This paper was converted on www.awesomepapers.org from LaTeX by an anonymous user.
Want to know more? Visit the Converter page.

Morphisms from projective spaces to G/PG/P

Sarjick Bakshi  and  A J Parameswaran
Abstract.

We study maps between projective spaces and flag varieties. Let G=SL(n,)G=SL(n,\mathbb{C}). We show that there is no non constant map from 2\mathbb{P}^{2} to full flag variety G/BG/B. We classify the minimal parabolic subgroups PP for which there is a map from 3\mathbb{P}^{3} to G/PG/P.

Tata Institute of Fundamental Research, Mumbai, India, [email protected], Corresponding author
Tata Institute of Fundamental Research, Mumbai, India, [email protected]

1. Introduction

Let G=SL(n,)G=SL(n,\mathbb{C}). Let BB denote the Borel subgroup of upper triangular matrices and TT be the maximal torus of diagonal matrices in GG. The projective homogeneous variety G/BG/B is called the generalised flag variety (or simply flag variety). Let WW denote the Weyl group for (G,T)(G,T). In our case, it is SnS_{n}, the symmetric group in nn letter. For wWw\in W, let X(w)X(w) be the corresponding Schubert variety inside G/BG/B which is defined to be the closure of BB orbit passing through the TT-fixed point wBwB. The BB-orbit is called a Schubert cell and is denoted by C(w)C(w). The study of cohomology ring of the generalised flag variety H(X,)H^{\bullet}(X,\mathbb{Z}) goes back to Ehresmann [6]. He also showed that the flag variety has a partition in terms of Schubert cells. In [3], Borel gave a presentation of the ring. Let x1,x2,,xnx_{1},x_{2},\ldots,x_{n} denote nn indeterminates. Let [x1,x2,,xn]\mathbb{Z}[x_{1},x_{2},\ldots,x_{n}] denote the graded polynomial ring with degree of xix_{i} being 22 for each ii. SnS_{n} acts on indeterminates by permuting the indices. Let I=[x1,x2,,xn]\plusSn{\mathcal{}{I}}=\mathbb{Z}[x_{1},x_{2},\ldots,x_{n}]_{\plus}^{S_{n}} denote the ideal of [x1,x2,,xn]\mathbb{Z}[x_{1},x_{2},\ldots,x_{n}] generated by SnS_{n} invariants in positive degrees. In [3], Borel showed that there is an graded isomorphism

H(G/B,)[x1,x2,,xn]/I.H^{\bullet}(G/B,\mathbb{Z})\cong\mathbb{Z}[x_{1},x_{2},\ldots,x_{n}]/{\mathcal{}{I}}.

In [4], Chow showed that there is a basis of H(G/B,)H^{\bullet}(G/B,\mathbb{Z}) given by the classes of the Schubert varieties inside the Grassmannian. From [8], we know that the cohomology ring H(G/B,)H^{\bullet}(G/B,\mathbb{Z}) is also isomorphic to the Chow ring A(G/B)A^{\bullet}(G/B).

Let PP be a parabolic subgroup containing BB, let WPW_{P} denote subgroup of WW generated by simple reflections that appears in the Bruhat decomposition of PP. Let WP=W/WPW^{P}=W/W_{P} denote the set of minimal length coset representives. In [13], Reiner–Woo–Yong gave a presentation of the cohomology ring H(G/P,)H^{\bullet}(G/P,\mathbb{Z}). They show that

H(G/P,)H(G/B,)WP,H^{\bullet}(G/P,\mathbb{Z})\cong H^{\bullet}(G/B,\mathbb{Z})^{W_{P}},

where H(G/B,)WPH^{\bullet}(G/B,\mathbb{Z})^{W_{P}} denotes the ring of invariants of H(G/B,)H^{\bullet}(G/B,\mathbb{Z}) under the WPW_{P} action. In the same paper, they also study a presentation of the cohomology ring H(X(w),)H^{\bullet}(X(w),\mathbb{Z}) and get a relatively shorter presentation of the cohomology ring of the Schubert varieties appearing in the flag varieties.

Let knk\leq n be two positive integers. Let Gr(k,n)Gr(k,n) denote the Grassmannian of kk-dimensional subspaces of an nn dimensional complex vector space. Note that the Grassmannian variety is also a partial flag variety G/PG/P for a maximal parabolic subgroup PP. Maps from projective spaces to Grassmannian have been studied extensively by Tango in a series of papers [16, 18, 19]. In [16], he showed that there is no map from m\mathbb{P}^{m} to Gr(k,n)Gr(k,n) for mnm\geq n. In [17], he also showed that there is an indecomposable vector bundle of rank n1n-1 on n\mathbb{P}^{n}. While answering Lazarsfeld’s problems [9], under the assumption 1<k<nk1<k<n-k Paranjape–Srinivas in [12] show that there is a finite surjective morphism Gr(k,n)Gr(k,n) to Gr(l,m)Gr(l,m) if and only if k=lk=l and n=mn=m, in which case they are isomorphic. This naturally led to the study of maps between various projective homogeneous spaces.

The study of cohomology ring plays a crucial role in attempting many of these questions. Tango’s original proof compared the Chow ring of the two varieties. In [10], Muñoz–Occhetta–Conde showed that a weaker property of the cohomology ring is required to obtain Tango type result which they called effective good divisibility which was an improvement of the notion of effective good divisibility introduced by Pan in [11]. Using effective good divisibility, Naldi–Occhetta in [1] were able to extend Tango’s result and they show that every morphism between Gr(k,n)Gr(k,n) to Gr(l,m)Gr(l,m) for n>mn>m is constant. In the paper they show that the effective good divisibility of the Grassmannian is nn. In [14], Muñoz–Occhetta–Conde studied the maps between rational homogenous variety using effective good divisibility. They show that there is no map from a projective variety to a rational homogeneous variety where the effective good divisibility of the projective variety is higher than that of a rational homogeneous variety.

In our paper, we study maps from projective spaces to certain partial flag varieties. We first observe the following:

Theorem 1.1.

There is no non constant map from 2\mathbb{P}^{2} to G/BG/B.

Let PP be a parabolic subgroup of GG containing BB. Let mm denote the rank of the Picard group Pic(G/P)Pic(G/P). We call it the rank of G/PG/P. For instance, rank of the Grassmannian Gr(k,n)Gr(k,n) is 11 and the rank of the full flag variety G/BG/B is n1n-1. Similarly, we have rank of G/PG/P is n2n-2, for a minimal parabolic subgroup PP. From 1.1 and Tango’s result it is natural to ask the following question:

Question 1.2.

Does there exist a map from m\mathbb{P}^{m} to G/PJG/P_{J} where m=n+1rank(G/P)m=n+1-\rm{rank}(G/P)?

We study the question when PP is a minimal parabolic subgroup. Let α1,α2,,αn1\alpha_{1},\alpha_{2},\ldots,\alpha_{n-1} denote the set of simple roots. Let sαjs_{\alpha_{j}}, 1jn11\leq j\leq n-1 denote the corresponding simple roots. To each such αj\alpha_{j} we can associate a minimal parabolic subgroup Pαj:=BBsαjBP_{\alpha_{j}}:=B\cup Bs_{\alpha_{j}}B. We obtain the following two theorems.

Theorem 1.3.

There is no non constant map from 3\mathbb{P}^{3} to G/PαiG/P_{\alpha_{i}} for i{1,n1}i\in\{1,n-1\}.

Theorem 1.4.

There is a map from 3\mathbb{P}^{3} to G/PαjG/P_{\alpha_{j}} for 1<j<n11<j<n-1.

After we had put up the first version in the arXiv, it was communicated to us via an email from Yanjie Li that the proof of Theorem 5.1 has an error. Formula (8) in the proof of the theorem is wrong. To overcome that, we are giving a completely elementary proof of theorem 1.4. Due to the same error, we have now removed Theorem 6.1 from the arxiv version where we state that there is no morphism from 4\mathbb{P}^{4} to G/PαjG/P_{\alpha_{j}} for 1jn11\leq j\leq n-1. We have some evidence for the statement and we are working on it currently. After putting up our first version we have been communicated of a generalisation of some of our results by Fang and Ren (see, [7]).

Acknowledgement We are very thankful to Shrawan Kumar for informing us about his conjecture, and various helpful discussions. We are thankful to Rohith Varma for stimulating discussions. We are also thankful to Senthamarai Kannan for suggesting the article [14].

2. Kumar’s conjecture

In this section we discuss Kumar’s conjecture. This conjecture is due to Shrawan Kumar (see, [15]). Let XX be a complex connected homogeneous projective variety. Then XX can be written as H/PH/P for a complex semisimple connected algebraic group HH and PHP\subset H a parabolic subgroup. Observe that the HH and PP are not necessarily unique. For instance, the projective space 2n1\mathbb{P}^{2n-1} is a homogeneous space for SL(2n,)SL(2n,\mathbb{C}) as well as Sp(2n,)Sp(2n,\mathbb{C}). Assume also that XX is indecomposable, ie XX cannot be written as X1×X2X_{1}\times X_{2} where none of X1X_{1} or X2X_{2} is singleton.

Definition 2.1.

Let HH be a complex connected indecomposable semisimple algebraic group with PP a parabolic subgroup. The semisimple rank of the pair (H,P)(H,P) is defined to be the rank of the semisimple part of the Levi subgroup of PP.

Definition 2.2.

Let XX be complex connected indecomposable homogeneous projective variety. The minimum semisimple stabilizer rank (respectively, maximum semisimple stabilizer rank) of XX is the minimum (respectively, maximum) of the semisimple ranks for all possible realisations of XX as a H/PH/P, with HH simple and connected algebraic group and PP a parabolic subgroup. If the minimum semisimple stabilizer rank of XX coincides with the maximum semisimple stabilizer rank of XX we call it the rank of XX. Denote the minimum semisimple rank (respectively, maximum semisimple rank) by minss rank (respectively, maxss rank).

For instance, 2n1\mathbb{P}^{2n-1} has minss rank n1n-1 because it is realisable as a homogeneous space of Sp(2n,)Sp(2n,\mathbb{C}) whereas the maxss rank is 2n22n-2 when it is realised as a quotient of SL(2n,)SL(2n,\mathbb{C}). Note that we have minss rank X=0X=0 if and only if XX has a realisation of the form H/BH/B for a Borel subgroup BB inside HH. In this case, the maxss rank is also 0. We also observe that the minss rank XX is same as maxss rank XX in most cases, however the cases where it doesn’t hold can be found in [5, §2].

Conjecture 2.3.

[15, Conjecture 5] Let XX and XX^{\prime} be two connected indecomposable homogeneous projective varieties.

  • (a)

    Assume that XX is different from 2n\mathbb{P}^{2n} (for n1n\geq 1) and

    minss rank X>maxss rank X.\text{minss rank }X>\text{maxss rank }X^{\prime}.

    Then, there does not exist any nonconstant algebraic map from XX to XX^{\prime}.

  • (b)

    If X=2nX=\mathbb{P}^{2n} (for n1n\geq 1) and there exists a non-constant regular map from XXX\to X^{\prime}, then

    minss rank 2n1=n1maxss rank X.\text{minss rank }\mathbb{P}^{2n-1}=n-1\leq\text{maxss rank }X^{\prime}.

Kumar in [15] has proved the conjecture when XX^{\prime} can be written as H/BH^{\prime}/B^{\prime} for a simple algebraic group HH^{\prime} with BB^{\prime} a Borel subgroup. Note that this is equivalent to the case when semisimple rank of XX^{\prime} is 0. However, to what generality the conjecture is solved remains unclear, and our paper is a step towards understanding this conjecture. We look at the conjecture when XX^{\prime} has rank 0 or XX^{\prime} is a rank 11 homogeneous space of SL(n,)SL(n,\mathbb{C}). In the former case, we verify conjecture (a) when XX is homogeneous spaces of SL(n,)SL(n,\mathbb{C}). In the later, we verify the conjecture for any projective space. Some of the major computations in this paper analyses the map from 3\mathbb{P}^{3} which has minss rank 11 to SL(n,)/PSL(n,\mathbb{C})/P for a minimal parabolic subgroup PP (which has maxss rank 11 as well).

3. Preliminaries

Let G=SL(n,)G=SL(n,\mathbb{C}) denote the set of all n×nn\times n matrices with determinant 11. Let BB denote the Borel subgroup of upper triangular matrices, and TT denote the maximal torus consisting of diagonal matrices inside GG. Denote RR the root system of (G,T)(G,T). Let R+R^{+} denote the subset of RR consisting of positive roots. Let ϵi\epsilon_{i} denote the character of TT which sends diag(t1,t2,,tn)diag(t_{1},t_{2},\ldots,t_{n}) to tit_{i}. Let αi=ϵiϵi+1\alpha_{i}=\epsilon_{i}-\epsilon_{i+1}. Then a subset S={α1,,αn1}S=\{\alpha_{1},\ldots,\alpha_{n-1}\} of R+R^{+} gives a set of simple roots. The Weyl group WW is the group generated by the simple reflections sαs_{\alpha}, αS\alpha\in S. In our case, WW is the symmetric group in nn letters SnS_{n}. The simple reflections sαis_{\alpha_{i}} can be thought of as the transposition of ii-th and i+1i+1-th letter. We would use the one-line notation (w(1),w(2),,w(n))(w(1),w(2),\ldots,w(n)) to denote the permutation ww in SnS_{n}.

Let JJ be a subset of SS. Let WJW_{J} denote the subgroup of WW generated by sαs_{\alpha}, αJ\alpha\in J. For every JJ we associate a parabolic subgroup PJP_{J} as follows

PJ=wWJBwB.P_{J}=\bigsqcup_{w\in W_{J}}BwB.

The set WJ=W/WJW^{J}=W/W_{J} is called the set of minimal length coset representatives. Alternatively, we have (see, [2, Section 2.5])

WJ={wWw(α)>0 for all αJ}.W^{J}=\{w\in W\mid w(\alpha)>0\text{ for all }\alpha\in J\}.

The full flag variety is by definition the variety G/BG/B. The projective homogeneous space G/PJG/P_{J} is called a partial flag variety and its Bruhat decomposition is given by

G/PJ=wWJBwPJ.G/P_{J}=\bigsqcup_{w\in W^{J}}BwP_{J}.

Whenever WJW_{J} is generated by one element sαs_{\alpha} for αS\alpha\in S, we call the associated parabolic subgroup a minimal parabolic subgroup and we denote it as PαP_{\alpha}. Note that, Pα/BP_{\alpha}/B is isomorphic to 1\mathbb{P}^{1}. Whenever JJ is obtained from SS by removing one simple root αk\alpha_{k}, we call the associated parabolic subgroup a maximal parabolic subgroup and we denote it by Pαk^P_{\hat{\alpha_{k}}}. We recall that the Grassmannian variety Gr(k,n)Gr(k,n) of kk dimensional subspaces of a nn-dimensional complex vector space is isomorphic to G/Pαk^G/{P_{\hat{\alpha_{k}}}}. Let

I(k,n)={(i1,i2,,ik)|1i1<i2<<ikn}.I(k,n)=\{(i_{1},i_{2},\ldots,i_{k})|~{}1\leq i_{1}<i_{2}<\cdots<i_{k}\leq n\}.

Let w=(i1,i2,,ik)I(k,n)w=(i_{1},i_{2},\ldots,i_{k})\in I(k,n). Let e1,e2,,ene_{1},e_{2},\ldots,e_{n} be the standard basis of n\mathbb{C}^{n}. Let MiM_{i} denote the vector space spanned by e1,e2,,eie_{1},e_{2},\ldots,e_{i}. The Schubert cell C(w)C(w) in the Grassmannian is defined as

C(w)={UGr(k,n)|dim(UMij)=j,1jk}.C(w)=\{U\in Gr(k,n)|~{}\rm{dim}(U\cap M_{i_{j}})=j,1\leq j\leq k\}.

The dimension of such a Schubert cell C(w)C(w) is given by j(ijj)\sum_{j}(i_{j}-j). The Schubert variety X(w)X(w) which is the closure of C(w)C(w) in Grassmannian can be seen to be

X(w)={UGr(k,n)|dim(UMij)j,1jk}.X(w)=\{U\in Gr(k,n)|~{}\rm{dim}(U\cap M_{i_{j}})\geq j,1\leq j\leq k\}.

Let RR denote the polynomial ring [x1,x2,,xn]\mathbb{Z}[x_{1},x_{2},\ldots,x_{n}] in nn variables with degree of xix_{i} being 22. We recall that SnS_{n} acts on the variables as

σ(xi)=xσ(i).\sigma(x_{i})=x_{\sigma(i)}.

The action extends to an action of SnS_{n} on RR. A polynomial f(x1,x2,,xn)f(x_{1},x_{2},\ldots,x_{n}) in RR is symmetric if and only if

f(x1,x2,,xn)=f(σ(x1),σ(x2),,σ(xn))f(x_{1},x_{2},\ldots,x_{n})=f(\sigma(x_{1}),\sigma(x_{2}),\ldots,\sigma(x_{n}))

for all σSn\sigma\in S_{n}. The power sum symmetric polynomial pk(x1,x2,xk)p_{k}(x_{1},x_{2}\ldots,x_{k}) is defined as

pk(x1,x2,xk)=i=1nxik.p_{k}(x_{1},x_{2}\ldots,x_{k})=\sum^{n}_{i=1}x_{i}^{k}.

We recall that the subring of invariants RSnR^{S_{n}} of RR is a graded subring and is generated by symmetric polynomials. Let I{\mathcal{}{I}} denote the ideal generated by symmetric polynomials in positive degree. The power sum symmetric polynomials pk(x1,x2,,xn)p_{k}(x_{1},x_{2},\ldots,x_{n}) for 1kn1\leq k\leq n form a set of generators for I{\mathcal{}{I}}.

Let XX be a projective variety. Let H(X)=d=1nHd(X)H^{\bullet}(X)=\bigoplus\limits_{d=1}^{n}H^{d}(X) denote the cohomology ring of the variety with integer coefficients. Let A(X)=d=1nAd(X)A^{\bullet}(X)=\bigoplus\limits_{d=1}^{n}A^{d}(X) denote its Chow ring. We recall from [8, Chapter 19] that there exists a cycle map

cy:A(X)H(X).cy:A^{\bullet}(X)\longrightarrow H^{\bullet}(X).

Whenever XX is a partial flag variety the map cycy is an isomorphism (see, [8, Example 19.1.11]) and the cohomologies in odd degrees vanish. When XX is the full flag variety G/BG/B we recall

Theorem 3.1.

[6, Ehresmann] H2d(G/B)H^{2d}(G/B) has a basis consisting of classes of Schubert varieties [X(w0w)][X(w_{0}w)] where l(w)=dl(w)=d where w0w_{0} is the longest word in WW.

In [3], Borel, gave a presentation of the cohomology ring using the polynomial ring RR and the ideal II

Theorem 3.2.

[3, Borel] H(G/B)R/IH^{\bullet}(G/B)\cong R/{\mathcal{}{I}}.

The results were extended for G/PG/P, where PP is a parabolic subgroup of GG containing BB in [13]. Let JSJ\subset S such that P=PJP=P_{J}. We have WJW_{J} the subgroup of Weyl group generated by JJ as above. Since WJW_{J} is subgroup of WW it also acts on H(G/B)H^{\bullet}(G/B). Reiner–Woo–Yong show that,

Theorem 3.3.

[13][Reiner–Woo–Yong] H(G/P)H(G/B)WJH^{\bullet}(G/P)\cong H^{\bullet}(G/B)^{W_{J}}.

We observe that

H(G/P)H(G/B).H^{\bullet}(G/P)\hookrightarrow H^{\bullet}(G/B).
Remark 3.4.

Under this inclusion we recall from [13], the cohomology classes [X(w)][X(w)] where wWJw\in W^{J} lies in H(G/B)WJH^{\bullet}(G/B)^{W_{J}} and forms a basis of H(G/P)H^{\bullet}(G/P). More precisely, a basis of H2d(G/B)WJH^{2d}(G/B)^{W_{J}} consists of the Schubert classes [X(w)][X(w)], where wWJw\in W^{J} and X(w)X(w) is a codimension dd Schubert subvariety of G/PG/P. This can be thought of as a generalisation of Ehresmann’s theorem.

4. Maps from 2\mathbb{P}^{2} to G/BG/B

As in the previous section, we have G=SL(n,)G=SL(n,\mathbb{C}), BB denotes the Borel subgroup consisting of the diagonal matrices in GG. We will begin this section by proving the following:

Theorem 4.1.

There exists no nonconstant map from 2\mathbb{P}^{2} to G/BG/B.

Proof.

Let ϕ\phi be such a map and

ϕi:Hi(G/B)Hi(2)\phi^{*i}:H^{i}(G/B)\longrightarrow H^{i}(\mathbb{P}^{2})

be the map induced at the level of cohomology. We have from 3.2

H(G/B)[x1,x2,,xn]/IH^{\bullet}(G/B)\cong\mathbb{Z}[x_{1},x_{2},\ldots,x_{n}]/{\mathcal{}{I}}

where I{\mathcal{}{I}} is the proper ideal of [x1,x2,,xn]\mathbb{Z}[x_{1},x_{2},\ldots,x_{n}] consisting of elementary symmetric polynomials. We have xix_{i} lies in H2(G/B)H^{2}(G/B). In other words, degree of xix_{i} is 22. And we have,

H(2)[t]/t3.H^{\bullet}(\mathbb{P}^{2})\cong\mathbb{Z}[t]/t^{3}.

where degree of tt is 22. Since ϕ(H2(G/B))H2(2)\phi^{*}(H^{2}(G/B))\subseteq H^{2}(\mathbb{P}^{2}), we can assume

ϕ(xi)=ait\phi^{*}(x_{i})=a_{i}t

for some aia_{i}\in\mathbb{Z}. Since I{\mathcal{}{I}} is generated by power sum symmetric polynomials, we have

xi2=0\sum x_{i}^{2}=0

in H(G/B)H^{\bullet}(G/B). Thus in the image we will have,

ai2=0.\sum a_{i}^{2}=0.

Since aia_{i} are all integer we have ai=0a_{i}=0 for all ii. Therefore ϕi=0\phi^{*i}=0 for all i>0i>0. Hence, the map ϕ\phi is a constant map. ∎

Corollary 4.2.

Let HH be a reductive group and BHB_{H} be a Borel subgroup of HH. Then there is no non constant morphism from 2\mathbb{P}^{2} to H/BHH/B_{H}.

Proof.

Choose a faithful representation of HH in SL(m,)SL(m,\mathbb{C}) such that BHB_{H} maps to a Borel subgroup BB of SL(m,)SL(m,\mathbb{C}). So we get a embedding of H/BHH/B_{H} inside SL(m,)/BSL(m,\mathbb{C})/B. We now use theorem 4.1 to conclude the proof. ∎

Corollary 4.3.

A morphism from Gr(r,s)Gr(r,s) where s3s\geq 3 to G/BG/B is constant.

Proof.

Since Pic(Gr(r,s))\rm{Pic}(Gr(r,s)) is \mathbb{Z}, we have every map from Gr(r,s)Gr(r,s) to a projective variety is either finite or constant. Since 2\mathbb{P}^{2} sits inside Gr(r,s)Gr(r,s) whenever s3s\geq 3 and we have only constant morphism from 2\mathbb{P}^{2} to G/BG/B, the maps from Gr(r,s)Gr(r,s) to G/BG/B must be constant as well. ∎

Let VV be a vector space of dimension nn and

1i1<i2<<ik=n.1\leq i_{1}<i_{2}<\cdots<i_{k}=n.

be a sequence of integers. We define G(i1,i2,,ik)G(i_{1},i_{2},\ldots,i_{k}) the partial flag variety G/PG/P consisting of linear subspaces Li1,Li2,,LikL_{i_{1}},L_{i_{2}},\ldots,L_{i_{k}} of VV such that LijLij+1L_{i_{j}}\subset L_{i_{j+1}} and dim(Lij)=ij\mathrm{dim}(L_{i_{j}})=i_{j}.

Remark 4.4.

If k=2k=2 and i1=di_{1}=d we obtain G(i1,i2)G(i_{1},i_{2}) as the Grassmannian variety Gr(d,n)Gr(d,n). The full flag variety G/BG/B is obtained by choosing ij=ji_{j}=j. And any partial flag variety G/PG/P where PP contains BB can be obtained this way.

Lemma 4.5.

There exists a Gr(r,s)Gr(r,s) with s3s\geq 3 passing through each point of G/PG/P where PP is a parabolic subgroup which is not a Borel subgroup.

Proof.

Since PP is not a Borel subgroup we have n3n\geq 3. We are already done for the case of Grassmannian variety Gr(d,n)Gr(d,n). So we can assume k>2k>2 and G/P=G(i1,i2,,ik)G/P=G(i_{1},i_{2},\ldots,i_{k}). If PP is not BB then there either i1>1i_{1}>1 or i1=1i_{1}=1 and there exists a smallest jj such that ij+1>ij+1i_{j+1}>i_{j}+1. If i1>1i_{1}>1, then we have the fibers of the projection

G(i1,i2,,ik)G(i2,i3,,ik)G(i_{1},i_{2},\ldots,i_{k})\longrightarrow G(i_{2},i_{3},\ldots,i_{k})

is Gr(i1,i2)Gr(i_{1},i_{2}) with i23i_{2}\geq 3, hence we are done.

If i1=1i_{1}=1, choose the smallest jj such that ij=ji_{j}=j and ij+1>j+1i_{j+1}>j+1. If j=1j=1, ie. i2>2i_{2}>2, we have the fibres of the projection

(1) G(1,i2,,ik)G(i2,i3,,ik)G(1,i_{2},\ldots,i_{k})\longrightarrow G(i_{2},i_{3},\ldots,i_{k})

is r1\mathbb{P}^{r-1} where r=i212r=i_{2}-1\geq 2. If j2j\geq 2, then we have the fibre of

(2) G(i1,i2,,ik)G(i1,i2,,ij1,ij+1,,ik)G(i_{1},i_{2},\ldots,i_{k})\longrightarrow G(i_{1},i_{2},\ldots,i_{j-1},i_{j+1},\ldots,i_{k})

is r1\mathbb{P}^{r-1} where r=ij+1ij13r=i_{j+1}-i_{j-1}\geq 3.

This proves the lemma. ∎

Remark 4.6.

Let PP be a parabolic subgroup which is not a maximal parabolic or a Borel. The proof of the above lemma provides a Gr(r,s)Gr(r,s)-fibration G/PG/PG/P\rightarrow G/P^{\prime} for some s3s\geq 3 where PP^{\prime} is a parabolic subgroup containing PP

Corollary 4.7.

Let HH be a reductive group and BHB_{H} be a Borel subgroup of HH. Fix a parabolic subgroup PP of GG and a non constant morphism ϕ:G/PH/BH\phi:G/P\rightarrow H/B_{H}. Then PP is a Borel subgroup.

Proof.

We know that any H/BHH/B_{H} embeds inside a G/BG/B where BB is a Borel subgroup of GG. So we are reduced to the case where H=GH=G and BHB_{H} is a Borel subgroup of GG.

We assume on the contrary that PP is not a Borel subgroup. If PP is a maximal parabolic subgroup then by lemma 4.3 the map ϕ\phi must be constant.

We can therefore assume PP not a Borel or a maximal parabolic subgroup. From corollary 4.6 we obtain a parabolic subgroup PP^{\prime} containing PP such that G/PG/PG/P\rightarrow G/P^{\prime} is Gr(r,s)Gr(r,s)-fibration with s3s\geq 3. Since ϕ\phi is constant on Gr(r,s)Gr(r,s) we have ϕ\phi factors through G/PG/P^{\prime}. Repeating the argument we can assume that ϕ\phi factors through a G/QG/Q where QQ is a maximal parabolic subgroup and hence we conclude that ϕ\phi is constant.

Remark 4.8.

Shrawan Kumar [15] has extended corollary 4.7 to an arbitrary simple group GG.

5. Maps from 3\mathbb{P}^{3} to G/PG/P for a minimal parabolic subgroup

We assume the notations from the previous sections. We thus have PαP_{\alpha} the minimal parabolic subgroup BBsαBB\cup Bs_{\alpha}B. When α=α1\alpha=\alpha_{1} we will show that there is no non constant map from 3\mathbb{P}^{3} to G/PαG/{P_{\alpha}}. Since G/Pα1G/Pαn1G/P_{\alpha_{1}}\cong G/P_{\alpha_{n-1}} we conclude that there is no non constant map from 3\mathbb{P}^{3} to G/Pαn1G/{P_{\alpha_{n-1}}} as well. However, when we have any other minimal parabolic subgroup PαjP_{\alpha_{j}}, j1,n1j\neq 1,n-1 we will show that there are non constant maps from 3\mathbb{P}^{3} to G/PαjG/{P_{\alpha_{j}}}.

Fix a basis e1,e2,,ene_{1},e_{2},\ldots,e_{n} of VV. Define the subspaces MiM_{i} to the span of e1,e2,eie_{1},e_{2}\ldots,e_{i}. Let DkD_{k} denote the Schubert divisor in the Gr(k,n)Gr(k,n) which is defined as

Dk={FGr(k,n)|FMnk0}.D_{k}=\{F\in Gr(k,n)|F\cap M_{n-k}\neq 0\}.

We define the following two codimension 2 Schubert subvarieties of the Grassmannian Gr(k,n)Gr(k,n):

Dk,k+1:={FGr(k,n)|FMnk10}D_{k,k+1}:=\{F\in Gr(k,n)|~{}F\cap M_{n-k-1}\neq 0\}
Dk,k1:={FGr(k,n)|dim(FMnk+12)}.D_{k,k-1}:=\{F\in Gr(k,n)|~{}\rm{dim}(F\cap M_{n-k+1}\geq 2)\}.

We note that Dn,n+1D_{n,n+1} and D1,0D_{1,0} are empty sets. We prove the following lemmas.

Lemma 5.1.

Let 1kn1\leq k\leq n. We have the following relation in H4(Gr(k,n))H^{4}(Gr(k,n))

Dk.Dk=Dk,k1+Dk,k+1.D_{k}.D_{k}=D_{k,k-1}+D_{k,k+1}.
Proof.

To prove the lemma we would intersect the Schubert divisors fixing two different complimentary nkn-k dimensional vector subspaces. Let MnkM^{\prime}_{n-k} is the vector space generated by Mnk1M_{n-k-1} and enk+1e_{n-k+1}. Let Dk:={FGr(k,n)|FMnk0}D^{\prime}_{k}:=\{F\in Gr(k,n)|~{}F\cap M^{\prime}_{n-k}\neq 0\} be the divisor linearly equivalent to DkD_{k} defined with respect to MnkM^{\prime}_{n-k}. Then we can see that

DkDk={FGr(k,n)|FMnk0}{FGr(k,n)|FMnk0}D_{k}\cap D^{\prime}_{k}=\{F\in Gr(k,n)|~{}F\cap M_{n-k}\neq 0\}\cap\{F\in Gr(k,n)|~{}F\cap M^{\prime}_{n-k}\neq 0\}
={FGr(k,n)|FMnk10}{dim(FMnk+1)2}=\{F\in Gr(k,n)|F\cap M_{n-k-1}\neq 0\}\cup\{\rm{dim}(F\cap M_{n-k+1})\geq 2\}

which by definition is Dk,k+1Dk,k1D_{k,k+1}\cup D_{k,k-1}. Hence, the lemma follows. ∎

Lemma 5.2.

We have the following relation in the cohomology H4(G(k,k+1,n))H^{4}(G(k,k+1,n))

Dk.Dk+1=Dk,k+1+Dk+1,k.D_{k}.D_{k+1}=D_{k,k+1}+D_{k+1,k}.
Proof.

We note that the intersection of DkD_{k} with Dk+1D_{k+1} is happening at G(k,k+1,n)G(k,k+1,n). Dk+1D_{k+1} is linearly equivalent to {(F,E)G(k,k+1,n)|EMnk10}\{(F,E)\in G(k,k+1,n)|~{}E\cap M_{n-k-1}\neq 0\} in G(k,k+1,n)G(k,k+1,n). We observe that both Dk,k+1D_{k,k+1} and Dk+1,kD_{k+1,k} lie in the intersection of DkD_{k} and Dk+1D_{k+1}. If we choose a FF from the intersection not in Dk,k+1D_{k,k+1} we observe that FMnk1=0F\cap M_{n-k-1}=0 and FMnk0F\cap M_{n-k}\neq 0. Then FMnkF\cap M_{n-k} and EMnk1E\cap M_{n-k-1} are non zero and linearly independent, so they span atleast two dimensional vector space and it is contained in EMnkE\cap M_{n-k}. So we have,

{FGr(k,n)|FMnk0}{(F,E)G(k,k+1,n)|EMnk10}\{F\in Gr(k,n)|~{}F\cap M_{n-k}\neq 0\}\cap\{(F,E)\in G(k,k+1,n)|~{}E\cap M_{n-k-1}\neq 0\}
={FGr(k,n)|FMnk10}{EGr(k+1,n)|dim(EMnk)2}.=\{F\in Gr(k,n)|~{}F\cap M_{n-k-1}\neq 0\}\cup\{E\in Gr(k+1,n)|~{}\rm{dim}(E\cap M_{n-k})\geq 2\}.

Hence, the lemma follows. ∎

Let E1E_{1} denote the codimension 33 Schubert cycle defined by the Schubert variety {FGr(2,n)|FMn30andFMn1}\{F\in Gr(2,n)|~{}F\cap M_{n-3}\neq 0~{}\rm{and}~{}F\subset M_{n-1}\}. Let E2E_{2} be the codimension 33 Schubert cycle defined by the Schubert variety {FGr(2,n)|FMn40}\{F\in Gr(2,n)|~{}F\cap M_{n-4}\neq 0\}.

Lemma 5.3.

We have the following relations in H6(Gr(2,n))H^{6}(Gr(2,n)) :

  • (i)

    D2,1.D2=E1D_{2,1}.D_{2}=E_{1}.

  • (ii)

    D2,3.D2=E1+E2D_{2,3}.D_{2}=E_{1}+E_{2}.

Proof.

(i) Let Mn2′′M_{n-2}^{\prime\prime} be the n2n-2 dimensional vector space spanned by e1,e2,,en3,ene_{1},e_{2},\ldots,e_{n-3},e_{n}. Let D2′′D_{2}^{\prime\prime} be the divisor linearly equivalent to D2D_{2} defined by {FGr(2,n)|FMn2′′0}\{F\in Gr(2,n)|~{}F\cap M^{\prime\prime}_{n-2}\neq 0\}. We have FMn1F\subset M_{n-1} as it is in D2,1D_{2,1}. It follows that FMn2′′Mn1Mn2′′=Mn3F\cap M_{n-2}^{\prime\prime}\subset M_{n-1}\cap M_{n-2}^{\prime\prime}=M_{n-3} is nonzero. Therefore, FMn30F\cap M_{n-3}\neq 0.

(ii) Let Mn2′′′M_{n-2}^{\prime\prime\prime} be the n2n-2 dimensional vector space spanned by e1,e2,,en4,en2,en1e_{1},e_{2},\ldots,e_{n-4},e_{n-2},e_{n-1}. Let D2′′′D_{2}^{\prime\prime\prime} be the divisor linearly equivalent to D2D_{2} defined by {FGr(2,n)|FMn2′′′0}\{F\in Gr(2,n)|~{}F\cap M^{\prime\prime\prime}_{n-2}\neq 0\}. Let FF be in the intersection of D2,3D_{2,3} and D2′′′D^{\prime\prime\prime}_{2}. If FMn40F\cap M_{n-4}\neq 0 then FF is the component E2E_{2}. So we assume FMn4=0F\cap M_{n-4}=0. Notice that Mn4=Mn2Mn2′′′M_{n-4}=M_{n-2}\cap M_{n-2}^{\prime\prime\prime}. But on the other hand FMn20F\cap M_{n-2}\neq 0 and FMn2′′′0F\cap M^{\prime\prime\prime}_{n-2}\neq 0, therefore FF is contained in the span of Mn2M_{n-2} and Mn2′′′M_{n-2}^{\prime\prime\prime} which is Mn1M_{n-1}. Hence FF is contained in Mn1M_{n-1}, i.e FE1F\in E_{1}. Hence the lemma.

Since the map H(Gr(k,n))H^{*}(Gr(k,n)) to H(G/B)H^{*}(G/B) is injective the above relations holds in H(G/B)H^{*}(G/B) as well. We use the same notations DiD_{i} and Di,jD_{i,j} to define the Schubert classes in H(G/B)H^{*}(G/B). Note that the above relations can also be deduced from Monk’s formula.

Theorem 5.4.

There is no nonconstant map from 3\mathbb{P}^{3} to G/Pα1G/{P_{\alpha_{1}}}.

Proof.

Let P=Pα1P=P_{\alpha_{1}}. So G/PG/P is G(2,3,,n)G(2,3,\ldots,n). Let

ϕ:3G/P\phi:\mathbb{P}^{3}\longrightarrow G/P

be a map. Let

ϕ:H(G/P)H(3)\phi^{*}:H^{\bullet}(G/P)\longrightarrow H^{\bullet}(\mathbb{P}^{3})

be the map at the level of cohomology. Let

ϕi:Hi(G/P)Hi(3)\phi^{*i}:H^{i}(G/P)\longrightarrow H^{i}(\mathbb{P}^{3})

be the map at degree ii. We know that H(3)[t]/(t4)H^{*}(\mathbb{P}^{3})\cong\mathbb{Z}[t]/(t^{4}). We will show that ϕi=0\phi^{*i}=0 for all i>0i>0.

We know that the divisors in G(2,3,,n)G(2,3,\ldots,n) are D2,D3,,Dn1D_{2},D_{3},\ldots,D_{n-1}. From 5.1 and 5.2 we have the following relations in H(G/P)H^{*}(G/P)

D2D2\displaystyle D_{2}D_{2} =D2,1+D2,3\displaystyle=D_{2,1}+D_{2,3}
D2D3\displaystyle D_{2}D_{3} =D2,3+D3,2\displaystyle=D_{2,3}+D_{3,2}
D3D3\displaystyle D_{3}D_{3} =D3,2+D3,4\displaystyle=D_{3,2}+D_{3,4}
\displaystyle\;\;\vdots
Dn2Dn1\displaystyle D_{n-2}D_{n-1} =Dn2,n1+Dn1,n2\displaystyle=D_{n-2,n-1}+D_{n-1,n-2}
Dn1Dn1\displaystyle D_{n-1}D_{n-1} =Dn1,n2\displaystyle=D_{n-1,n-2}

Letting ϕ(Di)=ait\phi^{*}(D_{i})=a_{i}t in H2(3)H^{2}(\mathbb{P}^{3}) and ϕ(Di,j)=bi,jt2\phi^{*}(D_{i,j})=b_{i,j}t^{2} in H4(3)H^{4}(\mathbb{P}^{3}) we obtain the following relations in H(3)H^{*}(\mathbb{P}^{3}).

a22\displaystyle a_{2}^{2} =b2,1+b2,3\displaystyle=b_{2,1}+b_{2,3}
a2a3\displaystyle a_{2}a_{3} =b2,3+b3,2\displaystyle=b_{2,3}+b_{3,2}
\displaystyle\;\;\vdots
an12\displaystyle a_{n-1}^{2} =bn1,n2\displaystyle=b_{n-1,n-2}

So rewriting bi,jb_{i,j} in terms of ai,ja_{i,j} we obtain

bn1,n2\displaystyle b_{n-1,n-2} =an12\displaystyle=a_{n-1}^{2}
bn2,n1\displaystyle b_{n-2,n-1} =an1an2an12\displaystyle=a_{n-1}a_{n-2}-a_{n-1}^{2}
\displaystyle\;\;\vdots
bi,i1\displaystyle b_{i,i-1} =(ai2+ai+12+an12)(aiai+1+ai+1ai+2+an2an1)\displaystyle=(a_{i}^{2}+a_{i+1}^{2}+\ldots a_{n-1}^{2})-(a_{i}a_{i+1}+a_{i+1}a_{i+2}\ldots+a_{n-2}a_{n-1})
bi1,i\displaystyle b_{i-1,i} =(ai1ai+aiai+1++an2an1)(ai2+ai+12+an12)\displaystyle=(a_{i-1}a_{i}+a_{i}a_{i+1}+\ldots+a_{n-2}a_{n-1})-(a_{i}^{2}+a_{i+1}^{2}+\ldots a_{n-1}^{2})
\displaystyle\;\;\vdots
b2,3\displaystyle b_{2,3} =(a2a3+a3a4++an2an1)(a32++an12)\displaystyle=(a_{2}a_{3}+a_{3}a_{4}+\ldots+a_{n-2}a_{n-1})-(a_{3}^{2}+\ldots+a_{n-1}^{2})
b2,1\displaystyle b_{2,1} =(a22++an12)(a2a3+a3a4++an2an1).\displaystyle=(a_{2}^{2}+\ldots+a_{n-1}^{2})-(a_{2}a_{3}+a_{3}a_{4}+\ldots+a_{n-2}a_{n-1}).

Let ϕ(E1)=c1t3\phi^{*}(E_{1})=c_{1}t^{3} and ϕ(E2)=c2t3\phi^{*}(E_{2})=c_{2}t^{3}. Therefore from 5.3 we get

b2,1a2=c1b_{2,1}a_{2}=c_{1}
b2,3a2=c1+c2.b_{2,3}a_{2}=c_{1}+c_{2}.

We know Schubert classes are represented by algebraic cycles and hence their pullbacks are algebraic cycles in the projective space. Therefore, Schubert polynomials are mapped to non negative classes in the cohomology of projective spaces. So c20c_{2}\geq 0. Therefore, we obtain b2,3a2b2,1a2b_{2,3}a_{2}\geq b_{2,1}a_{2}. We have a20a_{2}\geq 0. If a2>0a_{2}>0 we observe b2,3b2,1b_{2,3}\geq b_{2,1}. Hence,

(a2a3+a3a4++an2an1)(a32++an12)(a22++an12)(a2a3+a3a4++an2an1)(a_{2}a_{3}+a_{3}a_{4}+\ldots+a_{n-2}a_{n-1})-(a_{3}^{2}+\ldots+a_{n-1}^{2})\geq(a_{2}^{2}+\ldots+a_{n-1}^{2})-(a_{2}a_{3}+a_{3}a_{4}+\ldots+a_{n-2}a_{n-1})

which implies that

(a2a3)2+(a3a4)2+(an2an1)2+a220(a_{2}-a_{3})^{2}+(a_{3}-a_{4})^{2}+\ldots(a_{n-2}-a_{n-1})^{2}+a_{2}^{2}\leq 0

which forces ai=0a_{i}=0 for all ii.

If a2=0a_{2}=0 we have b2,3=b2,1=0b_{2,3}=b_{2,1}=0. Then we obtain

a32+a42++an12=a3a4+an2an1.a_{3}^{2}+a_{4}^{2}+\ldots+a_{n-1}^{2}=a_{3}a_{4}+\ldots a_{n-2}a_{n-1}.

which implies

(a3a4)2+(a4a5)2++(an2an1)2+a32+an12=0(a_{3}-a_{4})^{2}+(a_{4}-a_{5})^{2}+\ldots+(a_{n-2}-a_{n-1})^{2}+a_{3}^{2}+a_{n-1}^{2}=0

which forces all ai=0a_{i}=0. Therefore, we conclude that ϕi=0\phi^{*i}=0 for all i>0i>0. ∎

Corollary 5.5.

There is no nonconstant map from 3\mathbb{P}^{3} to G/Pαn1G/{P_{\alpha_{n-1}}}.

Proof.

Since G=SL(n,)G=SL(n,\mathbb{C}), we have an automorphism of GG which is induced by the Dynkin involution taking αi\alpha_{i} to αni\alpha_{n-i} for all 1in11\leq i\leq n-1. Under this automorphism we have Pα1P_{\alpha_{1}} isomorphic to Pαn1P_{\alpha_{n-1}}. We have G/Pα1G/P_{\alpha_{1}} isomorphic to G/Pαn1G/P_{\alpha_{n-1}}.

Lemma 5.6.

There is a morphism from 3\mathbb{P}^{3} to G(1,3,4)G(1,3,4).

Proof.

Let VV be a vector space of dimension 44. Let 3\mathbb{P}^{3} be the projective space of lines in VV. Fix a non-degenerate alternating bilinear form. Because the form is non-degenerate and alternating it follows that for every line LL the orthogonal compliment LL^{\perp} of LL is a 33 dimensional subspace of VV containing LL. Hence, (L,L,V)(L,L^{\perp},V) is an element of G(1,3,4)G(1,3,4) and the map L(L,L,V)L\mapsto(L,L^{\perp},V) defines the required morphism. ∎

Theorem 5.7.

There are maps from 3\mathbb{P}^{3} to G/PαG/P_{\alpha} for all minimal parabolic subgroup PαP_{\alpha} with α{α1,αn1}{\alpha}\notin\{{\alpha_{1},\alpha_{n-1}}\}.

Proof.

Let α=αj\alpha=\alpha_{j} where 2jn22\leq j\leq n-2. Fix a flag

L1L2Lj2Lj+2Lj+3Ln1LnL_{1}\subset L_{2}\cdots\subset L_{j-2}\subset L_{j+2}\subset L_{j+3}\cdots\subset L_{n-1}\subset L_{n}

where dimension of Lj=jL_{j}=j. Then the fiber over this flag of the map

G(1,2,,j1,j+1,j+2,,n1,n)G(1,2,,j2,j+2,,n1,n)G(1,2,\ldots,j-1,j+1,j+2,\ldots,n-1,n)\longrightarrow G(1,2,\ldots,j-2,j+2,\ldots,n-1,n)

is isomorphic to G(1,3,4)G(1,3,4) which is identified as the flags (Lj1/Lj2,Lj+1/Lj2,Lj+2/Lj2)(L_{j-1}/L_{j-2},L_{j+1}/L_{j-2},L_{j+2}/L_{j-2}). So we have a map from G(1,3,4)G(1,3,4) to G/PαG/P_{\alpha}. And using lemma 5.6 we proof the theorem.

6. Declaration

Conflicts of interest The authors declare that they have no conflicts of interest.

References

  • [1] Gianluca Occhetta Angelo Naldi, Morphisms between Grassmannians., arXiv:2202.11411.
  • [2] Sara Billey and V. Lakshmibai, Singular loci of Schubert varieties, Progress in Mathematics, vol. 182, Birkhäuser Boston, Inc., Boston, MA, 2000. MR 1782635
  • [3] Armand Borel, Sur la cohomologie des espaces fibrés principaux et des espaces homogènes de groupes de Lie compacts, Ann. of Math. (2) 57 (1953), 115–207. MR 51508
  • [4] Wei-Liang Chow, Algebraic varieties with rational dissections, Proc. Nat. Acad. Sci. U.S.A. 42 (1956), 116–119. MR 78006
  • [5] M. Demazure, Automorphismes et déformations des variétés de Borel, Invent. Math. 39 (1977), no. 2, 179–186. MR 435092
  • [6] Charles Ehresmann, Sur la topologie de certains espaces homogènes, Ann. of Math. (2) 35 (1934), no. 2, 396–443. MR 1503170
  • [7] Xinyi Fang and Peng Ren, Morphisms from m\mathbb{P}^{m} to flag varieties, arXiv:2310.08151 (2023).
  • [8] William Fulton, Intersection theory, second ed., Ergebnisse der Mathematik und ihrer Grenzgebiete. 3. Folge. A Series of Modern Surveys in Mathematics [Results in Mathematics and Related Areas. 3rd Series. A Series of Modern Surveys in Mathematics], vol. 2, Springer-Verlag, Berlin, 1998. MR 1644323
  • [9] Robert Lazarsfeld, Some applications of the theory of positive vector bundles, Complete intersections (Acireale, 1983), Lecture Notes in Math., vol. 1092, Springer, Berlin, 1984, pp. 29–61. MR 775876
  • [10] Roberto Muñoz, Gianluca Occhetta, and Luis E. Solá Conde, Splitting conjectures for uniform flag bundles, Eur. J. Math. 6 (2020), no. 2, 430–452. MR 4104563
  • [11] Xuanyu Pan, Triviality and split of vector bundles on rationally connected varieties, Math. Res. Lett. 22 (2015), no. 2, 529–547. MR 3342245
  • [12] K. H. Paranjape and V. Srinivas, Self-maps of homogeneous spaces, Invent. Math. 98 (1989), no. 2, 425–444. MR 1016272
  • [13] Victor Reiner, Alexander Woo, and Alexander Yong, Presenting the cohomology of a Schubert variety, Trans. Amer. Math. Soc. 363 (2011), no. 1, 521–543. MR 2719692
  • [14] Luis E. Solá Conde Roberto Muñoz, Gianluca Occhetta, Maximal disjoint Schubert cycles in rational homogeneous varieties., arXiv:2211.08216.
  • [15] Kumar Shrawan, Nonexistance of regular maps between homogeneous projective varieties, arXiv:2307.07018 (2023).
  • [16] Hiroshi Tango, On (n1)(n-1)-dimensional projective spaces contained in the Grassmann variety Gr(n, 1){\rm Gr}(n,\,1), J. Math. Kyoto Univ. 14 (1974), 415–460. MR 371915
  • [17] by same author, An example of indecomposable vector bundle of rank n1n-1 on PnP^{n}, J. Math. Kyoto Univ. 16 (1976), no. 1, 137–141. MR 401766
  • [18] by same author, On morphisms from projective space PnP^{n} to the Grassmann variety Gr(n,d)G{\rm r}(n,d), J. Math. Kyoto Univ. 16 (1976), no. 1, 201–207. MR 401787
  • [19] by same author, On morphisms from projective space 𝐏n{\bf P}^{n} to the Grassmann variety Gr(n,d){\rm Gr}(n,\,d). II, Bull. Kyoto Univ. Ed. Ser. B (1984), no. 64, 1–20. MR 759949