Apmaths's Blog2026-07-20T23:25:57+07:00http://apmaths.github.io/blogApmathsapmaths@gmail.comComplete Set of Step By Step Listening 1, 2, 3 [Full Ebook+Audio+Answer Key]2024-01-24T00:00:00+07:00http://apmaths.github.io//blog/Complete-Set of Step-By-Step-Listening-1-2-3<p>Complete Set of Step By Step Listening 1, 2, 3 [Full Ebook+Audio+Answer Key]
Step by Step Listening is an extremely useful series in the process of learning and practicing English, especially the Listening skill. Today, we will introduce to readers the content of “Complete Set of Step By Step Listening 1, 2, 3 [Full PDF+Audio+Answer Key].” Let’s explore it together!</p>
<!--more-->
<p>Each Step By Step Listening book includes 3 parts: book+audio+answer key, used for listening practice in all three corresponding levels: starter, mover, flyer. Each book has a total of 24 lessons, with each lesson covering a specific topic. In the Step By Step Listening series, students practice listening to various forms of communication: casual conversation, giving directions, making requests, describing, instructing, apologizing, and suggesting. Alongside these, listening skills are practiced throughout the lessons, including listening for key words, details, and main ideas, listening to questions and providing answers, listening to different viewpoints, listening and drawing conclusions, and listening to identify and grasp information. Link to download the complete set of Step By Step Listening 1,2,3 [Full Ebook + Audio + Answer Key]:</p>
<p><a href="https://drive.google.com/drive/folders/1KHErs7Eb4wAiItVyV4WeVK9eyBFJLpEw">Link</a></p>
<p>Password: 88888996688 </p>
<p>Download links for each part: Step By Step Listening 1 [PDF + Audio + Answer Key]: HERE Step By Step Listening 2 [PDF + Audio + Answer Key]: HERE Step By Step Listening 3 [PDF + Audio + Answer Key]: HERE Above are the related information about Step By Step Listening.</p>
<p>It is hoped that this article will provide readers with useful and valuable information. Good luck with your studies! FAQs: How many books are there in the Step By Step Listening series? The Step By Step Listening series consists of 3 books, including Step By Step Listening 1, Step By Step Listening 2, and Step By Step Listening 3. What is the purpose of compiling the Step By Step Listening series? The Step By Step Listening series helps students practice listening through various forms of communication: casual conversation, giving directions, describing, making requests, instructing, apologizing, and suggesting.</p>
<p>In addition, listening skills are consistently practiced throughout the lessons, including listening for key words, listening to questions and providing answers, listening to different viewpoints, listening and drawing conclusions, and listening to identify and grasp information. Is there a download file for the Step By Step Listening series included in the article? Yes, JES has included a link to download the complete set of Step By Step Listening 1,2,3 [Full Ebook + Audio + Answer Key] - Google Drive for readers to easily study and refer to.”</p>
Phương trình vi phân tuyến tính cấp 12023-12-20T00:00:00+07:00http://apmaths.github.io//blog/phuong-trinh-vi-phan-tuyen-tinh-cap-1<h2>Dạng phương trình vi phân tuyến tính cấp 1 không thuần nhất</h2>
<p>$$y'+p(x)y=q(x)(*)$$</p>
<p>Trường hợp \(q(x)=0\), ta có thể đưa phương trình trở về dạng phương trình vi phân tách biến và giải quyết nó.</p>
<!--more-->
<p>Trường hợp\(q(x)\neq 0\), ta có thể giải phương trình này bằng phương pháp sau</p>
<h3>Phương pháp giải thừa số tích phân</h3>
<p>Đối với phương pháp này ta phải nhớ nhân tử tích phân\(I(x)=e^{\int{p(x)dx}}\) </p>
<p>Ta nhân hai vế của\((*)\) với nhân tử tích phân\(I(x)\), ta được</p>
<p>$$y' e^{\int{p(x)dx}}+p(x)e^{\int{p(x)dx}}=q(x)e^{\int{p(x)dx}}$$</p>
<p>Dễ thấy vế trái sẽ là\(\left( ye^{\int{p(x)dx}} \right)'\), nghĩa là ta có</p>
<p>$$\left( ye^{\int{p(x)dx}} \right)'=q(x)e^{\int{p(x)dx}}$$</p>
<p>Tới đây ta chỉ cần lấy tích phân hai vế là xong. Tức là </p>
<p>$$y=\displaystyle \frac{1}{I(x)}\int{q(x)I(x)dx}$$</p>
<hr>
<h2>Ví dụ</h2>
<p>Ta sẽ đi qua phần ví dụ để nắm rõ cách làm này</p>
<p><b>Ví dụ 1.</b> Giải phương trình\(y'=x-2xy\)</p>
<p><b>Giải</b></p>
<p>Bằng một phép biến đổi đơn giản ta đưa phương trình về thành</p>
<p>$$y'+2xy=x(1)$\)</p>
<p>Rõ ràng đây là phương trình vi phân tuyến tính cấp 1 với \(p(x)=2x\) và\(q(x)=x\)</p>
<p>OK, ta đi tìm thằng đệ thừa số tích phân\(e^{\int{p(x)dx}}\) chính là\(e^{\int{2xdx}}=e^{x^2}\)</p>
<p>Ta nhân hai vế của phương trình\((1)\) cho\(e^{x^2}\), ta được</p>
<p>$\displaystyle y'e^{x^2}+2xye^{x^2}=xe^{x^2}\)</p>
<p>$\displaystyle \Rightarrow \left( ye^{x^2} \right)'=xe^{x^2}\)</p>
<p>Tới đây ta lấy tính phân hai vế, khi đó ta được</p>
<p>$\displaystyle y e^{x^2}=\int{xe^{x^2}dx}\)</p>
<p>$\displaystyle \Rightarrow ye^{x^2}=\frac{1}{2}e^{x^2}+C\)</p>
<p>Hay \(\displaystyle y=\frac{1}{2}+Ce^{-x^2}\)</p>
<p>Giờ thì mọi thứ đã sáng tỏ đúng không nào</p>
<hr>
<p><b>CÁC BÀI VIẾT LIÊN QUAN VỀ PHƯƠNG TRÌNH VI PHÂN</b></p>
<p><a href="/blog/mo-dau-ve-phuong-trinh-vi-phan.html">MỞ ĐẦU VỀ PHƯƠNG TRÌNH VI PHÂN</a></p>
<p><a href="/blog/phuong-trinh-vi-phan-tach-bien.html">PHƯƠNG TRÌNH VI PHÂN TÁCH BIẾN</a></p>
<p><a href="/blog/phuong-trinh-vi-phan-tuyen-tinh-cap-1.html">PHƯƠNG TRÌNH VI PHÂN TUYẾN TÍNH CẤP 1</a></p>
<p><a href="/blog/phuong-trinh-vi-phan-tuyen-tinh-cap-2.html">PHƯƠNG TRÌNH VI PHÂN TUYẾN TÍNH CẤP 2</a></p>
<p><a href="/blog/phuong-trinh-vi-phan-dang-cap.html">PHƯƠNG TRÌNH VI PHÂN ĐẲNG CẤP</a></p>
<p><a href="/blog/phuong-trinh-vi-phan-toan-phan.html">PHƯƠNG TRÌNH VI PHÂN TOÀN PHẦN</a></p>
<div id="fb-root"></div>
<div class="fb-comments" data-href="https://apmaths.github.io/blog/phuong-trinh-vi-phan-tuyen-tinh-cap-1" data-width="auto" data-numposts="10"></div>
Phương trình vi phân toàn phần2023-12-20T00:00:00+07:00http://apmaths.github.io//blog/phuong-trinh-vi-phan-toan-phan<p>Đầu tiên, sẽ trình bày tóm gọn nhất về dạng và cách giải và sau đó là đưa ra ví dụ, để khi cần thì chỉ cần mở ra nhớ và làm, cái này giống như là delta của phương trình bậc hai, nhớ và làm thôi, còn việc tìm hiểu chuyên sâu như thế nào đó là nhu cầu của mỗi người. OK! Let's Go!</p>
<!--more-->
<hr />
<!-------------------------------------------------->
<!----------------DẠNG CỦA PHƯƠNG TRÌNH VI PHÂN TOÀN PHẦN------------->
<h2>Dạng phương trình vi phân toàn phần</h2>
<p>Phương trình vi phân dạng</p>
<p>$$M(x,y)dx+N(x,y)dy=0\quad (1)$$</p>
<p>được gọi là <code>phương trình vi phần toàn phần</code> khi nó thỏa mãn điều kiện: vế trái của phương trình $(1)$ phải là vi phân toàn phần của một hàm khả vi nào đó. Tức là tồn tại một hàm $U(x,y)$ khả vi nào đó sao cho</p>
<p>$$dU(x,y)=M(x,y)dx+N(x,y)dy$$</p>
<p>Điều kiện để một phương trình vi phân dạng $(1)$ trở thành phương trình vi phân toàn phần (hay cách nhận biết phương trình vi phân toàn phần) là:</p>
<p>$$\displaystyle \frac{\partial M}{\partial y}=\frac{\partial N}{\partial x}$$</p>
<p><b>Ví dụ.</b> Phương trình vi phân $(3x^2+6xy^2)dx+(6x^2y+4y^3)dy$ là phương trình vi phân toàn phần vì</p>
<p>$M(x,y)=(3x^2+6xy^2), N(x,y)=(6x^2y+4y^3)$</p>
<p>$\displaystyle \frac{\partial M}{\partial y}=\frac{\partial N}{\partial x}=12xy$</p>
<hr />
<!-------------------CÁCH GIẢI-------------------->
<h2>Cách giải</h2>
<p>Ở đây ta chỉ nêu cách giải và áp dụng, nên mới nói là nó cũng như công thức delta mà thôi. Việc tìm hiểu thêm là nhu cầu của mỗi người.</p>
<p>Nếu phương trình $(1)$ là phương trình vi phân toàn phần thì tích phân tổng quát của phương trình $(1)$ là:</p>
<p>$$\displaystyle U(x,y)=\int_{x_0}^{x}M(x,y_0)dx+\int_{y_0}^{y}N(x,y)dy=C\quad (2.1)$$</p>
<p>hoặc</p>
<p>$$\displaystyle U(x,y)=\int_{x_0}^{x}M(x,y)dx+\int_{y_0}^{y}N(x_0,y)dy=C \quad (2.2)$$</p>
<p>với $(x_0,y_0)$ là một điểm điểm bất kỳ mà khi thay vào các hàm $M(x,y_0), N(x_0,y)$ xác định. Thường thì ta sẽ chọn sao cho thuận tiện trong việc tính tích phân nhất.</p>
<p>Cách nhớ: Công thức này cực kỳ dễ nhớ nếu ta để ý một chút. Trước tiên một cách hình thức ta lấy tích phân hai vế của của phương trình $(1)$, lưu ý chỗ có đuôi vi phân là $dx$ thì cận chạy từ $x_0$ đến $x$, chỗ có phần đuôi là $dy$ thì cận chạy từ $y_0$ đến $y$.</p>
<p>Điều quan trọng nhất là hãy nhớ chỉ được thay $x_0$ hoặc $y_0$ vào <span style="color:red">một trong hai</span> vị trí: phần có đuôi $dx$ thì chỉ được thay $y_0$. Phần có đuôi $dy$ thì chỉ được thay $x_0$.</p>
<hr />
<!-----------------------VÍ DỤ ÁP DỤNG------------------>
<h2>4. Ví dụ áp dụng</h2>
<p>Ví dụ luôn là thứ giải đáp mọi thắc mắc mà phần lý thuyết nếu đọc ta vẫn chưa nắm vững.</p>
<p><b>Ví dụ 1.</b> Giải phương trình:</p>
<p>$$(3x^2+6xy^2)dx+(6x^2y+4y^3)dy \quad (3.1)$$</p>
<p>Trước tiên ta phải kiểm tra điều kiện để phương trình đã cho là phương trình vi phân toàn phần hay không. Nếu có thì ta mới áp dụng được công thức của bài viết này, nếu không thì bài viết này coi như vứt, chẳng giúp ích được gì. (Mách nhỏ là ở trên, lúc nãy hình như kiểm tra là phương trình vi phân toàn phần rồi hay sao đó)</p>
<p>$M(x,y)=(3x^2+6xy^2), N(x,y)=(6x^2y+4y^3)$</p>
<p>$\displaystyle \frac{\partial M}{\partial y}=\frac{\partial N}{\partial x}=12xy$</p>
<p>Vậy $(3.1)$ là phương trình vi phân toàn phần. Ta chọn $(x_0;y_0)=(0,0)$. Khi đó theo công thức $(2.2)$ ta được</p>
<p>$\displaystyle \int_{0}^{x}(3x^2+6xy^2)dx+\int_{0}^{y}4y^3dy=C$</p>
<p>Hay tích phân tổng quát của $(3.1)$ là</p>
<p>$x^3+3x^2y^2+y^4=C$</p>
<hr />
<h2>Một số bài tập có lời giải</h2>
<p><b>Bài tập 1.</b> Giải phương trình $$(y-x)dx+(y^3+x)dy=0$$</p>
<p>$M(x;y)=y-x$</p>
<p>$N(x;y)=y^3+x$</p>
<p>$\dfrac{\partial M}{dy}=\dfrac{\partial N}{dx}=1$ (thỏa mãn điều kiện của phương trình vi phân toàn phần)</p>
<p>Chọn $(x_0;y_0)=(0;0)$. Tích phân tổng quát</p>
<p>$\displaystyle \int_0^x (0-x)dx+\int_0^y (y^3+x)dy=C$</p>
<p>$\displaystyle \Rightarrow \int_0^x(-x)dx+\int_0^y (y^3+x)dy=C$</p>
<p>$\displaystyle \Rightarrow -\frac{x^2}{2}+\frac{y^4}{4}+xy=C$</p>
<p><b>Bài tập 2.</b> Giải phương trình $$\left[ \left( 1+x+y \right)e^x+e^y \right]dx+\left( e^x+xe^y \right)dy=0$$</p>
<p>$\displaystyle M(x;y)=(1+x+y)e^{x}+e^y$</p>
<p>$\displaystyle N(x;y)=e^x+xe^y$</p>
<p>$\displaystyle \frac{\partial M}{dy}=\frac{\partial N}{dx}=e^x$</p>
<p>Chọn $(x_0;y_0)=(0;0)$</p>
<p>Suy ra tích phân tổng quát</p>
<p>$\displaystyle \int\limits_{0}^{x}{\left[ (1+x+0)e^x+e^0 \right]dx}+\int\limits_{0}^{y}{\left( e^x+xe^y \right)dy}=C$</p>
<p>$\displaystyle \Leftrightarrow \int\limits_{0}^{x}{\left[ (1+x)e^x+1 \right]dx}+\int\limits_{0}^{y}{\left( e^x+xe^y \right)dy}=C$</p>
<p>$\displaystyle \Leftrightarrow \left[ (1+x)e^x-e^x+x \right]\left| _{0}^{x} \right.+\left( ye^x+xe^y \right)\left| _{0}^{y} \right.=C$</p>
<p>$\displaystyle \Leftrightarrow e^x(x+y)+xe^y=C$</p>
<div class="fb-comments" data-href="/blog/phuong-trinh-vi-phan-toan-phan" data-width="" data-numposts="10"></div>
PHƯƠNG TRÌNH VI PHÂN TÁCH BIẾN2023-12-20T00:00:00+07:00http://apmaths.github.io//blog/phuong-trinh-vi-phan-tach-bien<hr />
<h2>DẠNG</h2>
<p><code>Phương trình vi phân tách biến</code> (hay còn có thể goi là <code>biến phân li</code>) là phương trình vi phân có dạng
$$M(x)dx+N(y)dy=0 \quad (1)$$</p>
<p>hay $$M(x)N(y)dx+P(x)Q(y)dy=0 \quad (2)$$</p>
<p>hay $$y'=f(x)g(y) \quad (3)$$</p>
<!--more-->
<p>Nhìn thì có vẻ khác nhau nhưng thực chất chỉ là một. Một cách hình thức ta có thể xem $y'=\dfrac{dy}{dx}$ và một vài phép biến đổi ta có thể chuyển qua lại giữa các dạng nói trên.</p>
<p>Có thể nói tóm gọn lại một câu là phương trình vi phân tách biến thì có dạng <code>hàm nhân</code> (tức là tích của hai hàm biến $x,y$ rời nhau).</p>
<p>Biến đổi một chút ở 3 dạng trên để dễ hình dung hơn.</p>
<p>$$(1)\Leftrightarrow y'=\frac{dy}{dx}=-M(x)\times \frac{1}{N(y)} \quad \text{có dạng hàm nhân}$$</p>
<p>$$(2)\Leftrightarrow y'=\frac{dy}{dx}=-\frac{M(x)}{P(x)}\times \frac{N(y)}{Q(y)} \quad \text{có dạng hàm nhân}$$</p>
<p>$$(3)\Leftrightarrow y'=\frac{dy}{dx}=f(x)\times g(y) \quad \text{có dạng hàm nhân}$$</p>
<p>Khi hiểu và hình dung được thì ghi nhớ cái này sẽ hay hơn (mới học thì có thể không quan tâm cũng được)</p>
<p>phương trình vi phân tách biến (biến phân ly) là phương trình vi phân có dạng:</p>
<p>$$M(x,y)dx+N(x,y)dy=0$$</p>
<p>Trong đó các hàm $M(x,y), N(x,y)$ có dạng hàm nhân. Dễ dàng thấy 3 trường hợp trên đều nằm trong này hết. Trong bài tập thì nhìn vào phương trình vi phân ta sẽ dễ thấy 1 trong 3 dạng trên hơn.</p>
<hr />
<h2>2. PHƯƠNG PHÁP GIẢI</h2>
<p>Cách giải phương trình vi phân dạng này rất dễ, vì hai biến $x,y$ là độc lập với nhau nên ta đưa $x$ về một vế, $y$ về một vế rồi ta lấy tích phân hai vế là xong.</p>
<p>Ở đây ta chỉ bàn về hướng giải quyết, còn việc trong quá trình giải đôi khi ta chia cho một hàm nào đó thì dĩ nhiên ta phải xét hàm đó khác không rồi. Bằng không sao mà chia được @@.</p>
<p>Cụ thể:</p>
<p>$$(1) \Leftrightarrow M(x)dx=-N(y)dy \Rightarrow \int{M(x)dx}= -\int{N(y)dx}$$</p>
<p>$$(2) \Leftrightarrow \frac{M(x)}{P(x)}dx=-\frac{Q(y)}{N(y)}dy \Rightarrow \int{\frac{M(x)}{P(x)}dx}=- \int{\frac{Q(y)}{N(y)}dy}$$</p>
<p>$$(3) \Leftrightarrow f(x)dx=\frac{1}{g(y)}dy \Rightarrow \int{f(x)dx}=\int{\frac{1}{g(y)}dy}$$</p>
<hr />
<h2>3. Ví dụ.</h2>
<p>Nhìn trên thì có vẻ rườm rà, nhưng thử bài ví dụ xem, thấy dễ ngay và liền :)</p>
<p><b>Ví dụ 1.</b> Giải phương trình: $$x(y^2-1)dx+y(x^2-1)dy=0\quad (1)$$</p>
<p>Thật không có gì phải bàn để kết luận đây là phương trình vi phân tách biến. Tuy nhiên ta vẫn phân tích một chút. Sẽ là thừa với một số người, nhưng sẽ là cực kỳ dễ hiểu với một số người khác. Caolac nằm trong tốp thứ 2 kakaka...</p>
<p>$M(x,y)=x(y^2-1)$ có dạng hàm nhân.</p>
<p>$N(x,y)=y(x^2-1)$ cũng có dạng hàm nhân.</p>
<p>Theo như phương pháp giải ở trên thì ta cứ quăng hết thằng $x$ về một vế, quăng hết thằng $y$ về một vế và sau đó lấy tích phân.</p>
<p>Với $x^2-1\ne 0, y^2-1 \ne 0$ ta có:</p>
<p>$$(1)\Leftrightarrow \frac{xdx}{x^2-1}=-\frac{ydy}{y^2-1} \Leftrightarrow \int{\frac{xdx}{x^2-1}}=-\int{\frac{ydy}{y^2-1}}\Leftrightarrow \frac{1}{2}\int{\frac{d(x^2-1)}{x^2-1}}=-\frac{1}{2}\int{\frac{d(y^2-1)}{y^2-1}}$$</p>
<p>$$\Leftrightarrow \ln|x^2-1|=-\ln|y^2-1|+\ln C \Leftrightarrow (x^2-1)(y^2-1)=C^2$$</p>
<p>Vậy nghiệm của phương trình đã cho là: $(x^2-1)(y^2-1)=C^2$</p>
<p><b>Nhận xét.</b> Việc giải phương trình vi phân tách biến hay còn gọi là biến phân ly này thực chất chỉ là tính tích phân. Do vậy cần phải nắm vững cách tính tích phân của một số hàm, lớp hàm đặc biệt. Khi đó thì tốc độ làm của chúng ta sẽ tăng lên đáng kể. Caolac sẽ viết riêng một bài về cách tính tích phân của một số lớp hàm đặc biệt và đính kèm link ở đây khi bài viết hoàn thành.</p>
<p><b>Ví dụ 2.</b> Giải phương trình </p>
<p> $$\frac{xdx}{1+x^2}+\frac{ydy}{1+y^2}=0,\quad\left( 1 \right)$$ </p>
<p>Dễ dàng thấy $\left( 1 \right)$ là phương trình vi phân tách biến</p>
<p>$\displaystyle \left( 1 \right)\Rightarrow \frac{xdx}{1+x^2}=-\frac{ydy}{1+y^2}$</p>
<p>Lấy tích phân hai vế ta được</p>
<p>$\displaystyle \Rightarrow \int{\frac{xdx}{1+x^2}}=-\int{\frac{ydy}{1+y^2}}$</p>
<p>$\displaystyle \Rightarrow \frac{1}{2}\int{\frac{d\left( 1+x^2 \right)}{1+x^2}}=-\frac{1}{2}\int{\frac{d\left( 1+y^2 \right)}{1+y^2}}$</p>
<p>$\Rightarrow \ln \left( 1+x^2 \right)=-\ln \left( 1+y^2 \right)+C_1$</p>
<p>$\Rightarrow \ln \left( 1+x^2 \right)+\ln \left( 1+y^2 \right)=\ln C,\left( \ln C=C_1 \right)$</p>
<p>$\Rightarrow \ln \left[ \left( 1+x^2 \right)\left( 1+y^2 \right) \right]=\ln C$</p>
<p>$\Rightarrow \left( 1+x^2 \right)\left( 1+y^2 \right)=C$</p>
<p>Vậy nghiệm tổng quát của phương trình là $\left( 1+x^2 \right)\left( 1+y^2 \right)=C$</p>
<p><b>Ví dụ 3.</b> Giải phương trình $$xy'+y=y^2,\quad \left( 1 \right)$$</p>
<p>$\left( 1 \right)\Rightarrow x\dfrac{dy}{dx}=y^2-y$</p>
<p>$\Rightarrow \dfrac{dy}{y^2-y}=\dfrac{dx}{x}$</p>
<p>Lấy tích phân hai vế</p>
<p>$\displaystyle \Rightarrow \int{\dfrac{dy}{y^2-y}}=\int{\dfrac{dx}{x}}$</p>
<p>$\displaystyle \Rightarrow \int{\left( \dfrac{1}{y-1}-\dfrac{1}{y} \right)dy}=\int{\dfrac{dx}{x}}$</p>
<p>$\Rightarrow \ln \left| y-1 \right|-\ln \left| y \right|=\ln \left| x \right|+C_1$</p>
<p>$\Rightarrow \ln \left| \dfrac{y-1}{y} \right|=\ln \left| x \right|+\ln C=\ln \left| Cx \right|,\left( \ln C=C_1 \right)$</p>
<p>$\Rightarrow Cx=\dfrac{y-1}{y}$</p>
<p>$\Rightarrow Cxy=y-1$</p>
<p>Vậy nghiệm tổng quát là $Cxy=y-1$</p>
<p><b>Ví dụ 4.</b> Giải phương trình $$y'y^2-x^2=0,\quad\left( 1 \right)$$ với điều kiện ban đầu $y\left( 0 \right)=2$</p>
<p>$\left( 1 \right)\Rightarrow y^2\dfrac{dy}{dx}-x^2=0$</p>
<p>$\Rightarrow y^2dy=x^2dx$</p>
<p>Lấy tích phân hai vế</p>
<p>$\displaystyle \Rightarrow \int{y^2dy}=\int{x^2dx}$</p>
<p>$\displaystyle \Rightarrow \frac{y^3}{3}=\frac{x^3}{3}+C_1$</p>
<p>$\Rightarrow y^3=x^3+C,\left( C=3C_1 \right)$</p>
<p>Với điều kiện ban đầu $y\left( 0 \right)=2$, suy ra</p>
<p>$2^3=0^3+C\Rightarrow C=8$</p>
<p>Vậy nghiệm riêng của phương trình ứng với điều kiện ban đầu $y\left( 0 \right)=2$ là: $y^3=x^3+8$</p>
<p><b>Ví dụ 5</b> Giải phương trình $$y'=3x^2y,\quad\left( 1 \right)$$</p>
<p>$\left( 1 \right)\Rightarrow \dfrac{dy}{dx}=3x^2y$</p>
<p>$\Rightarrow \dfrac{dy}{y}=3x^2dx$</p>
<p>Lấy tích phân hai vế</p>
<p>$\displaystyle \Rightarrow \int{\frac{dy}{y}}=\int{3x^2dx}$</p>
<p>$\Rightarrow \ln \left| y \right|=x^3+C_1$</p>
<p>$\Rightarrow \left| y \right|=e^{x^3+C_1}$</p>
<p>$\Rightarrow y=Ce^{x^3},\left( C=e^{C_1} \right)$</p>
<p>Vậy nghiệm tổng quát của phương trình là: $y=Ce^{x^3}$</p>
<p><b>Ví dụ 6.</b> Giải phuơng trình $$y'=x-y+5,\quad \left( 1 \right)$$</p>
<p>$\left( 1 \right)\Rightarrow \dfrac{dy}{dx}=x-y+5$</p>
<p>$\Rightarrow dy=\left( x-y+5 \right)dx,\left( 2 \right)$</p>
<p>Đặt $z=x-y+5\Rightarrow y=x-z+5$</p>
<p>$\Rightarrow dy=dx-dz$</p>
<p>Thay vào $\left( 2 \right)$ ta được</p>
<p>$\Rightarrow dx-dz=zdx$</p>
<p>$\Rightarrow \dfrac{dz}{1-z}=dx$</p>
<p>Lấy tích phân hai vế ta được</p>
<p>$\displaystyle \Rightarrow \int{\dfrac{dz}{1-z}}=\int{dx}$</p>
<p>$\Rightarrow \ln \left| 1-z \right|=x+C_1$</p>
<p>$\Rightarrow \left| 1-z \right|=e^{x+C_1}$</p>
<p>$\Rightarrow 1-z=Ce^{x},\left( C=e^{C_1} \right)$</p>
<p>$\Rightarrow 1-\left( x-y+5 \right)=Ce^{x}$</p>
<p>$\Rightarrow y=Ce^{x}+x+4$</p>
<p>Vậy nghiệm tổng quát của phương trình là: $y=Ce^{x}+x+4$</p>
<hr />
<p><b>CÁC BÀI VIẾT LIÊN QUAN VỀ PHƯƠNG TRÌNH VI PHÂN</b></p>
<p><a href="/blog/mo-au-ve-phuong-trinh-vi-phan.html">MỞ ĐẦU VỀ PHƯƠNG TRÌNH VI PHÂN</a></p>
<p><a href="/blog/phuong-trinh-vi-phan-tach-bien.html">PHƯƠNG TRÌNH VI PHÂN TÁCH BIẾN</a></p>
<p><a href="/blog/phuong-trinh-vi-phan-tuyen-tinh-cap-1.html">PHƯƠNG TRÌNH VI PHÂN TUYẾN TÍNH CẤP 1</a></p>
<p><a href="/blog/phuong-trinh-vi-phan-tuyen-tinh-cap-2.html">PHƯƠNG TRÌNH VI PHÂN TUYẾN TÍNH CẤP 2</a></p>
<p><a href="/blog/phuong-trinh-vi-phan.html">PHƯƠNG TRÌNH VI PHÂN ĐẲNG CẤP</a></p>
<p><a href="/blog/phuong-trinh-vi-phan-toan-phan.html">PHƯƠNG TRÌNH VI PHÂN TOÀN PHẦN</a></p>
First post2023-12-09T00:00:00+07:00http://apmaths.github.io//blog/first-post<p>Well. Finally got around to putting this old website together.
<!--more-->
Neat thing about it - powered by <a href="http://jekyllrb.com">Jekyll</a> and I can use Markdown to author my posts. It actually is a lot easier than I thought it was going to be.</p>
IMO 2023 Problem 52023-08-08T00:00:00+07:00http://apmaths.github.io//blog/imo-2023-p5<p>Let \(n\) be a positive integer. A Japanese triangle consists of \(1 + 2 + \dots + n\) circles arranged in an equilateral triangular shape such that for each \(i = 1\) , \(2\) , \(\dots\) , \(n\) , the \(i^{th}\) row contains exactly \(i\) circles, exactly one of which is coloured red. A ninja path in a Japanese triangle is a sequence of \(n\) circles obtained by starting in the top row, then repeatedly going from a circle to one of the two circles immediately below it and finishing in the bottom row. Here is an example of a Japanese triangle with \(n = 6\) , along with a ninja path in that triangle containing two red circles.
<img width="25%" src="https://i.postimg.cc/8PjB4KhN/38277e856b8643307b2384bfba157420ed1a5fb0.png" border="0" alt="38277e856b8643307b2384bfba157420ed1a5fb0" />
In terms of \(n\) , find the greatest \(k\) such that in each Japanese triangle there is a ninja path containing at least \(k\) red circles.
<!--more-->
<strong>Solution 1.</strong> The answer is \(\left \lfloor \log_2 n \right \rfloor + 1\) . For convenience, we label the rows of the triangle from the top, starting the count from \(1\) . For each red circle \(A\) , we assign it a number corresponding to the maximal amount of red circles that can appear in a ninja path from the top of the triangle to \(A\) . We can see that in a Japanese triangle, there is a ninja path with \(t\) red circles iff there is a red circle assigned with the number \(t\) . We are left to find the maximum number \(k\) that can appear in any Japanese triangle.</p>
<p>Claim 1: If \(A,B\) get assigned by the same number \(t\) , then the sub-triangle with \(A\) as the top can not contain \(B\) and vice versa.</p>
<p>Proof: Kinda straightforward. If the sub-triangle with \(A\) as the top contains \(B\) , then there is a ninja path that goes to \(A\) , and then \(B\) , with at least \(t+1\) red triangles, and \(B\) can not be assigned the number \(t\) .</p>
<p>Claim 2: there are at most \(2^i\) red circles assigned with the number \(i\) .</p>
<p>Proof: We proceed by induction. \(i=1\) is trivial. Assume that for any \(i \le t\) , there are at most \(2^i\) red circles assigned with the number \(i\) . Then, there are at most \(2^{t+1}-1\) red circles assigned with a number less than or equal to \(t\) . Hence, there exists a red circle \(A\) assigned with the number \(t+1\) that lies in a row \(h \le 2^{t+1}\) . By Claim 1, the sub-triangle with \(A\) as the top can not contain any other red circles with the number \(t+1\) . It also means that there are at most \(2^{t+1}-1\) rows either oriented “/” or “" that can contain a red circle of number \(t+1\) . By Claim 1 again, each of these rows can have at most one red circle with the number \(t+1\) . So there can be at most \(2^{t+1}\) red circles assigned with the number \(t+1\) , proving the induction.</p>
<p>The rest of the problem should be straightforward. Due to Claim 2, there is always a red circle assigned with the number \(\left \lfloor \log_2 n \right \rfloor + 1\) in any Japanese triangle with \(n\) rows. Using the similar idea, we can also construct a Japanese triangle where the maximal number assigned to a red circle is \(\left \lfloor \log_2 n \right \rfloor + 1\) . Hence, \(\left \lfloor \log_2 n \right \rfloor + 1\) is the maximal number.
\(\square\)</p>
<p><strong>Solution 2</strong> Let’s denote the circles as \((x,y)\) - the \(y\) th circle on the \(x\) th row, and let \(f(x,y)\) denote the most red we can visit in a path finishing in \((x,y)\) . We observe that adding a new row means that \(f(x,y) = \max(f(x-1,y-1),f(x-1,y))\) , and we need to find the smallest possible value of the largest number on the \(n\) th row.</p>
<p>Claim: On row \(2^z\) , we have a circle with \(f \geq z+1\) , and the sum of all \(f\) is at least \(z \cdot 2^z + 1\) .</p>
<p>We will prove this by induction:</p>
<p>Base: \(n=1\) and \(n=2\) \(\Rightarrow\) OK</p>
<p>Induction step: WLOG(let the largest \(f\) on row \(2^z\) be \((2^z,1)\) ). Adding \(2^z\) rows means that \(f(2^{z+1},i) = z+1\) , where \(i\) is from \(1\) to \(2^z+1\) . (If \(f(2^{z+1},i) \geq z+2\) , then the induction is done.) The rest have a sum of at least \((z \cdot 2^z - z) + 2^z\) (We add \(2^z\) because we have added that many rows, and each row we add, we increase the sum of a \(f\) by \(1\) because of a red circle).</p>
<p>But by pigeonhole principle, we have \(2^z-1\) numbers with a sum of \((z+1) \cdot 2^z - z = (z+1) \cdot (2^z-1) + 1\) , which means at least one circle has \(f \geq z+2\) . Then, the sum of all the circles on row \(2^{z+1}\) is at least \((z+1) \cdot (2^z-1) + 1 + (z+1) \cdot (2^z+1) = (z+1) \cdot 2^{z+1} + 1\) , which satisfies the induction.</p>
<p>Note: If we ever get \(f \geq z+2\) earlier, then the same logic applies, and the sum still ends up \(\geq (z+1) \cdot 2^{z+1} + 1\) .</p>
<p>This means for row \(n\) , we know the answer is \(\geq \left\lfloor \log_2(n) \right\rfloor +1\) . The example below shows that the answer can’t be \(\geq \left\lfloor \log_2(n) \right\rfloor +2\) , so it’s \(\boxed{\left\lfloor \log_2(n) \right\rfloor + 1}\)</p>
<p><img width="50%" src="https://i.postimg.cc/B62YphDm/5019322c458dc32d23ec4aa93069c021e41e47ca.png" border="0" alt="5019322c458dc32d23ec4aa93069c021e41e47ca" /></p>
<div class="fb-comments" data-href="https://hoangnamphan.github.io/blog/imo-2023-p5" data-width="" data-numposts="10"></div>
IMO 2023 Problem 42023-08-01T00:00:00+07:00http://apmaths.github.io//blog/imo-2023-p4<p><strong>Problem 4:</strong> Let \(x_1,x_2,\dots,x_{2023}\) be pairwise different positive real numbers such that</p>
\[a_n=\sqrt{(x_1+x_2+\dots+x_n)\left(\frac{1}{x_1}+\frac{1}{x_2}+\dots+\frac{1}{x_n}\right)}\]
<p>is an integer for every \(n=1,2,\dots,2023.\) Prove that \(a_{2023} \geqslant 3034.\)</p>
<p><strong>Hint:</strong>
Motivation. It is obvious to see that \(3034=2022\times\frac 32+1,\) which leads us to the lemma below \(.\)
<!--more-->
Lemma. For \(\forall n\in\mathbb Z_+,a_{n+2}\geq a_n+3.\)</p>
<p>Lemma Proof. As \(a_n,a_{n+2}\in\mathbb Z_+,\) we only need to prove that \(a_{n+2}>a_n+2.\) Using Cauchy inequality \(,\)
\(\begin{aligned}a_{n+2}=\sqrt{\sum\limits_{k=1}^{n+2}x_k\sum\limits_{k=1}^{n+2}\frac 1{x_k}}&\geqslant\sqrt{\sum\limits_{k=1}^{n}x_k\sum\limits_{k=1}^{n}\frac 1{x_k}}+\sqrt{(x_{n+1}+x_{n+2})\left(\frac 1{x_{n+1}}+\frac 1{x_{n+2}}\right)}\\&=a_n+\sqrt{4+\frac{(x_{n+1}-x_{n+2})^2}{x_{n+1}x_{n+2}}}>a_n+2.\blacksquare\\\end{aligned}\)
Proof. Using Lemma we have \(a_{2023}\geq a_{2021}+3\geq\cdots\geq a_1+3\times 1011=3034.\blacksquare\)</p>
<p><strong>Solution 1:</strong>
Obviously, \(a_n\) should be an increasing (strictly) sequence of positive integers with \(a_1=1\) . We will show that \(a_{n+1}=a_n+1\) and \(a_{n+2} = a_{n+1}+1\) doesn’t hold at the same time. Assume it does.</p>
<p>We know that
\(a_{n+1}^2-a_n^2=x_{n+1}\left(\frac{1}{x_1}+\ldots +\frac{1}{x_{n}}\right) + \frac{1}{x_{n+1}} \left(x_1+\ldots +x_n\right)+1 = 2a_n+1\) and \(2a_n = x_{n+1}\left(\frac{1}{x_1}+\ldots +\frac{1}{x_{n}}\right) + \frac{1}{x_{n+1}} \left(x_1+\ldots +x_n\right) \ge 2a_n\)</p>
<p>by AM-GM, so we should have</p>
\[x_{n+1}\left(\frac{1}{x_1}+\ldots +\frac{1}{x_{n}}\right) = \frac{1}{x_{n+1}} \left(x_1+\ldots +x_n\right) = a_n.\]
<p>Similarly, we can obtain
\(x_{n+2}\left(\frac{1}{x_1}+\ldots +\frac{1}{x_{n+1}}\right) = \frac{1}{x_{n+2}} \left(x_1+\ldots +x_{n+1}\right) = a_{n+1}.\)
That means
\(a_nx_{n+1}+x_{n+1}=x_1+x_2+\ldots +x_n+x_{n+1} = a_{n+1}x_{n+2}\)
and since \(a_{n+1}=a_n+1\) we get that \(x_{n+1}=x_{n+2}\) , which is a contradiction.</p>
<p>Hence, if \(a_{n+1}-a_n=1\) then \(a_{n+2}-a_{n+1}>1\) , or vice versa. In any case, \(a_{n+2}-a_n\ge 3\) and \(a_{2023} \ge a_1+3\cdot 1011 = 3034\) . \(\square\)</p>
<p><strong>Solution without using AM-GM</strong>
Lets prove \(a_{n+1}\) \geq \(a_{n-1}+3\)</p>
\[s = x_1 + x_2 + ... + x_{n-1}\]
\[h= \frac{1}{x_1} + \frac{1}{x_2} + ... + \frac{1}{x_{n-1}}\]
<p>\(a^2_{n-1}\) = hs</p>
\[a^2_{n} = (s+x_n)(h+\frac{1}{x_n})\]
\[a^2_{n} = hs + \frac{s}{x_n} + {x_n}h + 1\]
\[a^2_{n-1} = a_{n-1}^2 + \frac{s}{x_n} + xh + 1\]
<p>assume \(a^2_{n}\) = \(a_{n-1}+1\) ; \(a^2_{n+1}\) = \(a_{n}+1\)</p>
<p>otherwise its trivial since obviously \(a_{n-1}<a_{n}\)</p>
\[2a_{n-1}=\frac{s}{x_n}+xh\]
\[h=\frac{a^2}{s}\]
<p>2 \(a_{n-1}=\frac{s}x+\frac{x}s{a_{n-1}^2}\)</p>
<p>let \(k = \frac{s}x\)</p>
<p>\(2a_{n-1}\) =k+ \(\frac{1}k{a_{n-1}^2}\)</p>
<p>solving for \(a_{n-1}\) , we get:</p>
<p>\(a_{n-1}\) = \(\frac{s}x\)</p>
<p>s= \(a_{n-1}\) x</p>
\[a_{n-1}+1=\frac{s+x}x'\]
\[x'(a+1)=x(a+1)\]
<p>\(x'=x\)
contradiction, so \(a_{n+1}\) \geq \(a_{n-1}+3.\)</p>
<div style="text-align:right;"> <a href="https://artofproblemsolving.com/community/c6t390457f6h3107339_mount_inequality_erupts_on_a_sequence_o
" target="_blank"> Source.</a></div>
<div class="fb-comments" data-href="https://hoangnamphan.github.io/blog/imo-2023-p4" data-width="300" data-numposts="10"></div>
IMO 2023 Problem 32023-07-31T00:00:00+07:00http://apmaths.github.io//blog/imo-2023-p3<p><strong>Problem 3:</strong> For each integer \(k \geqslant 2\), determine all infinite sequences of positive integers \(a_1, a_2, \ldots\) for which there exists a polynomial \(P\) of the form \(P(x)=x^k+c_{k-1} x^{k-1}+\cdots+c_1 x+c_0\), where \(c_0, c_1, \ldots, c_{k-1}\) are non-negative integers, such that</p>
\[P\left(a_n\right)=a_{n+1} a_{n+2} \cdots a_{n+k}\]
<p>for every integer \(n \geqslant 1\).
<!--more--></p>
<p><strong>Solution 1.</strong> The answer is all constant or ascending arithmetic sequences, i.e. \(a_n = b + (n-1)d\) for positive integers \(b\) and nonnegative integers \(d\). These work since we can take \(P(x) = (x + d)(x + 2d) \cdots (x + kd)\). Now we show that no others work.</p>
<p>Before we begin, we note that given \(P\) and \(k\) consecutive values \(a_{n}, \ldots, a_{n+k-1}\) of the sequence, the rest of the sequence can be fully determined, since we have
\(a_{n+k} = \frac{P(a_n)}{a_{n+1}\cdots a_{n+k-1}}\)
and for \(n > 1\), \(a_{n-1} = P^{-1}(a_n\cdots a_{n+k-1})\) (note that \(P\) is increasing and hence injective on the positive integers). Call these two relations \((\star)\).</p>
<p>First, let’s handle the special case \(P(x) = x^k\); we aim to prove that the \(a_n\) are constant. To see this, let \(b\) be the minimum value attained by any \(a_n\). Then, if \(a_n = b\), we have that \(a_{n+1} \cdots a_{n+k} = b^k\) but \(a_{n+i} \geq b\) for all \(i\); therefore \(a_{n+1} = b\). Therefore the sequence is eventually constant at \(b\); by applying \((\star)\) we get that \(a_n = b\) for all \(n\), as desired.</p>
<p>Now assume that some non-leading coefficient of \(P(x)\) is positive. We prove a bunch of lemmas.</p>
<p>Lemma 1. The sequence \((a_n)\) achieves arbitrarily large values.
Proof. Since \(P(x) > x^k\) for all positive integers \(x\), we conclude that \(\max\{a_{n+1},\ldots,a_{n+k}\} > a_n\) for all \(n\). Iterating this proves the claim. \(\blacksquare\)</p>
<p>Lemma 2. For every \(m\), there are only finitely many \(n\) with \(a_n = m\).
Proof. For every such \(n\), we must have \(a_{n+1},\ldots,a_{n+k} \leq P(m)\). Thus, if there are infinitely such \(n\), we can find some \(n < n'\) with \(a_{n+i} = a_{n'+i}\) for all \(1 \leq i \leq k\). By iterating \((\star)\) we get that \(a_n\) is periodic, which contradicts Lemma 1. \(\blacksquare\)</p>
<p>Lemma 3. \(a_{n+1} \geq a_n\) for all \(n\).
Proof. Suppose not. Let \(m\) be the minimum value such that there exists an \(n\) such that \(a_n > a_{n+1} = m\). But then,
\(1 > \frac{P(a_{n+1})}{P(a_n)} = \frac{a_{n+k+1}}{a_{n+1}},\) so \(a_{n+k+1} < a_{n+1}\). Thus, if we let
\(m' = \min \{a_{n+2},\ldots,a_{n+k+1}\},\)
which must be less than \(m\), and \(n'\) be the minimal \(n+2 \leq n' \leq n+k+1\) with \(a_{n'} = m'\), we find that \(a_{n'-1} > a_{n'} = m'\), contradicting the minimality of \(m\). \(\blacksquare\)</p>
<p>Lemma 4. \(a_{n+1} - a_n = O(1)\)
Proof. By Lemma 3, \(a_{n+1}^k \leq a_{n+1} \cdots a_{n+k} = P(a_n)\), so \(a_{n+1} - a_n \leq P(a_n)^{1/k} - a_n\). But this is bounded above. \(\blacksquare\)</p>
<p>By Lemma 4, there are now finitely many possibilities for the tuple \((a_{n+1} - a_n, a_{n+2}-a_n,\ldots,a_{n+k+1} - a_n)\), so there must be one, call it \((d_1,\ldots,d_{k+1})\), which appears infinitely many times. Then we must have
\(P(a_n) = (a_n + d_1) \cdots (a_n + d_k) \text{ and } P(a_n+d_1) = (a_n+d_2)\cdots (a_n + d_{k+1})\)for infinitely many \(n\). By Lemma 2, this means that
\(P(x) = (x + d_1) \cdots (x + d_k) \text{ and } P(x+d_1) = (x+d_2)\cdots (x + d_{k+1})\)for infinitely many \(x\), meaning that the relations above have to hold as polynomials. Now, by Lemma 3 we have \(d_1 \leq \cdots \leq d_{k+1}\). Moreover, since a polynomial can be factored into linear factors in only one way, we conclude that \(d_i = d_{i+1} - d_i\) for all \(1 \leq i \leq k\), i.e. \(d_i = id_1\) for all \(i\). As a result, \(P(x) = (x + d_1)\cdots(x + kd_1)\), and there exists an \(n\) with \(a_{n+i} = a_n + id_1\) for all \(0 \leq i \leq k+1\). By iterating \((\star)\) forwards, we get that there exists an integer \(b\) with \(a_n = b + (n-1)d_1\) for all large \(n\). If \(b > 0\), we can iterate \((\star)\) backwards and conclude. Otherwise, we can iterate \((\star)\) backwards to get some \(n > 1\) with \(a_n \leq d_1\). But then \(a_n \cdots a_{n+k-1} < P(1)\), so there is no possibility for \(a_{n-1}\). So we are done in this case as well. \(\square\)</p>
<p><strong>Solution 2.</strong> The answer is all non-decreasing arithmetic sequences, i.e. \(a_n=b+(n-1)d\) for a positive integer \(b\) and a nonnegative integer \(d\). Clearly those work since we can choose
\(P(x)=\prod_{i=1}^k(x+id).\)
The main claim is as follows.</p>
<p>Claim 1. \(a_n\le a_{n+1}\) for all positive integers \(n\).
Proof. Suppose not. Pick positive integers \(m,n\) such that \(a_n>a_{n+1}=m\) and \(m\) is minimal. As \(c_{k-1},\ldots,c_0\) are all nonnegative, the polynomial \(P(x)\) is strictly increasing over positive integers so
\(1>\frac{P(a_{n+1})}{P(a_n)}=\frac{a_{n+k+1}}{a_{n+1}}.\)It follows that \(m=a_{n+1}>a_{n+k+1}\). Let \(n'\in[n+1,n+k]\) is the maximal integer such that \(a_{n'}\ge m\) and let \(m'=a_{n'+1}\). then \(m'<m\) and \(a_{n'}>a_{n'+1}=m'\), contradiction to the minimality of \(m\). \(\square\)</p>
<p>Now we divide into cases.
Case 1. The sequence \(\{a_n\}\) is bounded. Since \(\{a_n\}\) is weakly increasing, then it is eventually constant, so there exists \(t,c\in\mathbb N\) such that
\(a_n=c\) for all \(n\ge t\). Then we have
\(c^k\le P(c)=P(a_t)=\prod_{i=1}^k P(a_{t+i})=c^k.\)Since equality holds we have \(c_{k-1}=\dots=c_0=0\) so \(P(x)=x^k\). From backwards induction it follows that \(P(t-j)=c\) for all \(1\le j\le t-1\). Hence \(\{a_n\}_{n\in\mathbb N}\) is constant.
Case 2. \(\{a_n\}\) is unbounded. Let \(c\doteqdot\sum_{i=0}^{k-1}c_i\). Observe that for every positive integer \(n\) we have
\(P(n)\le n^k+cn^{k-1}=(n+c)n^{k-1}\)Thus, for every positive integer \(n\),
\((a_n+c)a_n^{k-1}\ge P(a_n)=\prod_{i=1}^ka_{n+i}\ge a_{n+k}a_n^{k-1}.\)Hence \(a_{n+k}\le a_n+c\).</p>
<p>Claim 2. Let \(e_1,e_2,\ldots,e_k\) be nonnegative integers such that there exist infinitely positive integers \(n\) for which \(a_{n+i}-a_n=e_i\) for all integer \(i\in[1,n]\). Then \(P(x)=\prod_{i=1}^n (x+e_i)\).
Proof. By assumption, there exists an infinite subset \(X\subset\mathbb N\) such that
\((a_{n+1}-a_n,a_{n+2}-a_n,\ldots,a_{n+k}-a_n)=(e_1,\ldots,e_k)\)for every \(n\in X\). Then for every \(n\in X\) we have
\(P(a_n)=\prod_{i=1}^k a_{n+i}=\prod_{i=1}^k (a_n+e_i).\)Hence the equation \(P(x)=\prod_{i=1}^n (x+e_i)\) has a solution set \(\{a_n\mid n\in X\}\). Moreover, as \(\{a_n\}_{n\in\mathbb N}\) is nondecreasing and unbounded, the set \(\{a_n\mid n\in X\}\) is unbounded as well, so \(P(x)=\prod_{i=1}^n (x+e_i)\) in \(\mathbb Z[x]\). \(\square\)</p>
<p>Claim 3. Let \(d\) be a nonnegative integer such that there exist infinitely many positive integers \(n\) for which \(a_{n+1}-a_n=d\). Then \(d\) is the smallest nonnegative integer such that \(P(-d)=0\).
Proof. By assumption, there exists an infinite subset \(X\subset\mathbb N\) such that \(a_{n+1}-a_n=d\) for every \(n\in X\). Define a function \(f\colon X\to\mathbb Z_{\ge 0}^k\) by
\(f(n)\doteqdot (a_{n+1}-a_n,a_{n+2}-a_n,\ldots,a_{n+k}-a_n).\)As \(a_{n+i}-a_n\le c\) for all \(n\in X\) and \(i\in[1,k]\), the image of \(f\) is finite. Hence by pigeonhole, there exists an infinite subset \(Y\subset X\) and nonnegative integers \(e_1,e_2,\ldots,e_k\) such that \(f(n)=(e_1,\ldots,e_k)\) for every \(n\in Y\). By claim 2 we have \(P(x)=\prod_{i=1}^n (x+e_i)\). As \(d=e_1\le e_2\ldots\le e_k\), the number \(d\) is the smallest with property \(P(-d)=0\). \(\square\)</p>
<p>Now define \(d_n\doteqdot a_{n+1}-a_n\) for each positive integer \(n\). As \(0\le d_n\le c\), the sequence \(\{d_n\}_{n\in\mathbb N}\) is bounded, so there exists a nonnegative integer \(d\) which occurs infinitely many in \(\{d_n\}_{n\in\mathbb N}\). By claim 3, \(d\) is unique so by boundness again, the sequence \(\{d_n\}_{n\in\mathbb N}\) is eventually constant with value \(d\). Formally, there exists a positive integer \(m\) such that \(d_n=d\) for all \(n\ge m\). Pick such a minimal \(m\). If \(m=1\) then we are done, so assume \(m\ge 2\) for contrary. Claim 2 implies that \(P(x)=\prod_{i=1}^k(x+id)\). Let \(Q(x)\doteqdot \prod_{i=1}^k(x+(i-1)d)\), hence \(P(x)=Q(x+d)\). By hypothesis we have
\(Q(a_{m-1}+d)=P(a_{m-1})=\prod_{i=1}^k a_{m-1+i}=\prod_{i=1}^k(a_m+(i-1)d)=Q(a_m).\)As the polynomial \(Q(x)\) is strictly increasing over positive integers, we have \(a_{m-1}+d=a_m\) or \(d_{m-1}=d\), contradiction.
To summarize, in both cases we have proved that \(a_1,a_2,\ldots\) is an arithmetic nondecreasing sequence, hence we are done.</p>
<p><strong>Solution 3.</strong> The answer is nondecreasing arithmetic sequences only. These work, since if the common difference is \(d\), we can write the polynomial \((x+d)\ldots(x+kd)\). We now prove that they are the only ones.</p>
<p>First, if all the \(c_i\) are zeroes, we may pick some prime \(p \mid a_1\) and examine the \(p\)-adic valuation of the sequence, which satisfies \(k\nu_p(a_n)=\nu_p(a_{n+1})+\cdots+\nu_p(a_{n+k})\). Since these should all be nonnegative integers, pick \(n\) with \(\nu_p(a_n)\) minimal; then \(\nu_p(a_{n+1}),\ldots\) are all minimal by induction, so working backwards we find that \(\nu_p(a_{n-1}),\nu_p(a_{n-2}),\ldots\) are all minimal by induction as well, i.e. the sequence is constant. Thus suppose at least one of the \(c_i\) is positive.</p>
<p>Claim 1: There does not exist a constant \(C\) such that we can find arbitrarily large \(n\) with</p>
\[a_{n+1},\ldots,a_{n+k} \leq C.\]
<p>Proof: For every \(n\) there exists some \(1 \leq i \leq k\) such that \(a_{n+i}>a_n\), else \(P(a_n)\leq a_n^k\). Then, we can essentially repeat this to find some \(a_m>C\), and then repeat it more to get some sequence of terms separated by distance at most \(k\), all exceeding \(C\). \(\blacksquare\)</p>
<p>Claim 2: There does not exist a constant \(C\) such that we can find arbitrarily large \(n\) with \(a_n \leq C\), i.e. every term in the sequence occurs finitely many times.
Proof: If such a \(C\) exists, pick \(n\) large with \(a_n \leq C\). By claim \(1\), we can find some \(1 \leq i \leq k\) such that \(a_{n+i}\) is massive, which is a contradiction by taking \(a_{n+i}>P(C)\). \(\blacksquare\)</p>
<p>The second claim allows us to find infinitely many \(n\) such that \(a_n<a_{n+1},a_{n+2},\ldots\); call such \(n\) endangered. Clearly, \(a_n\) across \(n\) endangered can grow arbitrarily large.</p>
<p>For every endangered \(n\), consider the differences \(a_{n+1}-a_n,\ldots,a_{n+k}-a_n\), which must be at least one. On the other hand, if \(a_n\) is sufficiently large, they must be at most \(100c_{k-1}+100\), so they are all bounded in some interval. Therefore, by Pigeonhole, some multiset of differences \(\{d_1,\ldots,d_k\}:=S\) with \(0<d_1\leq \cdots \leq d_k\) must occur infinitely many times. Writing the RHS as a polynomial in \(a_n\), we get that</p>
\[P(a_n)=(a_n+d_1)\ldots(a_n+d_k)\]
<p>for infinitely many choices of \(a_n\), hence this equality must hold in the polynomial sense.</p>
<p>Now we consider \(a_{n+1}\), where \(n\) is endangered and \(\{a_{n+1}-a_n,\ldots,a_{n+k}-a_n\}=S\). Let \(\{a_{n+2}-a_{n+1},\ldots,a_{n+k}-a_{n+1},a_{n+k+1}-a_{n+1}\}=S'\) (both sides are multisets), where \(d_1' \leq \cdots \leq d_k'\). The elements \(a_{n+2}-a_{n+1},\ldots,a_{n+k}-a_{n+1}\) are bounded (in both directions), since they are in \(S-S\), hence for \(a_{n+1}\) sufficiently large, \(a_{n+k+1}-a_{n+1}\) must be bounded in both directions as well, otherwise we get a size contradiction. Therefore by Pigeonhole some choice of differences \(\{d_1',\ldots,d_k'\}:=S'\) appears infinitely many times, for infinitely many \(a_{n+1}\) (since \(a_{n+1}>a_n\) grows arbitrarily large), so we get that \(P(a_{n+1})=(a_{n+1}+d_1')\ldots(a_{n+1}+d_k')\) holds infinitely many times, hence it must be a polynomial equality. But since polynomials have unique factorizations, it follows that \(S=S'\). It is then clear that \(a_{n+1}-a_n=d_1\), since otherwise \(S' \ni d_1-(a_{n+1}-a_n)\leq 0\), but \(S\) contains only positive integers. From here, we find that \(d_2-d_1=d_1'\), since if it is less then it can’t be in \(S'\) and if it is more then there is no way we can find \(d_1'\) in \(S-d_1\), and by induction we have \(d_3-d_1=d_2'\) and so on, until \(d_k-d_1=d_{k-1}'\), implying that \(a_{n+k+1}-a_{n+1}=d_k'\). This actually implies that \(d_i=id_1\) for all \(1 \leq i \leq k\).</p>
<p>Let \(d=d_1\) for convenience. We will perform a “parallel” induction on the two conditions:</p>
\[\{a_{m+1},\ldots,a_{m+k}\}=\{a_m+d,\ldots,a_m+kd\}\]
\[a_{m+1}=a_m+d\]
<p>For \(m=n\), where \(n\) is endangered, we just proved that these two conditions are true. Pick \(n\) to be large enough as a base case.
Now, if the two conditions are true for \(m\), since \(a_{m+1}=a_m+d\) by hypothesis, we have
\(\{a_{m+2},\ldots,a_{m+k}\}=\{a_m+2d,\ldots,a_m+kd\}=\{a_{m+1}+d,\ldots,a_{m+1}+(k-1)d\}.\)By dividing out both sides of the given equation by \(a_{m+2}\ldots a_{m+k}\), it follows that \(a_{m+k+1}=a_{m+1}+kd,\) so the first condition is true for \(m+1\). Let \(a_{m+2}=a_{m+1}+id\), so then</p>
\[\{a_{m+3},\ldots,a_{m+k+1}\}=\{a_{m+1}+(1-i)d,\ldots,a_{m+1}+(-1)d,a_{m+1}+d,\ldots,a_{m+1}+(k-i)d\}.\]
<p>If \(i \geq 2\), then \(a_{m+1}-d\) is actually in the above set. But by the Euclidean algorithm, since \(d\) is not a root of \(P\), we can write \(P(x)=(x-d)Q(x)+R(x)\) where \(R\) is a nonzero constant polynomial (not dependent on \(n,m\)), so then for \(a_{m+1}\) large, \(\tfrac{P(a_{m+1})}{a_{m+1}-d}\) cannot be an integer, but the product of all the other elements in the above set is clearly integral: contradiction. Hence \(i=1\), so \(a_{m+2}=a_{m+1}+d\), completing the induction.</p>
<p>It now follows that the sequence \(a_1,a_2,\ldots\) is eventually arithmetic (with common difference \(d\)). On the other hand, since the coefficients of \(P\) are all nonnegative, and at least one is positive, it is strictly increasing in \(\mathbb{R}^+\). Therefore, given \(a_{n+1},\ldots,a_{n+k}\), there is exactly one possible choice for \(a_n\) (which is not necessarily an integer). On the other hand, if \((a_{n+1},\ldots,a_{n+k})=(a+d,\ldots,a+kd)\), then it is clear that setting \(a_n=a\) will work, so we must have \(a_n=a\). Therefore, by inducting “downwards”, with base case \(m=n\) and inductive step \(m \to m-1\), it follows that the entire sequence is an arithmetic progression with common difference \(d\), finishing the problem. \(\blacksquare\)</p>
<div style="text-align:right;"> <a href="https://artofproblemsolving.com/community/c6t390457f6h3106754_imo_2023_p3" target="_blank"> Source.</a></div>
<div class="fb-comments" data-href="https://hoangnamphan.github.io/blog/imo-2023-p3" data-width="300" data-numposts="10"></div>
IMO 2023 P12023-07-28T00:00:00+07:00http://apmaths.github.io//blog/imo-2023-p1<p>
<b>Question:</b> Determine all composite integers \(n>1\) that satisfy the following property: if $$d_1, d_2, \ldots, d_k$$ are all the positive divisors of \(n\) with \(1= d_1 < d_2< \cdots < d_k = n\), then \(d_i\) divides $$d_{i+1}+d_{i+2}$$ for every \(1 \leqslant i \leqslant k-2.\)
</p>
<p> Read full article <a href="https://artofproblemsolving.com/community/c6h3106752" target="_blank">here </a>. </p>
<!--more-->
<h2>Solution 1:</h2>
<p>
Note that the three largest divisors of \(n\) are either $$\left \{ \frac{n}{q}, \frac{n}{p}, n \right \}$$ for distinct prime divisors \(p\) and \(q\) of \(n\), or $$\left \{ \frac{n}{p^2}, \frac{n}{p}, n \right \}$$ for some prime divisor \(p\) of \(n.\)</p>
<p>
In the former case we have a contradiction, since</p>
<p>
$$\frac{\frac{n}{p} + n}{\frac{n}{q}} = \frac{q(p+1)}{p},$$</p>
<p>
obviously not an integer.</p>
<p>
So, the three largest divisors of \(n\) are \(\frac{n}{p^2}, \frac{n}{p}\) and \(n\).</p>
<p>
With similar reasoning, now consider the fourth largest divisor. It is either \(\frac{n}{q}\) or \(\frac{n}{p^3}\); the former option fails, since</p>
<p.>
$$\frac{\frac{n}{p} + \frac{n}{p^2}}{\frac{n}{q}} = \frac{q(p+1)}{p^2},$$</p>
obviously not an integer either.
</p>
<p>
We repeat this reasoning to deduce that \(n\) is just \(p^{k-1}\).</p>
<h2>Solution 2: </h2>
<p>
Observe that we have \(d_i | d_{i+1}+d_{i+2} , 1 \leq i \leq k-2\), then put \(i=k-2\) then \(d_{k-2} | d_{k-1}+d_{k}=d_{k-1}+n\) but $$d_{k-2} |n \Longrightarrow d_{k-2} | d_{k-1} .$$</p>
<p>
Again putting \(i=k-3\) we get \(d_{k-3} | d_{k-2}+d_{k-1}\) but on other hand $$d_2=\frac{n}{d_{k-1}} | \frac{n}{d_{k-2}} =d_3 , d_2|d_3+d_4 \Longrightarrow d_2|d_4.$$ So $$d_{k-3}|d_{k-1} \Longrightarrow d_{k-3}|d_{k-2}.$$ </p>
<p>
Now using Induction we get $$d_1|d_2|d_3.......d_k.$$ Hence \(n=p^{k-1}\) is the solution.
</p>
<div class="fb-comments" data-href="https://hoangnamphan.github.io/blog/imo-2023-p1" data-width="350" data-numposts="10"></div>