January 16
Topics
- Introduction to the course.
- The Spaces \(C^r(\Omega)\), multivariate Taylor series
January 18
Topics
- Normed Spaces
- Function Spaces
Readings
- Ch. 1 of Hunter's notes (today I covered only normed spaces, next lecture I will go over metric ones)
- Function Spaces, by Terry Tao (UCLA)
- An interesting discussion on why the natural topology on \(C^\infty([a,b])\) cannot come from a norm
- An interesting discussion on the infinite dimensional vector space \(\mathbb R_\Bbb Q\).
- If you are curious enough, read the following notes on Hamel bases of vector spaces by K. Smith (UMich).
Homework #1, due Friday, Feb 1
- Consider the sequence of continuous piecewise linear functions in $C^0([0,1])$ defined by
$$
f_n(x) = \begin{cases}0,&x\in[0,\frac{1}{2}-\frac{1}{n}]\\ a_nx+b_n,&x\in[\frac{1}{2}-\frac{1}{n},\frac{1}{2}+\frac{1}{n}]\\ 1,&x\in[\frac{1}{2}+\frac{1}{n},1]\\\end{cases}
$$
where the coefficients $a_n$ and $b_n$ are chosen so that $f_n$ is continuous.
(a) Find the values of $a_n$ and $b_n$;
(b) Show that this is not a Cauchy sequence with respect to the $L^\infty$ norm $$ \|f\|_{L^\infty} = \sup_{x\in[0,1]}|f(x)|\,; $$ (c) Show that, on the contrary, this is a Cauchy sequence with respect to the $L^p$ norm $$ \|f\|_{L^p} = \left[\int\limits_0^1|f(x)|^p\,dx\right]^{1/p} $$ for any $1\leq p<\infty$.
(d) Finally show that, considering $C^0([0,1])$ as a subset of the normed vector space $(L^p([0,1]),\|\cdot\|_{L^p})$ of all integrable functions with finite $L^p$ norm, the sequence $\{f_n\}$ converges to the (integrable but discontinuous) function $$ f(x) = \begin{cases}0,&x\in[0,\frac{1}{2})\\ 1,&x\in(\frac{1}{2},1]\\\end{cases} $$ (the value of $f$ at $x=1/2$ is irrelevant since functions in $L^p$ spaces are considered the same whenever they differ in a set of measure zero). - Prove that $(C^0([0,1]),\|\cdot\|_{L^\infty})$ is a complete normed vector space (in short, a Banach space), namely prove that every uniformly Cauchy sequence $\{f_n\}\subset C^0([0,1])$ converges to a continuous function over $[0,1]$.
- Prove that $(C^1([0,1]),\|\cdot\|_{L^\infty})$ is not complete by building
a Cauchy sequence $\{f_n\}\subset C^1([0,1])$ that does not converge to a $C^1([0,1])$ function in the $L^\infty$ norm (no need to write
analytical formulas, a good picture would be enough).
In fact, by the Stone-Weierstrass theorem, we know that $$ \overline{C^r([0,1])}^{\|\cdot\|_{L^\infty}}=C^0([0,1]) $$ for any $r=0,1,2,\dots,\infty,\omega$ since every continuous function is the uniform limit of a sequence of polynomials. - Prove that, on the contrary, $(C^1([0,1]),\|\cdot\|_{C^1})$, with $$ \|f\|_{C^1} = \sup_{x\in[0,1]}|f(x)| + \sup_{x\in[0,1]}|f'(x)|\,, $$ is a Banach space, namely show that, given any uniformly Cauchy sequence $\{f_n\}$ of $C^1$ functions such that also the sequence of their derivatives $\{f'_n\}$ is uniformly Cauchy in $[0,1]$, $\{f_n\}$ does converge to a $C^1$ function over $[0,1]$.
January 23
Topics
- The Banach spaces $(C^k([0,1],\|\cdot\|_{C^k}), k=0,1,2,\dots$
- Metric spaces.
- Natural topologies on $C^\infty([0,1])$ and $C^k(\Bbb R)$.
Readings
- Metric Spaces by J.Hunter (UCDavis)
- The Banach spaces $C^k([a,b])$ by P. Garrett (UMN).
January 25
Topics
- Again about the $C^k$ spaces.
- Discussion on the completeness of $C^1(K)$.
Readings
- See Thm 5.18 in Sequences and series of functions by J.Hunter (UCDavis)
Homework
- Prove that $(C^2([0,1]),\|\cdot\|_{C^2})$ is a Banach space.
January 28
Topics
- Definition of the spaces $W^{k,p}$ and $L^p_{loc}$.
- Continuity of the differential operators.
Readings
- My (very short) notes on Function Spaces
- Evans, Appendix C
January 30
Topics
- Definition of the spaces $W^{k,p}$ and $L^p_{loc}$.
- Continuity of the differential operators.
- Mollifiers
Readings
- My (very short) notes on Function Spaces
- Evans, Appendix C
February 1
Topics
- Definition of the spaces $W^{k,p}$ and $L^p_{loc}$.
- Continuity of the differential operators.
Readings
- My (very short) notes on Function Spaces
- Evans, Appendix C
February 4
Topics
- Definition of a general PDE of order $k$.
- Digression: The dual of a finite-dimensional linear vector space.
Readings
- Evans, Chapter 1.
- Vector and coVectors on math.stackexchange.
- Covariance and contravariance of vectors on wikipedia.
February 6
Topics
- Solving the linear 1st order PDE $$ X^\alpha(x)\partial_\alpha u(x) + \mu(x) u(x)= v(x)\,. $$ The case $X=\partial_x$ and $\Omega=\Bbb R^2$.
Readings
- My (very short) notes on PDEs
- Linear 1st order PDEs on NPTEL (India)
February 8
Topics
- Digression: action on vectors and covectors induced by a map $f:\Bbb R^n\to\Bbb R^m$; $\{\partial_\alpha|_{x=x_0}\}$ and $\{dx^\alpha\}$ as bases of, respectively, the spaces of tangent and cotangent vectors at $x_0$.
- Influence of the topology of the integral trajectories of a vector field $X(x)$ on the regularity of the solutions of the linear 1st order PDE $$ X^\alpha(x)\partial_\alpha u(x) = v(x)\,. $$ The case $X=\partial_x$ and $\Omega=\Bbb R^2\setminus\{0\}\times[0,\infty)$.
Readings
- My (very short) notes on PDEs
- Here are some notes on flows of vector fields by J. Marsden (CalTech).
Homework #2, due Wednesday, Feb 20
- Solve the PDE $$ -yu_x+xu_y = 0,\;u(x,x^2)=x^3\;\;\hbox{ for }\;\;x>0, $$ in $\Omega=\Bbb R^2\setminus\{(0,0)\}$.
- Solve the PDE $$ 2 u_x+3 u_y + 8u = 0,\;u(x,0)=\sin x $$ in $\Omega=\Bbb R^2$.
- Solve the PDE $$ xu_x+yu_y = \sqrt{x^2+y^2},\; u=\frac{1}{2}\;\hbox{ on the circle }\;\{x^2+y^2=1\} $$ in $\Omega=\Bbb R^2\setminus\{(0,0)\}$.
- Show that the PDE $$ yu_x-xu_y = 1,\;u(0,y)=0 $$ in $\Omega=\Bbb R^2\setminus\{(0,0)\}$ does not admit a $C^0$ solution (Hint: assume such $u$ exists and study how its value changes along the integral trajectories of $X$).
- Show that the PDE above does admit, on the contrary, a $C^\infty$ solution in $\Omega=\Bbb R\times(0,\infty)$.
February 11 - March 8
Topics
- Quasi-linear 1st order PDEs
- Linear and non-linear wave equations
- Fully nonlinear 1st order PDEs
- A very short introduction to Lagrangian and Hamiltonian mechanics and the least action principle
- The Hamilton-Jacobi equation
Readings
- My short notes on PDEs
- The whole Chapter 3 of Evans. His notation is different from the one I use in the notes but it's healthy for you all to see different ones. We did not cover (yet) all topics in that chapter, here is a list of sections you can skip (although I suggest you to give them a look at a second read): 3.1.2 (envelopes), 3.3.2 (Legendre transformation, Hopf-Lax formula), 3.3.3 (weak solutions), 3.4 (Conservation laws)
- Schrodinger's train of thoughts, a very nice and readable article on the interplay between classical and quantum mechanics and the fundamental role in it of the Hamilton-Jacobi equation and the action functional.
Take-home Midterm, due Monday, Mar 18
- Solve the quasilinear PDE $$ u_x+uu_y = 2x,\;u(0,y)=y^2 $$ in $\Omega=\Bbb R^2$.
- Solve the fully nonlinear PDE $$ u^{\color{red}{2}}_x+u_y = u,\;u(x,0)=1+x^2 $$ in $\Omega=\Bbb R^2$.
- Solve the Hamilton-Jacobi PDE $$ u_xu_y = xy,\;u(x,0)=\color{red}{x} $$ in $\Omega=\Bbb R^2$.
-
Let $u$ be the solution of the Burgers quaslilinear 1st order PDE
$$
u_t+uu_x=0,\;u(0,x)=g(x)\in C^2(\Bbb R^2)
$$
and suppose that $\|g\|_{C^1(\Bbb R)} < \infty$ and that $u\in C^2((-T,T)\times\Bbb R)$.
Prove that:
- $g\geq0\implies u\geq0$ ;
- if $g$ has compact support, then so does $u(t,\cdot)$ for all $t\in(-T,T)$;
- if $g\geq0$ and has compact support, then $$ \|u(t,\cdot)\|_{L^p} = \|g\|_{L^p(\Bbb R)}\,,\;\;\forall t\in(-T,T) $$
- if $g\geq0$ and has compact support, then $$ \int_\Bbb R f(u(t,x))dx = \int_\Bbb R f(g(x))dx $$ for all $f\in C^1([0,\infty))$.
-
Let $u$ be the solution of the Burgers quaslilinear 1st order PDE
$$
u_t+uu_x=0,\;u(0,x)=g(x)\in C^2(\Bbb R^2)
$$
and suppose that $g'\geq0$ and has compact support (i.e. $g$ is constant
outside some compact set).
Prove that:
- the local solution $u(t,x)$, $t\in(-T,T)$, can be extended to all $t\geq0$;
- $u_x(t,x)\geq0$, and $u_x(t,\cdot)$ has compact support for all $t\geq0$;
- the function $f(t)=\|u(t,\cdot)\|_{L^2(\Bbb R)}$ is decreasing.
- I often mentioned in class that we can always assume locally that every smooth
hypersurface $\Gamma\subset\Bbb R^n$ can be written locally as $x^n=0$.
Prove this fact on the plane, namely let $\Gamma\subset\Bbb R^2$
be the regular level set of a function $h(x,y)$ (namely $dh\neq0$ on every point of $\Gamma$)
and show that, close to any point of $\Gamma$, we can find coordinates $(x',y')$ where
$\Gamma$ writes as $y'=0$.
Hint: you will need to use the implicit function theorem. - Now that you have some experience with 1st order PDEs, go on and prove the
following local existence and uniqueness theorem in two variables:
Let $F\in C^3(J^1(\Bbb R^2,\Bbb R))$, $\Gamma_0=\Bbb R\times\{0\}$,
$g\in C^3(\Bbb R)$ and $(x_0,0,g(x_0),g'(x_0),p_{y,0})$ a point of $F_0=\{F=0\}$
where $\partial_{p_y}F\neq0$. Then there exists a neighborhood $U\subset\Bbb R^2$ of $(x_0,0)$
and a unique function $u\in C^2(U)$ such that:
- $F(x,y,u(x,y),u_x(x,y),u_y(x,y))=0\hbox{ for all }(x,y)\in U$;
- $u(x,0)=g(x)\hbox{ for all }(x,0)\in U$;
- $u_y(x_0,0)=p_{y,0}$.
Hint: Existence. The only thing we did not cover is showing that the surface $\hat\Gamma=\cup_{t}\hat\Gamma_t$ in the notes is really the 1-graph of a function $u(x,y)$, namely you need to show that $p_\alpha(t)=\partial_x u(x(t),y(t))$ for all $t$. Uniqueness. Assume there is a second $C^2$ solution $v$ and check what happens to the restrictions of $u$ and $v$ along the projection $(x(t),y(t))$ of the integral trajectories of $X_F$. Essentially, the uniqueness for this PDE comes from the uniqueness of ODEs solutions.
March 18
Topics
-
A special case of quasilinear 1st order PDEs: Conservation Laws.
- Weak solutions
- Rankine-Hugoniot Condition
Readings
- notes on conservation laws by S. Baskar
- notes on PDEs by Y. Gorb
- Evans, pp. 135-139
March 20
Topics
-
A special case of quasilinear 1st order PDEs: Conservation Laws.
- Shock waves.
Readings
- notes on conservation laws by S. Baskar
- notes on PDEs by Y. Gorb
- Evans, pp. 139-143
March 22
Topics
-
A special case of quasilinear 1st order PDEs: Conservation Laws.
- The Lax-Oleinik formula
Readings
- notes on conservation laws by S. Baskar
- notes on PDEs by Y. Gorb
- Evans, pp. 143-148
March 25
Topics
- General 2nd order PDEs and the Cauchy-Kovaleskaya theorem.
Readings
- notes on PDEs by L. Erdos
- Evans, Section 4.6
March 27
Topics
- General 2nd order PDEs and the Cauchy-Kovaleskaya theorem.
Readings
- notes on PDEs by L. Erdos
- Evans, Section 4.6
March 29
Topics
- Classification of PDEs
Readings
- notes on PDEs by L. Erdos
- The Cauchy-Kovalevskaya Theorem by D. Gaidashev
- On the local solvability of PDEs with simple characteristics by Yu. Egorov
April 1-5
Topics
- More about the Cauchy-Kovalevskaya theorem
Readings
- The Cauchy-Kovalevskaya Theorem by D. Gaidashev
April 8-12
Topics
-
Laplace and Poisson equations:
- Fundamental solution of the Laplace equation
- Solution of the Poisson equation through convolution
- Mean-value formulas
- Maximum principle
- Uniqueness of the solution
- Smoothness of the solution
- Estimates on the derivatives
- Liouville's theorem
- Analiticity of harmoni functions
- Harnack's inequality
Readings
- Evans textbook, Section 2.2
April 15-19
Topics
-
Heat equation:
- Fundamental solution of the homogeneous problem
- Solution of the non-homogeneous problem (Duhamel principle)
- Mean-value formula
- Maximum principle
- Uniqueness of the solution
- Regularity of the solution
Readings
- Evans textbook, Section 2.3
April 22-26
Topics
-
Wave Equation:
- D'Alembert formula
- Domain of dependence and Domain of influence
- Solving the Poisson equation in 1+1 dimensions via coordinate change
- Solving the Poisson equation in 1+n dimensions via Duhamel's principle
Readings
- notes on Hyperblic PDEs by J. Shatah
- Evans textbook, Section 2.4
Take-home Final, due Tuesday, April 30
- Let $v\in C^2(\Omega)\cap C^0(\bar\Omega)$ be a subharmonic function, namely a solution of the partial differential inequality $$\Delta v\geq0.$$ Modify the proof in Evans for harmonic functions to prove that $$ v(x) \leq \frac{1}{Vol(B_r(x))}\int_{B_r(x)}v(y)dy $$ for all $x\in\Omega$ and $r>0$ such that the ball $B_r(x)$ is entirely contained in the open set $\Omega\subset\mathbb R^n$. Then show that, even for subharmonis functions, $$ \max_{x\in\bar\Omega}v(x) = \max_{x\in\partial\Omega}v(x). $$
- Prove the non-uniqueness of solutions of the Heat equation in 1+1 dimensions by showing that every series $$ u_k(x,t) = \sum_0^\infty \frac{g_k^{(n)}(t)}{(2n)!}x^{2n}, $$ with $$ g_k(t) = \begin{cases}e^{-\frac{1}{t^k}},&t>0\\ 0,&t=0\end{cases},\;k>1, $$ converges and solves the boundary value problem $$ u_t=u_{xx},\;u(0,x) = 0. $$ Help: verifying that the series formally solves the heat equation is easy. In order to prove the convergence of the series, use the fact that $$ |g_k^{(n)}(t)|\leq\frac{n!}{(\theta t)^n}e^{-\frac{1}{t^k}} $$ for some $\theta$ depending only on $k$ to find an upper bound for the series $$ \sum_0^\infty \left|\frac{g_k^{(n)}(t)}{(2n)!}x^{2n}\right|. $$
- Consider the 1D wave equation $$ u_{tt}-u_{xx}=0,\;u(0,x)=g(x),\;u_t(,x)=h(x), $$ where $g$ and $h$ have compact support. Prove that the function $$ E(t) = \int_{\mathbb R}\left[u_t^2(t,x) + u^2_x(t,x)\right]dx $$ is constant for any solution of the PDE above. Then prove also that, for $t$ large enough, both terms $$ K(t) = \int_{\mathbb R}u_t^2(t,x) dx, \;\; P(t) = \int_{\mathbb R}u_x^2(t,x) dx $$ are actually constant and equal to each other (equipartition of energy).