(𝑥+𝑦)≤𝑝(𝑥)+𝑝(𝑦). Moreover, by the interior property, 𝑝(𝑥)<1 for all 𝑥 in a neighborhood of 0 within aff(𝐴), while for any 𝑏∈𝐵, the disjointness 𝐴∩𝐵=∅ ensures 𝑝(𝑏)≥1 [1]. Phase IV: Linear functional construction via HahnBanach. Select an arbitrary element 𝑏0∈𝐵. On the onedimensional subspace span{𝑏0}, define the linear map 𝜑:span{𝑏0}→ℝ by 𝜑(𝑡𝑏0)=𝑡⋅𝑝(𝑏0). By the HahnBanach extension theorem (analytic form), there exists a linear extension 𝑓:𝔼→ℝ satisfying 𝑓(𝑥)≤𝑝(𝑥) for all 𝑥∈𝔼. The boundedness of 𝑝 on the unit sphere (via the interior condition) implies continuity of 𝑓, hence 𝑓∈𝔼∗ [1]. Phase V: Verifying separation inequalities. For any 𝑎∈𝐴, the definition of the gauge gives 𝑝(𝑎)≤1, hence 𝑓(𝑎)≤ 𝑝(𝑎)≤1. For any 𝑏∈𝐵, the disjointness condition ensures 𝑝(𝑏)≥1, yielding 𝑓(𝑏)≥𝑝(𝑏)≥1. Therefore, setting 𝛼=1 achieves the separation: 𝑓(𝑎)≤ 1≤𝑓(𝑏) for all 𝑎∈𝐴 and 𝑏∈𝐵 [2]. Phase VI: Translating to original coordinates. If the original configuration had 𝑎0≠0, define 𝑓˜(𝑥)=𝑓(𝑥−𝑎0) to obtain the separating functional for the untranslated sets 𝐴 and 𝐵 [2, 4]. This completes the proof. Table 2: Separation Outcome under the Interior-Point Conditions Separation Type Set Conditions Hyperplane Type Applications Basic One set has interior point Closed General feasibility Strict One compact, one closed Closed with gap Optimization duality Mazur Convex vs. point Closed Weak topology analysis Example 3.6 In sequence spaces, the Mazur theorem explains why weakly convergent sequences that don't converge strongly can be separated from their limits by appropriate linear functionals, providing geometric insight into the distinction between weak and strong topologies. (IV) SUPPORTING FUNCTIONALS AND HYPERPLANES (A) Existence of Supporting Hyperplanes Supporting hyperplanes represent a refined geometric concept where separation occurs at the boundary between sets, providing crucial information about the local structure of convex sets [14, 3]. Definition 4.1 (Supporting Hyperplane at a Boundary Point). Let 𝔼 be a real normed space and ℂ⊆𝔼 a nonempty convex set. A point 𝑥0∈ℂ is on the boundary of ℂ if no open neighborhood of 𝑥0 lies entirely within ℂ. A hyperplane 𝐻={ 𝑥∈𝔼:𝑓(𝑥)=𝑓(𝑥0)} determined by a nonzero continuous linear functional 𝑓∈𝔼∗ is said to support ℂ at 𝑥0 precisely when 𝑓(𝑥) ≤ 𝑓(𝑥0) for every 𝑥∈ℂ. In this situation, 𝐻 touches ℂ at 𝑥0 and does not intersect the interior of ℂ. This existence of supporting hyperplanes connects local boundary properties with global separation phenomena, forming a bridge between differential and geometric approaches to convex analysis [15, 14]. Theorem 4.2 (Existence of Supporting Hyperplanes at Boundary Points). Alike the above Definition 4.1, Let 𝔼 be a real normed space and ℂ⊂𝔼 a nonempty closed convex set. For any point 𝑥0 on the boundary of ℂ, one can construct a nonzero continuous linear functional 𝑓∈𝔼∗ and scalar 𝛼∈ℝ such that 𝑓(𝑥) ≤ 𝛼 for all 𝑥∈ℂ, with equality holding at the boundary point: 𝑓(𝑥0)=𝛼. This establishes that every boundary point of a closed convex set admits at least one supporting hyperplane in the sense of Definition 4.1 [1]. Proof. Since 𝑥0 lies on the boundary 𝜕ℂ, it does not belong to the “Separation Theorems and Supporting Functionals in Normed Spaces: Structure, Duality, and Applications” 7642 ETJ Volume 10 Issue 10 October 2025, Santosh Kumar interior int(ℂ). Consider the singleton set 𝐴={𝑥0} and the open set 𝐵=int(ℂ). These are disjoint: 𝐴∩𝐵=∅. By Theorem 3.1 (basic separation with open set), there exists a nonzero continuous linear functional 𝑓∈𝔼∗ and a constant 𝛽∈ℝ satisfying 𝑓(𝑥0) ≤ 𝛽 ≤ 𝑓(𝑦) for all 𝑦∈int(ℂ). By continuity of 𝑓 and the fact that ℂ=int(ℂ) (closure of interior for convex sets with nonempty interior), the inequality extends to the entire set: 𝑓(𝑥) ≤ 𝛽 for all 𝑥∈ℂ. Moreover, since 𝑥0∈ℂ, we have 𝑓(𝑥0)≤𝛽. The separation condition 𝑓(𝑥0)≤𝛽 combined with 𝛽≤𝑓(𝑦) for 𝑦∈int(ℂ) implies that 𝑓(𝑥0)=𝛽 must hold (otherwise the singleton would be strictly separated from the closure, contradicting 𝑥0∈ℂ). Setting 𝛼:=𝑓(𝑥0) yields the desired supporting hyperplane: 𝑓(𝑥) ≤ 𝛼 for all 𝑥∈ℂ,𝑓(𝑥0)=𝛼. This confirms that the hyperplane 𝐻={𝑥∈ 𝔼:𝑓(𝑥)=𝛼} supports ℂ at 𝑥0 in the sense of Definition 4.1. Example 4.3 (Supporting Hyperplane for the Unit Ball). Let 𝔼 be an inner-product space with norm ‖⋅‖ induced by ⟨⋅,⋅⟩. Define the closed unit ball 𝑈={ 𝑥∈𝔼:‖𝑥‖≤1}. Select any boundary vector 𝑢∈𝔼 with ‖𝑢‖=1. Consider the continuous linear functional 𝑔:𝔼→ℝ,𝑔(𝑥)=⟨𝑥,𝑢⟩. By the Cauchy–Schwarz inequality, 𝑔(𝑥)≤1 for every 𝑥∈ 𝑈, while 𝑔(𝑢)=1. Therefore the hyperplane 𝐻={ 𝑥∈𝔼:𝑔(𝑥)=1} meets 𝑈 exactly at the point 𝑢 and does not intersect the interior of 𝑈. Hence 𝐻 serves as a supporting hyperplane to 𝑈 at 𝑢 in the sense of Definition 4.1. (B) Support Functions Support functions provide a powerful analytical tool for studying convex sets through their supporting hyperplanes, creating a duality between geometric objects and their functional representations[14, 3]. Definition 4.4 (Support Function of a Set). Having characterized supporting hyperplanes at individual boundary points (Definition 4.1), we now develop a dualspace functional that encodes the complete family of all such supporting structures simultaneously. Taking 𝔼 to be a real normed space and ℂ⊂𝔼 a nonempty subset. For each continuous linear functional 𝑓∈𝔼∗, the support function 𝜎ℂ assigns the value 𝜎ℂ(𝑓) = sup 𝑥∈ℂ 𝑓(𝑥), representing the maximum value that 𝑓 attains over ℂ. When ℂ is unbounded in the direction of 𝑓, we allow 𝜎ℂ(𝑓)=+∞ [1, 2]. Connection to Supporting Hyperplanes: With reference to Theorem 4.2 we can see that the support function 𝜎ℂ provides a complete dual-space representation of the supporting hyperplane structure introduced in Definition 4.1. Specifically, for any 𝑓∈𝔼∗∖{0} and any boundary point 𝑥0∈𝜕ℂ where 𝑓(𝑥0)=𝜎ℂ(𝑓), the hyperplane 𝐻𝑓 = { 𝑥∈𝔼:𝑓(𝑥)=𝜎ℂ(𝑓) } is a supporting hyperplane to ℂ at 𝑥0 in the sense of Definition 4.1. Conversely, every supporting hyperplane at 𝑥0 arises from some 𝑓∈𝔼∗ satisfying 𝑓(𝑥0)=𝜎ℂ(𝑓). Thus 𝜎ℂ captures all supporting hyperplanes through a single functional defined on the dual space [3-5]. Geometric Interpretation: The support function encodes complete information about the boundary structure of closed convex sets through dual space evaluations. For any 𝑓∈𝔼∗ with ‖𝑓‖=1, the value 𝜎ℂ(𝑓) gives the signed distance from the origin to the supporting hyperplane with outward normal 𝑓. This establishes a one-toone correspondence between closed convex sets and their support functions, providing a powerful analytical framework for studying convex geometry through functional representations [1, 5, 6]. Thus, Support functions capture all geometric information of closed convex regions and furnish a functional-analytic toolkit for investigating their structural properties[16, 14] Proposition 4.5 The support function ℎ𝐴 satisfies: 1. Convexity: ℎ𝐴 is convex on 𝔼 2. Positive homogeneity: ℎ𝐴(𝜆𝑓)=𝜆ℎ𝐴(𝑓) for λ≥0 3. Subadditivity: ℎ𝐴+𝐵(𝑓)=ℎ𝐴(𝑓)+ℎ𝐵(𝑓) These properties reveal the support function as a sublinear functional on the dual space, connecting geometric operations on sets with analytical operations on functions[14, 3]. Example 4.7 (Duality Between 𝓵𝒑−𝓵𝒒 Unit Balls). Consider the ℓ𝑝 sequence space with 1<𝑝<∞, and let 𝑞 denote its conjugate exponent satisfying 1 𝑝+1 𝑞=1. Define the closed unit ball 𝐵𝑝 = { 𝑥=(𝑥𝑛)∈ℓ𝑝:‖𝑥‖𝑝≤1 }. For any sequence 𝑓=(𝑓𝑛)∈ℓ𝑞, we compute the support function 𝜎𝐵𝑝 introduced in Definition 4.4: 𝜎𝐵𝑝(𝑓) = sup ‖𝑥‖𝑝≤1 ∑ ∞ 𝑛=1 𝑓𝑛𝑥𝑛. Geometric interpretation: This calculation demonstrates that the support function of the ℓ𝑝 unit ball, viewed as a functional on the dual space ℓ𝑞, coincides precisely with the ℓ𝑞 norm. This provides a concrete manifestation of the natural isometric duality between ℓ𝑝 andℓ𝑞, showing how geometric properties (support functions) encode analytic structures (dual norms) [2, 3, 1]. (C) Exposed and Extreme Points The relationship between supporting hyperplanes and the geometric structure of convex sets leads naturally to the study of exposed and extreme points [13, 2]. Definition 4.8 (Exposed Boundary Points). Let ℂ be a nonempty closed convex body in a real normed space 𝔼. A boundary point x∈ ℂ is called exposed if there “Separation Theorems and Supporting Functionals in Normed Spaces: Structure, Duality, and Applications” 7643 ETJ Volume 10 Issue 10 October 2025, Santosh Kumar exists a nonzero continuous linear functional 𝑓∈𝔼∗ whose maximum over ℂ is attained only at x. In other words, 𝑓(𝑥)=max𝑐∈𝐶 𝑓(𝑐) and this maximum point is unique. This condition is equivalent to the existence of a single supporting hyperplane that touches ℂ precisely at x [3, 9, 2]. Theorem 4.9(Extreme vs. Exposed Points): Every exposed point is an extreme point, but the converse may fail in infinite-dimensional spaces. Specifically, within any real normed space 𝔼, each point lying on the boundary of a closed convex body that admits a unique supporting functional must be an extreme element of that body. However, in infinite-dimensional settings, there may exist extreme elements that cannot be exposed in this manner. This distinction reflects the fact that, although Kreĭn–Milman-type principles ensure a rich supply of extreme elements for compact convex bodies, not all of these elements admit unique supporting hyperplanes [1, 13, 2]. Example 4.10 (Support Functional on ℂ Example): Let ℂ([1]) denote the Banach space consisting of all realvalued functions continuous on the closed unit interval [1], equipped with the supremum norm ‖𝑓‖∞=sup 𝑡∈[1] |𝑓(𝑡)|. Select an arbitrary point 𝑡0∈[1] and define the closed convex set 𝐵={𝑓∈ℂ([1]):−1≤𝑓(𝑡)≤1 for all 𝑡∈[1]} [1] The evaluation functional 𝛿𝑡0:ℂ([1])→ℝ given by 𝛿𝑡0(𝑓)= 𝑓(𝑡0) represents a continuous linear functional on this space. Computing the support function of 𝐵 at 𝛿𝑡0, we obtain [1] ℎ𝐵(𝛿𝑡0)=sup𝑓∈𝐵 𝛿𝑡0(𝑓)=sup𝑓∈𝐵 𝑓(𝑡0)=1, demonstrating that 𝛿𝑡0 acts as a supporting functional for 𝐵 at the constant function 𝑓≡1. This illustrates that while every extreme point corresponds to a point mass (Dirac measure in the dual space), not all such point masses are exposed points—a subtle distinction characteristic of infinitedimensional spaces [12]. (V) DUALITY CONNECTIONS (A) Dual Cones and Polar Sets The geometric theory of separation extends naturally to the study of dual cones and polar sets, providing a comprehensive framework for understanding duality relationships in convex analysis[17, 8]. Definition 5.1 (Polar of a Set): Let 𝔼 be a real normed space and let 𝐴⊆𝔼 be any subset. The polar of 𝐴 is the subset 𝐴∘⊆𝔼∗ given by 𝐴∘ = { 𝑓∈𝔼∗∶ ∀ 𝑥∈𝐴,𝑓(𝑥)≤1 }. This construction provides a fundamental duality link between 𝐴 and the continuous dual space 𝔼∗, encapsulating all linear functionals that are uniformly bounded by one on 𝐴 [Rockafellar 1970, Conway 1990, Bauschke et al 2011] [9, 5, 16]. Theorem 5.2 (Bipolar Representation): Consider a nonempty convex subset 𝐴 of a real normed space 𝔼 that contains the origin. Referring to definition 5.1 define the polar of 𝐴 by 𝐴∘={ 𝑓∈𝔼∗: 𝑓(𝑥)≤1 and the bipolar by 𝐴∘∘ ={ 𝑥∈𝔼: 𝑓(𝑥)≤1 Then 𝐴∘∘ equals the closure of the convex hull of 𝐴 in the norm topology; that is, 𝐴∘∘ =conv(𝐴). This identification establishes a dual equivalence between a set and its polar’s polar [Rockafellar 1970, Aliprantis–Border 2006] [9, 6]. This theorem reveals the profound connection between double polarity and convex hull operations, showing how geometric operations in finite dimensions extend to topological closures in infinite dimensions[17, 8]. Example 5.3 (Dual Correspondence for the Closed Unit Sphere). Within a real normed space 𝔼, consider the set of all vectors with norm at most one: 𝔹={ 𝑥∈𝔼:‖𝑥‖≤1 }. This canonical convex body—the norm ball of radius one— has polar given by 𝔹∘={ 𝑓∈𝔼∗:𝑓(𝑥)≤1 for all 𝑥∈𝔹}. A direct consequence of the geometric Hahn–Banach theorem[1, 2] establishes that 𝔹∘={ 𝑓∈𝔼∗:‖𝑓‖𝔼∗≤1}, demonstrating that the polar of the primal norm ball coincides precisely with the norm ball of radius one in the dual space 𝔼∗[Rockafellar 1970, Conway 1990] [9, 5]. This duality correspondence between 𝔹 and 𝔹∘ exemplifies the fundamental symmetry inherent in the polar operation when applied to normalized convex bodies. (B)Separation and Duality in Optimization The connection between separation theorems and optimization duality provides one of the fundamental applications of geometric functional investigations[5, 7]. Theorem 5.4 (Generalized Farkas Alternative in Normed Spaces). Consider two systems of linear inequalities, denoted 𝑆1 and 𝑆2, formulated within a real normed space framework. These systems satisfy a mutual exclusion property: exactly one of the following alternatives holds: 1. System 𝑆1 admits a feasible solution—that is, one can find a vector 𝑥 fulfilling all constraints in 𝑆1. 2. The infeasibility of 𝑆1 can be certified by a dual witness: specifically, a nonzero bounded linear functional 𝑓 exists with the property that: • 𝑓 is non-positive on the entire constraint cone generated by 𝑆1, meaning 𝑓(𝑥)≤0 holds for every 𝑥 in this conical hull, and • 𝑓 attains a strictly positive value on at least one vector 𝑦 corresponding to system 𝑆2, i.e., 𝑓(𝑦)>0. This embodies a fundamental duality principle: the absence of feasible primal solutions corresponds precisely to the existence of a separating functional that witnesses this infeasibility through its sign pattern on the constraint structures[Rockafellar 1970, Boyd & Vandenberghe 2004] [9, 11]. “Separation Theorems and Supporting Functionals in Normed Spaces: Structure, Duality, and Applications” 7644 ETJ Volume 10 Issue 10 October 2025, Santosh Kumar Proof Sketch (Geometric Interpretation): Assume system 𝑆1 admits no feasible solution. The geometric Hahn-Banach separation principle guarantees the existence of a nonzero bounded linear map 𝑓 that distinguishes the constraint cone generated by 𝑆1 from vectors associated with 𝑆2. Specifically, this functional 𝑓 satisfies: • 𝑓(𝑥)≤0 for every vector 𝑥 compatible with the constraints in 𝑆1, and • 𝑓(𝑦)>0 for at least one vector 𝑦 linked to system 𝑆2. This sign pattern establishes that 𝑓 acts as a dual certificate of infeasibility: it witnesses the impossibility of satisfying 𝑆1 by exhibiting strict positivity on elements that would contradict feasibility[8, 5]. Consequently, the absence of primal solutions corresponds precisely to the existence of such a separating functional in the dual space—embodying the fundamental alternative structure of the Farkas lemma [7]. Corollary 5.5 The strong duality theorem in linear programming follows directly from the Farkas lemma, showing that geometric separation underlies the fundamental duality relationships in optimization theory. (C)Weak-Star Separation In infinite-dimensional spaces, notions of convergence and compactness become subtler. The weak-star topology offers a minimal yet powerful framework, preserving continuity of evaluation maps from the dual space. This topology enables separation theorems to hold where stronger topologies may fail, providing crucial insights into dual space structures and the behavior of convex sets[13, 2]. Theorem 5.6 (Weak-Star Separation in Dual Spaces): Let 𝔼 be a real normed space, and let ℂ⊂𝔼∗ be a nonempty convex set that is compact in the weak-∗ topology. If a functional 𝑓0∈𝔼∗ lies outside ℂ, one can find a vector 𝑥∈𝔼 satisfying ℝ⟨𝑓0,𝑥⟩>max 𝑓∈ℂ ℝ⟨𝑓,𝑥⟩. In other words, evaluation at 𝑥 provides a continuous linear form on 𝔼∗ (with respect to the weak-∗ topology) that strictly separates 𝑓0 from the convex body ℂ [Aliprantis–Border 2006, Conway 1990] [6, 5] .Proof Sketch Assuming 𝑓0∉ℂ, compactness under the weak-star topology allows the use of a continuous functional—given by evaluation at 𝑥∈𝔼 —to distinguish 𝑓0 from ℂ. The inequality on real parts guarantees a strict hyperplane-type separation. This leverages the duality between 𝔼 and its dual, where geometric notions of separation correspond to functional evaluations, directly linking algebraic duality with topological compactness through the lens of evaluation maps. This result shows how separation in dual spaces naturally involves evaluation functionals from the primal space, creating a geometric interpretation of the biduality relationship[4, 2]. Example 5.7 In ℓ∞=(ℂ0)∗, weak-star compact convex sets can be separated from external points using finite linear combinations of coordinate functionals, providing concrete representations for abstract separation results. Table 3: Summary of Duality Concepts and their Corresponding Primal and Dual Objects in Convex Analysis. Duality Concept Primal Object Dual Object Connection Polar Sets Convex set A Polar 𝐴∘ Bipolar theorem Support Functions Convex set ℂ Function ℎℂ Boundary characterization Farkas Lemma Linear system Dual system Alternative theorem Weak-star separation Dual space point Primal evaluation Biduality (VI) APPLICATIONS (A) Linear Programming Applications The geometric foundation provided by separation theorems directly translates to practical algorithms and theoretical results in linear programming[7, 8]. Application 6.1 (Dual Infeasibility Certificates in LP). Examine the linear program min 𝑥∈ℝ𝑛 ℂ𝑇𝑥 such that 𝐴 𝑥=𝑏,𝑥⪰0, with 𝐴∈ℝ𝑚×𝑛, 𝑏∈ℝ𝑚, and cost vector ℂ∈ℝ𝑛. If no primal vector 𝑥 satisfies these constraints, then one can exhibit a dual vector 𝑦∈ℝ𝑚 such that 𝐴𝑇𝑦 ⪰ ℂ,𝑏𝑇𝑦 < 0,𝑦⪰0. This vector 𝑦 serves as an explicit certificate of primal infeasibility: the hyperplane defined by 𝑦 strictly separates the origin from the infeasible region of the primal program, reflecting the alternative between primal feasibility and dual feasibility in linear programming[Boyd–Vandenberghe 2004] [11]. This interplay between primal constraints and dual certificates forms the mathematical foundation for many algorithms and termination criteria in linear programming. This geometric interpretation reveals that duality gaps correspond to separation distances, and strong duality equivalent to the absence of strict separation between primal and dual feasible regions[8]. Example 6.2 Consider the following system of linear inequalities that has no solution: {2𝑥1+𝑥2≤0, 3𝑥1+4𝑥2≤0, 𝑥1,𝑥2≥0. Given that no vector (𝑥1,𝑥2) can satisfy all these simultaneously, we use separation theory to construct a certificate of infeasibility—a vector 𝑦=(1,1,2) satisfying 𝑦≥0, 𝑀𝑇𝑦≥0, 𝑦𝑇𝑏<0, where matrix 𝑀 and 𝑏 come from the system's coefficients and right-hand sides[11]. “Separation Theorems and Supporting Functionals in Normed Spaces: Structure, Duality, and Applications” 7645 ETJ Volume 10 Issue 10 October 2025, Santosh Kumar (B) Convex Optimization Framework Modern convex optimization relies heavily on separationbased algorithms, where supporting hyperplanes guide iterative procedures toward optimal solutions[9, 7]. Application 6.3 (Ellipsoid and Cutting Plane Approaches): Modern convex optimization frequently employs iterative schemes that refine the search space to efficiently find optimal solutions. A prominent technique involves starting with an initial ellipsoidal region believed to contain the optimizer. At each iteration, the current center is tested for feasibility; if it fails, a hyperplane that separates infeasible regions from feasible ones is identified. This separating hyperplane, known as a cutting plane, is then used to truncate the search space, resulting in a smaller ellipsoid that still encloses the feasible region. The process repeats, with successively shrunk ellipsoids converging to the optimal solution. By iteratively eliminating infeasible portions using these cutting planes, the method elegantly balances precision and computational efficiency, with theoretical convergence guaranteed by the controlled decrease in ellipsoid volume. This approach underpins a wide range of algorithms for complex convex problems, providing a geometric perspective for optimization. The theoretical convergence of these methods depends directly on separation theorems, which guarantee the existence of appropriate cutting hyperplanes[7]. Example 6.4 (Support Vector Machines): According to Cortes, C., & Vapnik, V. (1995). Supportvector networks. Machine Learning, 20(3), 273–297, retrieved from https://doi.org/10.1007/BF00994018 “The support vector machine seeks a hyperplane that optimally separates two classes by maximizing the distance (margin) between the closest points of each class. This translates into solving the quadratic optimization problem: min 𝑤,𝑏,𝜉1 2‖𝑤‖2+ℂ∑𝜉𝑖 𝑖 subject to 𝑦𝑖(𝑤𝑇𝑥𝑖+𝑏)≥1−𝜉𝑖, 𝜉𝑖≥0 where 𝑤 and 𝑏 define the separating hyperplane, 𝜉𝑖 are slack variables allowing some misclassifications, and 𝐶 balances the trade-off between margin maximization and classification error.” (C)Control and Approximation Theory Separation theorems find important applications in control theory and approximation problems, where geometric characterizations lead to computational algorithms[2, 13]. Application 6.5 (Metric Projection in Hilbert Spaces). In a real Hilbert space 𝐻, fix any nonempty closed convex subset ℂ. For an arbitrary vector 𝑥∈𝐻, define its projection onto ℂ, denoted 𝑝=Πℂ(𝑥), as the unique minimizer of the distance to 𝑥: 𝑝=arg min 𝑦∈ℂ ‖𝑥−𝑦‖. Equivalently, 𝑝 is characterized by the inequality ⟨𝑥− 𝑝,𝑦− 𝑝⟩≤0 ∀ 𝑦∈ℂ, which states that the error vector 𝑥−𝑝 is orthogonal to the tangent cone at 𝑝. The existence, uniqueness, and variational condition of Πℂ follow from Hilbert space structure and the parallelogram identity combined with a separation argument in the dual space [1, 2]. Example 6.6 In approximation theory, the Remez exchange algorithm for polynomial approximation uses supporting hyperplane principles to identify optimal approximating polynomials by characterizing extremal points through separation properties. Table 4: Comparative Analysis of Separation Theorem Applications across Mathematical Domains Application Domain Separation Use Practical Impact Linear Programming Feasibility certificates Algorithm termination conditions Machine Learning Margin maximization Classification accuracy Control Theory Reachability analysis System verification Approximation Optimality conditions Numerical algorithms (VII) ADVANCED TOPICS AND EXTENSIONS (A) Infinite-Dimensional Considerations The extension of separation theorems to infinite-dimensional spaces reveals subtle phenomena that have no finitedimensional analogues[4,2]. Remark 7.1 In non-locally convex spaces, separation theorems may fail completely. The existence of separating hyperplanes depends critically on the availability of sufficient continuous linear functionals, which may not exist in spaces without local convexity[4]. Example 7.2 (Separation Failure in QuasiBanach Spaces). Let 0<𝑝<1 and consider the quasi-Banach space 𝐿𝑝(Ω) equipped with the quasi-norm ‖𝑓‖𝑝=(∫|𝑓|𝑝)1/𝑝. Since 𝐿𝑝 fails to be locally convex in this regime, one can construct two nonempty, closed convex subsets 𝐴 and 𝐵 in 𝐿𝑝(Ω) such that no continuous linear functional on 𝐿𝑝 yields a strict separating hyperplane between them. This phenomenon underscores that the existence of linear separators hinges critically on local convexity[Conway 1990, Rudin 1991] [5, 3]. The role of weak compactness becomes crucial in infinite dimensions, where the Eberlein-Šmulian theorem provides