Contents lists available atScienceDirect
Journal
of
Pure
and
Applied
Algebra
www.elsevier.com/locate/jpaa
On
higher
Gauss
maps
✩Pietro De Poia, Giovanna Ilardib,∗
a DipartimentodiMatematicaeInformatica,UniversitàdegliStudidi Udine,viadelleScienze,206,
33100Udine,Italy
bDipartimentodiMatematicaeApplicazioni,UniversitàdegliStudidiNapoli“FedericoII”,
80126 Napoli,Italy
a r t i cl e i n f o a b s t r a c t
Articlehistory:
Received8December2014 Receivedinrevisedform10April 2015
Availableonline27May2015 CommunicatedbyA.V.Geramita
MSC:
Primary:53A20;secondary:14N15; 51N35;53B99
Weprovethatthegeneralfibreofthei-thGaussmaphasdimensionm ifandonly ifatthegeneralpointthe(i+ 1)-thfundamentalformconsistsofconeswithvertex afixedPm−1,extendingaknowntheoremfortheusualGaussmap.Weprovethis viaarecursiveformulaforexpressinghigherfundamentalforms.Wealsoshowsome consequencesoftheseresults.
© 2015ElsevierB.V.All rights reserved.
1. Introduction
Let V ⊂ PN be a projective embeddedalgebraic variety and P ∈ V a point of it;with P it associates the i-th osculating space TP(i)(V ) ⊂ PN generated by the (i+ 1)-th infinitesimal neighbourhood (see e.g.
[9,Example II.3.2.5]);letdi be itsdimension.Werecallthatthei-th Gaussmap γi:V G(di,N ) isthe
rationalmapwhichassociatestothegeneralpointP itsi-thosculatingspaceTP(i)(V ).Fori= 1 thisreduces totheusualGauss map.
The study of the Gauss map has been rather intense, both in classical (see for example [5,2,13]) and modernalgebraicgeometry (seeforexample[8,1]).Instead,thestudyof higherorderGaussmap hasbeen muchmorescarce; asfaras weknow,we canonlycite theclassicalpaperofM. Castellani[3],reviewedin modernformin[7,11]andtheveryrecentpaper[6].CloselyrelatedresultshavebeenobtainedbyR. Piene onhigherdualvarieties,seeforexample[12].WenotemoreoverthatadifferentnotionofhigherGaussmap hasbeengivenbyF. Zak:seeforexample[15].
✩ TheauthorswerepartiallysupportedbyMIUR,PRIN2010–11project“Geometriadellevarietàalgebricheedeilorospazidi
moduli”.
* Correspondingauthor.
E-mailaddresses:pietro.depoi@uniud.it(P. De Poi),giovanna.ilardi@unina.it(G. Ilardi). http://dx.doi.org/10.1016/j.jpaa.2015.05.009
The main basicdifference between the first and higher Gauss maps is—as shownin[3,7,12]—thatthe fibresoftheGaussmap are(Zariskiopensubsetsof)linearspaces,whileforhigherGaussmapsthisisnot trueingeneral.
Inthispaper,thestudyofhigherGaussmapsistakenupfurther:inparticular,weanalysethe infinites-imal behavioursof these maps. Surprisingly enough,it turns out, see Theorem 3.6,that thegeneral fibre of thei-th Gaussmaphasdimensionm ifandonlyifatthegeneralpoint the(i+ 1)-thfundamentalform consistsofconeswithvertexafixedPm−1,asithappensfortheusualGauss map:see[8,(2.6)].Thisresult is obtainedbyvirtueofarecursiveformulaforexpressingthe(i+ 1)-th fundamentalformintermsofthe
i-th fundamental one:see Lemma 3.1. This last result isinteresting initsown right: forexample, wecan proveimmediatelyas acorollarythe(well-known)factthatJacobiansystemofthe(i+ 1)-thfundamental form is containedin thei-th fundamentalform. Finally,we show some resultsthatfollow from themain resultofthepaper:inparticular, asanexampleoftheconsequencesofthetheorem,westudy thevarieties whose image of ahigher Gauss map has as image a curve, describingcompletely theirvarieties of higher osculatingspaces.
2. Notationandpreliminaries
Weusenotationasin[8,9],andmostofthepreliminariesaretakenfrom[4].LetV ⊂ PN beaprojective variety of dimension k over C that will be always irreducible. For any point P ∈ V we use the following notation:TP(V )⊂ PN istheembeddedtangentprojectivespacetoV inP andTP(V ) istheZariskitangent
space.
As in[8],we abusenotation byidentifyingtheembeddedtangentspaceinPN withtheaffinecone over it inCN +1. Withthis conventionT
P(V )∼=
TP(V )
C .Wedenote by G(t,N ) theGrassmannian oft-planes of
PN.
TostudythebehaviourofV inP ,following[8](andreferences[2,6,7,10]therein),weconsiderthemanifold
F(V ) of framesinV .Anelementof F(V ) isaDarboux frame centredinP .This meansan(N + 1)-tuple
{A0; A1, . . . , Ak; . . . , AN}
whichisabasisofCN +1suchthat,ifπ :CN +1\ {0}→ PN isthecanonicalprojection,
π(A0) = P, and π(A0), π(A1), . . . , π(Ak) span TP(V ).
LetthisframemoveinF(V );thenwehavethefollowingstructureequations(intermsoftherestrictions to V oftheMaurer–Cartan1-formsωi, ωi,j onF(PN))fortheexteriorderivativesofthis movingframe
⎧ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎨ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎩ ωμ= 0 ∀μ > k dA0= k i=0ωiAi dAi= N j=0ωi,jAj i = 1, . . . , N dωj= k h=0ωh∧ ωh,j j = 0, . . . , k dωi,j= N h=0ωi,h∧ ωh,j i = 1, . . . , N, j = 0, . . . , N. (2.1)
Remark 2.1. Geometrically, the frame {Ai} definesacoordinate simplexin PN. The 1-forms ωi,ωi,j give
the rotationmatrixwhen thecoordinate simplexisinfinitesimally displaced; inparticular, moduloA0,as dA0 ∈ TP∗(PN) (the cotangent space), the1-forms ω1,. . . ,ωk give abasisfor the cotangentspace TP∗(V ),
thecorresponding π(Ai)= vi∈ TP(V ) giveabasisforTP(V ) suchthatvi istangentto thelineA0Ai,and
Wecandefinenow thet-thfundamentalform andthet-thosculatingspaceat P ∈ V ,fort≥ 2,see [8, §1.(b)and(d)]and[4,§1]:
Definition 2.2. Let P ∈ V , let t ≥ 2 be an integer and let I = (i1,. . . ,ik) be such that |I| ≤ t. The
t-th osculating space to V at P is the subspace T˜P(t)(V ) ⊂ PN spanned by A
0 and by all the derivatives d|I|A0 dvi1 1 · · · dv ik k ,wherev1,. . . ,vk spanTP(V ).
IfthepointP ∈ V isgeneral,wewillput
dt:= dim( ˜TP(t)(V )).
Remark2.3.Obviously,the“expected”dimensionofthet-thosculatingspace atageneralpointP for our
k-dimensional variety V ∈ PN is the minimum among N and k t :=
k+t
k
− 1;indeed, for the “general” variety,dt hasthis expecteddimension,but,asthenextexampleshows,thereareexceptions.
Example2.4. Letusconsider theso-called Togliatti surface S⊂ P5,whichisaparticularprojectionof the delPezzosexticsurfaceembeddedinP6,see[14]and[10];itistheclosureoftheimageoftherationalmap
ψ:P2 P5
(x : y : z) → (x2y : xy2: x2z : xz2: y2z : yz2).
FortheTogliattisurfacewehaved2= 4:thiscanbeeasilyseenfromthefactthattheparametrisationofS
given above,restrictedto theopen affineset x= 0, satisfiesthefollowing second order partial differential equation(classicallycalledLaplaceequation):
u2∂ 2x ∂u2 + uv ∂2x ∂u∂v + v 2∂2x ∂v2 − 2u ∂x ∂u− 2v ∂x ∂v + 2x = 0,
whereu= yx, v =zx,aretheaffinecoordinates,etc.
Example 2.5. It is immediate to see that for the Veronesesurface V2 ⊂ P5 we have dim ˜TP(2)(V2) = 5 for
everypointP ∈ V2: thereisnotahyperplanesectionofV2 withatriplepoint.
Definition 2.6.Lett ≥ 2 andV0 ⊆ V bethequasiprojective varietyofpoints where T˜P(t)(V ) hasmaximal
dimension.Thevariety
Tant(V ) :=
P∈V0 ˜
TP(t)(V )
iscalled thevariety ofosculating t-spacestoV .
Remark2.7.Obviouslywehave
dt≤ dt−1+ k− 1 + t t ≤ · · · ≤ t i=1 k + i− 1 i = kt.
Following [8]wegive
Definition 2.8. The t-th fundamental form of V in P is the linear system |It| in the projective space
P(TP(V ))∼=Pk−1 ofhypersurfacesof degreet definedsymbolicallybytheequations:
dtA0= 0.
Moreintrinsically,wewriteItas themap
It: Sym(t)T (V )→ Nt(V ) where Nt(V ) is the bundle defined locally as Nt
P(V ) :=
CN +1
˜
TP(t−1)(V ) and the map I
t is defined locally on eachv∈ TP(V ) as vt→ d tA 0 dvt mod ˜T (t−1) P (V ).
Example2.9.Withanexplicitcalculation(orfromExample 2.11)wecanshowthatthethirdfundamental form at thegeneral point P of theTogliatti surfaceS ⊂ P5 is given bya singlecubic inP(T
P(V ))∼=P1,
while for a “general” surface (i.e. if d2 = 5) of P5 the third fundamental form is the empty set; this is
obviously truefortheVeronesesurfaceV2⊂ P5.
Wewill denotethedimensionofthet-thfundamentalformbyΔt:
Δt:= dim(|It|),
i.e. |It|∼=PΔt.
Werecallthat(see[4,Corollary1.15]) Proposition 2.10.Wehave that
dt= dt−1+ Δt+ 1,
and viceversa:if dt= dt−1+ Δ+ 1,then thet-th fundamental formhas dimensionΔ.
Example 2.11.FortheTogliattisurfaceS⊂ P5 wehavethatd
3= 5 andd2= 4,and thereforeΔ3= 0.
Analogously, for a “general” surfaceof P5, d3 = d2 = 5, and Δ3 = −1, i.e. |III | = ∅; this is true for
examplefortheVeronesesurfaceV2⊂ P5.
From nowon,wewill supposethatourDarboux frame
{A0; A1, . . . , Ak; Ak+1, . . . , Ad2; Ad2+1, . . . , Ads; . . . , Adt; . . . , AN} (2.2)
is suchthatA0,A1,. . . ,Ads spanT˜ (s)
P (V ) foralls= 1,. . . ,t,withd1:= k.
Definition2.12.Lett≥ 1.Thet-th (projective) Gaussmap istherationalmap
γt: V G(dt,PN)
Werecallthat(seeforexample[4,Theorem1.18])
Theorem2.13.The firstdifferential of γt atP isthe(t+ 1)-thfundamental formatP .
Theideaof theproof isthefollowing:wehave,bythedefinitionofγt,that
dγPt: TPV TT˜(t)
P VG(dt,P
N),
andwe recallthatTT˜(t) P VG(dt
,PN)∼= Hom( ˜T(t)
P V,N t+1
P (V ));moreoverifwechooseaDarbouxframe asin
(2.2),wehavethatdA0∈ ˜TPV ⊂ ˜TP(t)V and
˜ TP(t)V CA0 = TP(t)V andthereforedγt P ∈ Hom(TPV ⊗ TP(t)V,NPt+1(V )).
Now,weremark that,inourDarbouxframe,wecaninterpretγtas
γt(P ) = A0∧ · · · ∧ Adt, andthereforeby(2.1), dγtP ≡ 1≤i≤dt dt+1≤j≤N (−1)dt−i+1ω i,jA0∧ · · · ∧ ˆAi∧ · · · ∧ Adt∧ Aj, mod ˜T (t) P V ;
now,abasisforTPV ⊗ TP(t)V canbeexpressed by(Aα⊗ Aμ)α=1,...,k μ=1,...,dt , and dγPt(Aα⊗ Aμ) = dt+1≤j≤N ωμ,j(Aα)Aj ∈ NPt+1(V );
ontheotherhand,forthe(t+ 1)-thfundamentalform wehave
dAμ
dvα ≡
dt+1≤j≤N
ωμ,j(Aα)Aj mod ˜TP(t)(V ).
3. Fibresof higherGauss maps
Westartbywritinghigherfundamentalformsexplicitly:chooseaDarbouxframe
{A0; A1, . . . , Ak; Ak+1, . . . , Ad2; Ad2+1, . . . , Ads; . . . , Adt; . . . , AN}
suchthatA0,A1,. . . ,Ads spanT˜ (s)
P (V ) for alls= 1,. . . ,t,with d1:= k, T˜P(1)(V )= ˜TP(V ) (and d0 := 0).
Weuseindices1≤ i(1),i(1)1 ,. . . ,i(1)h ≤ k = d1,ds−1+ 1≤ i(s)≤ ds fors= 2,. . . t anddt+ 1≤ i(t+1)≤ N.
Withthesenotations,thesecond fundamentalform isgivenbythequadrics
Vi(2)() = i(1)1 ,i (1) 2 qi(1) 1 ,i (1) 2 ;i()ωi (1) 1 ωi (1) 2 , (3.1) with> k,whereq i(1)1 ,i (1) 2 ;i() (= q i(1)2 ,i (1) 1 ;i() ) aredefinedby
ωi(1) 1 ,i()= i(1)2 qi(1) 1 ,i (1) 2 ;i()ωi (1) 2 , (3.2)
whichareobtainedviatheCartanlemma from 0 = dωi()= i(1)1 ω i(1)1 ∧ ωi (1) 1 ,i(), sinceωi()= 0 onT˜P(V ).
Thehigherfundamentalformscanbeexpressed asinthefollowing Lemma3.1.The (s+ 1)-th fundamental formisgivenby thepolynomials
Vi(s+1)() = k i(1)1 ,...,i (1) s+1=1 qi(1) 1 ,...,i (1) s+1;i()ωi (1) 1 · · · ωi (1) s+1, (3.3)
with ≥ s+ 1,whichinductively satisfy therelations
ds i(s)=d s−1+1 qi(1) 1 ,...,i (1) s ;i(s)ωi(s),i()= k i(1)s+1=1 qi(1) 1 ,...,i (1) s+1;i()ωi (1) s+1, (3.4) where qi(1) 1 ,...,i (1)
s ;i(s) are the coefficients of the s-th fundamental form (and with thenatural symmetries of
theindices:q...,i(1) j ,...,i (1) k ...;i() = q...,i(1) k ,...,i (1) j ,...;i()
),andthebasisof theinduction isthesecond fundamental form (i.e.s= 1)forwhichrelations(3.3)and (3.4)aretobe readas, respectively,(3.1)and(3.2).
Proof. BythedefinitionofT˜P(s)(V ),where s∈ {1,. . . ,t− 1},we havethat
dAi(s−1) ≡ 0 mod ˜TP(s)(V ) (3.5)
and from(2.1)wecanwrite
dAi(s−1) =
N
j=ds−1+1
ωi(s−1),jAj
from whichwededuce,by(3.5)
ωi(s−1),i(s+1) = ωi(s−1),i(s+2) =· · · = ωi(s−1),i(t) = ωi(s−1),i(t+1) = 0. (3.6) These equationsimplythat
0 = dωi(s−1),i()=
i(s)
ωi(s−1),i(s)∧ ωi(s),i() ≥ s + 1;
now, bytheinductivehypothesis,wehave i(s−1) qi(1) 1 ,...,i (1) s−1;i(s−1)ωi(s−1),i() = i(1)s qi(1) 1 ,...,i (1) s ;i()ωi(1)s (3.7)
whereqi(1) 1 ,...,i
(1)
s−1;i(s−1) arethecoefficientsofthe(s− 1)-thfundamentalformandqi(1)1 ,...,i (1)
s ;i()arethoseof
thes-th fundamentalform (with theconventionthatqi(1) 1 ,...,i
(1)
s ;i() = 1 if s≤ 1,and ifs= 2, (3.7)is just (3.2)i.e.theCartan lemmausedtoobtainthesecondfundamentalform(3.1))andthen,applyingrelations
(3.7)to(a linearcombination of)(3.6),wededuce i(s−1),i(s) qi(1) 1 ,...,i (1) s−1;i(s−1)ωi(s−1),i(s)∧ ωi(s),i()= i(1)s ,i(s) qi(1) 1 ,...,i (1) s ;i(s)ωi(1)s ∧ ωi(s),i()
then,bytheCartan lemma i(s) qi(1) 1 ,...,i (1) s ;i(s)ωi(s),i()= i(1)s+1 qi(1) 1 ,...,i (1) s+1;i()ωi (1) s+1;
andthe(s+ 1)-th fundamentalform is
Vi(s+1)() = i(1)1 ,...,i (1) s+1 qi(1) 1 ,...,i (1) s+1;i()ωi (1) 1 · · · ωi (1) s+1. 2
Remark3.2. Alternatively,we can write—for example writingback relations (3.4)in (3.3) or by adirect computation
Vi(s+1)() =
i(1),...,i(s)
ωi(1)ωi(1),i(2)ωi(1),i(3)· · · ωi(s−1),i(s)ωi(s),i(),
with≥ s+ 1 andthesumvaries,as above,for1≤ i(1)≤ k = d
1,. . . ,ds−1+ 1≤ i(s)≤ ds.
Werecall
Definition3.3. LetΣ bethelinearsystemofdimension d ofhypersurfacesof degreen> 1 inPN (N > 1),
generatedbythed+ 1 hypersurfacesf0= 0,. . . ,fd= 0.TheJacobiansystem J (Σ) ofΣ isthelinearsystem
definedbythepartial derivativesoftheforms f0,. . . ,fd:
J (Σ) := (∂fi/∂xj) i = 0, . . . , d; j = 0, . . . , r.
Remark 3.4. Obviously, theJacobian system J (Σ) does not depend on the choice of f0,. . . ,fd, butonly
on Σ.
Corollary 3.5. The (s+ 1)-th fundamental form is a linear system of polynomials of degree s+ 1 whose
Jacobiansystemis containedinthes-thfundamental form.
Proof. Wehave,byLemma 3.1
∂Vi(s+1)() ∂ωi(1) j = i(1)1 ,...,i (1) s ,i(s) qi(1) 1 ,...,i (1) s ;i()ωi(1)1 · · · ωi (1) s ∂ωi(s),i() ∂ωi(1) j . 2 (3.8)
Wearenowabletoprovethefollowing fact(see[8,(2.6)])
Theorem 3.6. The s-th Gaussmap γs (s≥ 1) has fibres of dimension m if and only if at a general point
P ∈ V theformsofdegrees+ 1 ofthe(s+ 1)-thfundamental form|Is+1| areconesoveraPm−1⊂ PT P(V ).
Proof. We chooseaDarbouxframeas above,i.e.
{A0; A1, . . . , Ak; Ak+1, . . . , Ad2; Ad2+1, . . . , Ads; . . . , Adt; . . . , Adt+1; . . . , AN}
suchthatA0,A1,. . . ,Ads spanT˜ (s)
P (V ) andindicesds−1+ 1≤ i(s)≤ dsforalls= 1,. . . ,t+ 1,withd1:= k,
etc.
Incoordinates,thes-thGauss mapγs is
γs(P ) = A0∧ · · · ∧ Ads,
and thereforeby(2.1), itsdifferential, whichisthe(s+ 1)-thfundamentalform,isgivenby
dγPs ≡ 1≤i≤ds ds+1≤j≤N (−1)ds−i+1ω i,jA0∧ · · · ∧ ˆAi∧ · · · ∧ Ads∧ Aj, mod ˜T (s) P V ;
therefore, sinceγshasfibres ofdimensionm if andonly ifdγs
P hasrankk− m (recallthatP isageneral
point) γshasfibresofdimensionm ifandonlyifthespace
U∗:=ωi,j 1≤i≤ds
ds+1≤j≤N
⊂ T∗ P(V )
has dimension k− m. Dually, this space defines a subspace U ⊂ TP(V ) of dimension m, defined by the
equations
ωi(h),i() = 0, h = 1, . . . , s; ≥ s + 1.
Now, we prove that Vi(s+1)() are cones with vertex P(U): let us suppose to choose a frame such that
ωm+1,. . . ,ωk form abasisforU∗,thatis
ωi,j 1≤i≤ds
ds+1≤j≤N
=ωm+1, . . . , ωk. (3.9)
Moreover, wechooseindices 1≤ j1(1),. . . ,jh(1)≤ m,andm+ 1≤ k(1)1 ,. . . ,k(1)h ≤ k.
From (3.4),thepointsofU satisfyalso 0 = i(h) qi(1) 1 ,...,i (1) h ;i(h)ωi (h),i()= i(1)h+1 qi(1) 1 ,...,i (1) h+1;i()ωi (1) h+1, (3.10)
whichimplies,foreverychoiceoftheindicesasabove,that
qi(1) 1 ,...,j
(1) h+1;i()
= 0. From thisweinfer,from (3.3), that
Vi(s+1)() = k(1)1 ,...,k (1) s+1 qk(1) 1 ,...,k (1) s+1;i()ωk (1) 1 · · · ωk (1) s+1, (3.11) thatis,V(s+1)
i() isaconewithvertex P
Viceversa,let ussuppose thatthe polynomials Vi(s+1)() aresingular alongaPm−1 =P(U);by choosing ourframe,wecansupposethatU isdefinedby
ωj = 0, j = m + 1, . . . , k,
orthat(3.9)holds.Then,we have
0 = dωi(h),i()= ds i=1 ωi(h),i∧ ωi,i()+ N j=ds+1 ωi(h),j∧ ωj,i() (3.12)
and ωi,i(),ωi(h),j ∈ U∗ and thereforethe Frobeniusintegrability conditionsfor ωm+1,. . . ,ωk are satisfied; then,from dAi(s) = N j=ds+1 ωi(s),jAj
we conclude that the s-th osculating space remains constant along the leaves of the foliation defined by
ωm+1,. . . ,ωk. 2
Remark3.7.Aswerecalledintheintroduction,onecandeduce,fromTheorem 3.6,that,fors= 1 (i.e.the usual Gaussmap), thefibres ofγ1 are(Zariskiopensubsets of)linearspaces: see[8, (2.10)]. Moreover,in [8,§3]they giveageneraldescriptionofthese varieties;we will pursuefurtherthis kindofdescription for highers inSubsection 3.1.
Example 3.8. Thefirst nontrivial exampleof applicationof the precedingtheorem with s≥ 2 is given by surfaceswhosesecondGaussmapγ2hasas imageacurve;thiscasehasbeenstudiedin[3]andin[7]:they proved thatthe curveswhich are fibres of γ2 are indeedcontained in3-dimensional projective spaces; by Theorem 3.6,wededucethatthethirdfundamentalform atthegeneralpointisgivenbyaperfectcube.
Letusseethefollowing consequenceoftheprecedingresult.
Corollary 3.9. If thes-th Gaussmapγs (s≥ 1)for ak-dimensionalvariety V has fibresof dimension m,
then ds+1≤ ds+ k− m + s s + 1 , (3.13)
andtheequalityisreachedifandonlyifthe(s+ 1)-thfundamental form|Is+1| isgivenbythelinearsystem
of allthecones overaPm−1⊂ PTP(V ).
Proof. By Proposition 2.10 we have to estimate the dimension Δs+1 of the linear system given by the
(s+ 1)-th fundamental form, butthis, by Theorem 3.6, is formed bydegree (s+ 1) hypersurfaces which are cones over a Pm−1 in a Pk−1(= P(TP(V ))). Since the linear systems of these cones has (projective)
dimensionk−m+ss+1 −1 (isequaltothecompletelinearsystemofdegree(s+ 1) hypersurfacesinaPk−m−2), wehavetheassertion. 2
3.1. Varietieswith acurveastheimage of thes-thGaussmap
Letus startwiththefollowing
Lemma 3.11. If ds+1 = ds+ 1, then ds+2 ≤ ds+ 2. Moreover, if ds+2 = ds+ 2, then |Is+1| and|Is+2| are
generatedbypowersofthesamelinearform,andifds+ 1= ds+2,thenourvarietyV iscontainedinPds+2.
Proof. We apply Proposition 2.10 to deducethat the(s+ 1)-th fundamentalform |Is+1| is generated by onlyoneform.Callthisform f ;byCorollary 3.5,ifg isaform in|Is+2|, thenallitspartialderivativesare
in|Is+1|;therefore,byEuler’sformula
g = 1 s + 2 k i=1 ωi ∂g ∂ωi = 1 s + 2 k i=1 ωiaif
ai ∈ C,i= 1,. . . ,k,andwededucethateitherg iszeroorg = f ,where is alinearform.Ifg = f ,bya
changeofcoordinateswecansupposethat= ω1;then,wehave ∂g ∂ω1 = f + ω1 ∂f ∂ω1 ∈ |I s+1| = (f) ∂g ∂ωj = ω1 ∂f ∂ωj ∈ |I s+1| = (f) j = 2, . . . , k
whichimplythateitherf doesnotdependonω1,buttheng isthezeroform,orf isreducibleoftheform f = ω1h.Wecannowproceedbyinductiontoprovethatf = ω1s+1:iff = ω1φ,then
∂g ∂ω1 = ( + 1)ωφ + ω+11 ∂φ ∂ω1 = nω1φ ∂g ∂ωj = ω1+1 ∂φ ∂ωj = njω1φ j = 2, . . . , k
n,nj ∈ C,j = 1,. . . ,k, and,as above,eitherφ doesnotdependonω1,buttheng isthezeroform,or φ is
reducible oftheformφ= ω1φ.
Ifg = 0, thenwecanconcludefrom Theorem 3.6(or, better,from Example 3.10). 2
Corollary3.12.Ifds+2= ds+1+ 1= ds+ 2,thenthes-thand(s+ 1)-thGaussmapshavefibresofdimension
k− 1.
Proof. ByLemma 3.11wehavethatthe|Is+1| and|Is+2| areconesoveraPk−2,therefore,byTheorem 3.6,
thethesisfollows. 2
Thecaseweareinterestedinnowistheonewithm= k− 1 (i.e. thenextafterExample 3.10),therefore, by(3.13),(ds≤)ds+1≤ ds+ 1;ifds+1< ds+ 1,thends+1= ds,butinthiscasewewouldhavenoelements
in |Is+1|, and thereforewe would be in Example 3.10. We deduce thatd
s+1 = ds+ 1, and we are inthe
caseof Lemma 3.11, andbyTheorem 3.6theimage ofthes-thGauss map hasdimension 1;the complete description ofsuch caseswill followfrom [8, (2.4)].To usethis result,we needsomenotations andresult: we startbyrecallingthenotationintroducedin[8,§2(a)];letB beanr-dimensionalvarietyand let
beamorphism;lety∈ B beageneralpoint,andletSy ⊂ V (wherePN=P(V ))bethe(d+ 1)-dimensional
vectorspacewhichcorresponds tof (y).Then,thedifferentialoff iny can bethoughtof
dyf : TyB → Hom(Sy, Ny),
where Ny := SVy. Moreexplicitly, ify1,. . . ,yr are local coordinatesofB near y,and ife0(y),. . . ed(y) is a
basisforthe (d+ 1)-dimensionalvectorspacesnearSy, wehave
dyf ∂ ∂yi (ej(y))≡ ∂(ej(y)) ∂yi mod Sy
fori= 1,. . . ,r andj = 0,. . . , d.Then, fixedw∈ TyB,onecandefinethe“infinitelynear”space
Sy
dw := Im dyf (w)⊂ Ny,
andifwedenoteby[v] theclassof v∈ V inNy, weconsiderthefollowingsubspacesof V :
Sy+ Sy dw := v∈ V | v = s + t, s ∈ Sy, [t]∈ Sy dw Sy∩ Sy dw := s∈ Sy| ([0] =)[s] ∈ Sy dw = ker(dyf (w)). Obviously,we have rk(dyf ) = ρ ⇐⇒ dim(Sy+ Sy dw) = d + 1 + ρ ⇐⇒ dim(Sy∩ Sy dw) = d + 1− ρ.
WealsodenotetheprojectivesubspacesofPN associatedtothevectorspacesjustdefinedby
Pd y+ Pd y dw :=P(Sy+ Sy dw) Pd y∩ Pd y dw :=P(Sy∩ Sy dw).
Thecaseofr = 1 iscompletely solvedin[8];asinthecitedarticle,withabuseofnotationwedenote by
y thelocalcoordinateofB nearourgeneralpoint y:
Proposition 3.13.(See[8, (2.4)].) Withnotationsas above,if r = 1 anddim(Pd y∩
Pd y
dw)= d− ρ,thenthere
existρ curves inPN,C
1,. . . ,Cρ parametrised byy, a1,. . . ,aρ ∈ N, andafixed (d−ρi=1ai)-dimensional
subspace H ⊂ PN such that Pd
y is the span of H together with the (ai − 1)-osculating (ai − 1)-planes
˜ T(ai−1) P (Ci)⊂ PN,that is Pd y= ˜T (a1−1) P (C1), . . . , ˜T (aρ−1) P (Cρ), H.
Letusnowusethis resultinourcase:we takeB astheimage ofthes-thGaussmap:
γs: V → Im(γs) =: B→ G(di s, N ),
dim B = 1 and Sy denotes just the ds-dimensionalvector subspace T˜P(s)(V ) ⊂ CN +1, where y = γs(P ),
havethatB definesalsoaruledvariety,whichisindeedthevarietyofosculatings-spacestoV :Tans(V ):=
∪y∈BPds
y ofdimensionds+ 1.ByCorollary 3.12,ifT˜P(s)(V )= ˜T
(s+1)
P (V ),thesamehappenswithT˜
(s+1)
P (V ):
we have,on B, theSy’s,whichare theds+1-dimensionalvector subspacesT˜P(s+1)(V )⊂ CN +1,they define
the(ruled)varietyofosculating(s+ 1)-spacestoV Tans+1(V ):=∪y∈BPdys+1 ofdimensionds+ 2 suchthat
V ⊂ Tans(V )⊂ Tans+1(V ).
Since, bydefinitionofthe(s+ 1) osculating space,wehavethat,fory∈ B,
Pds y + Pds y dy ⊂ P ds+1 y , whichimplies dim(Pds y ∩ Pds y dy ) = ds− 1. Then, byProposition 3.13
Proposition 3.14. If the s-th Gauss map has dimension one, then there exists a curve C ⊂ PN having
(a− 1)-osculating(a− 1) planes Pa−1y := ˜Ty(a−1)(C) andafixed PN−a−1 suchthat
Pds
y =Pa−1y +PN−a−1.
In otherwords,Tans(V ),istheconeoverPN−a−1 ofTana−1(C) (andthesameholdstrueforhigherGauss
maps).
Clearly, onecandeducemoregeneralresultson thevarietiesof osculatings-spacesinthe sameveinas at theendof[8,§3].
Acknowledgement
Wewouldlike tothanktheanonymousrefereeforvaluablesuggestionsand remarks. References
[1]M.A.Akivis,V.V.Goldberg,DifferentialGeometryofVarietieswithDegenerateGaussMaps,CMSBooksMath.,vol. 18, Springer-Verlag,NewYork,2004.
[2]E.Bertini,Introduzioneallageometriaprojettivadegliiperspaziconappendicesullecurvealgebricheelorosingolarità, E.Spoerri,Pisa,1907.
[3]M.Castellani,Sullesuperficieicuispaziosculatorisonobiosculatori,Rom.Acc.L.Rend.(5)31 (1)(1922)347–350. [4]P.DePoi,R.DiGennaro,G.Ilardi,Onvarietieswithhigherosculatingspaces,Rev.Mat.Iberoam.29 (4)(2013)191–1210. [5]P.delPezzo,Suglispazi tangentiadunasuperficieoadunavarietàimmersainunospaziodipiúdimensioni,Nap.Rend.
XXV(1886)176–180.
[6]S.DiRocco,K.Jabbusch,A.Lundman,AnoteonhigherorderGaussmaps,arXiv:1410.4811[math.AG]. [7]D.Franco,G.Ilardi,Onmultiosculatingspaces,Commun.Algebra29 (7)(2001)2961–2976.
[8]P.Griffiths,J.Harris,Algebraicgeometryandlocaldifferentialgeometry,Ann.Sci.Éc.Norm.Super.(4)12 (3)(1979) 355–452.
[9]R.Hartshorne,AlgebraicGeometry,Grad.TextsMath.,vol. 52,Springer-Verlag,NewYork,1977. [10]G.Ilardi,Togliattisystems,OsakaJ.Math.43 (1)(2006)1–12.
[11]J.M.Landsberg,Onsecondfundamentalformsofprojectivevarieties,Invent.Math.117 (2)(1994)303–315.
[12]R.Piene,Anoteonhigherorderdualvarieties,withanapplicationtoscrolls,in:Singularities,Part2,Arcata,CA,1981, 1983,pp. 335–342.
[13]C.Segre,Preliminaridiunateoriadellevarietàluoghidispazi,PalermoRend.30(1910)87–121.
[14]E.Togliatti,Alcuniesempîdisuperficiealgebrichedegliiperspazîcherappresentanoun’equazionediLaplace,Comment. Math.Helv.1(1929)255–272.
[15]F.L.Zak,TangentsandSecantsofAlgebraicVarieties,Transl.Math.Monogr.,vol. 127,AmericanMathematicalSociety, Providence,RI,1993,translatedfromtheRussianmanuscriptbytheauthor.