I already proved (a) and (b) but can't conclude (c). All I can conclude is that $\zeta$ has infinitely many zeros in $\Bbb C$ not in the critical strip.
@onepotatotwopotato I'm not very good with analytical stuff, but the zeroes of $\xi$ should be exactly the non-trivial zeroes of $\zeta$, because the Gamma function has poles that cancel with the trivial zeroes (i.e. the ones outside the critical strip)
(this is assuming that you know about the poles of the Gamma function and the location of the zeroes of $\zeta$ outside the critical strip)
@LukasHeger If I want to show that subsets of Q that has more than 1 point are not connected , can I argue like this? we can find x<z<y and take two nbhd that of right radius that do not intersect , hence our subset can be written as union of two seperated subsets that are nonempty
I notice the question has been posted in the forum but they argue that z is irrational
The following is Exercise 6.11 in Stein Complex analysis.
Let $f(z) = e^{az}e^{-e^z}$ where $a>0$. Observe that in the strip $\{x+iy:|y|<\pi\}$ the function $f(x+iy)$ is exponentially decreasing as $|x|$ tends to infinity. Prove that
$$\hat{f}(\xi) = \Gamma(a+i\xi),\quad\text{for all}\ \xi\in\Bb...
The weird one that I had was: let $c=\int f$. Define $x=\frac {\bar c}{|c|}$ so that $xc=|c|\implies |c|= x\int f= \int (xf)=\int Re (xf)\le \int |xf|\le \int |f|$
the linked one is intuitive because one knows that $z= |z| e^{i\theta}$
@Raffallo $\cosh(x) = \cosh(-x)$, so there will generally be two solutions for $p$. Using your numbers & my equation, I get $p\approx$ 225.2939 and -71.0939
The problem is from Exercise 7.11 in Stein complex analysis
Let
$$\varphi(x) =\sum_{p\leq x}\log p$$
where the sum is taken over all primes $\leq x$. Prove the following are equivalent as $x\to\infty$:
(i) $\varphi(x)\sim x,$
(ii) $\pi(x)\sim x/\log x,$
(iii) $\psi(x)\sim x,$
(iv) $\psi_1(x)\sim...
@PM2Ring also I'm not sure i I know x value. I know destination delta y and delta x, and value of a. Now I need to calculate parameter p to know how much my curve needs to be moved in the X axis direction to meet my delta x and delta y requirements
@Raffallo In that case, I don't understand your question. I thought you knew x, and p is basically the delta x. If you just have delta y, then x could be anywhere, and there are two delta x values that will correspond to the given delta y. Maybe you need to draw a diagram.
I have two points, where one is always on Y axis and I calculated delta x and delta y between them
also I have my curve without p parameter, which will always have her minimum value at Y axis and also on point 1. Now I need to add p parameter to move my curve along X axis to fit it to my two points. @PM2Ring
https://www.desmos.com/calculator/3xsgjkzpy7 There is and diagram in the online viewer. When you will turn on drawing second curve you will see, that he is almost fitting. The shift on the Y axis is negligible, because I can easily calculate it.
@Raffallo So your point 1 is $x=0$, $y=a$. And so you want to solve for $\Delta x$ in $a+\Delta y = a \cosh\left(\frac{x+\Delta x}{a}\right)$. That gives us $\Delta x = -x \pm a \cosh^{-1}\left(1+\frac{\Delta y}{a}\right)$
BTW, $\cosh^{-1}(x)\equiv \ln(x+\sqrt{x^2-1})$
@NotTfue It looks like there's a $\pi/2$ error. If two lines are perpendicular, the product of their slopes is $-1$.
I've tried to use x+49.506713 and x-203.706713 in the viewer and its not looking good. When you check curve intersection point with x=0 and x=77.10, the delta y is not equal to 7.05. https://www.desmos.com/calculator/3xsgjkzpy7
@PM2Ring I've used a*cosh((x+49.506713)/a) and here is the result. Curve passes through the point 2, but not through the point 1, so the delta y is wrong
@PM2Ring yes, you are right, that Y value is correct for point 2, but in the point 1 curve is no longer have correct y
@PM2Ring hmmm, yea, sounds correct. It can be even the absolute value there, since both results for p should be the same number but one negative and one possitive
I've used wolframalpha.com to calculate p and I've got 65.3441, which looks correct in my example data :) How would look like my final equation to solve? wolfram is giving me very complex formula :/
Let $\mathcal{F}$ be a family of functions of the form $\sum_{n=0}^\infty a_nz^n$ where $|a_n|\leq n$. Then $\mathcal{F}$ is a normal family such that every uniform limit of a given arbitrary sequence $\{f_n\}$ in $\mathcal{F}$ is contained in $\mathcal{F}$.
I wrote something wrong. If $\{f_n(z) = \sum_{k=1}^\infty a_{n,k}z^k\}$ is a sequence of functions contained in $\mathcal{F}$ and $f_n(z)\to f(z) = \sum_{k=1}^\infty a_kz^k$ uniformly, then $\lim_{n\to\infty}a_{n,k} = a_k$ and $|a_{n,k}|\leq k$ for all $n$ implies $|a_k|\leq k$. The family being normal is easy.
@geocalc33 Ok. That's easy. Even up to a million (or ten million) a naive method would be fine.
Summing primes is kinda related to counting primes. But (obviously) the sum of primes is significantly larger than the number of primes.
There's actually a fairly nice sieve-like algorithm for prime sums, so you can get the sum of primes < x without actually having to generate all the primes < x. I've often used it as a way of testing my prime generating code.
We can create the sequence by zipping the primes with an endless cycle of (1, -1). And to make a plot, we don't just want the final sum of the sequence, we want all the cumulative sums of the series, so we use accumulate. Like this:
Suppose $f \in L^{p_0} \cap L^{\infty}$ for some $p_0 < \infty$.
Prove that:
$\lim_{p \rightarrow \infty} \|f\|_p=\|f\|_{\infty}$
My attempt:
So the first step seems to be showing that the limit exists at all. I'm having trouble showing this, I feel like I'm getting a massive case of tunnel ...
$\int |f|^p =\int (|f|^q |f|^{p-q})\le \||f|^q\|_1 \||f|^{q-p}\|_\infty$ (Because Ho:lder's inequality is true here). But not sure how to get what I want to prove.
@TedShifrin Let me check that.
We can for sure say that $\||f|^r\|_\infty \le \|f\|_\infty^r$.
@Koro You can always choose the next point so that any subsequent points are less than ${1 \over 2^{n+1}}$ away, if yoyu see what I mean, so the sum is bounded.
We can for sure say that $\||f|^r\|_\infty \le \|f\|_\infty^r$.
@copper.hat yes, I'd understood that as well. Thanks.
We can take any convergent series of positive numbers $c_n$ and can get a subsequence $(x_{n_j})$ of Cauchy $(x_n)$ such that $|x_{n_{j+1}}-x_{n_j}|\lt c_j$
@TedShifrin I had a question about some notation. In a paper I was reading there was a situation where they wrote $\mathcal{L}_u(f\,dx) = (\nabla f\cdot u + f\nabla \cdot u)dx = (\nabla\cdot fu)dx$, where $u$ is a vector field, $f$ a non-vanishing function, and $dx$ the volume form on $\mathbb R^n$. I am unsure how they did the first equality. The second one is a product rule, but I don't quite understand their use of $\nabla$ alone. I thought the cdot was contraction with u, but I am unsure.
I tried going about it with Cartan's magic formula a few ways, but that gets me $\mathcal L_u(f\,dx) = \iota_u d(f\,dx) + d(\iota_u f\,dx) = \iota_u(\nabla f \wedge dx) + d(f\,\iota_udx)$.
The first term looks like something I'd call $(\nabla f\cdot u)\,dx$, but the rest seems harder to reconcile.
@anak it is product rule. First term gives directional derivative of $f$. Second term is Lie derivative of the volume form, which is div $u$ times volume form. This is often called the divergence theorem. You can get it out of Cartan. $d(\iota_u dx)$ is divergence.
Let $A$ be a Borel set such that $\lambda(A) = 0$, $\delta > 0$ as in definition of absolute continuity and fix $\varepsilon > 0$. There is a decreasing sequence of open sets $U_n$ with $\lambda(U_n) < \delta$ such that $\lambda_F(U_n)\to \lambda_F(A)$
Everyone should learn differential forms and Stokes’s Theorem. That’s undergraduate analysis. Manifolds are ubiquitous. Differential topology with bundles, intersection theory, degree …. Depends on what analysis.
@TedShifrin Okay so I seem to be getting not what I want still. With the div suggestion, clearly I get the $f(\nabla\cdot u)dx$ part. However, the other part was $df\wedge(\iota_udx)$, which seems different from $(\nabla f\cdot u)dx$ in general?
Or does it not matter where you contract a vector field with the interior product?
It's easier to see that $L_uf = \nabla f\cdot u$ without Cartan.
But you can work out the first term. Extend $u$ to a basis $u=v_1,\dots,v_n$ and let $\theta^1,\dots,\theta^n$ be the dual basis. Then write $dx = \phi \,\theta^1\wedge\dots\wedge\theta^n$.
Consider the relation on $\mathbb{R}$ given by $x \sim y$ if $xy \ge 0$. It is correct to say that this is reflexive because $x \sim x$ if $x^2 \ge 0$ (which is true for any $x\in\mathbb{R}$), it is symmetric because $x \sim y$ implies $xy \ge 0$, hence $yx \ge 0$ (because commutativity of product on $\mathbb{R}$) but it is not transitive because for $x=1, y=0$ and $z=-1$ we have $xy=0 \ge 0$, $yz=0 \ge 0$ but $xz=-1<0$?