Multivariate chain rules

-


Next, we consider differentiating a composite function such as \(f(\varphi(t), \psi(t))\) or \(f(\varphi(u, v), \psi(u, v))\). In the case of univariate functions, we have the chain rule. That is,

\[\frac{dg(f(x))}{dx} = g'(f(x))\cdot f'(x).\]

We have corresponding multivariate versions.

Theorem [Chain rule (1)]

Let \(f(x,y)\) be a totally differentiable function on an open region \(U (\subset \mathbb{R}^2)\). Let \(x = \varphi(t), y = \psi(t)\) be differentiable functions on an open interval \(I\). Suppose that \((\varphi(t), \psi(t))\in U\) for all \(t \in I\). Then the function \(z = f(\varphi(t), \psi(t))\) of \(t\) on \(I\) is differentiable on \(I\), and its derivative is given by

\[\frac{d}{dt}f(\varphi(t),\psi(t)) = \frac{\partial f}{\partial x}(\varphi(t),\psi(t))\frac{d\varphi}{dt}(t) + \frac{\partial f}{\partial y}(\varphi(t),\psi(t))\frac{d\psi}{dt}(t).\tag{Eq:ChainRule1}\]

Remark.  Eq. (Eq:ChainRule1) may be a little too cluttered and difficult to read. We could simplify the notation by using dependent variables:

\[\frac{dz}{dt}(t) = \frac{\partial z}{\partial x}(x,y)\cdot \frac{dx}{dt}(t) + \frac{\partial z}{\partial y}(x,y)\cdot \frac{dy}{dt}(t).\]

Or, even more simply,

\[\frac{dz}{dt} = \frac{\partial z}{\partial x}\cdot\frac{dx}{dt} + \frac{\partial z}{\partial y}\cdot\frac{dy}{dt}.\]

But, Eq. (Eq:ChainRule1) is the most accurate expression. □

Proof. For an arbitrary \(t_0 \in I\), let \(x_0 = \varphi(t_0)\) and \(y_0 = \psi(t_0)\). For an arbitrary real number \(\delta\) such that \(t_0 + \delta \in I\), let \(h(\delta) = \varphi(t_0 + \delta) - \varphi(t_0)\) and \(k(\delta) = \psi(t_0 + \delta) - \psi(t_0)\). We have

\[\begin{eqnarray} \lim_{\delta \to 0}\frac{h(\delta)}{\delta} &=& \frac{d\varphi}{dt}(t_0),\\ \lim_{\delta \to 0}\frac{k(\delta)}{\delta} &=& \frac{d\psi}{dt}(t_0). \end{eqnarray}\]

Since \(f(x,y)\) is totally differentiable,

\[\begin{eqnarray*} f(\varphi(t_0+\delta),\psi(t_0+\delta)) -f(\varphi(t_0),\psi(t_0)) &=& f(x_0 + h, y_0 + k) - f(x_0, y_0)\\ &=&\frac{\partial f(x_0,y_0)}{\partial x}\cdot h + \frac{\partial f(x_0,y_0)}{\partial y}\cdot k\\ && + o(\sqrt{h^2 + k^2}). \end{eqnarray*}\]

As \(\delta \to 0\),

\[\frac{o(\sqrt{h^2 + k^2})}{\delta} = \frac{h}{\delta}\cdot\frac{o(\sqrt{h^2 + k^2})}{h} \to \frac{d\varphi(t_0)}{dt}\cdot 0 = 0.\]

Therefore,

\[\begin{eqnarray} &&\lim_{\delta\to 0}\frac{f(\varphi(t_0+\delta),\psi(t_0+\delta)) - f(\varphi(t_0),\psi(t_0))}{\delta}\nonumber\\ && = \frac{\partial f}{\partial x}(x_0,y_0)\cdot \frac{d\varphi}{dt}(t_0) + \frac{\partial f}{\partial y}(x_0,y_0)\cdot \frac{d\psi}{dt}(t_0).\tag{Eq:Dt0} \end{eqnarray}\]

This shows that the function \(z = f(\varphi(t), \psi(t))\) is differentiable at \(t = t_0\) and its differential coefficient is given by the right-hand side of Eq. (Eq:Dt0). ■

Example. If \(z = f(x,y), x = at + b, y = ct + d\), then

\[\begin{eqnarray} \frac{dz}{dt} &=& \frac{\partial z}{\partial x}\cdot\frac{d x}{dt} + \frac{\partial z}{\partial y}\cdot\frac{d y}{dt}\\ &=& af_x(at+b, ct+d) + cf_y(at + b, ct + d). \end{eqnarray}\] □

Theorem [Chain rule (2)]

Let \(z = f(x,y)\) be a totally differentiable function on an open region \(U \subset \mathbb{R}^2\). Let \(x = \varphi(u, v), y = \psi(u,v)\) be functions on an open region \(V \subset \mathbb{R}^2\) such that \((\varphi(u,v), \psi(u,v)) \in U\) for all \((u,v)\in V\). Suppose that the partial derivatives \(\varphi_u(u,v), \varphi_v(u,v)\), \(\psi_u(u,v), \psi_v(u,v)\) exist on \(V\). Then, the function \(z = f(\varphi(u,v), \psi(u,v))\) on \(V\) has the partial derivatives with respect to \(u\) and \(v\) given by

\[\begin{eqnarray*} \frac{\partial}{\partial u}f(\varphi(u,v), \psi(u,v)) &=&\frac{\partial f}{\partial x}(\varphi(u,v), \psi(u,v))\frac{\partial \varphi}{\partial u}(u,v)\\ && +\frac{\partial f}{\partial y}(\varphi(u,v), \psi(u,v))\frac{\partial \psi}{\partial u}(u,v) \ ,\\ \frac{\partial}{\partial v}f(\varphi(u,v), \psi(u,v)) &=&\frac{\partial f}{\partial x}(\varphi(u,v), \psi(u,v))\frac{\partial \varphi}{\partial v}(u,v)\\ &&+\frac{\partial f}{\partial y}(\varphi(u,v), \psi(u,v))\frac{\partial \psi}{\partial v}(u,v) \ . \end{eqnarray*}\]

Remark. In a more simplified notation, the above partial derivatives may be written as

\[\begin{eqnarray*} \frac{\partial z}{\partial u} &=& \frac{\partial z}{\partial x}\cdot\frac{\partial x}{\partial u} + \frac{\partial z}{\partial y}\cdot\frac{\partial y}{\partial u},\\ \frac{\partial z}{\partial v} &=& \frac{\partial z}{\partial x}\cdot\frac{\partial x}{\partial v} + \frac{\partial z}{\partial y}\cdot\frac{\partial y}{\partial v}. \end{eqnarray*}\]

Proof. The partial differential coefficient

\[\frac{\partial z(u_0, v_0)}{\partial u}\]

at \((u,v) = (u_0,v_0)\) is the same as the differential coefficient of the "one-variable" function \(z = f(\varphi(u, v_0), \psi(u, v_0))\) at \(u = u_0\) where \(v = v_0\) is fixed. Then, we can apply the above Theorem [Chain Rule (1)]. Noting that the differential coefficients of the one-variable functions \(\varphi(u,v_0)\) and \(\psi(u, v_0)\) are calculated as \(\frac{\partial \varphi(u_0,v_0)}{\partial u}\) and \(\frac{\partial \psi(u_0,v_0)}{\partial u}\), respectively, we have the claimed result. The same argument applies to the partial derivative with respect to \(v\). ■

Example. Let \(f(x,y) = e^{x^2 + y^2}\) and \(\varphi(u,v) = u\cos v, \psi(u,v) = u\sin v\). Let \(g(u,v) = f(\varphi(u,v), \psi(u,v))\). Find \(g_u(u,v)\) and \(g_v(u,v)\). Let \(x = u\cos v\) and \(y = u\sin v\). Then, \(x^2 + y^2 = u^2\). Thus,

\[\begin{eqnarray*} g_u(u,v) &=& \frac{\partial}{\partial u}f(\varphi(u,v),\psi(u,v))\\ &=& \frac{\partial f}{\partial x}(\varphi(u,v), \psi(u,v))\frac{\partial \varphi}{\partial u}(u,v) + \frac{\partial f}{\partial y}(\varphi(u,v), \psi(u,v))\frac{\partial \psi}{\partial u}(u,v)\\ &=& \left[e^{u^2}\cdot 2u\cos v\right] \cos v +\left[e^{u^2}\cdot 2u\sin v\right] \sin v = 2ue^{u^2},\\ g_v(u,v) &=& \frac{\partial}{\partial v}f(\varphi(u,v),\psi(u,v))\\ &=& \frac{\partial f}{\partial x}(\varphi(u,v), \psi(u,v))\frac{\partial \varphi}{\partial v}(u,v) + \frac{\partial f}{\partial y}(\varphi(u,v), \psi(u,v))\frac{\partial \psi}{\partial v}(u,v)\\ &=& \left[e^{u^2}\cdot 2u\cos v\right]\cdot (-u \sin v) + \left[e^{u^2}\cdot 2u\sin v\right]\cdot(u\cos v) = 0. \end{eqnarray*}\] □





Comments

Post a Comment

Popular posts from this blog

Open sets and closed sets in \(\mathbb{R}^n\)

-
Image
Open sets In \(\mathbb{R}\), we have the notion of an open interval such as \((a, b) = \{x \in \mathbb{R} | a < x < b\}\). We want to extend this idea to apply to \(\mathbb{R}^n\). We also introduce the notions of bounded sets and closed sets in \(\mathbb{R}^n\). Recall that the \(\varepsilon\)-neighbor of a point \(x\in\mathbb{R}^n\) is defined as \(N_{\varepsilon}(x) = \{y \in \mathbb{R}^n | d(x, y) < \varepsilon \}\) where \(d(x,y)\) is the distance between \(x\) and \(y\). Definition (Open set) A subset \(U\) of \(\mathbb{R}^n\) is said to be an  open set  if the following holds: \[\forall x \in U ~ \exists \delta > 0 ~ (N_{\delta}(x) \subset U).\tag{Eq:OpenSet}\] That is, for every point in an open set \(U\), we can always find an open ball centered at that point, that is included in \(U\). See the following figure. Perhaps, it is instructive to see what is  not  an open set. Negating (Eq:OpenSet), we have \[\exists x \in U ~ \forall \delta > 0 ~ (N_{\delta}(x) \not
Read more

Euclidean spaces

-
Image
We would like to study multivariate functions (i.e., functions of many variables), continuous multivariate functions in particular. To define continuity, we need a measure of "closeness" between points. One measure of closeness is the Euclidean distance. The set \(\mathbb{R}^n\) (with \(n \in \mathbb{N}\)) with the Euclidean distance function is called a Euclidean space. This is the space where our functions of interest live. The real line is a geometric representation of \(\mathbb{R}\), the set of all real numbers. That is, each \(a \in \mathbb{R}\) is represented as the point \(a\) on the real line. The coordinate plane , or the \(x\)-\(y\) plane , is a geometric representation of \(\mathbb{R}^2\), the set of all pairs of real numbers. Each pair of real numbers \((a, b)\) is visualized as the point \((a, b)\) in the plane. Remark . Recall that \(\mathbb{R}^2 = \mathbb{R}\times\mathbb{R} = \{(x, y) | x, y \in \mathbb{R}\}\) is the Cartesian product of \(\mathbb{R}\) with i
Read more

Newton's method

-
Image
Newton's method (the Newton-Raphson method) is a very powerful numerical method for solving nonlinear equations.  Suppose we'd like to solve a non-linear equation \(f(x) = 0\), where \(f(x)\) is a (twice) differentiable non-linear function. Newton's method generates a sequence of numbers \(c_1, c_2, c_3, \cdots\) that converges to a solution of the equation. That is, if \(\alpha\) is a solution (i.e., \(f(\alpha) = 0\)) then,  \[\lim_{n\to\infty}c_n = \alpha,\] and this sequence \(\{c_n\}\) is generated by a series of linear approximations of the function \(f(x)\). Theorem (Newton's method) Let \(f(x)\) be a function that is twice differentiable on an open interval \(I\) that contains the closed interval \([a, b]\) (i.e., \([a,b]\subset I\)) and satisfy the following conditions: \(f(a) < 0\) and \(f(b) > 0\); For all \(x\in [a, b]\), \(f'(x) > 0\) and \(f''(x) > 0\). Let us define the sequence \(\{c_n\}\) by \[ \begin{eqnarray} c_1 &
Read more