--- title: Method of Interior Parallels url: https://www.emergentmind.com/topics/method-of-interior-parallels type: topic --- # Method of Interior Parallels Searching arXiv for the cited papers to ground the article in current records. In the cited literature, the expression *method of interior parallels* is used for several geometric procedures that share a common structural motif: a global property is accessed through an interior or inward construction. In Euclidean geometry, the phrase refers to Euclid’s construction of the line through a point \(P\notin \ell\) by copying an angle so that alternate interior angles are equal, together with the constructive justification supplied by Proposition I.29 and the Fifth Postulate [2208.10835]. In convex geometry, it refers to the deformation \(K_t=K\sim tB\) or \(K\sim tC\) by inner parallel bodies and to the resulting bounds and monotonicity formulas for perimeter, anisotropic perimeter, quermassintegrals, and isoperimetric quotients [1508.06414]. A separate planar construction uses interior angle bisection and compositions of rotations to extract a parallelogram from any quadrilateral, and a square from any parallelogram [1203.4167]. This suggests a unifying theme: an interior finite operation is used to encode a more global geometric relation. ## 1. Meanings and formal settings The term is not univocal. In the material considered here, it appears in three distinct settings: classical line geometry, convex-geometric inner offsets, and a polygonal rotation construction. | Setting | Basic object | Principal operation | |---|---|---| | Euclidean geometry | A line through \(P\notin \ell\) parallel to \(\ell\) | Copy an angle so alternate interior angles are equal | | Convex geometry | Inner parallel body \(K_t\) or \(\Omega_t\) | Minkowski difference \(K\sim tB\) or \(K\sim tC\) | | Planar quadrilaterals | A derived quadrilateral \(P^e\) | Interior angle bisection and compositions of rotations | In Euclid’s case, the operative relation is the alternate-interior-angle criterion. In convex geometry, the central definition is inward erosion by a ball or by a fixed convex body. In the quadrilateral construction, the interior data are the half-angles at the vertices and the fixed points of certain rotation compositions. The commonality is therefore not terminological accident alone; it lies in the passage from interior auxiliary structure to a statement about parallelism, inward motion, or a derived polygonal form. ## 2. Euclid’s construction through alternate interior angles The Euclidean problem is: given a straight line \(\ell\) in the plane and a point \(P\) not on \(\ell\), construct by ruler-and-compass a new line \(m\) through \(P\) such that \(m\parallel \ell\). Proposition I.31 proceeds in three moves: choose an arbitrary point \(A\) on \(\ell\), draw the transversal \(PA\), and then construct at \(P\) a second line \(PC\) so that the alternate interior angles with respect to \(\ell\) are equal [2208.10835]. In the detailed reconstruction, one lets \(a=PC\), \(b=\ell\), and \(c=PA\). The relevant finite property is \(P_{b,c}(a)\), “\(a\) meets \(b\) at \(c\) making alternate interior angles equal,” and the target infinite property is \(Q_b(a)\), “\(a\parallel b\).” With the notation \[ \alpha:=\angle APC,\qquad \beta:=\angle PAC, \] Euclid I.27 gives the implication \[ \alpha=\beta \;\Longrightarrow\; a\parallel b. \] The angle-copy step is Euclid I.23. At \(P\), one constructs a ray \(PC\) such that \[ \angle APC=\angle BAC. \] The exposition specifies the standard compass construction: draw a circle centered at \(P\) meeting \(PA\) at \(D\); draw a circle centered at \(A\) of the same radius meeting \(AB\) at \(E\); draw arcs centered at \(D\) and \(E\) with radius \(DE\); let them meet at \(C\); join \(P\) to \(C\). By construction, the angle at \(P\) equals the angle on \(\ell\), so the alternate interior angles are equal, and I.27 yields \(PC\parallel \ell\). At the level of geometric practice, the method is entirely finite: one chooses one point on \(\ell\), one transversal, and one copied angle. The deeper issue is not how to draw \(PC\), but how this finite act can count as a construction of the infinite relation expressed in Euclid’s Definition 23, namely that parallel lines, “produced indefinitely, do not meet.” ## 3. Constructive legitimacy and the Fifth Postulate Petrakis’s reconstruction assigns a genuinely constructive role to the Fifth Postulate. The key point is that Euclid I.31 by itself produces the finite property \(P_{b,c}(a)\), not yet the full infinite property \(Q_b(a)\). The converse implication, \[ a\parallel b \;\Longrightarrow\; \alpha=\beta, \] is Proposition I.29, and Euclid proves it by invoking the Fifth Postulate in its original form: if a straight line falling on two straight lines makes the interior angles on one side less than two right angles, then the two lines, if produced indefinitely, meet on that side [2208.10835]. Only after both directions are available, \[ P_{b,c}(a)\Rightarrow Q_b(a),\qquad Q_b(a)\Rightarrow P_{b,c}(a), \] may one infer the equivalence \[ P_{b,c}(a)\Leftrightarrow Q_b(a). \] Petrakis formulates this through four constructive principles. \(K1\) requires constructing “as accurately as possible,” so the target is the definition of parallel itself. \(K2\) requires that there be no gaps: each auxiliary step must appeal only to earlier postulates or already established propositions. \(K3\) distinguishes the generality hierarchy, according to which “parallel” is an infinite property and “alternate interior angles equal” is a finite property, hence strictly weaker until equivalence is proved. \(K4\) permits equivalent-property interchange: once \(P_{b,c}(a)\Leftrightarrow Q_b(a)\) is established, a construction of the finite property can count as a construction of the infinite one. On this reading, the Fifth Postulate is not an external consistency condition appended to an otherwise complete construction. It is the ingredient that legitimates the passage from a copied angle to a genuinely constructed parallel line. The construction is therefore finite in execution but epistemologically dependent on the equivalence supplied by I.29. The same paper contrasts this with Bolyai’s construction of limiting parallels in a hyperbolic plane with Lobachevsky’s Hyperbolic Axiom \(L\). Bolyai erects perpendiculars, constructs a right-angled transversal, chooses a point \(R\) on \(\ell\), erects from \(R\) a perpendicular to the transversal, and then uses a circle centered at \(P\) through \(Q\) to obtain a point \(X\), declaring \(PX\) to be the limiting parallel. By an elementary continuity principle, the circle meets the relevant line; but the construction presupposes the existence of a limiting parallel ray via \(L\). In Petrakis’s analysis, this violates \(K2\) and does not supply a finite equivalence analogous to I.27 and I.29. ## 4. Inner parallel bodies and sharp perimeter bounds In convex geometry, the basic object is the inner parallel body, also called the interior parallel set. For a convex body \(\Omega\subset\mathbb R^n\) with nonempty interior and for \(t\ge 0\), \[ \Omega_t:=\{x\in\Omega:\dist(x,\Omega^c)\ge t\}=\Omega\sim tB, \] where \(B\) is the unit ball and \(\sim\) denotes Minkowski difference, \[ A\sim C=\{x:x+C\subseteq A\}. \] The inradius is \[ r=\sup\{\lambda\ge 0:\Omega\sim \lambda B\neq\varnothing\}. \] Larson’s main theorem gives a sharp lower bound for the perimeter of inner parallel sets: \[ |\partial\Omega_t|\ge \Bigl(1-\frac tr\Bigr)^{n-1}_+\,|\partial\Omega|. \] For \(t\in[0,r]\), this is \[ |\partial\Omega_t|\ge \Bigl(1-\frac tr\Bigr)^{n-1}|\partial\Omega|, \] and the bound depends only on the perimeter and inradius of the original body; it is independent of regularity properties of \(\partial\Omega\) [1508.06414]. A principal structural lemma is the inradius drop: \[ r_t=r-t \] for the inradius \(r_t\) of \(\Omega_t\). The proof strategy reduces the estimate to an extremal reverse problem. For fixed \(\Omega\) and \(t\), one considers the family \[ \mathcal F_t(\Omega)=\{\widehat\Omega\subset\mathbb R^n\text{ convex}:\widehat\Omega_t=\Omega\}, \] and shows that the unique maximizer of perimeter in this family is \[ \widetilde\Omega=\bigcap_{u\in\mathcal U(\Omega)}\{x:\langle x,u\rangle\le h(\Omega,u)+t\}, \] where \(h(\Omega,u)\) is the support function and \(\mathcal U(\Omega)\) is the set of outward unit normals at the regular boundary points of \(\Omega\). Mixed volumes enter through the characterization \[ |\partial K|=n\,W(B,K,\dots,K), \] together with multilinearity, symmetry, translation invariance, and monotonicity under inclusion. The proof compares the extremal body with a dilation of \(\Omega\) by using the support-function inequality \[ h(\widetilde\Omega,u)=h(\Omega,u)+t\le \Bigl(1+\frac tr\Bigr)h(\Omega,u). \] A notable feature is that no direct analog of Steiner’s formula for inner sets is used, because none holds in general. Equality occurs for some \(t\in(0,r)\) if and only if \(\Omega\) is homothetic to its form-body \[ \Omega_*=\bigcap_{u\in\mathcal U(\Omega)}\{x:\langle x,u\rangle\le 1\}. \] In that case equality holds for all \(t\ge 0\). The paper also notes that, combined with Steinhagen’s inequality, the result yields width-based perimeter-drop bounds, and in the planar case this resolves a conjecture of Geisinger–Laptev–Weidl. ## 5. Anisotropic interior parallels and eikonal abrasion Crasta extends the isotropic theory to an anisotropic setting. Let \(K\subset\mathbb R^n\) be a convex body and let \(\|\cdot\|\) be a norm with unit ball \(B\). Then \[ K_t=\{x\in K:\dist(x,\partial K)\ge t\}=K\smallsetminus tB. \] More generally, for a convex body \(C\) containing \(0\) in its interior, \[ K_t^C=\{x:x+tC\subset K\}=K\smallsetminus tC. \] The anisotropy is encoded by a strictly positive, even, convex, one-homogeneous gauge \(\phi\), equivalently the support function of a convex body \(K_0\): \[ \phi(u)=h_{K_0}(u)=\sup\{x\cdot u:x\in K_0\}. \] For a convex body \(\Omega\) with outer unit normal \(\nu_\Omega(x)\), the \(\phi\)-anisotropic perimeter is \[ P_\phi(\Omega)=\int_{\partial\Omega}\phi(\nu_\Omega(x))\,d\mathcal H^{n-1}(x), \] and the associated anisotropic isoperimetric quotient is \[ Q_\phi(\Omega)=\frac{P_\phi(\Omega)^n}{|\Omega|^{\,n-1}}. \] The principal inequality is the anisotropic inner-parallel perimeter bound \[ P_\phi(K_t)\ge (1-t/r)^{\,n-1}P_\phi(K) \] for every \(t\in[0,r]\), where \(r\) is the inradius of \(K\) [2101.03307]. The proof uses mixed volumes and Brunn–Minkowski concavity. With \[ v_i(t):=V^{(i)}(K_t,K_0),\qquad i=0,1,\dots,n, \] one has \(v_0(t)=|K_t|\), \(v_1(t)=(1/n)P_\phi(K_t)\), and \(v_n(t)=|K_0|\). For each fixed \(i\), the function \[ f_i(t):=[v_i(t)]^{1/(n-i)} \] is concave in \(t\). In particular, \[ f_1(t)=\Bigl[\frac{P_\phi(K_t)}{n}\Bigr]^{1/(n-1)} \] is concave on \([0,r]\), vanishes at \(t=r\), and therefore satisfies \[ f_1(t)\ge (1-t/r)f_1(0), \] which yields the perimeter bound after raising to the \((n-1)\)-st power. The same framework gives monotonicity of the anisotropic isoperimetric quotient along inward motion. Writing \(I(t)=Q_\phi(K_t)\), the right derivative is computed as \[ I'(t)^+ = -\,n\,\frac{P_\phi(K_t)^2-(n-1)\,|K_t|\,P_\phi''(K_t)}{|K_t|^n}\le 0. \] An equivalent level-set viewpoint uses the dual gauge \(\phi^*\) and the anisotropic distance \[ d_\phi(x):=\inf\{\phi^*(y-x):y\in\partial K\}, \] so that \(K_t=\{x:d_\phi(x)\ge t\}\) evolves formally by the Hamilton–Jacobi equation \[ \partial_t x(t)+\phi(\nu(x(t)))=0. \] The paper also lists applications to quantitative bounds on eigenvalues of Dirichlet–Laplace operators on convex domains, geometric inequalities in convex analysis, and modeling of abrasion and shape evolution in materials science. ## 6. Quermassintegrals, form bodies, and corrected monotonicity results A further development studies inner parallel bodies relative to a fixed convex body \(E\subset\mathbb R^n\) with interior. The relative inradius is \[ r(K;E)=\sup\{r\ge 0:\exists x\in\mathbb R^n\text{ with }x+rE\subset K\}, \] and for \(-r(K;E)\le \rho\le 0\), \[ K_\rho:=K\sim |\rho|E=\{x\in\mathbb R^n:x+|\rho|E\subseteq K\}. \] Equivalently, \[ K_{\rho} =\Bigl\{x\in\mathbb R^n:\langle x,u\rangle\le h(K,u)-|\rho|\,h(E,u),\quad u\in S^{n-1}\Bigr\}. \] For outer parallels, the relative Steiner polynomial is \[ \Vol(K+\lambda E)=\sum_{k=0}^n\binom nk\,W_k(K;E)\,\lambda^k, \] where \(W_k(K;E)\) are the relative quermassintegrals. Re-indexing yields for the inner parallels \[ W_k(\rho)=W_k(K_\rho;E)=\sum_{j=0}^{n-k}\binom{n-k}{j}(-\rho)^j\,W_{k+j}(K;E). \] Thus each \(W_k(\rho)\) is a polynomial in \(\rho\) of degree \(n-k\) [1910.05367]. The corresponding isoperimetric quotient and deficit are \[ Q_{i,j}(\rho):=\frac{W_j(\rho)^{\,n-i}}{W_i(\rho)^{\,n-j}\,\Vol(E)^{\,j-i}}, \qquad \Delta_{i,j}(\rho):=W_j(\rho)^{\,n-i}-W_i(\rho)^{\,n-j}\,\Vol(E)^{\,j-i}, \] for \(0\le i