17.6. OPERATORS WITH CLOSED RANGE 469

Proof: Consider λ ̸= 0. The N(Rk

λ

)are increasing in k and Rλ

(N(Rk+1

λ

))⊆ N

(Rk

λ

).

This follows from the definition. (It isn’t necessary to assume in most of this that λ ∈ Λ,just a nonzero number will do.) Now

Rλ =−λ

(I− 1

λL)

If these things are strictly increasing for infinitely many k, then by Lemma 17.6.3, there isan infinite sequence xk,xk ∈ N

(Rk+1

λ

)\N(Rk

λ

), dist

(Lxk,LN

(Rk

λ

))≥ 1/2. Hence

∥Lxk−Lxk−1∥ ≥ 1/2

and this can’t happen because L is compact so {Lxk} has a Cauchy subsequence. Thereforethere exists a smallest k such that

N(

Rkλ

)= N

(Rm

λ

),m≥ k

On the other hand,{

Rkλ

X}

are decreasing in k. By similar reasoning using Lemma 17.6.3and the observation that Rλ

(Rk

λX)⊇ Rk+1

λX (in fact they are equal) it follows that the{

Rkλ

X}

are also eventually constant, say for m≥ l.Now if you have y ∈ N

(Rk

λ

)∩Rk

λX , then y = Rk

λw and also Rk

λy = 0. Hence R2k

λw = 0

and so, w ∈ N(R2k

λ

)= N

(Rk

λ

)which implies Rk

λw = 0 and so y = 0. It follows N

(Rk

λ

)∩

Rkλ

X = {0}.Now suppose l > k. Then there exists y ∈ Rl−1

λX \Rl

λX and so Rλ y ∈ Rl

λX = Rl+1

λX =

Rλ Rlλ

X . So Rλ y = Rλ z for some z ∈ Rlλ

X . Thus y− z ̸= 0 because y /∈ Rlλ

X but z is. How-ever, Rλ (y− z) = 0 and so

(y− z) ∈ N (Rλ )∩Rkλ

X ⊆ N(

Rkλ

)∩Rk

λX

which cannot happen from the above which showed that N(Rk

λ

)∩Rk

λX = {0}. Thus l ≤ k.

Next suppose l < k. Then you would have Rlλ

X = Rkλ

X and N(Rk

λ

)⫌ N

(Rl

λ

). Thus

there exists y ∈ N(Rk

λ

)but not in N

(Rl

λ

). Hence Rk

λy = 0 but Rl

λy ̸= 0. However, Rl

λy is in

Rkλ

X from the definition of l and so there is u such that Rlλ

y = Rkλ

u. Thus

0 = Rkλ

y = Rk−l+lλ

y = Rk−lλ

Rlλ

y = Rk−lλ

Rkλ

u = R2k−lλ

u

Now it follows that u ∈ N(R2k−l

λ

)= N

(Rk

λ

). This is a contradiction because it says that

Rkλ

u = 0 but right above the displayed equation, we had Rlλ

y = Rkλ

u and Rlλ

y ̸= 0. Thus,with the above paragraph, k = l.

What about the claim that Rλ restricted to Rkλ

X is a homeomorphism? It maps Rkλ

Xto Rk+1

λX = Rk

λX . Also, if Rλ (y) = 0 for y ∈ Rk

λX , then Rk

λy = 0 also and so y ∈ Rk

λX ∩

N(Rk

λ

). It was shown above that this implies y = 0. Thus Rλ appears to be one to one. By

assumption, it is continuous. Also from Lemma 17.6.2,

Rkλ

X is closed.

17.6. OPERATORS WITH CLOSED RANGE 469Proof: Consider A 4 0. The N(R‘) are increasing in k and Ry (N (R4*')) CN (RK).This follows from the definition. (It isn’t necessary to assume in most of this that A € A,just a nonzero number will do.) Now1R,=-A G 4)If these things are strictly increasing for infinitely many k, then by Lemma 17.6.3, there isan infinite sequence x;,x, € N (Ri*1) \N (Ri), dist (Lxx, LN (Ri )) > 1/2. Hence|| Lg — Lxx—1|| > 1/2and this can’t happen because L is compact so {Lx;} has a Cauchy subsequence. Thereforethere exists a smallest k such thatN (RL) =N (Rf), m>kOn the other hand, {REX } are decreasing in k. By similar reasoning using Lemma 17.6.3and the observation that Ry (R&X) D RV*'X (in fact they are equal) it follows that the{RU X } are also eventually constant, say for m > I.Now if you have y € N (Ri) ORX, then y = Ri w and also Ri y = 0. Hence R}*w = 0and so, w € N (R}*) = (Ri) which implies Ri w = 0 and so y = 0. It follows N (R4)NRX = {0}.Now suppose / > k. Then there exists y € Ri 'X \ Ri,.X and so Ryy € RX = RY'X =RR, X. So Ryy = Ryz for some z € RX. Thus y—z 40 because y ¢ RX but z is. How-ever, Ry (y—z) =0 and so(y—z) EN(R,) REX CN (Ri) AREXwhich cannot happen from the above which showed that N (Ri) ORK X = {0}. Thus 1 <k.Next suppose / < k. Then you would have R)X = R|X and N (Ri) 2 N(R) ). Thusthere exists y € N (Ri) but not in NV (Ri) . Hence Rhy = 0 but Rhy 4 0. However, Rhy is inR},X from the definition of / and so there is u such that R) y = Riu. ThusO=Riy=R, Ty =RL Ry = RY Riu = RuNow it follows that u € N (R5*') = N (R4). This is a contradiction because it says thatRiu = 0 but right above the displayed equation, we had Ri y= Riu and Ri, y #0. Thus,with the above paragraph, k = 1.What about the claim that Ry restricted to RX is a homeomorphism? It maps RKXto RYIX = RUX. Also, if Ry (y) = 0 for y € REX, then Rhy = 0 also and so ye RUX alN (Ri). It was shown above that this implies y = 0. Thus Ry appears to be one to one. Byassumption, it is continuous. Also from Lemma 17.6.2,RUX is closed.