Example. Real Fourier sectors of a cycle [fdip-004W]
Example. Real Fourier sectors of a cycle [fdip-004W]
Let \(n\ge 3\), \(V=E=\mathbb Z/n\mathbb Z\), with edge \(e_v\) from \(v\) to \(v+1\), uniform weight \(w>0\) and mass \(m>0\). The generator \(r\) of \(C_n\) sends \(v\) to \(v+1\) and \(e_v\) to \(e_{v+1}\), with every edge sign positive. Its action on fields is \((\rho (r)f)(v)=f(v-1)\). The cyclic characters in [serre1977linear, Section 5.1, p. 35] and the real projectors yield the following direct calculation.
Set \(\theta _k=2\pi k/n\). The normalized constant vector \(u_0(v)=1/\sqrt n\) spans the trivial sector. For \(1\le k<n/2\), the sector \(H_k\) has orthonormal basis \[c_k(v)=\sqrt {2/n}\cos (\theta _kv),\qquad s_k(v)=\sqrt {2/n}\sin (\theta _kv).\] When \(n\) is even there is also \(u_{n/2}(v)=(-1)^v/\sqrt n\). These give all sectors indexed by \(0\le k\le \lfloor n/2\rfloor \). In the ordered basis \((c_k,s_k)\), the rotation is \[\rho (r)|_{H_k}=\begin {pmatrix} \cos \theta _k&-\sin \theta _k\\ \sin \theta _k&\cos \theta _k \end {pmatrix}.\] For \(0<\theta _k<\pi \) it has no real eigenline, so \(H_k\) is real irreducible. Its complex constituents have characters \(r\mapsto e^{i\theta _k}\) and \(r\mapsto e^{-i\theta _k}\).
The precision and each sector eigenvalue are \[\begin {aligned} (Qf)(v)&=(m^2+2w)f(v)-wf(v-1)-wf(v+1),\\ Q|_{H_k}&=q_kI,\\ q_k&=m^2+2w(1-\cos \theta _k). \end {aligned}\] Substitution of each sine, cosine and exceptional one-dimensional mode proves the formula. The orthogonal projector kernels are \[\begin {aligned} P_0(v,u)&=1/n,\\ P_k(v,u)&=\frac 2n\cos (\theta _k(v-u))\quad (0<k<n/2),\\ P_{n/2}(v,u)&=(-1)^{v-u}/n\quad (n\text { even}). \end {aligned}\] They follow by summing the outer products of the displayed orthonormal basis vectors, so their sum is the identity. For \(n=3\), this reads \(P_0=\mathbf 1\mathbf 1^T/3\) and \(P_1=I-P_0\); the eigenvalues are \(m^2\) and \(m^2+3w\), the latter twice. The later circle comparison rescales these same finite modes; no continuum limit is used in this calculation.