Expansions and Characterizations of Sieved Random Walk Polynomials

Let k\geq2 first. We establish the expansion (2.1) via induction on n\in\mathbb{N}_0. It is clear from the recurrence relation for the Chebyshev polynomials of the first kind that P_i(x;k)=T_i(x) for all i\in\{0,\ldots,k\}, so (2.1) is true for n=0. Now let n\in\mathbb{N} be arbitrary but fixed and assume the validity of (2.1) for n-1. In particular, we then have

Due to (?) and (?), the latter equation can be rewritten as

In the same way, we obtain

We now use that P_{k m}(x;k)=P_m(T_k(x)), m\in\mathbb{N}_0 [GvA88, Theorem 1, Section VI], which yields

Combining (Proof), (Proof) and (Proof) with the relation 2x P_{k n-1}(x;k)=P_{k n}(x;k)+P_{k n-2}(x;k) and using the recurrence relation for the Chebyshev polynomials of the first kind, we obtain that r_n(j)=p_{n-1}(j-1)+q_{n-1}(j) for each j\in\{0,\ldots,\left\lfloor n/2\right\rfloor\}. Since

for each j\in\{0,\ldots,\left\lfloor n/2\right\rfloor\} (which makes use of the recursions (?) and (?), as well as another use of definition (?)), we therefore get

We now combine (Proof) with (Proof), write

use (?), (?) and definition (?) again and obtain

Thus if k=2, then (2.1) is shown. If k\geq3, we use induction on i to prove that

and have already shown the initial step i\in\{0,1\}; we hence assume i\in\{0,\ldots,k-3\} to be arbitrary but fixed and (Proof) to hold for i,i+1, and then calculate

This finishes the proof of (2.1) for k\geq2, and we have simultaneously established (2.1) and (2.1). (2.1) for the remaining case k=1 is an immediate consequence of (2.1). Finally, (2.1) can be seen as follows: let n\in\mathbb{N}_0 and j\in\{0,\ldots,\left\lfloor n/2\right\rfloor\}. By (2.1) and (2.1), we have

for all i\in\{0,\ldots,j\}. Taking the sum from 0 to j and using definition (?), we get

as desired.

The case k=1 is trivial because \mathrm{A}_1=\mathrm{id}, so let k\geq2 from now on (in particular, we then have |\cos(\pi/k)|=\cos(\pi/k)). “(i) \Rightarrow (ii)”: let n\in\mathbb{N}_0 and i\in\{0,\ldots,k-1\}. Using Theorem 2.1 and (?), we have

For each j\in\{0,\ldots,\left\lfloor n/2\right\rfloor\} (except the case n even and j=n/2, but then p_n(j)=0), we compute

In the same way, we have

for all j\in\{0,\ldots,\left\lfloor n/2\right\rfloor\}. Hence, we see that P_{k n+i}(x;k) is an eigenvector of \mathrm{A}_k which corresponds to the eigenvalue T_{k n+i}(\cos(\pi/k)). “(ii) \Rightarrow (iii)” is trivial from orthogonality. “(iii) \Rightarrow (i)”: the condition and Lemma 3.3 yield

It is immediate from (Proof) that c_1=1/2. Let n\in\mathbb{N} be such that k does not divide n+1. Moreover, assume that, for each j\in\{1,\ldots,n\}, c_j=1/2 if j is not a multiple of k. Decomposing n+1=k l+i with unique l\in\mathbb{N}_0, i\in\{1,\ldots,k-1\}, on the one hand, we get

from (Proof), which then simplifies to

On the other hand, we compute

Combining (Proof) and (Proof), we obtain that c_{n+1}=c_{k l+i}=1/2. This finishes the proof of “(iii) \Rightarrow (i)”. Concerning the remaining assertion, it suffices to show that

this can be seen as follows: let l\in\mathbb{N}_0. Since T_{k l}(\cos(\pi/k))=(-1)^l, and since

for l\neq0, we have

Consequently, the recurrence relations for the sequences

coincide.

First, note that in view of (?) (iii) is an obvious reformulation of (ii). Moreover, the case k=1 is just Theorem 1.1 because \mathrm{D}_1=\mathrm{d}/\mathrm{d}x (concerning (iv), we refer to the proof given in [LO08]), so let k\geq2 from now on. “(i) \Rightarrow (ii)”: we know from Theorem 3.1 “(i) \Rightarrow (ii)” and Lemma 3.4 that

and we now use induction on n to deduce that

for each n\in\mathbb{N}, which implies the assertion due to (?) and (?). It is obvious from Lemma 3.3 that (Proof) is satisfied for n\in\{1,2\}, so let n\in\mathbb{N}\backslash\{1\} be arbitrary but fixed now and assume (Proof) to hold for both n-1 and n. If j\in\{1,\ldots,\left\lfloor n/2\right\rfloor\}, then (Proof) implies

and we distinguish two cases: if k\,{\not|}\,n, then (Proof) yields

because a_{n-1}=c_{n-1}=1/2 (if k\,{\not|}\,(n-1)) or U_{n-2}(\cos(\pi/k))=0 (if k|(n-1)), which, by the recurrence relation for the Chebyshev polynomials of the second kind, simplifies to

If k|n, however, we compute from (Proof)

If one takes into account Lemma 3.3 for the remaining case “j=0” again, the proof of the direction “(i) \Rightarrow (ii)” is finished. “(ii) \Rightarrow (iv)” is trivial from (?). The direction “(iv) \Rightarrow (i)” is more involved; we first deal with c_1 and c_2: due to the conditions \kappa_3(0;k)=\sigma(3;k) and \kappa_4(1;k)=\sigma(4;k), a tedious but straightforward calculation based on Lemmas 3.3 and 3.4 yields

and

Combining these equations, we obtain

To get another equation for c_1, c_2, c_3, we first note that Lemma 3.3 and (?) imply that

which, since

due to (?), becomes

after another tedious but straightforward calculation which uses (?). We then take into account that Lemma 3.4 and the conditions \kappa_3(0;k)=\sigma(3;k), \kappa_4(1;k)=\sigma(4;k) and \kappa_5(0;k)=\kappa_5(2;k)(=\sigma(5;k)) yield \alpha_4(0;k)=\alpha_4(2;k), which can be rewritten as

as a consequence of (Proof) and the simplifications

Now, the combination of (Proof) with (Proof) gives

dividing this by 1-c_2, we finally obtain

From now on, we consider the nonlinear system (Proof), (Proof), (Proof), which depends on k. If k=2, then (Proof) immediately gives c_1=1/2 because U_2(0)=-1. In the following, let k\geq3; here, \cos(\pi/k)\neq0 and (Proof) reduces to

Combining (Proof) and (Proof), we see that

In the same way combining (Proof), (Proof) on the one hand and (Proof), (Proof) on the other hand, we get

which can be rewritten as

Moreover, we can deduce from (Proof) and (Proof) that

which now considerably simplifies to

If 4c_1c_2-4c_1+1=0, it is immediate from (Proof) that c_2=1/2. If 2c_1c_2-2c_1+c_2=0, then (Proof) reads U_2(\cos(\pi/k))[1-2c_2]=3-6c_2, and we can conclude that c_2=1/2, too (because U_2(\cos(\pi/k))=4(\cos(\pi/k))^2-1<3). If c_2=1/2, then (Proof) reduces to (1-2c_1)(1-c_1-c_3)=0, and if additionally c_3=1-c_1, then (Proof) implies that c_1=1/2. Finally, if c_1=1/2, then {c_2=1/2} because otherwise (Proof) would yield {-1\!=\!U_2(\cos(\pi/k))\!=\!\sin(3\pi/k)/\sin(\pi/k)\!\geq\!0}. Therefore, allowing k\geq2 to be arbitrary again, we have seen that c_1=1/2 in each possible case, and that c_2=1/2 if k\neq2. Now let n\in\mathbb{N}\backslash\{1\} be arbitrary but fixed and assume that, for all j\in\{1,\ldots,n\}, c_j=1/2 if j is not a multiple of k. Lemma 3.4 and the conditions \kappa_{n+1}(n-2;k)=\sigma(n+1;k) and \kappa_{n+2}(n-1;k)=\sigma(n+2;k) yield

Since \alpha_{n+1}(n-1;k) is uniquely determined by the recurrence coefficients c_1,\ldots,c_n (which is already a consequence of (?)), we have \alpha_{n+1}(n-1;k)=0 by Theorem 3.1 “(i) \Rightarrow (ii)” and the induction hypothesis.(Such arguments will occur several times in our proof, so we give some more details at this stage: compare (P_j(x))_{j\in\mathbb{N}_0} to the sequence \big(\widetilde{P}_j(x)\big)_{j\in\mathbb{N}_0} which is given by \widetilde{c}_j:=c_j for j\in\{1,\ldots,n\} and \widetilde{c}_j:=1/2 otherwise. Then \big(\widetilde{P}_j(x)\big)_{j\in\mathbb{N}_0} is k-sieved, so 0=\widetilde{\alpha}_{n+1}(n-1;k)=\alpha_{n+1}(n-1;k). We used similar ideas also in our paper [Ka16a].) Therefore, (Proof) simplifies to

and then multiplication with a_n c_n h(n), Lemma 3.3 and the recurrence relation for the Chebyshev polynomials of the second kind imply that

If k\geq3 and if n+1 is not a multiple of k, then \cos(\pi/k)U_n(\cos(\pi/k))\neq0; consequently, (Proof) tells c_{n+1}=1/2 because U_{n-1}(\cos(\pi/k))=0 or c_n=1/2, depending on whether k divides n or not. The case k=2 is more involved—here, (Proof) does not allow for any conclusion. Instead, we apply an idea which we similarly used in [Ka16a] (at that time for the operators \mathcal{D}_q and \mathcal{A}_q; cf. particularly [Ka16a, Lemma 3.1, proof of Theorem 2.3]) and proceed as follows: by the conditions, there is some m\in\{0,\ldots,\left\lfloor(n-2)/2\right\rfloor\} such that {\kappa_{n+3}(n-2-2m;2)=\sigma(n+3;2)}. Since c_1,\ldots,c_n fix the first n+2 polynomials P_0(x),\ldots,P_{n+1}(x), c_1,\ldots,c_n yield unique \lambda_0,\ldots,\lambda_{n+2},\nu_0,\ldots,\nu_{n-1}\in\mathbb{R} such that

Therefore, c_1,\ldots,c_n determine the integral

uniquely, so (?) and Theorem 3.1 “(i) \Rightarrow (ii)” imply that

as a consequence of the induction hypothesis. Hence, using Theorem 3.1 “(i) \Rightarrow (ii)” again, (?) and the fact that \alpha_n(n-2-2m;2) is uniquely determined by c_1,\ldots,c_n, too (use (?)), we see that

Furthermore, we have

which can be seen as follows: on the one hand, since \kappa_{n+1}(n-2-2m;2) is uniquely determined by c_1,\ldots,c_n (this is a consequence of (?)), one has \kappa_{n+1}(n-2-2m;2)=\sigma(n+1;2) by the already established direction “(i) \Rightarrow (ii)”; hence, by Lemma 3.4 we have

On the other hand, by Lemma 3.4 and the condition \kappa_{n+3}(n;2)=\sigma(n+3;2), we have

so (Proof) is established. Combining (Proof) with (Proof), we obtain that

Now following the proof of Theorem 3.1 “(iii) \Rightarrow (i)”, we can conclude that c_{n+1}=1/2 if n+1 is odd, which finishes the proof of Theorem 3.2.

(i) If one expands P_n(x) as in (?), this is obvious from the definitions (in particular, use (?)).

(ii) Using (?), (?) and (i), we have

for some R(x)\in\mathbb{R}[x] with \deg R(x)\leq n-3; thus a comparison of the coefficients of T_{n-2}(x) and (?) yield

Finally, since

(see the proof of [IO11, Lemma 5.1]) and a_{n-2}a_{n-1}h(n-2)=c_{n-1}c_n h(n), we obtain the assertion.

(iii) While on the one hand one has

by (?), on the other hand one has

with polynomials R(x),S(x)\in\mathbb{R}[x] with \deg R(x),\deg S(x)\leq n-2. Consequently, we have

and as obviously r_{n-1}(0)=(2-\delta_{n,1})a_{n-1}r_n(0) and a_{n-1}h(n-1)=c_n h(n), the proof is complete (note that, by definition, \kappa_n(n-1;k)=\sigma(n;k)).

Due to linearity, it clearly suffices to establish that

This, however, can easily be seen from the equations (?), (?), (?) and (?) by the computation

for m\leq n (the expansion T_m(x)T_n(x)=T_{m+n}(x)/2+T_{n-m}(x)/2 is well known). Now let n,j\in\mathbb{N}_0. Via (?), we compute

multiplication with P_j(x), integration with respect to \mu and the equations (?) and (?) yield the second assertion.