Highest Weight Modules Over The Quantum Periplectic Superalgebra of Type P

We will prove (a) and (c). The remaining parts can be deduced similarly.

First, we prove (a) for j=n. For every i we have

If i = n-2, then we have that [e_{n-1}, [f_{\overline{n-1}},e_{\overline{n-2}}]] = [e_{n-1}, [e_{n-1},e_{n-2}]] = 0. If i = n-1, then we have that

Otherwise, we have that [F_{\overline{n}}, e_{\overline{i}}] = 0.

Next we prove (a) for j < n. Using the relations in Proposition 1.1, we have that:

Note that [F_{\overline{j}}, e_{\overline{i}}] = 0 from above, unless |i-j| \leq 2. So, we need to check the three subcases i-j = 0,1,-1.

If j=i, then

If i = j+1, then

If j=i+1, then

Now, we prove (c). Note that [e_i,e_{\overline{i-1}}] = 0 for all 2\leq i \leq n, so [e_i, [e_i,e_{\overline{i-1}}]] = 0. Also,

Let U be the unital associative superalgebra over \mathbb{C}(q) generated by the elements e_i, f_i, e_{\overline{i}}, f_{\overline{i}} for i\in I, F_{\overline{i}} for i\in I \cup \{ n\}, and q^h for h\in P^{\vee} with defining relations given in the statement of the proposition above.

We first note that using (?) and particular choices for i,j,k,l in (?) one can establish on a case-by-case basis all relations in the proposition. Hence, we have an associative superalgebra homomorphism \psi: U\rightarrow \mathfrak{U}_q\mathfrak{p}_n. The relations in (?) immediately show that \psi is surjective.

It remains to show that \psi is injective. For this, we prove that (?) is obtained from the relations in the statement of the proposition by considering the following 26 cases:

The verification in each case uses the relations (?) and (?) and appropriate induction. In fact, for some cases, we apply useful identities that follow from (?) and (?), see Lemma 2.1 below. For example, in Case 21 we use Lemma 2.1(c). For reader’s convenience, we write detailed proofs for cases 2 and 25. The remaining cases are established using analogous reasoning.

Case 2. Suppose that \vert k \vert < \vert i \vert = \vert j \vert < \vert l \vert in (?). We prove that

is obtained from some of the relations in the proposition applying induction on |\ell| - |i| first, and then induction on |i| - |k|. We consider only the case of when i=-j >0 and k,\ell > 0 as the other cases follow similarly.

We start with the base case of the first induction, i.e., \ell = i+1. For the base case of the second induction, we have i = k+1. Then

The induction step for the second induction (i - k \geq 2) is established as follows:

For the induction step of the first induction (\ell - i \geq 2) we proceed as follows:

Case 25. Suppose that \vert i \vert < \vert j \vert = \vert k \vert = \vert l \vert in (?). We prove that

for j=\ell, and

for j=-\ell, using some of the relations in the proposition. We proceed by induction on |j| - |i| and consider only the case i > 0 and j=k > 0 as the other cases follow similarly.

For the base case j = i + 1, the relations

when j=\ell, and

when j=-\ell, follow from the relations q^{k_{i+1}}f_{i} = qf_{i}q^{k_{i+1}} and f_iF_{\overline{i+1}} - q^{-1}F_{\overline{i+1}}f_i=-2q^{k_{i+1}}f_{\overline{i}}. For the induction step (j - i \geq 2) we have:

for j=\ell, and

for j=-\ell.

We will prove the second isomorphism. Let \{f_\zeta \mid \zeta \in \Omega\} be a basis of \mathfrak{U}_q^- consisting of monomials in f_i’s, f_{\overline{i}}’s, and F_{\overline{j}}’s (1\leq i \leq n-1, 1\leq j \leq n), with \Omega being an index set. Consider the map \varphi: \mathfrak{U}_q^- \otimes \mathfrak{U}_q^0 \rightarrow \mathfrak{U}_q^{\leq 0} defined by \varphi(f_\zeta \otimes q^h) = f_\zeta q^h. The defining relations of \mathfrak{U}_q\mathfrak{p}_n imply that f_\zeta q^h span \mathfrak{U}_q^{\leq 0}, so \varphi is surjective. It remains to show that the set \{f_\zeta q^h \mid \zeta \in \Omega, h\in P^\vee\} is linearly independent over \mathbb{C}(q).

Suppose

for some C_{\zeta, h} \in \mathbb{C}(q). Then

Write \displaystyle \alpha = -\sum_{i=1}^{n-1}(m_i\alpha_i+n_i\gamma_i) - \sum_{i=1}^{n}r_i\beta_i, for m_i,n_i, r_i\in \mathbb{Z}_{\geq 0}, and let \displaystyle h_\alpha = \sum_{i=1}^{n-1}(m_i+n_i)k_{i+1} + \sum_{i=1}^{n}r_ik_{i} and \displaystyle h_\alpha^\prime = \sum_{i=1}^{n-1}(m_i+n_i)k_{i} +r_ik_{i}. From \displaystyle \mathfrak{U}_q\mathfrak{p}_n = \bigoplus_{\alpha\in Q} (\mathfrak{U}_q)_\alpha, we have that, for each \alpha\in Q_-,

Since f_\zeta is a monomial in f_i’s, f_{\overline{i}}’s, and F_{\overline{i}}’s, we have

Hence, the degree (\alpha, 0) term in the decomposition of \Delta(f_\zeta) equals f_\zeta \otimes q^{h_{\alpha}}. Applying the comultiplication to (?) gives

Collecting the terms of degree (\alpha, 0) gives that

Since for every \alpha, the set \{q^{h+h_{\alpha}} \mid h \in P^\vee\} is linearly independent, we have that, for all h\in P^\vee:

Due to the linear independence of f_\zeta, we conclude that C_{\zeta, h} =0 for all \zeta, h, as desired.

Let \{f_\zeta \mid \zeta \in \Omega\}, \{q^h \mid h \in P^\vee\}, and \{e_\tau \mid \tau \in \Omega^\prime\} be monomial bases of \mathfrak{U}_q^-, \mathfrak{U}_q^0, and \mathfrak{U}_q^+, respectively, where \Omega is the index set as in the proof of Lemma 2.4, and \Omega^\prime is another index set. Using the defining relations of \mathfrak{U}_q\mathfrak{p}_n in Proposition 2.2, we can express every monomial in \mathfrak{U}_q(\mathfrak{p}_n) as a linear combination of monomials each of which has e_i and e_{\overline{i}} on the right. By Lemma 2.4, the monomials f_{\zeta}q^he_{\tau} span \mathfrak{U}_q\mathfrak{p}_n. Hence, it remains to show that f_{\zeta}q^he_{\tau} are linearly independent over \mathbb{C}(q).

Suppose

where C_{\zeta, h,\tau} is some nonzero constant in \mathbb{C}(q). Due to the Q-grading of \mathfrak{U}_q\mathfrak{p}_n, we have that, for all \alpha \in Q:

Define a partial ordering on \mathfrak{h}^* by \lambda \leq \mu if and only if \lambda - \mu \in Q_- for \lambda,\mu \in \mathfrak{h}^*. We then choose \gamma = \deg f_{\zeta} and \beta = \deg e_{\tau}, which are minimal and maximal, respectively, among those for which \gamma + \beta = \alpha and C_{\zeta, h, \tau} \neq 0. If \displaystyle \gamma = -\sum_{i=1}^{n-1}(m_i\alpha_{i}+n_i\gamma_{i}) - \sum_{i=1}^{n}r_i\beta_{i}, set \displaystyle h_\gamma = \sum_{i=1}^{n-1}(m_i+n_i)k_{i+1} + \sum_{i=1}^{n}r_ik_{i}, and if \displaystyle \beta = \sum_{i=1}^{n-1}(m^\prime_i\alpha_{i}+n^\prime_i\gamma_{i}), set \displaystyle h_\beta = \sum_{i=1}^{n-1}(m^\prime_ik_{i}+n^\prime_ik_{i}), for m_i,m^\prime_i,n_i,n^\prime_i, r_i\in \mathbb{Z}_{\geq 0}.

The term of degree (0,\beta) in \Delta(e_\tau) is q^{h_\beta}\otimes e_{\tau} and the term of degree (\gamma,0) in \Delta(f_\zeta) is f_{\zeta} \otimes q^{h_\gamma}. Applying the comultiplication to the sum in (?), and looking at the terms of degree (\gamma, \beta), we have that

By Lemma 2.4, the elements f_\zeta q^h are linearly independent for \zeta\in\Omega, h\in P^\vee. Thus, for all h\in P^\vee, we have that

Due to the linear independence of e_\tau, we conclude that C_{\zeta, h, \tau} =0, leading to contradiction. Therefore all coefficients in (?) are zero.

The proof is of this is similar to that of Proposition 1.9 in [GJKK]. The main idea of the proof is that since e^2_{\overline{i}} = 0 and f^2_{\overline{i}} = 0, then, with the aid of the Poincaré-Birkhoff-Witt theorem, we show that \mathfrak{U}_q(\mathfrak{p}_n)_0v, where (\mathfrak{p}_n)_0 is generated by \{e_i, f_i, k_j \mid i\in I, j\in J\}, is finite generated, using the fact that f_i^{\lambda(k_i)-\lambda(k_{i+1}) + 1}v = 0 and v is a highest weight vector. For details, see the proof of Proposition 1.9 in [GJKK].

If f_i^{\lambda(k_i) - \lambda(k_{i+1})+1} v \neq 0 and \mathfrak{U}_q^+f_i^{\lambda(k_i) - \lambda(k_{i+1})+1} v = 0, then f_i^{\lambda(k_i) - \lambda(k_{i+1})+1} v generates a nontrivial proper submodule of V^q(\lambda), which would be a contradiction to the fact that V^q(\lambda) is irreducible. Therefore, it is enough to show that \mathfrak{U}_q^+f_i^{\lambda(k_i) - \lambda(k_{i+1})+1} v = 0.

Note that for j\neq i-1,i, we have that

We next prove that e_jf_i^{\lambda(k_i) - \lambda(k_{i+1})+1}v = 0 for j=i,i-1. In the case j=i, we use Lemma 3.5(a) for m= \lambda(k_i) - \lambda(k_{i+1})+1:

For j=i-1 we apply induction on m. The base case m=1 follows from

Assume that e_{i-1}f_i^{ m} v =0. Then we have

Therefore, e_jf_i^{ \lambda(k_i) - \lambda(k_{i+1})+1}v = 0 for all j\in I.

Lastly, we prove that e_{\overline{j}}f_i^{\lambda(k_i) - \lambda(k_{i+1})+1}v = 0 for all j \in I. Note that for j\neq i-1,

For j = i-1, using that e_{i-1}f_i^{ m} v = 0, we have

Thus we obtain e_{\overline{i-1}}f_i^mv = 0 by induction on m. Hence, \mathfrak{U}_q^+f_i^{\lambda(k_i) - \lambda(k_{i+1})+1} v = 0 and the proof is complete.

Consider the isomorphism \varphi : \mathfrak{U}_q\mathfrak{p}_n \longrightarrow \mathfrak{U}_q^-\otimes \mathfrak{U}_q^0\otimes \mathfrak{U}_q^+ from Theorem 2.5. Note that the following relations hold:

These relations, together with Lemma 4.2, imply that the image of the restriction \overline{\varphi} of \varphi to \mathfrak{U}_{\mathbf{A}_1} is a subset of \mathfrak{U}_{\mathbf{A}_1}^-\otimes \mathfrak{U}_{\mathbf{A}_1}^0 \otimes \mathfrak{U}_{\mathbf{A}_1}^+. To define an inverse of \overline{\varphi} we multiply the corresponding terms in the tensor product. This completes the proof.

In the light of Proposition 4.3, it suffices to show that \mathfrak{U}_{\mathbf{A}_1}^+ v = 0 and \mathfrak{U}_{\mathbf{A}_1}^0 v = \mathbf{A}_1v. The former identity follows from the fact that v is a highest weight vector. For the latter identity we use following:

The idea is standard but for reader’s convenience we present the proof. Let v = v_1+ \hdots + v_p \in V_{\mathbf{A}_1}, where v_j\in V_{\mu_j}^q and \mu_j\in P for each j\in \{1,2,\hdots, p\}. We can assume that \mu_j are distinct due to the weight decomposition of V^q. Therefore, it is enough to show that v_j\in V_{\mathbf{A}_1} for each j.

Fix an index j. For each i\neq j, choose h_i\in P^\vee such that \mu_j(h_i) \neq \mu_i(h_i). Let u\in \mathfrak{U}_{\mathbf{A}_1} be defined by

Then for each i \neq j:

where k\in \{ 1, 2,\hdots, p\}. Therefore uv_j = v_j and uv_i = 0. Hence, v_j = uv\in V_{\mathbf{A}_1}.

Since \mathbf{A}_1 is a principal ideal domain, every finitely generated torsion-free module over \mathbf{A}_1 is free. Notice that, for each \mu\in P, the weight space (V_{\mathbf{A}_1})_\mu is finitely generated as an \mathbf{A}_1-module. The weight space is also torsion free, as otherwise it would contradict the fact that \mathbf{A}_1 is an integral domain. Thus, for each \mu\in P, the weight space (V_{\mathbf{A}_1})_\mu is a free \mathbf{A}_1-module.

Since \mathbb{C}(q) is the field of quotients of \mathbf{A}_1, a set of vectors of a \mathbb{C}(q)-vector space is linearly independent if and only if it is \mathbf{A}_1-linearly independent. Thus \operatorname{rank}_{\mathbf{A}_1}(V_{\mathbf{A}_1})_\mu \leq \dim_{\mathbb{C}(q)}V^q_\mu. Let f_\zeta’s be some monomials in f_i’s, f_{\overline{j}}’s, and F_{\overline{\ell}}’s. Since \{f_\zeta\} forms a linearly independent set over \mathbb{C}(q), \{f_{\zeta}v \; | \; \zeta \in Z\} is a \mathbb{C}(q)-basis of V_\mu^q for an appropriate set Z. This basis is also contained in (V_{\mathbf{A}_1})_\mu by definition. So we have that \operatorname{rank}_{\mathbf{A}_1}(V_{\mathbf{A}_1})_\mu \geq \dim_{\mathbb{C}(q)}V^q_\mu, which completes the proof.

We first prove that \overline{v}_i, i=1,...,m, span V^1_\mu. For every v in V^1_\mu = (\mathbf{A}_1/\mathbf{J}_1) \otimes_{\mathbf{A}_1} (V_{\mathbf{A}_1})_\mu we have

where a_i = b_i + \mathbf{J}_1 \in \mathbf{A}_1/\mathbf{J}_1, \overline{v}_i = (1 \otimes_{\mathbf{A}_1} v_i). So the set \{1 \otimes_{\mathbf{A}_1} v_i\} \mathbb{C}-spans V^1_\mu. The linear independence follows from the fact that \{v_i\}, i=1,...,m, are \mathbf{A}_1-linearly independent.

We now prove part (b). Assume that \displaystyle \sum_{i=1}^m c_i(q)v_i = 0 for c_i(q) \in \mathbf{A}_1, with some c_j(q)\neq 0 for some j. Then, multiplying by an appropriate power of q-1, we may assume that c_j(1) \neq 0. Applying the classical limit gives \displaystyle \sum_{i=1}^m c_i(1)\overline{v}_i = 0. This contradicts the linear independence of \overline{v}_i, i=1,...,m.

The first assertion follows from Proposition 4.6, while the second assertion follows from Lemma 4.10.

To prove (a), we recall that by Theorem 4.1 in [AGG], there is a \mathbb{C}-superalgebra isomorphism

where \mathcal{A} is the localization of \mathbb{C}[q,q^{-1}] at the ideal generated by q-1, and \mathfrak{U}_{\mathcal{A}}\mathfrak{p}_n is the \mathcal{A}-subalgebra of \mathfrak{U}_q\mathfrak{p}_n generated by a set of elements \tau_{ij} (for the precise definition of \tau_{ij}, see §4 in [AGG]). The map \theta: \mathfrak{U}_{\mathcal{A}}\mathfrak{p}_n/(q-1)\mathfrak{U}_{\mathcal{A}}\mathfrak{p}_n \rightarrow \mathfrak{U}_{\mathbf{A}_1}/ \mathbf{A}_1\mathfrak{U}_{\mathbf{A}_1}, defined by

is also a \mathbb{C}-superalgebra isomorphism. Indeed, this follows from the definition of the corresponding generators and the fact that the classical limit of t_{ij} and \tau_{ij} coincide. Also, we have already established that \sigma: \mathfrak{U}_{\mathbf{A}_1}/\mathbf{A}_1\mathfrak{U}_{\mathbf{A}_1} \rightarrow U_1 is a \mathbb{C}-superalgebra isomorphism by the definition of U_1. Therefore, the map

is a \mathbb{C}-superalgebra isomorphism.

To prove (b), let w\in (V_{\mathbf{A}_1})_\mu and h\in P^\vee. Then we have that

Taking the classical limit of both sides gives \overline{h}\overline{w} = \mu(h)\overline{w}, as desired.

It remains to prove (c). From (b) we have that \overline{h}\overline{v} = \lambda(h)\overline{v}. Since v is highest vector of V_{\mathbf{A}_1} with highest weight \lambda, it follows that \overline{e_{{i}}} \overline{v} = 0 and \overline{e_{\overline{i}}} \overline{v} = 0 for each i\in I. Therefore, (a) and Proposition 4.5 imply that V^1 = U_1^-\overline{v} = \mathfrak{U}\mathfrak{p}_n^-\overline{v}. Hence, by definition, V^1 is a highest weight \mathfrak{U}\mathfrak{p}_n-module with highest weight \lambda and highest weight vector \overline{v}.

Propositions 4.7 and 4.11 imply that \dim_\mathbb{C} V^1_\mu = \dim_{\mathbb{C}(q)} V^q_\mu for each \mu\in P. This, along with Theorem 4.12(b), gives us the desired result.

Let V^q = V^q(\lambda) with highest weight vector v, and suppose that \lambda\in \Lambda^+. From Proposition 3.6, we have that f_i^{\lambda(k_i)-\lambda(k_{i+1}) + 1}v = 0 for all i\in I. Applying the classical limit leads to \overline{f}_i^{\lambda(k_i)-\lambda(k_{i+1}) + 1}\overline{v} = 0. Since V^1 is a highest weight module, Proposition 3.3 gives us that V^1 is finite dimensional. Thus, by Proposition 4.13, V^q is finite dimensional.

Conversely, suppose that V^q(\lambda) is finite dimensional. By Proposition 4.13, we have that V^1 is also finite dimensional. By Proposition 3.4, we have that \lambda\in \Lambda^+.

Let w = w_{\tau,\tilde{r},\tilde{s}} and \theta = y_{\tau}\mathsf{c}_{\tilde{r},\tilde{s}}w. Note that the weight of \theta is the same as the weight of w since y_{\tau}c_{\tilde{r},\tilde{s}} commutes with the action of \mathfrak{U}_q\mathfrak{p}_n. Note that the fact that \theta \neq 0 and the linear independence property follow by applying the classical limit and using Theorem 3.8 in [Mo].

To show that \theta is a maximal vector, it suffices to show that \theta is annihilated by each root vector e_{i} and e_{\overline{i}}, i\in I. The action of e_{i} and e_{\overline{i}} on w can be explicitly written as follows

Let x = e_i or x= e_{\overline{i}}. We look at how x acts on \mathsf{c}_{\tilde{r},\tilde{s}}w. Note that \mathsf{c} V^{\otimes 2} is the trivial \mathfrak{U}_q\mathfrak{p}_n-module \mathbb{C}(q). Therefore the sum of the terms with x acting on a pair of contracted tensor factors is zero. Also, from (?), we see that the action of x on non-contracted tensor factors, or in other words where w_i\in V_0, those specific terms in the summation will be zero, except when x=e_{j-1} for some j\geq 2. Thus x either annihilates \mathsf{c}_{\tilde{r},\tilde{s}}w or produces a sum of tensors which are obtained by applying x = e_{j-1}, for some j\geq 2, to a factor unaffected by \mathsf{c}_{\tilde{r},\tilde{s}}. Each of those tensors has one factor whose subscript is in (\tilde{r} \cup \tilde{s})^c and has been lowered by one, so u_j has been changed to u_{j-1}. Denote such a tensor by \mathsf{v}. We want to show that y_\tau\mathsf{v} = 0.

Fix \psi \in W_+. There are two factors of the tensor h(\sigma_{+}^\tau)h(\psi)h(\sigma_{+}^\tau)^{-1}\mathsf{v} that have the same u_j. Then there exists a transposition (a, b) = (\sigma_{-}^\tau) \rho\, (\sigma_{-}^\tau)^{-1} \in (\sigma_{-}^\tau)W_-(\sigma_{-}^\tau)^{-1} which permutes these two factors. Using that \ell (a,b) = \ell (\rho), we obtain

By Lemma 5.2, we have that for \rho\in W_-,

and hence

Thus we have

Hence,

for each \psi \in W_+. Therefore, we have that

This concludes the proof of y_\tau\mathsf{v} = 0.

Applying the classical limit to \theta^q_i yields the linearly independent maximal vectors \theta_i in [Mo] (after appropriate sign change for \theta^q_3, see Remark 5.1). Since the complete set of linearly independent maximal vectors of the \mathfrak{U}\mathfrak{p}_n-module V^{\otimes 2} has exactly three vectors, the result follows from Proposition 4.13. The second statement is subject to a direct verification.

We have that

However, since

it follows that \mathsf{c}_1 y_{\ytableaushort{1,2}}V^{\otimes 2} cannot be zero. Therefore, \mathsf{c}_1 y_{\ytableaushort{1,2}}V^{\otimes 2} \cong \mathbb{C}(q). Consider the restriction of the contraction map \mathsf{c}_1 to y_{\ytableaushort{1,2}}V^{\otimes 2}, and let \mathcal{N} denote the kernel of this restriction. Then

By Proposition 5.4, \theta_2^q \in \mathcal{N} and \theta_1^q, \theta_3^q \notin \mathcal{N}. Hence, by the same proposition, \theta_2^q is the only, up to a scalar multiple, maximal vector in \mathcal{N}, in particular, \mathcal{N} is simple. Thus

This implies the exact sequence (5.5). The sequence does not split because y_{\ytableaushort{1,2}}V^{\otimes 2} has a unique up to a scalar multiple maximal vector.

We have that

This implies that

Note that \theta_1^q \not\in \mathsf{c}_1 V^{\otimes 2}. However, (Proof) gives that \theta_3^q\in \mathsf{c}_1 (V^{\otimes 2}). By Proposition 5.4, \theta_3^q is the only, up to a scalar multiple, maximal vector in \mathsf{c}_1 V^{\otimes 2}. Thus, we have that

By Proposition 5.4, we have

which implies the exact sequence (5.6). Since \theta_1^q generates y_{\ytableaushort{12}}V^{\otimes 2}, the sequence does not split.

Let T: V^{\otimes 2} \longrightarrow V^{\otimes 2} be the \mathfrak{U}_q\mathfrak{p}_n-module homomorphism defined by

Note that

So, we have that y_{\ytableaushort{1,2}}V^{\otimes 2} \subset \ker T, which implies that y_{\ytableaushort{1,2}}V^{\otimes 2} = \ker T since the only maximal vector of V^{\otimes 2} in \ker T is \theta_2^q. Thus we have a short exact sequence

Using the embedding \iota: y_{\ytableaushort{12}} V^{\otimes 2} \longrightarrow V^{\otimes 2} and that T \circ \iota= \mathrm{id}, we see that the sequence above splits. The remaining part of the theorem follows from Propositions 5.5 and 5.6.

Consider the contraction map \mathsf{c}_1. Let \mathcal{N} be the kernel of the restriction of \mathsf{c}_1 to y_{\ytableaushort{1,2,3}} V^{\otimes 3}. We have that

and

By Proposition 5.8, \theta_7^q is the only maximal vector in \mathcal{N}. Hence,

Moreover,

which implies the exact sequence. The sequence does not split because y_{\ytableaushort{1,2,3}}V^{\otimes 3} has a unique up to a scalar multiple maximal vector.

Let \mathsf{K} = \mathsf{c}_1\mathsf{c}_2+q\mathsf{t}_2\mathsf{c}_1\mathsf{c}_2 +q^2\mathsf{c}_2. Note that

It follows that

Note that \theta_4^q \neq \mathsf{K} V^{\otimes 3}. However, \mathsf{K}(u_1\otimes u_1 \otimes u_{-1}) = \theta_1^q + q\theta_2^q + q^2\theta_3^q. By Proposition 5.8, \theta_1^q + q\theta_2^q + q^2\theta_3^q is the only, up to a scalar multiple, maximal vector in \mathsf{K} V^{\otimes 3}. Thus, we have that

By Proposition 5.8, we have

which implies the exact sequence (5.10). Since \theta_4^q generates y_{\ytableaushort{12}}V^{\otimes 2}, the sequence does not split.

Consider the \mathfrak{U}_q\mathfrak{p}_n-module homomorphism T: y_{\ytableaushort{12,3}}V^{\otimes 3} \rightarrow \mathsf{c}_2V^{\otimes 3} such that T(y_{\ytableaushort{12,3}}v) = \mathsf{c}_2y_{\ytableaushort{12,3}}v. Since

we have that T is surjective and

Note that

Thus, -\theta_1^q - (q-q^{-1})\theta_2^q + \theta_3^q\in y_{\ytableaushort{12,3}}\mathsf{c}_2V^{\otimes 3}. However, u_1\otimes u_1\otimes u_2 \not\in \mathsf{c}_2V^{\otimes 3}, which implies that \theta^q_5\in y_{\ytableaushort{12,3}}\mathsf{c}_2V^{\otimes 3}. So, by Proposition 5.8, we have that

Since \mathsf{c}_2V^{\otimes 3} \cong V^q(\epsilon_1), we have an inclusion map i: \mathsf{c}_2V^{\otimes 3} \rightarrow y_{\ytableaushort{12,3}}V^{\otimes 3} such that i\circ T = \text{id}. Therefore, the following short exact sequence splits:

By Proposition 5.8, since \theta_5^q\in \ker T, we have that \ker T = \mathfrak{U}_q\mathfrak{p}_n\theta_5^q \cong V^q(2\epsilon_1 + \epsilon_2). This implies the exact sequence (5.11).

Let v_1 = \theta_6^q, w_1 = \theta_5^q, v_2 = -q\theta_2^q + \theta_3^q, and w_2 = -\theta_1^q - (q-q^{-1})\theta_2^q + \theta_3^q. Consider the \mathfrak{U}_q\mathfrak{p}_n-module homomorphism

such that

By Propositions 5.8 and 5.11, S is an isomorphism. The short exact sequence (5.12) then follows from Proposition 5.11.

Let T: V^{\otimes 3} \longrightarrow V^{\otimes 3} be the \mathfrak{U}_q\mathfrak{p}_n-module homomorphism defined by

Note that

where \mathsf{K} = y_{\ytableaushort{12,3}} , y_{\ytableaushort{13,2}}, y_{\ytableaushort{1,2,3}}. By Proposition 5.8, this implies that -\theta_1^q - (q-q^{-1})\theta_2^q + \theta_3^q , -q\theta_2^q + \theta_3^q, \theta_q^5, \theta_q^6, \theta_q^7\in \ker T. We have the short exact sequence

Using the embedding \iota: y_{\ytableaushort{123}} V^{\otimes 3} \longrightarrow V^{\otimes 3} and that T \circ \iota= \mathrm{id}, we see that the sequence above splits.

Let T^\prime: \ker T \longrightarrow \ker T be the \mathfrak{U}_q\mathfrak{p}_n-module homomorphism defined by

Note that

where \mathsf{K} = y_{\ytableaushort{13,2}}, y_{\ytableaushort{1,2,3}}. By Proposition 5.8, this implies that -q\theta_2^q + \theta_3^q, \theta_q^6, \theta_q^7\in \ker T^\prime. Using the embedding \iota^\prime: y_{\ytableaushort{12,3}} V^{\otimes 3} \longrightarrow V^{\otimes 3} and that T^\prime \circ \iota^\prime = \mathrm{id}, we see that the short exact sequence

splits. Using similar arguments, we have that \ker T^\prime = y_{\ytableaushort{13,2}}V^{\otimes 3} \oplus y_{\ytableaushort{1,2,3}} V^{\otimes 3}, and thus the decomposition (?) follows.

The remaining part of the theorem follows from Propositions 5.9, 5.10, 5.11, and 5.12.

Fix k\geq 2, and assume for the sake of contradiction that V^{\otimes k} is completely reducible. For any r,s \in \{1,2\hdots, k-1\} such that r\neq s,

Consecutive applications of \mathsf{c}_{r,s} to V^{\otimes k} for appropriate r and s will lead to a submodule M of V^{\otimes k} that is isomorphic either to V^{\otimes 2} or to V^{\otimes 3}. By Theorems 5.7 and 5.13, V^{\otimes 2} and V^{\otimes 3} are not completely reducible. This leads to a contradiction as submodules of completely reducible modules are also completely reducible.