Gradient: The gradient of f at (x,y), when its first-order partial derivatives exist, is the vector gradf(x,y)=(βxβfβ(x,y),βyβfβ(x,y)).
Differentiability: The function f is differentiable at a = (x,y) if it has first-order partial derivatives there and lim(h,k)β(0,0)ββ₯(h,k)β₯f(x+h,y+k)βf(x,y)ββxβfβ(x,y)hββyβfβ(x,y)kβ=0.
Differential: The differential of f at (x,y) is the linear function df(x,y)β=βxβfβ(x,y)dx+βyβfβ(x,y)dy, where dx(h,k)=h and dy(h,k)=k.
β Must-know
π Formula - At a point where the limits exist, the first-order partial derivatives are βxβfβ(x,y)=limhβ0βhf(x+h,y)βf(x,y)β and βyβfβ(x,y)=limhβ0βhf(x,y+h)βf(x,y)β.
The gradient points in the direction of maximal increase of f and is normal to the contour lines of f.
π If a function f is of class C1 at a point a, then f is differentiable at a.
π Formula - If f is differentiable at (x_0,y_0), the tangent plane to its graph at (x_0,y_0,f(x_0,y_0)) has equation z=f(x0β,y0β)+βxβfβ(x0β,y0β)(xβx0β)+βyβfβ(x0β,y0β)(yβy0β).
Further detail
π Formula - The directional derivative of f at a = (xβ,yβ) along v = (xα΅₯,yα΅₯), when the limit exists, is dfaβ(v)=limtβ0βtf(xaβ+txvβ,yaβ+tyvβ)βf(xaβ,yaβ)β.
A function f is of class CΒΉ on U if its first-order partial derivatives are continuous on U.
Linear combinations, products, and quotients with a denominator that never vanishes of CΒΉ functions on U are CΒΉ on U, and if f: U β β and g: β β β are CΒΉ, then g β f is CΒΉ.
π Formula - At a differentiable point (x,y), the first-order TaylorβYoung expansion is f(x+h,y+k)=f(x,y)+βxβfβ(x,y)h+βyβfβ(x,y)k+β₯(h,k)β₯Ο(h,k), where Ο(h,k)β0 as (h,k)β(0,0).
For small increments h and k, the differential evaluated at (h,k) approximates the change in f: df(x,y)β(h,k)βf(x+h,y+k)βf(x,y).
π If f is of class CΒΉ at a, then f is differentiable at a.
π 2. Differential and Measurement Uncertainty
β Must-know
π Formula - For small variations, the differential estimates the function variation by ΞfββxβfβΞx+βyβfβΞy.
If the relative uncertainty in temperature is 0.3% and that in volume is 0.2%, the differential estimate gives a relative pressure uncertainty bounded by 0.5%.
Further detail
π Formula - For one mole of ideal gas, pressure satisfies P=f(T,V)=VRTβ, where R is the ideal gas constant.
π 3. Higher-Order Derivatives and Taylor Formula
π Key Concepts & Definitions
Hessian matrix: The Hessian matrix of a function of two variables is Hfβ(x,y)=(fxxβ(x,y)fxyβ(x,y)βfyxβ(x,y)fyyβ(x,y)β), wherever these derivatives exist.
Second-order partial derivatives: Second-order partial derivatives are obtained by differentiating a first-order partial derivative again; for a function of two variables they are βx2β2fβ, βxβyβ2fβ, βyβxβ2fβ, and βy2β2fβ.
Second-order remainder: The remainder o(β₯(h,k)β₯2) is a quantity of the form β₯(h,k)β₯2Ο(h,k), where Ο(h,k) tends to zero as (h,k) tends to (0,0).
β Must-know
A function is of class Ck when all its partial derivatives up to order k are continuous; it is of class Cβ when partial derivatives of every order are continuous.
π If a function is of class C2 on an open set, its mixed second-order partial derivatives are equal, so its Hessian matrix is symmetric.
π Formula - For a function of class C2 and a point (x,y) in its domain, the second-order TaylorβYoung formula is f(x+h,y+k)=f(x,y)+fxβ(x,y)h+fyβ(x,y)k+21β(fxxβ(x,y)h2+2fxyβ(x,y)hk+fyyβ(x,y)k2)+o(β₯(h,k)β₯2).
Further detail
Linear combinations, products, quotients with a nowhere-zero denominator, and compositions of functions of class Ck remain of class Ck on their respective domains.
π 4. Extrema and Critical Point Classification
π Key Concepts & Definitions
Critical point: A point is a critical point of a function when its differential is zero, equivalently when its gradient is zero and both first-order partial derivatives vanish there.
Local minimum: A function reaches a local minimum at a point when there is a neighborhood of that point on which its value is no greater than the function's value at any other point in that neighborhood.
Local maximum: A function reaches a local maximum at a point when there is a neighborhood of that point on which its value is no less than the function's value at any other point in that neighborhood.
Global minimum: A function reaches a global minimum at a point when its value there is no greater than its value at any point in its domain.
Global maximum: A function reaches a global maximum at a point when its value there is no less than its value at any point in its domain.
β Must-know
If a function is differentiable at a point and reaches a local extremum there, then that point is critical.
At a critical point of a function of class C2, if rtβs2>0 and r>0, the function has a local minimum there.
At a critical point of a function of class C2, if rtβs2>0 and r<0, the function has a local maximum there.
At a critical point of a function of class C2, if rtβs2<0, the function has neither a local minimum nor a local maximum and is at a saddle point.
Further detail
At a critical point of a function of class C2, let r=fxxβ, s=fxyβ, and t=fyyβ be the second-order partial derivatives evaluated at that point.
π The Monge criterion gives no conclusion at a critical point when rtβs2=0.
A saddle point can have a minimum along one direction and a maximum at the same point along an orthogonal direction.
π At a critical point of a function of class C2, positive Hessian determinant and positive trace imply a local minimum, while positive determinant and negative trace imply a local maximum; negative determinant implies a saddle point, and zero determinant gives no conclusion.
π‘ Memory Hook
A local extremum is a peak or valley; a saddle is a minimum in one direction and a maximum in an orthogonal direction.
π 5. Torus Altitude and Critical Points
β Must-know
π Formula - The altitude of a point on the torus is modeled by z(Ο,Ο)=RsinΟ+rsinΟsinΟ, where the torus has major radius R and minor radius r with R>r.
The critical points of the altitude function occur, modulo 2Ο, when Οβ{Ο/2,3Ο/2} and Οβ{Ο/2,3Ο/2}.
π The point with Ο=Ο/2,Ο=Ο/2 is a local maximum, the point with Ο=3Ο/2,Ο=Ο/2 is a local minimum, and the points with Ο=3Ο/2 are saddles.
Further detail
At each critical point, the second-order Taylor polynomial is obtained from the value and Hessian of z at that point.
π 6. Cabin Volume Optimization
β Must-know
The available wood surface area is 90m2, includes the ground-floor rectangle, and excludes the rectangle between the cuboid and the roof.
π Formula - After eliminating h using the surface-area constraint, the volume is V(x,y)=2x+2y90xyβ815βx2y2β.
Further detail
The cabin consists of a cuboid with dimensions x,y,h and a triangular-prism roof whose isosceles-triangle base angles are Ξ±=arctan(3/4).
π Assuming V(x,y) has a unique global maximum for x,y>0, the cabin dimensions and maximum volume are determined by locating that maximum and recovering h from the area constraint.
π 7. Glacier Temperature and Heat Equation
β Must-know
π Formula - The glacier temperature is modeled by T(x,t)=T0β+T1βeβΟxsin(Ξ»tβΞΌx), where t is measured in days and x in feet.
π The temperature model is to be shown to satisfy the heat equation βtβT=kβxxβT for some constant k.
Further detail
The model uses Ξ»=3652Οβ and positive constants for the other parameters.
π‘ Memory Hook
Exponential depth decay and periodic time variation produce a temperature model to test against the heat equation.
π 8. Machine Yield Optimization
β Must-know
π Formula - The machine yield is given by R(U,Ο)=1+U21+ΟββUΟ+Ο2, where U is tension and Ο is frequency.
Further detail
π The optimal machine yield is found by maximizing R(U,Ο) over the admissible values of tension and frequency.
π 9. Scalar and Vector Fields
π Key Concepts & Definitions
Scalar field: A function on an open subset of R3 that assigns a real number to each point.
Vector field: A function on an open subset of R3 that assigns a vector in R3 to each point.
π Essential Points
Examples of scalar fields include temperature, pressure, and fluid density; examples of vector fields include velocity, electric fields, and magnetic fields.
π 10. Gradient Curl and Divergence
π Key Concepts & Definitions
Gradient: For a scalar field f, the gradient is the vector field gradf=(βf/βx,βf/βy,βf/βz).
Curl: For a vector field F=(Fxβ,Fyβ,Fzβ), the curl is rotF=(βFzβ/βyββFyβ/βz,βFxβ/βzββFzβ/βx,βFyβ/βxββFxβ/βy).
Divergence: For a vector field F=(Fxβ,Fyβ,Fzβ), the divergence is the scalar field divF=βFxβ/βx+βFyβ/βy+βFzβ/βz.
π Essential Points
The nabla operator is β=(β/βx,β/βy,β/βz), so gradf=βf, rotF=βΓF, and divF=ββ F.
π 11. Potentials and Exactness
π Key Concepts & Definitions
Scalar potential: A vector field derives from a scalar potential if there is a scalar field f such that F=gradf.
β Must-know
In the operator sequence scalar fields gradβ vector fields rotβ vector fields divβ scalar fields, the composition of either pair of successive operators is zero: rot(gradf)=0 and div(rotF)=0.
The sequence of gradient, curl, and divergence operators is exact on open sets without holes, such as R3, a ball, or the interior of a cube; exactness can fail on domains with holes, such as a torus interior or a ball with its center removed.
Further detail
The electric field derives from the electric potential V up to sign, as expressed in the course by E=gradV.
Maxwell's equations were published in 1862 by James Clerk Maxwell and describe the electromagnetic field through four equations: divE=Ο/Ξ΅0β, rotE=ββB/βt, c2rotB=βE/βt+j/Ξ΅0β, and divB=0.
On R3βOx, the field F(x,y,z)=(0,βz/(y2+z2),y/(y2+z2)) has zero curl but does not derive from a scalar potential.
The study of when the gradientβcurlβdivergence sequence is exact, and its generalizations, contributed to the development of cohomology in topology in the 20th century.
π‘ Memory Hook
A curl-free field need not derive from a potential when the domain has a hole.
Test yourself
Take a 31-question practice quiz on Differential Calculus and Extrema, with every answer explained.
1. At a glacier depth of 2 feet and time 30 days, which expression gives the modeled temperature?
2. For one mole of an ideal gas with temperature T and volume V, which expression gives pressure as a function of those variables?