We now state the main theorem of this section:
Given any smooth positive function \(\alpha(s) : I \to (0, \infty)\), and a smooth real-valued function \(\beta(s) : I \to \R\), there exists an arc-length parametrized, non-degenerate curve \(\gamma(s) : I \to \R^3\) such that its curvature \(\kappa(s) \equiv \alpha(s)\) and its torsion \(\tau(s) \equiv \beta(s)\).
Moreover, if \(\bar{\gamma}(s) : I \to \R^3\) is another arc-length parametrized, non-degenerate curve whose curvature \(\bar{\kappa}(s) \equiv \alpha(s)\) and torsion \(\bar{\tau}(s) \equiv \beta(s)\), then there exists a \(3\times 3\) constant matrix \(A\) with \(A^T A = I\), and a constant vector \(p\), such that \(\bar{\gamma}(s) = A\gamma(s) + p\) for any \(s \in I\).
Step 1: Use the existence theorem of ODEs (Theorem 1.22) to show there exists a moving orthonormal frame \(\{\hat{e}_1(s), \hat{e}_2(s), \hat{e}_3(s)\}\) which satisfies an ODE system analogous to the Frenet-Serret System (1.6): \[(1.7)\qquad \begin{bmatrix}\hat{e}_1(s)\\\hat{e}_2(s)\\\hat{e}_3(s)\end{bmatrix}' = \begin{bmatrix}0&\alpha(s)&0\\-\alpha(s)&0&\beta(s)\\0&-\beta(s)&0\end{bmatrix}\begin{bmatrix}\hat{e}_1(s)\\\hat{e}_2(s)\\\hat{e}_3(s)\end{bmatrix}\] with initial conditions \(\hat{e}_1(0) = \hat{i}\), \(\hat{e}_2(0) = \hat{j}\), \(\hat{e}_3(0) = \hat{k}\). Since \(\alpha(s)\) and \(\beta(s)\) are smooth (in particular \(C^1\)), by Theorem 1.22 a solution exists on a maximal interval \(s \in (T_-, T_+)\).
We claim the frame \(\{\hat{e}_1(s), \hat{e}_2(s), \hat{e}_3(s)\}\) remains orthonormal as long as the solution exists. From (1.7) one derives (see Exercise 1.17): \[(1.8)\qquad \frac{d}{ds}\begin{bmatrix}\hat{e}_1\cdot\hat{e}_1\\ \hat{e}_2\cdot\hat{e}_2\\ \hat{e}_3\cdot\hat{e}_3\\ \hat{e}_1\cdot\hat{e}_2\\ \hat{e}_2\cdot\hat{e}_3\\ \hat{e}_3\cdot\hat{e}_1\end{bmatrix} = M(\alpha,\beta)\begin{bmatrix}\hat{e}_1\cdot\hat{e}_1\\ \hat{e}_2\cdot\hat{e}_2\\ \hat{e}_3\cdot\hat{e}_3\\ \hat{e}_1\cdot\hat{e}_2\\ \hat{e}_2\cdot\hat{e}_3\\ \hat{e}_3\cdot\hat{e}_1\end{bmatrix}\] for a certain matrix \(M(\alpha,\beta)\). The constant vector \((1,1,1,0,0,0)\) satisfies (1.8) with the given initial data. By the uniqueness theorem (Theorem 1.22), the dot products \((\hat{e}_i \cdot \hat{e}_j)\) remain equal to \((1,1,1,0,0,0)\) for all \(s \in (T_-, T_+)\), i.e. the frame stays orthonormal. Since the frame remains bounded, it extends to all \(s \in I\).
Step 2: Using \(\hat{e}_1(s)\) from Step 1, define: \[\gamma(s) = \int_0^s \hat{e}_1(s)\, ds.\] Then \(\mathbf{T}(s) := \gamma'(s) = \hat{e}_1(s)\) is unit, so \(\gamma(s)\) is arc-length parametrized. Computing the curvature: \(\gamma''(s) = \hat{e}_1'(s) = \alpha(s)\hat{e}_2(s)\), so \(\kappa(s) = |\gamma''(s)| = \alpha(s)\) and \(\mathbf{N}(s) = \hat{e}_2(s)\). By continuity of the orthonormal frame, \(\mathbf{B}(s) = \mathbf{T}(s)\times\mathbf{N}(s) = \hat{e}_3(s)\), and \(\mathbf{B}'(s) = \hat{e}_3'(s) = -\beta(s)\hat{e}_2(s) = -\beta(s)\mathbf{N}\), so \(\tau(s) = \beta(s)\).
Step 3 (Uniqueness): Suppose \(\bar{\gamma}(s)\) has the same curvature \(\alpha(s)\) and torsion \(\beta(s)\). Define the matrix \(A = [\bar{\mathbf{T}}(0)\; \bar{\mathbf{N}}(0)\; \bar{\mathbf{B}}(0)]\). By orthonormality, \(A^T A = I\). Both the Frenet-Serret Frame of \(A\gamma(s)+\bar{\gamma}(0)\) (namely \(\{A\mathbf{T}(s), A\mathbf{N}(s), A\mathbf{B}(s)\}\)) and of \(\bar{\gamma}(s)\) satisfy the same ODE (1.7), and they agree at \(s=0\). By Theorem 1.22, \(A\mathbf{T}(s)\equiv\bar{\mathbf{T}}(s)\). Setting \(f(s) := |\bar{\gamma}(s) - (A\gamma(s) - \bar{\gamma}(0))|^2\), we compute \(f'(s) = 2(\bar{\mathbf{T}}(s)-A\mathbf{T}(s))\cdot(\cdots) = 0\) and \(f(0)=0\), so \(f\equiv 0\) and \(\bar{\gamma}(s) \equiv A\gamma(s) + \bar{\gamma}(0)\).
Verify (1.8).
The existence part of Theorem 1.23 only shows a curve with prescribed curvature and torsion exists, but it is in general difficult to find such a curve explicitly. While the existence part does not have much practical use, the uniqueness part has some nice corollaries.
First recall that a helix is a curve of the form \(\gamma_{a,b}(t) = (a\cos t, a\sin t, bt)\) where \(a \neq 0\) and \(b\) can be any real number. Its arc-length parametrization is given by: \[\gamma_{a,b}(s) = \left(a\cos\frac{s}{\sqrt{a^2+b^2}},\; a\sin\frac{s}{\sqrt{a^2+b^2}},\; \frac{bs}{\sqrt{a^2+b^2}}\right).\] It can be computed that its curvature and torsion are both constants: \[\kappa_{a,b}(s) \equiv \frac{a}{a^2+b^2}, \qquad \tau_{a,b}(s) \equiv \frac{b}{a^2+b^2}.\] Conversely, given two constants \(\kappa_0 \gt 0\) and \(\tau_0 \in \R\), by taking \(a = \frac{\kappa_0}{\kappa_0^2+\tau_0^2}\) and \(b = \frac{\tau_0}{\kappa_0^2+\tau_0^2}\), the helix \(\gamma_{a,b}(s)\) with this pair of \(a\) and \(b\) has curvature \(\kappa_0\) and torsion \(\tau_0\). Hence, the uniqueness part of Theorem 1.23 asserts that:
A non-degenerate curve \(\gamma(s)\) has constant curvature and torsion if and only if \(\gamma(s)\) is congruent to one of the helices \(\gamma_{a,b}(s)\).
Two space curves \(\gamma(s)\) and \(\tilde{\gamma}(s)\) are said to be congruent if there exists a \(3\times 3\) orthogonal matrix \(A\) and a constant vector \(p \in \R^3\) such that \(\tilde{\gamma}(s) = A\gamma(s) + p\). In simpler terms, one can obtain \(\tilde{\gamma}(s)\) by rotating and translating \(\gamma(s)\).