On the ne structure and hierarchy of gradient catastrophes for multidimensional homogeneous Euler equation B.G.Konopelchenko1and G.Ortenzi2

2025-05-02 0 0 1.97MB 22 页 10玖币

侵权投诉

On the ﬁne structure and hierarchy of gradient catastrophes for

multidimensional homogeneous Euler equation

B.G.Konopelchenko 1and G.Ortenzi 2∗

1Dipartimento di Matematica e Fisica “Ennio De Giorgi”, Universit`a del Salento, 73100 Lecce, Italy

2Dipartimento di Matematica e Applicazioni, Universit`a di Milano-Bicocca, via Cozzi 55, 20126 Milano, Italy

2INFN, Sezione di Milano-Bicocca, Piazza della Scienza 3, 20126 Milano, Italy

October 11, 2022

Abstract

Blow-ups of derivatives and gradient catastrophes for the n-dimensional homogeneous Euler equation are discussed. It is

shown that, in the case of generic initial data, the blow-ups exhibit a ﬁne structure in accordance of the admissible ranks

of certain matrix generated by the initial data. Blow-ups form a hierarchy composed by n+ 1 levels with the strongest

singularity of derivatives given by ∂ui/∂xk∼ |δx|−(n+1)/(n+2) along certain critical directions. It is demonstrated that

in the multi-dimensional case there are certain bounded linear superposition of blow-up derivatives. Particular results for

the potential motion are presented too. Hodograph equations are basic tools of the analysis.

1 Introduction

Multidimensional quasi-linear partial diﬀerential equations and systems have been studied intensively during last decades.

The homogeneous n-dimensional Euler equation

ut+u· ∇u= 0 ,u:Rn→Rn(1.1)

is one of the most important representative of this class of equations. It is the basic equation in the theory of continuous

media in the case when eﬀects of dissipation, pressure etc. are negligible (see e.g. [14, 13, 20]). Equation (1.1) arises in

various branches of physics from hydrodynamics to astrophysics [14, 13, 20, 21, 17].

A remarkable feature of the homogeneous Euler equation (HEE) is that it is solvable by the multidimensional version

of the classical hodograph method [21, 17, 4, 5, 6]. Namely, solutions of the HEE are provided by the vector hodograph

equation

xi−uit−fi(u) = 0 , i = 1,...,n (1.2)

where fi(u) are the local inverse of the initial data ui(x,0) for equation (1.1). The hodograph equation (1.2) is also a

powerful tool for the analysis of the singularities of solutions for HEE, in particular, of blow-ups of derivatives and gradient

catastrophes [21, 17, 4, 11, 10]. It was demonstrated that features and properties of blow-ups and gradient catastrophes

(GC) for the multidimensional HEE are quite diﬀerent from those in the text-book one-dimensional examples [14, 20].

Existence or nonexistence of blow-ups in diﬀerent dimensions, boundedness of certain linear superpositions of derivatives,

the case of potential ﬂows have been discussed brieﬂy recently in paper [10].

In this paper the detailed analysis of the blow-ups and GC for HEE (1.1) is presented. It is shown that the blow ups

of derivatives, which occur on the hypersurface Γ deﬁned by the equation

det(M(t, u)) = 0 ,where Mil =tδil +∂fi

∂ul

, i, l = 1,...,n, (1.3)

form a hierarchy. Degree of singularities of derivatives and dimensions of subspaces in Γ at which they occur are diﬀerent

for diﬀerent members (levels) of such hierarchy.

Blow-up of the ﬁrst level have a ﬁne structure due to the diﬀerent admissible ranks r1of the degenerate matrix M.

It is shown that there are r1-dimensional subspaces in the spaces of variation δukand δxisuch that the corresponding

derivatives ∂uk

∂xiare bound. The rest of the derivatives blow-up on Γ as

∂uk

∂xi∼−1/2,  →0,(1.4)

∗Corresponding author. E-mail: giovanni.ortenzi@unimib.it, Phone: +39(0)264485725

arXiv:2210.03939v1 [nlin.SI] 8 Oct 2022

where is the distance from the catastrophe point in the most singular direction. Admissible values of the rank r1and the

dimensions dim Γr1of the corresponding subspaces in Γ are calculated for generic functions f(u), i.e. for generic initial

data u(x,0). For instance, for dimension n= 3, the admissible ranks r1are r1= 2,1 and correspondingly dim Γr1=2 = 3

and dim Γr1=1 = 0, while for n= 8, one has r1= 7,6,5 and dim Γr1= 8,5,0. On the other side, in the particular case of

potential ﬂows, dim Γr1assumes others values. This case is analysed in the paper too.

It is shown that in the generic case the hierarchy of blow-ups has n+ 1 levels. On the m-th level there are also the

subsections of smaller dimensions where superpositions of derivatives are bounded while the blowing-up derivatives behave

as ∂uk

∂xi∼−m/(m+1) ,  →0.(1.5)

So, the most singular behavior of derivatives for HEE in the generic case is given by

∂uk

∂xi∼−(n+1)/(n+2) ,  →0.(1.6)

Non-generic and Poincar´e cases for initial data are discussed too. Characteristic features of blow-ups of derivatives for

the two- and three-dimensional HEE are considered in details. Some concrete cases are analyzed. The results presented

in this paper conﬁrm an expected great diﬀerence between the properties of the one-dimensional and multi-dimensional

homogeneous Euler equations.

The paper is organized as follows. In Section 2 blow-ups of derivatives of the ﬁrst level are analyzed. Admissible

ranks of the matrix M(t, u) and corresponding dim Γr1are calculated in Section 3. Non-generic and Poincar´e cases are

considered in Section 4. The situation with potential motion is discussed in section 5. Blow-ups of higher level and

hierarchy are described in Section 6 for the maximal rank case and in Section 7 for the lower rank case. Two-dimensional

and three-dimensional cases are described in Sections 8 and 9, respectively. An explicit 2D example has been carried out

in Section 10.

2 Generic blow-ups of derivatives: ﬁrst level

The hodograph equations (1.2) will be our basic tool in the study of blow-ups of derivatives and gradient catastrophes for

the HEE (1.1).

Diﬀerentiating (1.2) with respect to xk, one obtains [6, 11, 10]

l=1

Mil

∂ul

∂xk

=δik ,with Mil =tδil +∂fi

∂ul

, i, k, l = 1,...,n, (2.1)

where δik is the Kronecker symbol. This relation implies that the derivatives ∂ul

∂xkand also ∂ul

∂t become unbounded at the

hypersurface Γ deﬁned by the equation

det(M(t, u)) = 0 .(2.2)

The degeneracy of the matrix Mis the central point in the analysis of blow-ups. General properties of the hypersurface

(2.2) and some of their implications have been discussed in [10].

For deeper study of blow-ups one needs to consider the inﬁnitesimal version of the hodograph mapping (1.2) around a

point u0on the blow-up hypersurface Γ (2.2) at ﬁxed time t, i.e. with the expansion

δxi=tδui+X

∂fi

∂uk

(u0)δuk+1

k,l

∂2fi

∂uk∂ul

(u0)δukδul+1

k,l,m

∂3fi

∂uk∂ul∂um

(u0)δukδulδum+. . .

Mik(u0)δuk+1

k,l

∂2fi

∂uk∂ul

(u0)δukδul+1

k,l,m

∂3fi

∂uk∂ul∂um

(u0)δukδulδum+. . . , I = 1,...,n.

(2.3)

For variable tone has the expansions (2.3) with the substitution δx→δ(x−u0t) in the l.h.s. after the subtraction of

the contribution from the inﬁnitesimal Galilean transformation.

An important feature of the multi-dimensional case is the invariance of HEE (1.1) under the group SO(n) of rotations

in Rn. It is easy to see that in order to ensure that the variables (u1,...,un) are the component of the n-dimensional

vector uit is suﬃcient to require that the functions f(u) = (f1(u),...,fn(u)) in (1.2) are transformed as the component

of a vector. In such a case Mik are components of a tensor and the hodograph equations (1.2), condition (2.2) and (2.3)

are invariant to the group of rotations SO(n) too. Consequently, it would be preferable to formulate the basic properties

of blow-ups in a covariant form.

Another important peculiarity of the multi-dimensional HEE is that the terms linear in δudo not disappear in the

expansion (2.3) in contrast to the one-dimensional case. The coeﬃcients Mik(u0) do not vanish on Γ, however they are

very special. They are elements of the degenerate n×nmatrix M, which is a key element for the subsequent analysis.

Let us assume that the rank of the matrix Mat the point u0∈Γ is equal to r1. It means that (see e.g. [7]) there exist

n−r1vectors R(α)(u0) and L(α)(u0) with α= 1,...,n−r1such that

k=1

Mik(u0)R(α)

k= 0 , i = 1, . . . , n , α = 1,...,n−r1,(2.4)

and n

i=1

L(α)

iMik(u0) = 0 , k = 1, . . . , n , α = 1,...,n−r1.(2.5)

It is well known (see e.g. [7]) that the requirements that the matrix Mhas rank r1and that there are n−r1linearly

independent vectors R(α)and L(α)satisfying the conditions (2.4) and (2.5) are equivalent. An advantage of conditions

(2.4) and (2.5) is that they are invariant under the rotation in Rn.

The properties (2.4) and (2.5) of the matrix M(u) have immediate consequences for the expansion (2.3). First, the

relations (2.4) imply that for the variations δuof the form

δ+uk=

n−r1

α=1

R(α)

kδaα,(2.6)

where δaαare arbitrary inﬁnitesimals, one has

k=1

Mik(u)δ+uk= 0 , i = 1,...,n. (2.7)

This fact suggests us to introduce r1linearly independent vectors ˜

R(β),β= 1,...,r1complementary to the vectors R(α),

α= 1,...,n−r1such that arbitrary variation δucan be decomposed as

δu=

n−r1

α=1

R(α)δaα+

β=1

R(β)δbβ,(2.8)

where δbβare arbitrary inﬁnitesimals. Substituting (2.8) into (2.3) one gets

δx=

β=1

Qβ

(01)δbβ+

β1,β2=1

Qβ1β2

(02) δbβ1δbβ2+

n−r1

α1,α2=1

Qα1α2

(20) δaα1δaα2+

n−r1

α=1

β=1

Qαβ

(11)δaαδbβ+. . . , (2.9)

where

Qβ

(01) i≡

k=1

Mik(u0)˜

R(β)

Qα1α2

(20) i≡1

k,l=1

∂2fi

∂uk∂ul

(u0)R(α1)

kR(α2)

Qβ1β2

(02) i≡1

k,l=1

∂2fi

∂uk∂ul

(u0)˜

R(β1)

k˜

R(β2)

Qαβ

(11) i≡1

k,l=1

∂2fi

∂uk∂ul

(u0)R(α)

k˜

R(β)

. . .

Qα1...αrβ1...βs

(rs)i≡1

(r+s)!

k1,...,kr,

l1...,ls=1

∂r+sfi

∂uk1. . . ∂ukr∂ul1. . . ∂uls

(u0)R(α1)

k1. . . R(αr)

R(β1)

l1. . . ˜

R(βs)

ls, r +s≥2

i= 1, . . . , n , α, α1, α2,··· = 1,...,n−r1, β, β1, β2,··· = 1,...,r1.

(2.10)

One observes that the variables δaαand δbβbehave diﬀerently at the blow-up points.

Further, the existence of the vectors L(α),α= 1,...,n−r1with the property (2.5) suggests to introduce new variables

δyα=L(α)·δx, δ˜yγ=˜

L(γ)·δx, α = 1,...,n−r1, γ = 1,...,r1(2.11)

where (L(1),...,L(n−r1),˜

L(1),...,˜

L(r1)) form a basis of Rn.

Note that the transformation from the variables δxito the variables δyα, δ ˜yγis invertible.

Taking the scalar product of (2.9) with L(α)and ˜

L(γ), one obtains respectively the expansions

δyα=

β1,β2=1 Q(α)β1β2

(02) δbβ1δbβ2+

n−r1

α1,α2=1 Q(α)α1α2

(20) δaα1δaα2+

n−r1

α=1

β=1 Q(α)αβ

(11) δaαδbβ+. . . ,

α= 1,...,n−r1

(2.12)

and

δ˜yγ=

β=1

Q(γ)β

(01) δbβ+

β1,β2=1

Q(γ)β1β2

(02) δbβ1δbβ2+

n−r1

α1,α2=1

Q(γ)α1α2

(20) δaα1δaα2+

n−r1

α=1

β=1

Q(γ)αβ

(11) δaαδbβ+. . . , γ = 1,...,r1

(2.13)

where the coeﬃcients in (2.12) and (2.13) are deﬁned as

Q(α)α1...αrβ1...βs

(rs)≡Lα·Qα1...αrβ1...βs

(rs),

Q(γ)α1...αrβ1...βs

(rs)≡˜

Lγ·Qα1...αrβ1...βs

(rs).(2.14)

It is noted that the coeﬃcients in these expansions depend on the point u0∈Γ and generically Q(α)β

01 =˜

L(α)·Bβ

(0) 6= 0.

It was brieﬂy mentioned in paper [10] that the relations (2.9) and (2.12) are essential for the analysis of blow-ups of

derivatives for HEE (1.1). As we will see, the expansions (2.9), (2.12) and (2.13) completely deﬁne the structures and

characteristic properties of the blow-up of the ﬁrst and higher levels as well as the corresponding hierarchy of blow-ups.

Blow-ups of derivatives of the ﬁrst level correspond to the situation when all families of bilinear forms in (2.9), (2.12)

and (2.13) and all their possible linear superpositions do not vanish. Let us denote the (n−r1)-dimensional subspace of

variation of uaround the point u0given by the formula (2.6) as H+

(1)u0and as H−

(1)u0its complement spanned by the

vectors ˜

R(β)

δ−u=

β=1

R(β)δbβ.(2.15)

Due to the invertibility of the transformations (2.11), the generic variation of xaround the point x0can also be

represented in the form

δx=δ+x+δ−x(2.16)

where

δ+x=

n−r1

α=1

Pαδyα, δ−x=

β=1

Pαδ˜yα(2.17)

and n−r1

α=1

PiαL(α)

β=1

Piβ ˜

L(β)

k=δik .(2.18)

We will denote the subspace of variations of xigiven by the ﬁrst of (2.17) as H+

(1)x0and that given by the second of (2.17)

as H−

(1)x0.

Now, it is easy to see that the expansions (2.12) and (2.13) deﬁne the behavior of derivations with the variations of δu

and δxbelonging to diﬀerent subspaces. Indeed, the expansion (2.13) implies that if

δu∈H−

(1)u0, δx∈H−

(1)x0,(2.19)

one has

δ˜yγ∼˜

Q(γ)β

(01) δbβ,(2.20)

and, consequently, ∂bβ

∂˜yγ∼O(1) .(2.21)

Analogously, if

δu∈H+

(1)u0, δx∈H±

(1)x0,(2.22)

the expansions (2.12) and (2.13) imply that

δaα∼ |δx|1/2(2.23)

and so ∂aα

∂xk∼ |δx|−1/2,|δx| → 0.(2.24)

Finally if

δu∈H−

(1)u0, δx∈H+

(1)x0,(2.25)

one has

δbβ∼ |δy|1/2,(2.26)

and, hence, ∂bβ

∂yα∼ |δy|−1/2,|δy| → 0.(2.27)

It is noted that

∂

∂yα

=Pα·∂

∂x=

i=1

Piα ·∂

∂xi

, α = 1,...,n−r1(2.28)

and

∂

∂˜yα

=˜

Pγ·∂

∂x=

i=1

Piγ ·∂

∂xi

, γ = 1,...,r1(2.29)

where the matrices Piα and ˜

Piγ are deﬁned by (2.17). Summarizing one has the following table for the behavior of

derivatives ∂ui

∂xkas δx→0

δu

δxH+

(1)u0H−

(1)u0

H+

(1)x0∞ ∞

H−

(1)x0∞O(1)

It is noted that the derivatives ∂ui

∂xkdo not blow-up in r1-dimensional subspace of variations δ−u,δ−x. The rank r1of

the matrix M(u0) and dimensions of corresponding subspaces may vary from point to point on the blow-up hypersurface

Γ. Subspaces H+

(1)u0,H−

(1)u0,H+

(1)x0, andH−

(1)x0are elements of the orbits generated by the group SO(n) of rotations. The

dimensions of all elements of corresponding orbits are, obviously, the same.

3 On the admissible ranks of the matrix M(u0)

The matrix M(u), playing the central role in the whole construction, has a rather special form (2.1). It is parametrized

by nfunctions fi(u), i= 1,...,n of nvariables u=u1,...,un. Functions fiare essentially local inverse to the initial

data u(x, t = 0) for the equation (1.1).

In the analysis of the admissible constraints for matrix Mone should distinguishes two cases. First, usually considered

case, corresponds to the so-called generic initial data u0and, consequently, to the generic functions f(u). In such a case

the elements of the matrix Mdepend on n+ 1 variables t, u1,...,unfor the generic functions f1,...,fn.

The second case corresponds to the situation when one considers not the ﬁxed initial data but a family of initial data.

In this case one can view the function fias depending not only on nvariables u1,...,un, but also on a certain number

(possibly inﬁnite) of parameters λi,i > 0. For the discussion of such situation in the one-dimensional case see e.g. [10].

As we will see, the situation with possible ranks of the matrix Mis quite diﬀerent in these two cases.

Let us begin with the generic case. The matrix Mdepends on n+ 1 variables t, u. The requirement that M(u0) has

the rank r1imposes (n−r1)2constraints of order (r1+ 1) (see e.g. [7]).

Hence, the dimension of the subspace at which the matrix M(u0) has rank r1is given by

dim Γr1=n+ 1 −(n−r1)2.(3.1)

Since dim Γr1cannot be negative, in the generic case, one has the following constraint on the rank r1

(n−r1)2≤n+ 1 ,(3.2)

i.e. ln−√n+ 1m≤r1< n , (3.3)

文档加载中……请稍候！
如果长时间未打开，您也可以点击刷新试试。

下载文档到电脑，查找使用更方便

10 玖币 0人已下载

立即下载

摘要：

OnthenestructureandhierarchyofgradientcatastrophesformultidimensionalhomogeneousEulerequationB.G.Konopelchenko1andG.Ortenzi2*1DipartimentodiMatematicaeFisica\EnnioDeGiorgi",UniversitadelSalento,73100Lecce,Italy2DipartimentodiMatematicaeApplicazioni,UniversitadiMilano-Bicocca,viaCozzi55,20126Milan...

展开>> 收起<<

On the ne structure and hierarchy of gradient catastrophes for multidimensional homogeneous Euler equation B.G.Konopelchenko1and G.Ortenzi2.pdf

共22页,预览5页

还剩页未读，继续阅读

声明：本站为文档C2C交易模式，即用户上传的文档直接被用户下载，本站只是中间服务平台，本站所有文档下载所得的收益归上传人(含作者)所有。玖贝云文库仅提供信息存储空间，仅对用户上传内容的表现方式做保护处理，对上载内容本身不做任何修改或编辑。若文档所含内容侵犯了您的版权或隐私，请立即通知玖贝云文库，我们立即给予删除！

On the ne structure and hierarchy of gradient catastrophes for multidimensional homogeneous Euler equation B.G.Konopelchenko1and G.Ortenzi2

相关推荐

开通VIP享超值会员特权

作者详情

相关内容

热门标签

举报选择: