The Cyclic Subspace
Definition 1. Let T∈L(V)T∈L(V), and let a nonzero vector x∈Vx∈V. The subspace W=⟨x,T(x),T2(x),...⟩W=⟨x,T(x),T2(x),...⟩ is called the TT-cyclic subspace of VV generated by xx.
Theorem 1
Theorem 1. Let T∈L(V)T∈L(V), and let WW be the TT-cyclic subspace of VV generated by 0≠x∈V0≠x∈V. Then
(a) WW is TT-invariant.
(b) Any TT-invariant subspace of VV containing xx also contain WW.
Proof. (a) ∀u∈T(W),∃y∈W∀u∈T(W),∃y∈W such that u=T(y)u=T(y). Then y=a0x+a1T(x)+⋯y=a0x+a1T(x)+⋯ for some a0,a1,...a0,a1,.... Thus u=T(y)=a0T(x)+a1T2(x)+⋯∈Wu=T(y)=a0T(x)+a1T2(x)+⋯∈W. Thus T(W)⊆WT(W)⊆W.
(b) Let KK be a TT-invariant subspace of VV. ∀v∈W,v=a0x+a1T(x)+⋯∀v∈W,v=a0x+a1T(x)+⋯ for some a0,a1,...a0,a1,.... Note that Tn(x)∈K,∀n∈NTn(x)∈K,∀n∈N, because x∈Kx∈K. Then v∈Kv∈K. Thus W⊆KW⊆K. ◼■
다시 말해 TT-순환 부분공간은 xx를 포함하는 VV의 가장 작은 TT-불변 부분공간이다.
Theorem 2
Theorem 2. Let T∈L(V)T∈L(V), and let WW denote the TT-cyclic subspace of VV generated by 0≠v∈V0≠v∈V. Let k=dim(W)k=dim(W). Then
(a) {v,T(v),...,Tk−1(v)}{v,T(v),...,Tk−1(v)} is a basis for WW.
(b) If a0v+a1T(v)+⋯+ak−1Tk−1(v)+Tk(v)=0a0v+a1T(v)+⋯+ak−1Tk−1(v)+Tk(v)=0, then the characteristic polynomial of TWTW is f(t)=(−1)k(a0+a1t+⋯+ak−1tk−1+tk)f(t)=(−1)k(a0+a1t+⋯+ak−1tk−1+tk).
Proof. (a) Since v≠0,{v}v≠0,{v} is linearly independent. Let jj be the largest positive integer such that β={v,T(v),...,Tj−1(v)}β={v,T(v),...,Tj−1(v)} is linaerly independent. Let Z=⟨β⟩Z=⟨β⟩. Then ββ is a basis for ZZ.
Let u∈Zu∈Z. Then !∃b0,...,bj−1∈F!∃b0,...,bj−1∈F such that u=b0v+⋯+bj−1Tj−1(v)u=b0v+⋯+bj−1Tj−1(v). Since Tj(v)∈ZTj(v)∈Z, T(u)=b0T(v)+⋯+bj−1Tj(v)T(u)=b0T(v)+⋯+bj−1Tj(v). Thus T(u)∈ZT(u)∈Z, so ZZ is TT-invariant.
Since ZZ is a subspace of VV containing vv, by Theorem 1 W⊆ZW⊆Z. Clearly Z⊆WZ⊆W. Thus W=ZW=Z, and j=kj=k.
(b) Suppose that a0v+a1T(v)+⋯+ak−1Tk−1(v)+Tk(v)=0a0v+a1T(v)+⋯+ak−1Tk−1(v)+Tk(v)=0. Then Tk(v)=(−a0)v+(−a1)T(v)+⋯+(−ak−1)Tk−1(v)Tk(v)=(−a0)v+(−a1)T(v)+⋯+(−ak−1)Tk−1(v).
Note that [TW]β=(00⋯0−a010⋯0−a1⋮⋮⋱⋮⋮00⋯0−ak−200⋯1−ak−1), where β={v,T(v),...,Tk−1(v)}. Then the characteristic polynomial of TW is f(t)=det([TW]β−tIk)=det(−t0⋯0−a01−t⋯0−a1⋮⋮⋱⋮⋮00⋯−t−ak−200⋯1−ak−1−t).
We claim that f(t)=(−1)k(a0+a1t+⋯+ak−1tk−1+tk). We use the induction on k. If k=2, then f(t)=det(−t−a01−a1−t) = −t(a1−t)+a0 = (−1)2(a0+a1t+t2).
Assume the statement is true for k−1≥2. Then f(t)=k∑j=1(−1)1+jA1jdet(~A1j)=(−t)det(−t⋯0−a1⋮⋱⋮⋮0⋯1−ak−1−t)+(−1)1+k(−a0)det(1−t⋯0⋮⋮⋱⋮00⋯1)=(−t)(−1)k−1(a1+⋯+ak−1tk−2+tk−1)+(−1)1+k(−a0)=(−1)k(a1t+⋯+ak−1tk−1+tk)+(−1)ka0=(−1)k(a0+a1t+⋯+ak−1tk−1+tk). Thus f(t)=(−1)k(a0+a1t+⋯+ak−1tk−1+tk). ◼