Topic 8

The Multivariable Chain Rule

Draw a tree from the outside function down to the final variables. A derivative is a sum over every path to that variable. Each path gives the product of the derivatives along it: dwdt=fx dxdt+fy dydt\dfrac{dw}{dt} = f_x\,\dfrac{dx}{dt} + f_y\,\dfrac{dy}{dt}. Always evaluate the partials of ff at the inner point (x(t),y(t))(x(t), y(t)).

4 kinds of question. Open the one you are stuck on.

1

Find dw/dt using the multivariable chain rule

e.g. “Let w=f(x,y)w = f(x, y) with xx and yy depending on tt. Find dwdt\dfrac{dw}{dt}”
  1. 1
    Draw the tree: ww on top, tt at the bottom

    w→x,y→tw \to x, y \to t. If tt also appears directly in ff, add a branch w→tw \to t.

  2. 2
    Write one term per path from ww to tt

    Each term is a product: a partial of ff times the ordinary derivative of the inner function. A direct branch contributes ftf_t.

  3. 3
    Compute each piece
  4. 4
    Substitute the point and the numbers

    Evaluate fxf_x and fyf_y at the inner point (x(t),y(t))(x(t), y(t)).

Remember: Every path, multiply along it, add them up.

See it in a problem3 worked examples

Let w=x2y−y3w = x^2y - y^3 with x=cos⁡tx = \cos t, y=sin⁡ty = \sin t. Find dwdt\dfrac{dw}{dt} at t=π4t = \tfrac{\pi}{4}.

  1. 1
    Draw the tree: ww on top, tt at the bottom

    w→x,y→tw \to x, y \to t; no direct tt.

  2. 2
    Write one term per path from ww to tt

    dwdt=wx dxdt+wy dydt\dfrac{dw}{dt} = w_x\,\dfrac{dx}{dt} + w_y\,\dfrac{dy}{dt}.

  3. 3
    Compute each piece

    wx=2xyw_x = 2xy,  wy=x2−3y2\ w_y = x^2 - 3y^2,  dxdt=−sin⁡t\ \dfrac{dx}{dt} = -\sin t,  dydt=cos⁡t\ \dfrac{dy}{dt} = \cos t.

  4. 4
    Substitute the point and the numbers

    At t=π4t = \tfrac{\pi}{4}, x=y=22x = y = \tfrac{\sqrt{2}}{2}: wx=1w_x = 1, wy=12−32=−1w_y = \tfrac{1}{2} - \tfrac{3}{2} = -1. So dwdt=1(−22)+(−1)22=−2\dfrac{dw}{dt} = 1\left(-\tfrac{\sqrt{2}}{2}\right) + (-1)\tfrac{\sqrt{2}}{2} = -\sqrt{2}.

Answer:

dwdt∣t=π/4=−2\dfrac{dw}{dt}\Big|_{t = \pi/4} = -\sqrt{2}

Common mistakes3
  • Confusing dwdt\dfrac{dw}{dt} with ftf_t: ftf_t holds xx and yy fixed, dwdt\dfrac{dw}{dt} lets them move with tt.
  • Evaluating fxf_x at the wrong point: at t=1t = 1 it is fx(x(1),y(1))f_x(x(1), y(1)).
  • Writing ∂x∂t\dfrac{\partial x}{\partial t} for a one-variable x(t)x(t): it is dxdt\dfrac{dx}{dt}.
2

Find partial derivatives of a composite function with the chain rule

e.g. “Use the chain rule to evaluate ∂w∂s\dfrac{\partial w}{\partial s} and ∂w∂t\dfrac{\partial w}{\partial t}”
  1. 1
    Draw the tree: ww, then x,yx, y, then s,ts, t
  2. 2
    For ∂w∂s\dfrac{\partial w}{\partial s}, add one product per path to ss

    Each product is (partial of the outer function) × (partial of the inner function in ss). Repeat for tt.

  3. 3
    Compute the inner partials and the inner point

    At (s,t)=(1,2)(s, t) = (1, 2), find xx and yy first: that is where fxf_x, fyf_y are evaluated.

  4. 4
    Substitute and simplify

Remember: One sum per variable at the bottom of the tree.

See it in a problem3 worked examples

Let w=x2+xyw = x^2 + xy with x=stx = st, y=s−ty = s - t. Find ∂w∂s\dfrac{\partial w}{\partial s} and ∂w∂t\dfrac{\partial w}{\partial t} at (s,t)=(1,2)(s, t) = (1, 2).

  1. 1
    Draw the tree: ww, then x,yx, y, then s,ts, t

    w→x,yw \to x, y; each of x,y→s,tx, y \to s, t.

  2. 2
    For ∂w∂s\dfrac{\partial w}{\partial s}, add one product per path to ss

    ws=wxxs+wyysw_s = w_x x_s + w_y y_s,  wt=wxxt+wyyt\ w_t = w_x x_t + w_y y_t.

  3. 3
    Compute the inner partials and the inner point

    xs=tx_s = t, xt=sx_t = s, ys=1y_s = 1, yt=−1y_t = -1. At (1,2)(1, 2): x=2x = 2, y=−1y = -1, so wx=2x+y=3w_x = 2x + y = 3, wy=x=2w_y = x = 2.

  4. 4
    Substitute and simplify

    ws=3⋅2+2⋅1=8w_s = 3 \cdot 2 + 2 \cdot 1 = 8,  wt=3⋅1+2⋅(−1)=1\ w_t = 3 \cdot 1 + 2 \cdot (-1) = 1.

Answer:

∂w∂s=8\dfrac{\partial w}{\partial s} = 8,  ∂w∂t=1\ \dfrac{\partial w}{\partial t} = 1

Common mistakes3
  • Evaluating fxf_x at (s,t)(s, t) instead of at (x(s,t),y(s,t))\big(x(s, t), y(s, t)\big).
  • Missing a path: if xx depends on both ss and tt, it appears in both sums.
  • Writing dxds\dfrac{dx}{ds} when xx depends on both ss and tt: it is ∂x∂s\dfrac{\partial x}{\partial s}.
3

Show a composite function satisfies an equation for any f

e.g. “Suppose that w=f(xz,yz)w = f(xz, yz), where ff is differentiable. Show that …”
  1. 1
    Name the inside: u=…u = \dots, v=…v = \dots
  2. 2
    Find each partial the equation needs with the chain rule

    fuf_u, fvf_v (or F′F') stay as symbols; only the inner derivatives become formulas.

  3. 3
    Substitute into the equation and collect like terms
  4. 4
    Show both sides match, or match coefficients

Remember: f stays a symbol; only the insides get differentiated.

See it in a problem4 worked examples

Let z=F(x−3y)z = F(x - 3y) for a differentiable FF. Show that 3zx+zy=03z_x + z_y = 0.

  1. 1
    Name the inside: u=…u = \dots, v=…v = \dots

    u=x−3yu = x - 3y, so z=F(u)z = F(u).

  2. 2
    Find each partial the equation needs with the chain rule

    zx=F′(u)⋅1z_x = F'(u) \cdot 1,  zy=F′(u)⋅(−3)\ z_y = F'(u) \cdot (-3).

  3. 3
    Substitute into the equation and collect like terms

    3zx+zy=3F′(u)−3F′(u)3z_x + z_y = 3F'(u) - 3F'(u).

  4. 4
    Show both sides match, or match coefficients

    =0= 0 for every FF. ✓

Answer:

3zx+zy=3F′(x−3y)−3F′(x−3y)=03z_x + z_y = 3F'(x - 3y) - 3F'(x - 3y) = 0

Common mistakes3
  • Giving ff a formula. ff is , so fuf_u stays a symbol.
  • Using one name for two functions: wxw_x (a function of x,y,zx, y, z) is not fxf_x.
  • Forgetting the product rule when ff is multiplied by something: ∂∂x[x2F(… )]\dfrac{\partial}{\partial x}\big[x^2F(\dots)\big] has two terms.
4

Find second derivatives with the chain rule

e.g. “Find all second order derivatives of g(s,t)=f(2s+3t,3s−2t)g(s, t) = f(2s + 3t, 3s - 2t)”
  1. 1
    Find the first derivative with the chain rule
  2. 2
    Differentiate again, using the product rule where needed

    Each term looks like (inner derivative) × fx(… )f_x(\dots). If the inner derivative is not constant, the product rule gives two terms.

  3. 3
    Chain again on each fx(x,y)f_x(x, y) and fy(x,y)f_y(x, y)

    ∂∂sfx=fxx ∂x∂s+fxy ∂y∂s\dfrac{\partial}{\partial s}f_x = f_{xx}\,\dfrac{\partial x}{\partial s} + f_{xy}\,\dfrac{\partial y}{\partial s}.

  4. 4
    Collect terms, using fxy=fyxf_{xy} = f_{yx}

Remember: f_x is a composite too: chain it again.

See it in a problem3 worked examples

Let g(s,t)=f(s+2t, 3s−t)g(s, t) = f(s + 2t,\ 3s - t). Find gstg_{st} in terms of the partials of ff.

  1. 1
    Find the first derivative with the chain rule

    gs=f1+3f2g_s = f_1 + 3f_2.

  2. 2
    Differentiate again, using the product rule where needed

    The inner derivatives 11 and 33 are constants, so no product-rule terms: gst=∂∂tf1+3 ∂∂tf2g_{st} = \dfrac{\partial}{\partial t}f_1 + 3\,\dfrac{\partial}{\partial t}f_2.

  3. 3
    Chain again on each fx(x,y)f_x(x, y) and fy(x,y)f_y(x, y)

    ∂∂tf1=2f11−f12\dfrac{\partial}{\partial t}f_1 = 2f_{11} - f_{12},  ∂∂tf2=2f21−f22\ \dfrac{\partial}{\partial t}f_2 = 2f_{21} - f_{22}.

  4. 4
    Collect terms, using fxy=fyxf_{xy} = f_{yx}

    gst=2f11−f12+6f21−3f22=2f11+5f12−3f22g_{st} = 2f_{11} - f_{12} + 6f_{21} - 3f_{22} = 2f_{11} + 5f_{12} - 3f_{22}.

Answer:

gst=2f11+5f12−3f22g_{st} = 2f_{11} + 5f_{12} - 3f_{22}, all at (s+2t, 3s−t)(s + 2t,\ 3s - t)

Common mistakes3
  • Treating fxf_x as a constant on the second pass. It depends on ss through xx and yy, so it needs the chain rule.
  • Dropping the product-rule term when the inner derivative is not constant (like dxdt=2t\dfrac{dx}{dt} = 2t).
  • Evaluating at the wrong point: fxxf_{xx} is taken at (x(t0),y(t0))\big(x(t_0), y(t_0)\big).

Need the formulas? Must memorize · Flashcards

Found this helpful? Leave us a review on Google ↗Spotted a mistake?

Which topics are you reviewing?

The topic list, Flashcards, Must memorize and Exam words will only show these. Change it any time with the Topics button above the topic list.

Vectors & geometry
Partial derivatives
Optimization
Multiple integrals
11 of 20 topics selected