Notice
Recent Posts
Recent Comments
Link
«   2026/06   »
1 2 3 4 5 6
7 8 9 10 11 12 13
14 15 16 17 18 19 20
21 22 23 24 25 26 27
28 29 30
Tags
more
Archives
Today
Total
관리 메뉴

개발일기

오일러-라그랑주 방정식, 최소 작용의 원리 본문

공부

오일러-라그랑주 방정식, 최소 작용의 원리

kimjw7815 2025. 4. 21. 15:10

오일러 라그랑주 방정식을 설명하려고 생각을 많이 해봤다.
임의의 경로를 나타내는 함수는 $x(t)$이고, 라그랑지안은 $L(x(t),\dot{x}(t),t)$로 나타나고, 작용을 미분할 때는 시간 $t$와는 또 다른 변수 $\epsilon$이 튀어나와서 미분 변수가 된다.
그러면 $t$에 대한 매개변수의 함수의 미분을 다루는 게 맞을까. 내가 보기엔 $f(x,y,t)$의 미분에 대해 먼저 설명하고, 매개변수 함수의 미분을 설명하고, 그걸 엮어서 라그랑지안 $\epsilon$ 미분을 설명하는 게 맞을 것 같은데.
 
우선 글을 읽기 전, 각각의 단락마다 함수와 변수를 다 다르게 설정하므로, 끊어서 읽기를 추천하고, 각 단락에서 말하고자 하는 바만 각 단락에서 이해하고 넘어간다면 글을 읽는 데에 큰 어려움은 없을 것이다.
이해가 안 된다면 gpt에게 물어보도록 하자. AI는 좋은 문명

다변수 함수의 미분은 어떻게 하는가?

$f(x,y,z)$의 미분에 대해 생각해보자.
만약 $f(x,y,z)$가 각각의 변수에 대하여 $f(x,y,z)=f_1(x)+f_2(y)+f_3(z)$로 나타낼 수 있다면, 각각의 미분은 다음과 같을 것이다.
$\displaystyle \frac{d}{dx}f(x,y,z)=\frac{d}{dx}f_1(x)$
$\displaystyle \frac{d}{dy}f(x,y,z)=\frac{d}{dy}f_2(y)$
$\displaystyle \frac{d}{dz}f(x,y,z)=\frac{d}{dz}f_3(z)$
$x$를 미분 변수로 두면, $y$나 $z$는 상수일 테니까.
 
만약 $f(x,y)$가 $f(x,y)=x^2y$처럼 두 다른 변수의 비선형 연산의 형태로 나타난다고 해도, 미분 변수가 아닌 다른 변수를 상수로 두고 똑같이 미분해주면 될 것이다.
$\displaystyle \frac{d}{dx} f(x,y)=2xy, \frac{d}{dy} f(x,y)=x^2$
사실 이거는 그냥 미분이 아니라 편미분이라고 한다. 교과서에서 알려주는 $y$를 $x$의 함수로 두고 미분하는 음함수의 미분법과는 다른 개념이다. 기호로는 이렇게 쓴다.
$\displaystyle \frac{\partial}{\partial x} f(x,y,z), \frac{\partial}{\partial y} f(x,y,z), \frac{\partial}{\partial z} f(x,y,z)$
각각 $x$, $y$, $z$에 대한 편미분을 나타낸다.
 

매개변수 함수의 미분은 어떻게 하는가?

이번엔 $x$와 $y$를 다른 변수에 대한 함수로 보고, 임의의 함수 $x(t), y(t)$를 받는 함수 $f(x(t),y(t))$가 있다고 치자.
$\displaystyle \frac{d}{dt} f(x(t),y(t))=\frac{\partial f}{\partial x} \frac{dx}{dt}+\frac{\partial f}{\partial y} \frac{dy}{dt}$
이게 성립한다. 이해가 잘 안 되면 예제로 $\displaystyle \frac{d}{dt}f(x,y)=\frac{d}{dt}(x^2+y^2)$을 떠올려보자. 한 눈에 보일 것이다.
$\displaystyle \frac{d}{dt} f(x,y) =2xx'+2yy'=\frac{\partial f}{\partial x} \frac{dx}{dt}+\frac{\partial f}{\partial y} \frac{dy}{dt}$
 
 

부분적분

$\displaystyle \int_{x_1}^{x_2} \ln{x} dx$를 풀려면 어떻게 할까. 이걸 어떻게 풀 수 있을까.
다시 미분으로 돌아가서, $\displaystyle \frac{d}{dx}f(x)g(x)=f'(x)g(x)+f(x)g'(x)$를 이용하면 쉽게 풀 수 있다.
저 수식에 인테그랄을 다시 끼워주면, $\displaystyle f(x)g(x)=\int f'(x)g(x)dx+\int f(x)g'(x)dx$이다.
$\ln{x}$가 적분 안에 들어가 있는게 귀찮으니까, 이 놈을 미분해주면 간단해지는 걸 이용해서 $g(x)=\ln{x}$로 가정해보자. 즉, 가운데 항인 $\displaystyle \int f'(x)g(x) dx$를 우리가 원래 구하던 식으로 두어보자. 그러면 자연스럽게 $g(x)$와 붙어있는 $f'(x)$가 $1$이 되야한다는 걸, 즉 $f(x)=x$가 되어야한다는 걸 알 수 있다.
이 때 다시 미분 식으로 돌아가면, $\displaystyle x\ln{x}=\int \ln{x}dx+\int 1dx$가 성립하므로, 간단하게 원래식에 대해서 정리해주면
$\displaystyle \int \ln{x}dx=x\ln{x}-x+C, \int_{x_1}^{x_2} \ln{x}dx=\Big[x\ln{x}-x\Big]_{x_1}^{x_2}$
임을 알 수 있다. 이게 바로 부분적분이다. 물론, $\ln{x}$가 아닌 다른 식에도 적용 가능하다. 보통은 다음과 같이 쓴다.
$\displaystyle \int f'(x)g(x)=f(x)g(x)-\int f(x)g'(x)dx$
 

오일러-라그랑주 방정식

여기서 약간의 조건들만 추가해서 오일러-라그랑주 방정식을 유도해보자.
우리는 $y(t)$를 $x(t)$의 $t$에 대한 미분인 $\dot{x}(t)$로 가정할 것이다.
$x(t)$ 자체도, 임의의 변수 $\epsilon$과 임의의 함수 $\eta(t)$에 대하여 $x(\epsilon,t)=x(0,t)+\epsilon\eta(t)$로 가정할 것이다.
또한 전체 함수도 $\displaystyle g[x(\epsilon, t)]=\int_{t_1}^{t_2} f(x(\epsilon,t),\dot{x}(\epsilon,t),t)dt$로 가정하고, $\eta(t_1)=\eta(t_2)=0$으로 끝점도 고정해줄 것이다.
조건이 좀 많고 난해해보일 수 있지만, 위의 내용들만 기억해두면 그렇게 어렵지 않을 것이다.
이 때 한번 $\displaystyle \frac{d}{d\epsilon}g[x(\epsilon, t)]$를 구하고, 그게 0이 되는 지점을 찾아보자. 이것이 이후 물리적인 직관으로 연결될 수 있는 부분이니, 잘 기억해두자.
 
우선 적분 안으로 $\displaystyle \frac{d}{d\epsilon}$를 옮겨주면, 식은 $\displaystyle \int_{t_1}^{t_2} \frac{d}{d\epsilon}f(x(\epsilon,t),\dot{x}(\epsilon,t),t)dt$가 될 것이다.
변수가 3개지만, 아까 했듯이 똑같이 하면 된다. 세번째 인수 $t$는 $\epsilon$과 독립 변수이기에 미분하면 0이니 고려하지 않으면 된다. 그러면 적분 안의 수식은 다음과 같이 나타날 것이다.
$\displaystyle \frac{\partial f}{\partial x} \frac{dx}{d\epsilon}+ \frac{\partial f}{\partial \dot{x}} \frac{d\dot{x}}{d\epsilon} $
아까 위에서 보여준 매개변수 함수의 미분을 그대로 따른 형태이다. 각 매개변수 함수 대한 전체 함수의 편미분에 각각 매개변수 함수의 미분이 곱해진 형태이다.
 
이 때, $\displaystyle \frac{dx}{d\epsilon}, \frac{d\dot{x}}{d\epsilon} $를 보자. $x(\epsilon,t)=x(0,t)+\epsilon\eta(t)$이므로, 미분해주면 남는 건 $\eta(t)$뿐일 것이다. $x(0,t)$에는 $\epsilon$이 없으니 상수 취급할 수 있고, $\eta(t)$를 상수로 보았을 때 $\epsilon\eta(t)$가 일차식의 형태이기 때문이다.
$\displaystyle\dot{x}(\epsilon,t)=\dot{x}(0,t)+\epsilon\dot{\eta}(t)$이기 때문에, 이것도 미분해주면 마찬가지로 $\dot{\eta}(t)$만 남으리라는 걸 알 수 있다.
이 때 적분 안의 수식을 다시 쓰면, $\displaystyle \frac{\partial f}{\partial x}\eta+\frac{\partial f}{\partial \dot{x}}\dot{\eta}$이다.
 
여기서 $\dot{\eta}$를 $\eta$로 나타내어보자. 식은 원래 적분 안에 있었으므로, 저 항만 적분식을 유지한 채 따로 떼어오면 $\displaystyle \int_{t_1}^{t_2} \frac{\partial f}{\partial \dot{x}} \frac{d\eta}{dt} dt$가 되겠다. 여기서 부분적분을 써줄 수 있겠다. $\displaystyle \frac{d\eta}{dt}$를 이미 미분이 되어진 $g'(t)$로 본다면, 부분적분에 대입했을 때 식은 다음과 같이 정리될 것이다.
$\displaystyle \int \frac{\partial f}{\partial \dot{x}} \frac{d\eta}{dt}=\left[\frac{\partial f}{\partial \dot{x}} \eta(t)\right]_{t_1}^{t_2}-\int_{t_1}^{t_2} \frac{d}{dt}\left(\frac{\partial f}{\partial \dot{x}}\right) \eta dt$
이 때, 위의 조건 $\eta(t_1)=\eta(t_2)=0$을 이용하면, 적분 부분만 남는 걸 알 수 있다. 마침내 다시 원래 식으로 돌아가보자.
 
지금까지 정리한 걸 모두 적용한다면, 식은 다음과 같이 전개할 수 있다.
$\displaystyle \frac{d}{d\epsilon}g[x(\epsilon, t)]=\frac{d}{d\epsilon} \int_{t_1}^{t_2} f(x(\epsilon,t),\dot{x}(\epsilon,t),t)dt = \int_{t_1}^{t_2} \frac{\partial f}{\partial x} \eta dt+ \int_{t_1}^{t_2}\frac{\partial f}{\partial \dot{x}} \dot{\eta} dt = $
$\displaystyle \int_{t_1}^{t_2} \eta(t) \left(\frac{\partial f}{\partial x}-\frac{d}{dt}\left(\frac{\partial f}{\partial \dot{x}}\right)\right)dt=0$
이 때, $\eta(t)$는 임의의 함수이므로, $\displaystyle \frac{d}{d\epsilon} g[x(\epsilon,t)]=0$가 항상 성립하려면, $\eta(t)$의 계수인 $\displaystyle  \frac{\partial f}{\partial x}-\frac{d}{dt}\left(\frac{\partial f}{\partial \dot{x}}\right)$가 항상 0이 되어야한다. 즉
$\displaystyle \frac{\partial f}{\partial x}-\frac{d}{dt}\left(\frac{\partial f}{\partial \dot{x}}\right)=0$이다.
 
이제 함수들의 모양만 바꾸어보자. $f$는 $L$로, $g$는 $S$로 바꾸어보자.
그럼 다음과 같이 써진다.

$\displaystyle S[x(\epsilon, t)]=\int_{t_1}^{t_2} L(x(\epsilon,t),\dot{x}(\epsilon,t),t)dt$에 대하여
$\displaystyle \frac{d}{d\epsilon} S[x(\epsilon,t)]=0$가 항상 성립하려면
$\displaystyle \frac{\partial L}{\partial x}-\frac{d}{dt}\left(\frac{\partial L}{\partial \dot{x}}\right)=0$이어야 한다.

이게 오일러-라그랑주 방정식이다. 이 때, $S[x(\epsilon,t)]$는 특정한 경로에 대한 작용량이고, $L(x(\epsilon,t),\dot{x}(\epsilon,t),t)$는 라그랑지안이다. $x(0,t)$는 최소 거리를 가지는 경로이고, $\eta{t}$는 임의의 경로이다. 경로의 변칙율에 대한 작용의 민감도가 극소점을 가져 경로가 안정될 때, 위의 수식이 성립해야한다는 것이다. 즉 위의 정리는 이와 같은 작용이 최소화 되는 조건을 수학적으로 서술한 것이다.