First Year Engineering is coming soon

We're still writing and reviewing the notes for every subject, and we'll open it when every subject is ready. Meanwhile you can practise aptitude, ask Aria about any subject, or see the branches that are ready.

Partial differentiation

Partial differentiation: definitions, rules (product/quotient/chain), higher-order and mixed partials with Clairaut’s theorem, interpretation (tangent plane/level curves), and exam-style computations (including units) to support later topics like total derivative, Jacobians, and maxima–minima.

Drafted with Aria, reviewed by the AiCanCode.org team · Updated . How we write and check our notes. Spotted an error? Use Give Feedback at the bottom of the page.

Why it matters

Partial differentiation lets us study how a multivariable quantity changes when one input varies and the others are held fixed. Engineers use it to compute heat flux from temperature fields, stresses from displacement fields, sensitivity of outputs to parameters, and to set up optimization with constraints (later topics).

Key ideas

∂f/∂x(a, b) = lim_{h→0} [f(a + h, b) − f(a, b)]/h, provided the limit exists. Intuition: rate of change along the x-direction while holding y constant.

  • Notation: ∂f/∂x, f_x, D₁f, and similarly ∂f/∂y, f_y, D₂f.
  • Higher-order partial derivatives: Differentiate again: f_xx = ∂²f/∂x², f_yy = ∂²f/∂y², and mixed partials f_xy = ∂²f/∂y∂x, f_yx = ∂²f/∂x∂y. At a point they may or may not be equal; see Clairaut’s theorem.
  • Exact statement (Clairaut/Schwarz theorem): If f is defined on an open set containing a point (a, b) and if f_xy and f_yx are continuous in a neighborhood of (a, b), then f_xy(a, b) = f_yx(a, b). Assumptions: existence and continuity of the second partial derivatives in a neighborhood.
  • Geometric meaning: For z = f(x, y), ∂f/∂x at (a, b) is the slope of the curve formed by intersecting the surface with the plane y = b; ∂f/∂y is slope along x = a. The pair (f_x, f_y) determines the tangent plane and direction of steepest ascent (via gradient; more in optimization topics).
  • Increment and linear approximation: For small Δx, Δy, f(a + Δx, b + Δy) ≈ f(a, b) + f_x(a, b) Δx + f_y(a, b) Δy, if f_x and f_y are continuous near (a, b). This is the first-order Taylor approximation (full Taylor series is treated in "Taylor's and Maclaurin's series"). Equivalently, the total differential at (a, b) is dz = f_x(a, b) dx + f_y(a, b) dy.
  • Product rule for partial derivatives: If u(x, y) and v(x, y) are differentiable, then ∂(uv)/∂x = u_x v + u v_x, and similarly for y.
  • Quotient rule for partial derivatives: If v ≠ 0, ∂(u/v)/∂x = (u_x v − u v_x)/v².
  • Chain rule (multivariable): If z = f(x, y) with x = x(u, v), y = y(u, v), and all functions are differentiable, then ∂z/∂u = f_x ∂x/∂u + f_y ∂y/∂u and ∂z/∂v = f_x ∂x/∂v + f_y ∂y/∂v, evaluated at the corresponding point. Assumptions: differentiability of the compositions.
  • Implicit differentiation (multivariable): If F(x, y, z) = 0 defines z as a function of x and y near a point and F_z ≠ 0 at that point, then ∂z/∂x = −F_x/F_z and ∂z/∂y = −F_y/F_z. Assumptions: F continuously differentiable and F_z ≠ 0 (Implicit Function Theorem conditions).
  • Second derivative test for shape (local curvature only): f_xx and f_yy describe concavity along coordinate directions; mixed partial f_xy couples the directions. Full maxima–minima tests and Hessian use are treated under "Maxima and minima of functions of two variables".

Typical exam derivations universities ask:

  1. Derivation of the chain rule for z = f(x, y) with x = x(u, v), y = y(u, v):
    • Start from total increment: Δz = f(x + Δx, y + Δy) − f(x, y).
    • For small changes, use linear approximation: Δz ≈ f_x Δx + f_y Δy.
    • But Δx = x_u Δu + x_v Δv and Δy = y_u Δu + y_v Δv.
    • Collect Δu, Δv terms: Δz ≈ (f_x x_u + f_y y_u) Δu + (f_x x_v + f_y y_v) Δv.
    • Compare coefficients of Δu, Δv to get ∂z/∂u and ∂z/∂v.
  2. Proof outline of Clairaut’s theorem under continuity of second partials:
    • Define ϕ(h, k) = [f(a + h, b + k) − f(a + h, b) − f(a, b + k) + f(a, b)]/(hk) for h, k ≠ 0.
    • By the mean value theorem in one variable applied twice, show that ϕ(h, k) equals f_xy evaluated at some intermediate point (a + θh, b + ηk), θ, η ∈ (0, 1).
    • Similarly, swapping the order yields ϕ(h, k) equals f_yx at another intermediate point.
    • If f_xy and f_yx are continuous near (a, b), then as h, k → 0, both expressions tend to f_xy(a, b) and f_yx(a, b) respectively, but the limit of ϕ(h, k) is unique; hence f_xy(a, b) = f_yx(a, b).

What to draw and label (typical exam diagram requests):

  • For z = f(x, y), draw 3D axes with surface patch near point P(a, b, f(a, b)).
    • Draw the lines x = a and y = b on the domain plane.
    • Show two curves on the surface formed by intersecting planes x = a and y = b; their slopes at P are ∂f/∂y and ∂f/∂x respectively.
    • Sketch the tangent plane at P with equation z = f(a, b) + f_x(a, b)(x − a) + f_y(a, b)(y − b). Label P, axes, f_x, f_y, and tangent plane.

Formulas

∂f/∂x(a, b) = lim_{h→0} [f(a + h, b) − f(a, b)]/h — f: ℝ² → ℝ, x is first variable; unit of ∂f/∂x is unit of f per unit of x; applies when limit exists. ∂f/∂y(a, b) = lim_{k→0} [f(a, b + k) − f(a, b)]/k — Similar meaning and units for y. f_xx = ∂²f/∂x², f_yy = ∂²f/∂y², f_xy = ∂²f/∂y∂x, f_yx = ∂²f/∂x∂y — Second-order partials; units: unit of f per (unit of x)² for f_xx, per (unit of y)² for f_yy, and per (unit of x)(unit of y) for f_xy and f_yx; applies if derivatives exist. Clairaut’s theorem: if f_xy and f_yx are continuous near (a, b), then f_xy(a, b) = f_yx(a, b) — Dimensionless statement; applies under continuity assumption. Product rule: ∂(uv)/∂x = u_x v + u v_x — u, v differentiable; units: consistent with product; applies per variable similarly. Quotient rule: ∂(u/v)/∂x = (u_x v − u v_x)/v² — v ≠ 0; units: consistent; applies per variable similarly. Chain rule: ∂z/∂u = f_x x_u + f_y y_u; ∂z/∂v = f_x x_v + f_y y_v — z = f(x, y), x = x(u, v), y = y(u, v); units: ∂z/∂u has unit of z per unit of u; applies when functions are differentiable. Implicit differentiation: ∂z/∂x = −F_x/F_z, ∂z/∂y = −F_y/F_z — F(x, y, z) = 0 with F_z ≠ 0; units: ∂z/∂x has unit of z per unit of x. Tangent plane: z ≈ f(a, b) + f_x(a, b)(x − a) + f_y(a, b)(y − b) — Linear approximation; units: unit of z; applies near (a, b).

Worked examples

Example 1: Compute first and second partials; verify mixed partials equality Given f(x, y) = x²y³ + 4xy + e^{xy}. All variables are dimensionless. Steps:

  1. First-order partial derivatives (treat the other variable as constant):
    • Formula: f_x = ∂/∂x (x²y³) + ∂/∂x (4xy) + ∂/∂x (e^{xy}). Computation: f_x = 2xy³ + 4y + y e^{xy}.
    • Formula: f_y = ∂/∂y (x²y³) + ∂/∂y (4xy) + ∂/∂y (e^{xy}). Computation: f_y = 3x²y² + 4x + x e^{xy}.
  2. Second-order and mixed partials:
    • f_xx = ∂/∂x (2xy³ + 4y + y e^{xy}) = 2y³ + y² e^{xy}.
    • f_yy = ∂/∂y (3x²y² + 4x + x e^{xy}) = 6x²y + x² e^{xy}.
    • f_xy = ∂/∂y (2xy³ + 4y + y e^{xy}) = 6xy² + 4 + (e^{xy} + x y e^{xy}).
    • f_yx = ∂/∂x (3x²y² + 4x + x e^{xy}) = 6xy² + 4 + (e^{xy} + x y e^{xy}).
  3. Verification: f_xy = f_yx for all (x, y), consistent with Clairaut’s theorem since second partials are continuous for polynomials and e^{xy}. Final answer: f_x = 2xy³ + 4y + y e^{xy}, f_y = 3x²y² + 4x + x e^{xy}; f_xy = f_yx = 6xy² + 4 + e^{xy}(1 + xy).

Example 2: Chain rule with polar change of variables Let z = x² y + e^{xy} where x = r cosθ, y = r sinθ. Variables r (meters, m) and θ (radians, rad); z is dimensionless (for illustration). Find ∂z/∂r and ∂z/∂θ at r = 1 m, θ = π/4 rad. Steps:

  1. Compute f_x and f_y for f(x, y) = x² y + e^{xy}:
    • f_x = 2x y + y e^{xy}.
    • f_y = x² + x e^{xy}.
  2. Compute x_r, x_θ, y_r, y_θ:
    • x_r = cosθ (m/m = dimensionless), x_θ = −r sinθ (m/rad),
    • y_r = sinθ (dimensionless), y_θ = r cosθ (m/rad).
  3. Chain rule:
    • ∂z/∂r = f_x x_r + f_y y_r.
    • ∂z/∂θ = f_x x_θ + f_y y_θ.
  4. Evaluate at r = 1, θ = π/4: x = (1)(√2/2), y = (1)(√2/2).
    • Compute e^{xy} = e^{(1/2)} = e^{0.5} ≈ 1.64872 (dimensionless).
    • f_x = 2xy + y e^{xy} = 2(1/2) + (√2/2)(1.64872) = 1 + 1.16603 ≈ 2.16603.
    • f_y = x² + x e^{xy} = (1/2) + (√2/2)(1.64872) = 0.5 + 1.16603 ≈ 1.66603.
    • x_r = cos(π/4) = √2/2 ≈ 0.70711; y_r = sin(π/4) = √2/2 ≈ 0.70711.
    • x_θ = −r sinθ = −(1)(√2/2) ≈ −0.70711 m/rad; y_θ = r cosθ = (1)(√2/2) ≈ 0.70711 m/rad.
  5. Calculate:
    • ∂z/∂r ≈ 2.70938 per meter (1/m).
    • ∂z/∂θ ≈ −0.35355 per radian (1/rad). Final answer: ∂z/∂r ≈ 2.70938 m⁻¹, ∂z/∂θ ≈ −0.35355 rad⁻¹.

Example 3: Partial differentiation in a physical field (with units) A steady-state temperature field T(x, y) in a thin plate is given by T(x, y) = 2x²y − 3y, where x, y are in meters (m), T in degrees Celsius (°C). Find: (a) ∂T/∂x and ∂T/∂y at (x, y) = (1 m, 2 m) with units. (b) The linearized change ΔT when moving from (1, 2) to (1.02, 1.98) meters. (c) Verify units of second partials at (1, 2). Steps:

  1. First-order partials:
    • Formula: ∂T/∂x = ∂/∂x (2x²y − 3y) = 4xy.
    • Evaluate at (1, 2): ∂T/∂x = 4(1)(2) = 8 °C/m.
    • Formula: ∂T/∂y = ∂/∂y (2x²y − 3y) = 2x² − 3.
    • Evaluate at (1, 2): ∂T/∂y = 2(1)² − 3 = −1 °C/m.
  2. Linearized change (first-order approximation):
    • Formula: ΔT ≈ (∂T/∂x) Δx + (∂T/∂y) Δy.
    • Δx = 1.02 − 1.00 = +0.02 m; Δy = 1.98 − 2.00 = −0.02 m.
    • Compute: ΔT ≈ (8 °C/m)(0.02 m) + (−1 °C/m)(−0.02 m) = 0.16 °C + 0.02 °C = 0.18 °C.
  3. Second partials and their units:
    • T_xx = ∂/∂x (4xy) = 4y; at (1, 2): T_xx = 8 °C/m².
    • T_yy = ∂/∂y (2x² − 3) = 0; at (1, 2): T_yy = 0 °C/m².
    • T_xy = ∂/∂y (4xy) = 4x; T_yx = ∂/∂x (2x² − 3) = 4x; at (1, 2): both 4 °C/m². Final answer: (a) ∂T/∂x = 8 °C/m, ∂T/∂y = −1 °C/m at (1, 2). (b) ΔT ≈ 0.18 °C. (c) T_xx = 8 °C/m², T_yy = 0 °C/m², T_xy = T_yx = 4 °C/m² at (1, 2).

Common mistakes

  • Applying Clairaut’s theorem without checking continuity assumptions (there are counterexamples where mixed partials exist but are not continuous and differ at a point).
  • Dropping units in applied problems; first partials have units of “output per input,” second partials have “output per input squared.”
  • In chain rule, omitting one branch (forgetting x’s dependence on u or y’s on v) or evaluating f_x, f_y at the wrong point (must substitute x(u, v), y(u, v) first).
  • Arithmetic slips in e^{xy}-type terms: forgetting that ∂/∂x e^{xy} = y e^{xy} and ∂/∂y e^{xy} = x e^{xy}.

For your exam

  • Derivations: chain rule for z = f(x, y) with x = x(u, v), y = y(u, v); outline proof of equality of mixed partials under continuity.
  • Computation questions: find first and second partials; evaluate at a point; verify f_xy = f_yx; apply product/quotient rules; chain rule under change of variables (Cartesian to polar); implicit differentiation to find ∂z/∂x from F(x, y, z) = 0.
  • Applied numericals: interpret units of partials; linear approximation to estimate change; prepare to connect to “Total derivative and change of variables,” “Jacobians,” and “Maxima and minima of functions of two variables” for follow-up topics.
  • Full-marks answers include: clear step-by-step differentiation with correct rules, substitution at the correct point, units shown in applied problems, proper statement of assumptions (e.g., continuity for Clairaut), and a brief concluding line.

Quick check

  1. Define ∂f/∂x at (a, b) in words and with the limit form.
  2. For f(x, y) = x e^{xy}, compute f_x and f_y.
  3. If F(x, y, z) = x² + y² + z² − 1 = 0, find ∂z/∂x at point where z ≠ 0. Answers: (1) It is the rate of change of f as x varies with y fixed; ∂f/∂x(a, b) = lim_{h→0} [f(a + h, b) − f(a, b)]/h. (2) f_x = e^{xy} + x y e^{xy}, f_y = x² e^{xy}. (3) ∂z/∂x = −F_x/F_z = −(2x)/(2z) = −x/z (valid when z ≠ 0).

Finished this topic? Mark it so your progress, study plan and readiness keep up.

Stuck on something here?