Part 1 — Functions, Domain, Range, Level Curves, Limits, Continuity

Functions of two variables

, .

  • = domain, set of -values = range; independent (inputs), dependent (output).
  • Typical domain constraints: → radicand ; → argument ; denominator .

Regions in the plane

  • Interior point: center of some disk lying entirely in .
  • Boundary point: every disk around it contains points inside and outside .
  • Open: all points are interior. Closed: contains all its boundary points.
  • Bounded: lies inside some disk of fixed radius; otherwise unbounded.

Domains

  • : , closed (contains the parabola) and unbounded.
  • : , i.e. ; open, boundary and .

Graphs and level curves

  • Level curve: set of points in where (drawn in the -plane).
  • Graph / surface: all points in space.
  • Contour curve: the curve in space where the plane cuts the surface (the level curve lifted to height ).

Paraboloid

Level curves are circles of radius . The contour curve is the circle lifted to height 75.

Limits

  • may approach from any direction.
  • Limit rules: sum, difference, constant multiple, product, quotient (), power, root ( for even roots).
  • Polynomials/rational functions (defined at the point): just substitute.
  • Indeterminate with roots: multiply by the conjugate.

Uniqueness: a limit, when it exists, is the same along every path.

Two-path test: if has different limits along two paths to , the limit does not exist (typical paths: , , , ).

Polar coordinates (at the origin): , , for all .

  • If the result is independent of → the limit exists.
  • If it depends on → the limit does not exist.
  • At : , .

Two paths, two limits

: along the limit is ; along it is . In polar form depends on ⇒ no limit at .

Continuity

is continuous at if

  1. is defined,
  2. exists,
  3. the limit equals .

Useful tool: sandwich theorem, e.g. .

Part 2 — Partial Derivatives, Chain Rule, Implicit Differentiation

Partial derivatives

  • To compute : treat (and other variables) as constants; same idea for , .
  • Geometrically: is the slope of the curve cut on the surface by the plane .

Second-order: , , , .

Schwarz (Clairaut) theorem: if are continuous on an open region around , then .

Slope of a slice

: , so = slope of the curve cut by the plane at .

Differentiability

is differentiable at if exist there and

(i.e. with ). Partials existing alone is not enough.

Chain rule

  • One variable: .
  • , , :
  • : add .
  • Functions on surfaces: with functions of :
  • Use a dependency (tree) diagram: multiply along each branch, add the branches.

Implicit differentiation

If defines as a differentiable function of and :

Part 3 — Directional Derivative, Gradient, Tangent Lines/Planes, Differentials

Vector review

  • , ; is a unit vector.
  • Dot product: ; .
  • Cross product: , perpendicular to both and ; = area of the parallelogram. , , .
  • Line through parallel to : , , .
  • Plane through with normal : . (Normal from 3 points: .)
  • Position vector ; velocity is tangent to the curve.

Directional derivative

For a unit vector :

(If is not unit, use .)

Gradient

  • , , , .
  • , so:
    1. increases fastest in the direction (rate );
    2. decreases fastest in the direction ;
    3. zero change in directions .
  • is normal to level curves (and level surfaces).

Tangent line to a level curve

Tangent to an ellipse

is the level curve of . At : .
Tangent: . The gradient is perpendicular to it.

Tangent plane & normal line to a surface at

  • Tangent plane:
  • Normal line: , ,
  • For at :

Estimating change & linearization

  • Change when moving in direction : .
  • Linearization at : , and (tangent-plane approximation). Three variables: add .
  • Error bound: if on a rectangle around :

Tangent plane = linearization

at : , so . Near the plane and the surface almost coincide.

Part 4 — Extreme Values and Saddle Points

Taylor’s formula (order 2) near


Order :

Local extrema

  • Local max at : on an open disk around ; local min: .
  • First derivative test: local extremum at an interior point where partials exist ⇒ (horizontal tangent plane).
  • Critical point: interior point where , or where or doesn’t exist.
  • Saddle point: critical point where every disk around it has points with and points with .

Three kinds of critical point at the origin

: minimum; : maximum; : saddle (up along , down along ). In all three, at .

Second derivative test

With , , and discriminant (Hessian) :

ConditionNature
, local maximum
, local minimum
saddle point
inconclusive — find another way

(Comes from completing the square: .)

Classifying 5 critical points

: critical points and .

  • : , , → saddle.
  • Same signs: , → local maxima. Opposite signs: → local minima.

Absolute extrema on a closed, bounded region

  1. Find interior critical points; evaluate there.
  2. On each boundary piece, reduce to one variable; find its critical points and the endpoints; evaluate .
  3. Largest value = absolute max, smallest = absolute min.

Triangle

. Interior: with . Edges: , , , .
Absolute max at ; absolute min at and .

Lagrange multipliers

To optimize subject to a constraint , solve

( = Lagrange multiplier; at the optimum the level curves of and are tangent, so their gradients are parallel.) Tip: to minimize a distance, minimize its square.

Closest point of to the origin

Minimize with : , ⇒ . gives (impossible), gives .
At those points the level circle is tangent to the hyperbola; distance .