Professional Documents
Culture Documents
Full Ebook of Analysis of The Navier Stokes Problem Solution of A Millennium Problem 2Nd Edition Alexander G Ramm Online PDF All Chapter
Full Ebook of Analysis of The Navier Stokes Problem Solution of A Millennium Problem 2Nd Edition Alexander G Ramm Online PDF All Chapter
https://ebookmeta.com/product/c20-recipes-a-problem-solution-
approach-2nd-edition-j-burton-browning/
https://ebookmeta.com/product/a-problem-book-in-real-analysis-
problem-books-in-mathematics-2010th-edition-aksoy/
https://ebookmeta.com/product/a-student-s-guide-to-the-navier-
stokes-equations-instructor-solution-manual-solutions-essential-
extras-1st-edition-justin-w-garvin/
https://ebookmeta.com/product/jakarta-ee-recipes-a-problem-
solution-approach-josh-juneau/
Common Lisp Recipes A Problem Solution Approach Edmund
Weitz
https://ebookmeta.com/product/common-lisp-recipes-a-problem-
solution-approach-edmund-weitz/
https://ebookmeta.com/product/php-and-mysql-recipes-a-problem-
solution-approach-2nd-edition-frank-m-kromann/
https://ebookmeta.com/product/geometry-of-the-phase-retrieval-
problem-graveyard-of-algorithms-1st-edition-alexander-h-barnett/
https://ebookmeta.com/product/java-17-recipes-a-problem-solution-
approach-4th-edition-josh-juneau-3/
https://ebookmeta.com/product/asyncio-recipes-a-problem-solution-
approach-1st-edition-mohamed-mustapha-tahrioui/
Synthesis Lectures on
Mathematics & Statistics
Alexander G. Ramm
Analysis of the
Navier-Stokes
Problem
Solution of a Millennium Problem
Second Edition
Synthesis Lectures on Mathematics &
Statistics
Series Editor
Steven G. Krantz, Department of Mathematics, Washington University, Saint Louis, MO, USA
This series includes titles in applied mathematics and statistics for cross-disciplinary
STEM professionals, educators, researchers, and students. The series focuses on new and
traditional techniques to develop mathematical knowledge and skills, an understanding of
core mathematical reasoning, and the ability to utilize data in specific applications.
Alexander G. Ramm
Analysis
of the Navier-Stokes
Problem
Solution of a Millennium Problem
Second Edition
Alexander G. Ramm
Department of Mathematics
Kansas State University
Oakland, CA, USA
This Springer imprint is published by the registered company Springer Nature Switzerland AG
The registered company address is: Gewerbestrasse 11, 6330 Cham, Switzerland
To Luba
Preface to the Second Edition
The second edition of my monograph “The Navier-Stokes problem” has a new title and
contains several new features: the Introduction is expanded, Appendices 4 and 5 are added,
and the contradictions of the Navier-Stokes equations are emphasized.
We solve the millennium problem concerning the Navier-Stokes equations by proving
that the Navier-Stokes problem (NSP) in R3 does not have a solution defined for all times
t ≥ 0. This follows from the paradox the author found and proved in Chap. 7 of this
monograph.
Appendix 4 contains a brief but essentially self-contained presentation of our basic
ideas and results on the NSP.
Mathematical tool we use is the theory of integral equations and inequalities with
hyper-singular kernels of special type. This theory was developed by the author for an
analysis of the NSP.
In Appendix 5, there are some examples of applications of this theory.
vii
Preface to the First Edition
sup( v + ∇v ) ≤ c, (1)
t≥0
where v = v(x,t) is the solution to problem (2), the velocity vector of the incompressible
viscous fluid, and by c > 0 various constants, independent of the data, are denoted.
Here and throughout the book, c > 0 stand for various constants independent of x
and t, the norm ||·|| is the L 2 (R3 ) norm, by ∇v the collection of the norms of the first
∂v
derivatives ∂xmj is understood, and H 1 (R3 ) is the Sobolev space W21 (R3 ).
The NSP consists of solving the equation
ix
x Preface to the First Edition
where v = v(x, t) is the velocity of the fluid, f = f (x, t) is the (exterior) force, v0 (x) is
the initial velocity, p = p(x, t) is the pressure, v = const > 0 is the viscosity coefficient,
v = ∇ 2 v is the Laplacean of v. The fluid is assumed non-compressible and viscous, its
density is constant ρ = 1, v := ∂v ∂t . We look for a solution to NSP (2) in the space W.
We prove that the NSP problem is equivalent to the integral equation
t
v=F− ds dyG(x − y, t − s)(v, ∇)v, (3)
0 R3
and
|x|2
e− 4vt
g(x, t) = . (5)
(4vtπ)3/2
The data F depends only on f (x,t) and v0 (x). The function G(x,t) is calculated
analytically in Chap. 4.
Let us define the Fourier transform
1
F (v) := ṽ = ṽ(ξ, t) = e−iξ·x v(x, t)d x. (6)
(2π)3 R3
Take the Fourier transform of equation (3) and use the known formula F (vw) =
(2π)3 F (v)F (w) to get another integral equation equivalent to the NSP:
t
ṽ = F̃ − (2π)3 ds G̃(ξ, t − s)ṽ(−iξ ṽ) (7)
0
| f˜h̃| ≤ f˜ h̃ , (9)
where · = · L 2 (R3 )
We prove that
1 ξ j ξm −|ξ|2 t
G̃(ξ, t) = δ jm − e , (10)
(2π)3 ξ2
and
g̃(ξ, t) = e−|ξ|
2 νt
. (11)
Therefore,
|G̃| ≤ ce−|ξ| t ,
2
(12)
ξ ξ
because |1 − |ξ| 2 | ≤ 2.
j m
sup ṽ ≤ c (14)
t≥0
and inequality (9) were used. Estimate (14) is a consequence of the estimate (1) and of
the Parseval identities:
(2π)3 ṽ 2
= ṽ 2 , (2π)3 |ξ|ṽ 2
= ∇v 2 . (15)
Let
The integral in this inequality diverges from the classical analysis point of view (we
write sometimes diverges classically). We have to define this integral and derive some
estimates for positive solutions to inequality (18). This is done in Chaps. 4 and 5.
Together with the inequality (18) we study the corresponding integral equation:
t
q(s)ds
q(t) = b0 (t) + c . (19)
0 (t − s)5/4
We solve equation (19) in closed form under suitable assumptions on the data b0 (t),
namely, we assume that b0 (t), is a smooth and rapidly decaying as t → ∞. Moreover,
we prove that
q(0) = 0 (21)
provided that the data b0 (t) are smooth and rapidly decaying as t → ∞
From Eqs. (20) and (21), it follows that b(0) = 0. This and the definition of b(t) =
|ξ|ṽ imply that v0 (x) = 0, although originally we assumed that v0 (x) = v(x, 0) ≡ 0.
This is the NSP paradox. It shows that the NSP problem is contradictory physically and
mathematically and that the NSP with the suitable data does not have a solution defined
for all t ≥ 0.
Therefore, the millennium problem concerning the Navier-Stokes equations is solved:
we prove that this problem does not have a solution.
We prove that the NSP with the zero data, that is, f = 0 and v0 (x) = 0 does have a
solution v(x, t) = 0 and this solution is unique in the space W .
These are our main results concerning the NSP. These results show that physically
correct equations should be found for the motion of incompressible viscous fluid.
1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
2 Brief History of the Navier–Stokes Problem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
Reference . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
3 Statement of the Navier–Stokes Problem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
Reference . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
4 Theory of Some Hyper-Singular Integral Equations . . . . . . . . . . . . . . . . . . . . . . 23
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
5 A Priori Estimates of the Solution to the NSP . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
6 Uniqueness of the Solution to the NSP . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
7 The Paradox and Its Consequences . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
8 Logical Analysis of Our Proof . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
xiii
About the Author
Alexander G. Ramm was born in Russia, emigrated to USA in 1979, and is a US citi-
zen. He is Professor Emeritus of Mathematics with broad interests in analysis, scattering
theory, inverse problems, theoretical physics, engineering, signal estimation, tomography,
theoretical numerical analysis, and applied mathematics. He is an author of 716 research
papers and 20 research monographs and an editor of 3 books. He has lectured at many
universities throughout the world, given more than 150 invited and plenary talks at vari-
ous Conferences, and supervised 11 Ph.D. students. He was Fulbright Research Professor
in Israel and Ukraine, distinguished visiting professor in Mexico and Egypt, Mercator
Professor in Germany, Research Professor in France, and invited plenary speaker at the
7-th PACOM; he won Khwarizmi international award in 2004 and received other honors.
A. G. Ramm was the first to prove the uniqueness of the solution to inverse scattering
problems with fixed-energy scattering data, the first to prove the uniqueness of the solu-
tion to inverse scattering problems with non-over-determined scattering data, and the first
to study inverse scattering problems with under-determined scattering data. He solved
many specific inverse problems and developed new methods and ideas in the area of
inverse scattering problems. He introduced the notion of Property C for a pair of differen-
tial operators and applied Property C for one-dimensional and multi-dimensional inverse
scattering problems.
A. G. Ramm solved the many-body wave scattering problem when the bodies are small
particles of arbitrary shapes, assuming that a d λ, where a is the characteristic size
of small particles, d is the distance between neighboring particles, and λ is the wavelength
in the material in which the small particles are embedded. Multiple scattering is essential
under these assumptions. He used this theory to give a recipe for creating materials with a
desired refraction coefficient and materials with a desired wave-focusing property. These
results attracted the attention of the scientists working in nanotechnology.
A. G. Ramm gave formulas for the scattering amplitude for scalar and electromagnetic
waves by small bodies of arbitrary shapes and formulas for the polarizability tensors for
such bodies.
A. G. Ramm gave a solution to the Pompeiu problem, proved Schiffer’s conjecture,
and gave the first symmetry results in harmonic analysis.
xv
xvi About the Author
A. G. Ramm has developed the Dynamical Systems Method (DSM) for solving linear
and non-linear operator equations, especially ill-posed.
A. G. Ramm developed a random fields estimation theory.
A. G. Ramm has developed a theory of convolution equations with hyper-singular
integrals.
A. G. Ramm has introduced a wide class of domains with non-compact boundaries.
He studied the spectral properties of Schrödinger’s operators in this class of domains and
gave sufficient conditions for the absence of eigenvalues on the continuous spectrum of
these operators.
A. G. Ramm has solved one of the millennium problems concerning the Navier-Stokes
problem (NSP) and proved the NSP paradox, which shows the contradictory nature of the
NSP and the non-existence of its solution on the interval t ∈ [0, ∞) for the initial data
v0 (x) ≡ 0 and f (x, t) = 0.
Introduction
1
In this work a proof of the author’s basic results concerning the Navier-Stokes problem
(NSP) is given. The NSP is:
where v = v(x, t) is the velocity of the fluid, f = f (x, t) is the (exterior) force, v0 (x) is
the initial velocity, p = p(x, t) is the pressure, ν = const > 0 is the viscosity coefficient,
v = ∇ 2 v is the Laplacean of v. The fluid is assumed non-compressible and viscous, its
density is constant ρ = 1, v := ∂v
∂t .
We look for a solution to NSP (0.2) in the space W = {v : v ∈ H 1 × C(R+ ), ∇ · v = 0},
where H 1 = W21 (R3 ) is the Sobolev space and C(R+ ) is the space of continuous functions
on R+ with the norm bC(R+ ) = supt≥0 |b(t)|.
Our results include:
(a) The a priori estimate
of the solution to the NSP. Here v is the velocity vector, · is the L 2 (R3 ) norm, by c > 0
various constants independent of t and x are denoted. This a priori estimate is proved in
Chap. 5.
(b) The theory of integral equations
t
5
b(t) = b0 (t) + (t − s)− 4 b(s)ds (1.3)
0
and inequalities t 5
b(t) ≤ b0 (t) + (t − s)− 4 b(s)ds, (1.4)
0
5 t 5
with hyper-singular kernels (t − s)− 4 is developed. Note that the integral 0 (t − s)− 4 b(s)ds
diverges classically, that is, from the classical analysis point of view. We define such integrals
and solve in closed form the corresponding integral equations with the data b0 (t) smooth
and rapidly decaying as t → ∞. We study also integral equations with other hyper-singular
kernels. This theory is developed in Appendix 5.
(c) This paradox and its consequences are discussed in Chap. 7.
This paradox says that if the data (the initial velocity v0 (x) and the force f (x, t) in
the Navier-Stokes equations, see Chap. 4), are smooth and rapidly decaying, f = 0 (for
simplicity only) but v0 (x) ≡ 0, and the solution to the NPS exists on R+ , then v0 (x) = 0.
This means that the solution to the NSP cannot exist.
The millennium problem concerning the NSP consists of finding whether the solution to
the NSP exists for all t ≥ 0 and is smooth, provided that the data, that is, the initial velocity
v(x, 0) and the force f (x, t), are smooth and rapidly decaying as |x| → ∞ and t → ∞.
Our proof of the NSP paradox demonstrates that the solution to the NSP with f (x, t) = 0
and u 0 (x) ≡ 0 cannot exist for all t ≥ 0.
In paper [1], p. 472, Theorem 2, there is a statement that, for f (x, t) = 0 and u 0 (x)
sufficiently small, the solution to the NSP exists for all t ≥ 0 if m ≤ q, where m is the
dimension of the space and the solution is in L q . In our case m = 3 and q = 2, so the
condition m ≤ q does not hold. Therefore, the claim in [1], p. 472, is not applicable.
(d) The uniqueness theorem for the solutions to the NSP is proved in Chap. 6. This
theorem says that if the data vanishes, that is, f = 0 and u 0 (x) = 0 then there exists a
solution v(x, t) = 0 of the NSP and this solution is unique in the space W := {v : v ∈
H 1 (R3 ) × C(R+ ), ∇ · v = 0}, where H 1 is the usual Sobolev space and C(R+ ) is the
space of continuous functions b(t) with the supt≥0 |b(t)| norm.
We do not mention here other results.
The earlier author’s results [2–5] are used, but our presentation in this monograph is
essentially self-contained.
The Navier-Stokes equations are discussed in many books and papers. We only mention
here [6], [7] and [2].
Our main goal is to present for broad audience the author’s result concerning the Navier-
Stokes problem in R3 without boundaries. This result can be briefly formulated as follows:
Assume (for simplicity only) that the exterior force f (x, t) = 0. If the initial velocity
v0 (x) := v(x, 0) ≡ 0, ∇ · v0 (x) = 0, v0 (x) is smooth and rapidly decaying as |x| → ∞ and
the solution v(x, t) of the NSP exists for all t ≥ 0, then v0 (x) = 0.
This result we call the NSP paradox (or just paradox), shows that:
1 Introduction 3
a’) The NSP is not a correct statement of the problem of motion of viscous incompress-
ible fluid; it is neither physically nor mathematically correct statement of the dynamics of
incompressible viscous fluid.
b’) The NSP does not have a solution unless v0 (x) = 0 and f (x, t) = 0; in this case the
solution v(x, t) = 0 for all t ≥ 0.
This result solves the millennium problem related to the Navier-Stokes equations. It
encourages the search for the correct equations describing the dynamics of viscous incom-
pressible fluid.
Let us explain the steps of our proof.
The NSP consists of solving the equations (1.1).
a) First we reduce the NSP to an equivalent integral equation:
t
v(x, t) = F − ds G(x − y, t − s)(v, ∇)vdy, (1.5)
0 R3
We assume (for simplicity only and without loss of generality) that f = f (x, t) = 0. Under
this assumption one has:
F(x, t) := g(x − y, t)v0 (y)dy, (1.6)
R3
where
|x|2
e− 4νt
, t > 0; g(x, t) = 0, t ≤ 0; g̃ = e−|ξ| νt .
2
g(x, t) = (1.7)
(4νπt)3/2
The tensor G = G(x, t) = G jm (x, t) is calculated explicitly:
−3 ξ p ξm ξ p ξm
eiξ·x δ pm − 2 e−νξ t dξ; G̃ = (2π)−3 δ pm − 2 e−νξ t .
2 2
G(x, t) = (2π)
R 3 ξ ξ
(1.8)
Let us define the Fourier transform:
ṽ := ṽ(ξ, t) := (2π)−3 v(x, t)e−iξ·x d x. (1.9)
R3
Take the Fourier transform of equation (1.2) and get the integral equation equivalent to this
equation:
4 1 Introduction
t
ṽ(ξ, t) = F̃(ξ, t) − (2π)3 ds G̃(ξ, t − s)ṽ(iξ ṽ), (1.11)
0
where denotes the convolution in R3 . The following inequality, that comes from the
Cauchy inequality, is useful:
|ṽ(iξ ṽ)| ≤ ṽ|ξ|ṽ. (1.12)
By c here and throughout the paper various positive constants, independent of t and x, are
denoted. We denote by c1 := |(− 41 )| > 0 the special constant from equation (4.4), and use
∂v
the following notations: v j,m := ∂xmj , := R3 .
Let us write equation (1.1) as
Proof of Claim 1. Multiply equation (1.14) by v j , integrate over R3 and sum up over j
to get
1
(v2 ),t ≤ |( f , v)| ≤ f v. (1.16)
2
where z ,t := ∂z
∂t . In deriving inequality (1.16) we have used integration by parts:
− p, j v j d x = pv j, j d x = 0, νv, j j v j d x = −ν v, j v, j d x ≤ 0,
and
1
vm v j,m v j d x = − vm,m v j v j d x = 0.
2
From inequality (1.16) it follows that v,t ≤ f . Consequently,
∞
v ≤ v0 + f dt < c.
0
1 Introduction 5
By the Parseval equality the desired estimate (1.13) follows. Claim 1 is proved.
Inequalities (1.13) and (1.12) imply
This inequality is important because it allows one to estimate the nonlinear term in the
middle side of equation (0.13) by the linear term on its right side.
Equations (1.11) and (1.12) imply inequality
t t
e−ν(t−s)ξ |ξ|ṽds ≤ | F̃| + e−ν(t−s)ξ b(s)ds, b(s) = |ξ|ṽ
2 2
ṽ(ξ, t) ≤ F̃(ξ, t) + c
0 0
(1.18)
We derive this inequality later.
From formula (1.8) it follows that
where
b0 (t) := |ξ| F̃(x, t), b(t) := |ξ|ṽ(x, t) ≥ 0. (1.21)
6 1 Introduction
Here and below the norm · is the L 2 (R3 ) norm. If the data v0 (x) is smooth and rapidly
decaying as |x| → ∞, then |ṽ(ξ)| ≤ c(1 + |ξ|2 )−m , m > 25 . Therefore, from the definition
of b0 (t) it follows that
∞
e−r νt r 4 (1 + r 2 )−m dr := cI ,
2
b02 (t) ≤ c (1.22)
0
where we have used the spherical coordinates. From this relation it follows that the Laplace
transform ∞
Lb0 := e− pt b0 (t)dt
0
can be estimated:
|Lb0 | ≤ c(1 + | p|)−1 , Re p > 0.
1
1 one has: I = 0 e−r νt r 4 (1 + r 2 )−m dr + O(e−νt ) := J + O(e−νt ), J =
2
Indeed, for t
5 νt ρ −m dρ 5 5
(νt)− 2 0 e−ρ ρ2 (1 + νt ) 2ρ1/2
= O(t − 2 ). Thus, b0 (t) = O(t − 4 ), t 1 and supt≥0
b0 (t) ≤ c. Consequently, |Lb0 | ≤ c(1 + | p|)−1 , Re p ≥ 0. This estimate is used later.
Since the convolution integral in (1.20) diverges classically, we give a new definition of
this integral in Section 3 and estimate the solution b(t) to integral inequality (1.20) by the
solution q(t) to the integral equation with the same hyper-singular kernel:
t
5
q(t) = b0 (t) + c (t − s)− 4 q(s)ds. (1.23)
0
We prove that Eq. (1.23) has a unique solution in the space C(R+ ), and the following
estimate holds:
sup q(t) ≤ c, (1.26)
t≥0
provided that the datum b0 (t) is smooth and rapidly decaying at infinity.
References 7
q(0) = 0. (1.27)
(d) We prove that any solution b(t) ≥ 0 of inequality (1.20), where b0 (t) a smooth rapidly
decaying function, satisfies inequality (1.24). Since q(0) = 0 and 0 ≤ b(t) ≤ q(t), it follows
that b(0) = 0.
This yields the NSP paradox mentioned at the beginning of this section. Indeed, the
initial data v0 (x) ≡ 0, so b(0) > 0, but we prove that b(0) = 0.
The NSP paradox implies the conclusions we have made:
The NSP is physically not a correct description of motion of the incompressible
viscous fluid in the whole space R3 without boundaries; the NSP does not have a
solution on the whole interval [0, ∞) unless the data are equal to zero; in this case the
solution to the NSP does exist on the whole interval [0, ∞) and is identically equal to
zero.
The uniqueness of the solution to NSP is proved in Chap. 6.
References
Let ρ = ρ(x, t) be the density of the fluid and v(x, t) be its velocity. If D ⊂ R3 is a bounded
domain with a smooth boundary S, then the fluid mass in D is the integral D ρ d x and the
amount of fluid flowing through the boundary is S ρv · N d S, where N is the outer unit
normal to S, and S = ∂ D is the boundary of D. The conservation of the mass requires to
have
∂
− ρ d x = ρv · N d S. (2.1)
∂t D S
By the divergence theorem one has
ρv · N d S = ∇ · (ρv) d x. (2.2)
S D
∇ · v = 0. (2.5)
dv
ρ = −∇ p + f (2.6)
dt
where f is the exterior force, f = f (x, t).
For incompressible fluid, ρ = const. For convenience and without loss of generality, we
assume that
ρ = 1. (2.7)
One has
dv ∂v
= + (v, ∇)v. (2.8)
dt ∂t
From (2.6)–(2.8) it follows that
∂v
+ (v, ∇)v = −∇ p + f . (2.9)
∂t
This is Euler’s equation (1752) for incompressible fluid. D’Alembert’s paradox (1752)
proved that Euler’s equation is physically incorrect.
Let us write
v = vjej, (2.10)
where {e j }3j=1 is an orthonormal basis in R3 , (ei , e j ) = δi j , and over the repeated indices
in (2.10) summation is understood.
Euler’s Equation (2.9) can be written as
v j + vi v j,i = − p, j + f j , (2.11)
where
∂v j ∂v j ∂p
v j := , v j,i := , p, j := . (2.12)
∂t ∂ xi ∂x j
The Navier–Stokes equation (1822) for viscous incompressible fluid is
3
∂ 2v j
v j := v j,ii := .
j=1
∂ xi2
Equation (2.7) indicates that the fluid is incompressible. Equation (2.5) should be considered
together with the Navier–Stokes equation (2.13).
The initial velocity
v(x, 0) = v0 (x) (2.14)
is given.
Reference 11
The NSP in the whole space R3 consists of solving Eqs. (2.13), (2.5), and (2.14) simul-
taneously:
We assume that v(x, t) decays as |x| → ∞ for any fixed t > 0 and as t → ∞ for any
fixed x.
We will show in Chap. 7 that the NSP in R3 is contradictory. This means that this problem
is physically incorrect. We also prove that although the solution to the NSP is unique, it does
not exist on the whole interval t ∈ [0, ∞) unless v0 (x) = f (x, t) = 0; in this case the
solution is equal to zero identically, v(x, t) = 0, so it does exist for all t ≥ 0.
Claude-Louis Navier (1785–1836) was a French engineer who introduces the Navier–
Stokes equations in 1822. George Gabriel Stokes (1819–1903) was Anglo-Irish physicist
and mathematician. Stokes graduated in 1841 from Cambridge University. He published the
Navier–Stokes equations in 1845 not being familiar with the earlier work of Navier. Among
his students are Lord Rayleigh (1842–1919) and H. Lamb (1849–1934).
The motion of the fluid presents many unsolved questions. Turbulent motions are an
example.
In [1] it is mentioned (in the Introduction) that there is no satisfactory theory of the
motion of fluids and there are paradoxes related to some concrete problems in this theory.
The paradox we construct in Chap. 7 is of a general nature and it shows that the NSP is not
physically acceptable.
Reference
∇ · v0 = 0. (3.4)
The functions v(x, t) and p = p(x, t), solving Eqs. (3.1)–(3.3), are to be found. These
functions decay as |x| → ∞, t → ∞.
Our task is to derive an integral equation for v that is equivalent to (3.1)–(3.3). To do this,
we construct the Green’s function G(x, t) that solves the linear problem:
G = G jm (x, t) (3.11)
is a tensor.
Let us define the Fourier transform:
1
F(v) := ṽ(ξ, t) = v(x, t)e−iξ·x d x, ξ · x = ξ j x j , (3.12)
(2π)3 R3
Let
G= H (ξ, t)eiξ·x dξ, G = G jm (x, t), (3.14)
R3
1
H (ξ, t) = G(x, t)e−iξ·x d x, H = H jm (ξ, t), H = F(G) = G̃. (3.15)
(2π)3 R3
Taking the Fourier transform of Equations (3.5)–(3.6) one gets:
δ(t)δ jm
H + νξ 2 H = − iξ Pm (ξ, t), (3.16)
(2π)3
ξ · H = 0, (3.17)
δ(t)ξm
= iξ 2 Pm (ξ, t). (3.20)
(2π)3
Therefore
δ(t)ξm
Pm (ξ, t) = −i , ξ2 = ξ j ξ j . (3.21)
(2π)3 ξ 2
3 Statement of the Navier-Stokes Problem 15
Indeed, let
1 ξ j ξm
νξ 2 := a, δ jm − := b. (3.25)
(2π)3 ξ2
Then formulas (3.22)–(3.23) can be written as
( p + a)L(H ) = b. (3.27)
Let us define
1
δ jm e−νξ
2 t+iξ·x
I1 := dξ (3.31)
(2π)3 R3
and
1 ξ j ξm −νξ 2 t+iξ·x
I2 := − e dξ. (3.32)
(2π)3 R3 ξ2
Integrals (3.31)–(3.32) can be calculated explicitly. Therefore, the function G(x, t) will be
calculated explicitly.
16 3 Statement of the Navier-Stokes Problem
q2
1 ∞
−νs 2 t e− 4νt
J= e cos(sq) ds = . (3.34)
π 0 (4νtπ)1/2
Integral J 3 = I1 , that is
I1 = δ jm g(x, t), (3.35)
where
|x|2
e− 4νt
g(x, t) = , t > 0; lim g(x, t) = g(x, 0) = 0. (3.36)
(4πνt)3/2 t→0+
Let us calculate I2 .
One has
e−νξ t+iξ·x
2
1
I2 = ∂ jm dξ
(2π)3 R3 ξ2
∞ (3.37)
1 −νtr 2 ir |x| cos θ
= ∂ jm dr e e sin θ dθ dφ ,
(2π)3 0 S2
∂2
where ∂ jm = ∂x j ∂xm . Therefore,
∞ 1
1
dr e−νtr eir |x|u du
2
I2 = ∂ jm
(2π)2 0 −1
∞
1 − e−ir |x| eir |x|
dr e−νtr ·
2
= ∂ jm
(2π)2
0 ir |x|
∞
2 −νtr 2 sin(r |x|)
= ∂ jm dr e . (3.38)
(2π)2 0 r |x|
where x
2
e−s ds.
2
Erf(x) = √ (3.40)
π 0
Using (3.38)–(3.40) one gets
|x| |x|
√ √
1 1 π 2 2 νt 1 1 2 νt
e−s ds e−s ds .
2 2
I2 = ∂ jm √ = ∂ jm (3.41)
2π 2 |x| 2 π 0 2π 3/2 |x| 0
3 Statement of the Navier-Stokes Problem 17
w − νw = 0 in R3 × R+ , (3.44)
w(x, 0) = v0 in R 3
(3.45)
w̃ + νξ 2 w̃ = 0, (3.47)
|x|2
−νtξ 2 1 −iξ·x e− 4νt
e = e d x, (3.50)
(2π)3 R3 (4νtπ)3/2
and then using the theorem about the Fourier transform of the convolution
F h(x − y) f (y) dy = (2π)3 h̃(ξ) f˜(ξ), (3.51)
R3
Proof Apply the operator ∂t − ν to Eq. (3.52) and use Eqs. (3.5) and (3.44) to get
t
vm − νvm = ds dy δ(x − y)δ(t − s)δ jm − ∇ pm [ f (y, s) − (v, ∇)v]
0 R3
where t
p= ds dy pm [ f − (v, ∇)v]. (3.54)
0 R3
Equation (3.6) yields
∇ ·v =∇ · g(x − y, t)v0 (y) dy = ∇ · g(z, t)v0 (x + z) dz = 0, (3.55)
R3 R3
∇x · v(x + z) = 0.
Corollary 3.3 Integral Equation (3.52) is equivalent to the equation obtained by the Fourier
transform of it:
t
ṽ(ξ, t) = ˜
d x G̃(ξ, t − s) f (ξ, s) − ṽm (ξ − η, s)ṽ j (η, s)iηm dη + ṽ0 (ξ)e−νtξ .
2
0 R3
(3.56)
Let
v = v(x, t) = ( |v(x, t)|2 d x)1/2 . (3.57)
R3
The norm (3.57) is generated by the inner product in the Hilbert space H = L 2 (R3 ),
3 Statement of the Navier-Stokes Problem 19
(v, w) = v w̄ d x. (3.58)
R3
∞
Lemma 3.4 Assume that v0 + 0 f dt < c. Then the solution to the NSP (3.1)–(3.3)
satisfies the a priori estimate:
sup v(x, t) < c. (3.59)
t≥0
Proof Multiply (3.1) by v in L 2 (R3 ) and assume that ∇v < ∞, to get
1
∂t v +
2
vm v j vm, j d x = ( f , v) − (∇ p, v) + ν(v, v).
2 R3
One has
(∇ p, v) = −( p, ∇ · v) = 0, (3.60)
where Eq. (3.2) was used.
Furthermore,
ν(v, v) = −ν∇v2 < 0, (3.61)
and
1 1
vm v j vm, j d x = v j (vm
2
), j = − v j, j vm vm d x = 0, (3.62)
R3 2 R3 2 R3
where Eq. (3.2) was used, v j, j = 0.
Therefore,
∂t v2 ≤ 2 f v, (3.63)
so t
v2 ≤ v0 2 + 2 f v ds. (3.64)
0
Consequently,
∞
sup v2 ≤ c1 + f d x · sup v ≤ c1 + c2 sup v. (3.64 )
t≥0 0 t≥0 t≥0
ξ j · ξm
|δ jm | ≤ c, ≤ c. (3.67)
ξ2
This result is well known and we omit its proof. Note that
h(ξ − η)g(η) dη ≤ h g. (3.69)
R3
Proof Let us calculate the first integral. The second one is calculated similarly.
One has for any a > 0
∞
−aξ 2 c
e−ar r 2 dr = 3/2 .
2
e dξ = 4π
R 3 0 a
Therefore,
c
e−aξ =
2
. (3.73)
a 3/4
Formula (3.72) is derived similarly. The parameter a = ν(t − s).
Since ν = const > 0, Lemma 3.6 is proved.
Assumption 3A. Let us assume in what follows that ṽ0 (ξ) and f˜(ξ, t) are smooth and
rapidly decaying functions of their arguments.
Reference 21
From Assumption (1.15) it follows that b0 (t) is a smooth rapidly decaying function as
t → ∞.
The integral in formula (3.75) diverges classically. We define this integral in Chap. 4.
Reference
where Re λ > 0, c > 0 stands for various constants, b0 (t) is a given function which is
assumed to be smooth and rapidly decaying as t → ∞. Taking the Laplace transform of
(4.1) one gets
(λ)
L(q) = L(b0 ) + cL(q) λ , (4.2)
p
where formula (A2.13) was used:
(λ)
L(t λ−1 ) = . (4.3)
pλ
Here (λ) is the gamma function and formula (4.3) is valid classically for Re λ > 0; this
formula is valid for all complex λ except λ = 0, −1, −2, . . ., by analytic continuation with
respect to λ because (λ) is analytic in C except for the points λ = 0, −1, −2, . . . and p −λ
is an entire function of λ; see Appendix 2. In Appendix 3, one finds the results on the theory
of the Laplace transform that are used in this book.
We are interested in the value λ = − 41 because it appears in inequality (3.75), λ − 1 =
− 4 = − 41 − 1, so λ = − 41 . The value (− 41 ) can be calculated:
5
1 3 3
(− ) = −4( ) := −c1 , c1 = 4( ) > 0. (4.4)
4 4 4
L(b0 )
L(q) = 1
, cc1 = const > 0. (4.5)
1 + cc1 p 4
© The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 23
A. G. Ramm, Analysis of the Navier-Stokes Problem, Synthesis Lectures
on Mathematics & Statistics, https://doi.org/10.1007/978-3-031-30723-2_4
24 4 Theory of Some Hyper-Singular Integral Equations
Equation (4.1) then has a unique solution q, q = q(t) is a bounded function on C([0, ∞)),
and
q(0) = 0. (4.8)
Proof The statement of Theorem 4.1 concerning continuity of q and estimate (4.7) follow
from Theorem 4.1 in Appendix 3. The conclusion (4.8) follows from Theorem A3.2 of
Appendix 3.
Theorem 4.1 is proved.
Let us establish a relation between the solution to Eq. (4.1) and solutions to inequality
(3.75).
Theorem 4.2 If b = b(t) solves inequality (3.75) and q(t) solves Eq. (4.1) with λ = − 41 ,
then
b(t) ≤ q(t). (4.9)
t
Lemma 4.3 For any T > 0 and for p > −1, the operator A f := 0 (t − s) p f (s) ds in the
Banach space X 0 = C([0, T ]) has spectral radius r (A) = 0.
Proof of Lemma 4.3. The spectral radius r (A) of a linear operator A is defined by the
formula (see, for example, [1]):
1
r (A) = lim An n . (4.10)
n→∞
n ( p + 1)
An f X 0 ≤ t n( p+1) f X0 . (4.11)
(n( p + 1) + 1)
For n = 0, estimate (4.11) is obvious. For n = 1, one has
4 Theory of Some Hyper-Singular Integral Equations 25
t
p+1
(t − s) p f (s) ds ≤ t f X0 . (4.12)
p+1
0
Suppose estimate (4.11) holds for some n. Let us prove that it holds for n + 1. One has
t
An+1 f X 0 ≤ (t − s) p |An f | ds
0
t n ( p + 1)
≤ (t − s) p s n( p+1) ds f X0 , (4.13)
0 (n( p + 1) + 1)
t 1
(t − s) s
p n( p+1)
ds = t p+n( p+1)+1
(1 − u) p u n( p+1) du
0 0
( p + 1)(n( p + 1) + 1)
= t ( p+1)(n+1) .
( p + 1 + n( p + 1) + 1)
Thus,
n+1 ( p + 1)
An+1 f X 0 ≤ t (n+1)( p+1) f X0 . (4.14)
((n + 1)( p + 1) + 1)
Applying formula (A2.9) to (4.11), one gets
1/n 1 1/n
lim An f X 0 = t p+1 ( p + 1) lim f X 0
n→∞ lim (n( p + 1) + 1) n→∞
n→∞
t p+1 ( p + 1)
= 1
= 0. (4.15)
lim e(n( p+1)+ 2 ) ln(n( p+1)+1)
n→∞
Lemma 4.4 Let A be a linear operator in a Banach space Z . If r (A) < 1, then equation
f = f0 + A f (4.16)
Proof Let us show that Eq. (4.15) is an equation with A < 1 in the Banach space Z . If
this is done then formula (4.17) follows and the series (4.17) converges in Z .
26 4 Theory of Some Hyper-Singular Integral Equations
If the series (4.17) converges in Z then its sum solves (4.16). Indeed, from (4.17) one
derives
∞
f = f0 + A A j f0 = f0 + A f . (4.18)
j=0
m−1
m−1
m−1
A j f 0 + Am A j f 0 + A2m A j f0 + · · · . (4.20)
j=0 j=0 j=0
Denote
m−1
h := A j f0 . (4.21)
j=0
Then
∞
∞
A j f0 = Amq h. (4.22)
j=0 q=0
Therefore the series (4.17) converges and Eq. (4.16) has a solution.
Let us prove that this solution is unique. Since A is a linear operator, it is sufficient to
prove that the equation
f = Af (4.24)
has only the trivial solution f = 0. From (4.24) one derives
f = Am f . (4.25)
t
If A f = 0 (t − s)λ−1 f (s) ds, λ > 0, then by Lemma 4.3, r (A) = 0. Therefore,
Eq. (4.16) with this A has a solution, this solution is unique and can be obtained by iterations
by formula (4.17).
4 Theory of Some Hyper-Singular Integral Equations 27
In the book [2] it is stated that one can construct a Banach space Z 1 in which A1 < 1
if A is a linear operator and r (A) < 1 in the original Banach space Z .
The proof of this statement in [2] is not clear to this author. By this reason we formulated
and proved Lemma 4.4.
Since this result is used below, let us give an independent proof of it using the specific
form of the operator A.
Let us prove that if r (A) = 0 then Eq. (4.16) has a solution, this solution is unique in Z ,
Eq. (4.16) is solvable by iterations
and f is given by formula (4.17). We took c A because in (3.75) c > 0 is not necessarily
equal to 1.
Using estimate (4.11) one concludes that the series (4.17) converges in Z . Clearly, its
sum solves Eq. (4.16) and is obtained as a limit
f = lim f n , (4.28)
n→∞
where f n are iterations defined by formula (4.27). To prove uniqueness of the solution to
Eq. (4.16) assume that f 0 = 0, so the Eq. (4.16) is of the form
h = c Ah. (4.29)
Continue with this argument and prove (4.32) for any T > 0.
Let us now introduce the function
t λ−1
λ = , (4.33)
(λ)
28 4 Theory of Some Hyper-Singular Integral Equations
where
0, t < 0,
t λ = t+
λ
= (4.34)
t λ, t ≥ 0.
Define the convolution: t (t − s)λ−1
λ ∗ f := f (s) ds. (4.35)
0 (λ)
The right side of Equation (4.35) is well defined classically if Re λ > 0 and f (s) is contin-
uous on R+ = [0, ∞).
Let us define the right side of (4.35) for Re λ < 0. To do this, consider the Laplace
transform
L(λ ∗ f ) = L(λ )L( f ) = L( f ) p −λ , (4.36)
where formula (A3.21) was used. Formula (4.36) is well defined classically for Re λ > 0
and for Re λ ≤ 0 it is defined by the analytic continuation with respect to λ since L( f ) does
not depend on λ and p −λ is an entire function of λ.
The convolution λ ∗ f for Re λ < 0 is defined by the formula
λ ∗ f = L −1 (L( f ) p λ ). (4.37)
We are interested in the value λ = − 41 because this value is important for the NSP, see
Eq. (3.75). If λ = − 41 then
1 3
(− ) = −4( ) := −c1 , (4.38)
4 4
and formula (4.36) for λ = − 41 takes the form
1
L(− 1 ∗ f ) = p 4 L( f ). (4.39)
4
1
L(λ ∗ μ ) = . (4.42)
p λ+μ
By formula (4.37) with f = μ one obtains
1
λ ∗ μ = L −1 ( ) = λ+μ . (4.43)
p λ+μ
4 Theory of Some Hyper-Singular Integral Equations 29
One can give a different proof of formula (4.40). Namely, let u = st . Then
(t − s)λ−1 s μ−1
t
λ ∗ μ = ds
0 (λ) (μ)
1
1
= t λ+μ−1 (1 − u)λ−1 u μ−1 du
(λ)(μ) 0
t λ+μ−1 (λ)(μ)
=
(λ)(μ) (λ + μ)
t λ+μ−1
=
(λ + μ)
= λ+μ . (4.46)
λ ∗ f ≥ 0. (4.49)
Proof For λ > 0 and f continuous the convolution (4.49) is defined classically and inequal-
ity (4.49) is obvious.
Lemma 4.6 is proved.
30 4 Theory of Some Hyper-Singular Integral Equations
Apply the operator 1 ∗ to inequality (4.47) and to Eq. (4.48) and use Eq. (4.41) to get
4
1 1
1 ∗ b ≤ 1 ∗ b0 − cc1 b, b≤ 1 ∗ b0 − 1 ∗ b, (4.50)
4 4 cc1 4 cc1 4
1 1
1 ∗ q = 1 ∗ b0 − cc1 q, q= 1 ∗ b0 − 1 ∗ q, (4.51)
4 4 cc1 4 cc1 4
Let us now prove Theorem 4.2.
Proof of Theorem 4.2. Denote the positive constant cc1 := c2 > 0. By Lemma 4.4 the
inequality (4.50) and Eq. (4.51) can be solved by iterations:
∞
n
1
b≤ (−1)n 1 ∗ b0 , (4.52)
c2 4
n=0
∞ n
1
q= (−1)n 1 ∗ b0 . (4.53)
c2 4
n=0
From (4.52) and (4.53) the conclusion of Theorem 4.2 follows, in particular inequality (4.9)
holds.
Theorem 4.2 is proved.
Theorem 4.7 Assume that b0 (t) in Eq. (4.48) is smooth and rapidly decaying as t → ∞.
Then Eq. (4.48) has a unique solution in C([0, ∞)) and
Therefore,
L(b0 )
L(q) = 1
. (4.56)
1 + cc1 p 4
Under our assumption about b0 (t) one has
c
|L(b0 )| ≤ , Re p > 0, (4.57)
1 + | p|
1
and L(b0 ) is an analytic function of p in the half-plan Re p > 0. Define p 4 in the half-plane
Re p > 0 by the formula:
1 1 π π
p 4 = | p| 4 eϕ/4 , ϕ = arg p, − ≤ϕ≤ . (4.58)
2 2
References 31
1
Then p 4 is analytic in the region Re p > 0; L(q), defined by formula (4.56), is an analytic
function in the half-plane Re z > 0 and, using inequality (4.57), one gets
c
|L(q)| ≤ , c2 := cc1 > 0. (4.59)
1
(1 + | p|) 1 + c2 | p| 4 eiϕ/4
By Theorem 4.1, see formulas (4.7) and (4.8), one gets inequality (4.54) and the relation
q(0) = 0.
Theorem 4.7 is proved.
References
One of the a priori estimates was formulated and proved in Lemma 3.4, namely, estimate
(3.59):
sup v(x, t) < c, (5.1)
t≥0
where v(x, t) is a solution to the NSP (3.1)–(3.3). It was proved under the assumption
∞
v0 (x, t) + f (x, s) ds < c. (5.2)
0
The other basic a priori estimate is formulated in Theorem 4.7. Let us use this result and
the Parseval’s equation to derive the following theorem.
Theorem 5.1 Denote b(t) := |ξ|ṽ(ξ, t) and assume that v0 (x) and f (x, t) are smooth
functions of their arguments rapidly decaying as |x| → ∞ and t → ∞. Then
Proof Under the assumption of Theorem 5.1 the Assumption 3A (see p. 19) is satisfied
and b(t) solves inequality (3.75). Inequality (3.75) is equivalent to inequality (4.47). By
Theorem 4.2 of Chap. 4 inequality (4.9) holds, where q(t) solves Eq. (4.48). By Theorem 4.7
of Chap. 4, not only does inequality (5.3) hold since
Theorem 5.2 Under the assumptions of Theorem 5.1 the following inequalities hold:
sup ξ 2 |ṽ(ξ, t)| < c, ∀t ≥ 0; sup |v(ξ, t)| ≤ c(1 + t). (5.8)
t≥0,|ξ|≥0 |ξ|>0
where
t
F̃(ξ, t) := ṽ0 (ξ)e + −νtξ 2
ds H (ξ, t − s) f˜(ξ, s), (5.10)
0
(v, ∇)v = ṽm (ξ − η, s)ṽ j (η, s)iη dη, (5.11)
R3
−νξ 2 (t−s)
|H (ξ, t − s)| ≤ ce , (5.12)
where inequality (3.66) is identical to inequality (5.12). It follows from (5.9) and (3.69) that
Thus,
t
ds e−νξ
2 (t−s)
|ṽ(ξ, t)| ≤ | F̃(ξ, t)| + c |ξ|ṽ(ξ, s) ṽ(ξ − η, s)
0
t
ds e−νξ
2 (t−s)
≤ | F̃(ξ, t)| + c , (5.13)
0
1 − e−νξ t
2
The function F̃(ξ, t) depends only on the data f (x, t) and v0 (x). We may assume that the
data are smooth and rapidly decaying and then
c
| F̃(ξ, t)| ≤ . (5.15)
1 + |ξ|2
The last term in (5.14) for |ξ| → 0 is of the order νt, so it is not bounded uniformly in
t ∈ [0, ∞).
Let us multiply both terms of (5.14) by |ξ|2 and get
c
|ξ|2 |ṽ(ξ, t)| ≤ |ξ|2 | F̃(ξ, t)| + . (5.16)
ν
By the estimate (5.15) one gets
One has
1 − e−νξ
2t
FIN DU JOURNAL.
FRAGMENTS
I
SOUVENIR [36] .
[37] Tout ce qui suit a été extrait d’un cahier rédigé par
Mlle de Guérin vers 1841, qu’elle avait apporté du Cayla
lors de son dernier voyage à Paris, avec l’intention vague
de l’insérer peut-être dans le recueil des œuvres
posthumes de Maurice ou de le laisser imprimer
séparément, mais toujours en souvenir de lui et pour
l’honneur de sa mémoire. Sauf les premières et les
dernières pages, qui paraissent empruntées à des lettres
écrites par la sœur à son frère pendant le séjour de celui-
ci en Bretagne, ce n’était qu’une reproduction souvent
littérale, par endroits un peu châtiée, du IIe cahier, que
nous avons préféré mettre à son rang dans la suite du
Journal et sous sa forme primitive. Nous avons
seulement tenu à reproduire ici les passages qui ne se
trouvent point ailleurs, et, à la date du 27 mai et du 21
août 1835, deux variantes qui nous ont semblé offrir un
véritable intérêt littéraire.
L’Imitation de Jésus-Christ.
L’Introduction à la vie dévote, de saint François de Sales.
Le Combat spirituel.
Les Méditations de Bossuet.
Méditations de Médaille.
Lettres spirituelles de Bossuet.
Heures de Fénelon.
Journée du chrétien.
Les Sages entretiens.
L’Ame élevée à Dieu.
L’Ame embrasée de l’amour divin.
Le Mois de Marie.
La Vie des Saints.
Entretiens d’un missionnaire et d’un berger.
Le Dogme générateur de la piété chrétienne, par M. Gerbet.
Le Froment des élus.
Élévations sur les mystères de Bossuet.
Le Guide du jeune âge, de M. Lamennais ; livre que je relis
souvent.