Contentssnovikov/math437.pdf · Differential 2-forms in Rn. Change of coordinates and restriction of 2-forms to 2-surfaces in Rn. Case n= 2. Integration. Case of any k-forms in Rn.
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
S.P.Novikov — Math 437Spring 2015
Differential forms
Textbook: D.Bachman, A geometric approach to differential forms, Birkhauser.Additional material: M.Spivak, Calculus on Manifolds.
1.Coordinates. Cartesian coordinates in a domain U ⊂ Rn. Important ex-amples (polar, cylindrical, spherical, hyperbolic, complex).
2.Differential 1-forms and vector fields, Jacoby matrix, change of coordi-nates, pull-back for functions and 1-forms, restriction to curves and integra-tion. Orientation.
3.Important examples, closed 1-forms, Newton–Leibnitz property for in-tegration, exact and closed 1-forms, 1-dimensional (co)homology of domainsU ⊂ R2. The differential of the angle function. Topology of planar domain,number of ”cuts” necessary to make it simply connected.
4.Complex coordinates z, z in R2. Cauchy 1-forms f(z) dz, the Cauchy–Riemann equation ∂f/∂z = 0. What is the value of the integral
∮|z|=1
zn dz?
Is the form f(z) dz closed if ∂f/∂z = 0?5.What is d(X) if X is a 1-form? Differential 2-forms in Rn. Change
of coordinates and restriction of 2-forms to 2-surfaces in Rn. Case n = 2.Integration. Case of any k-forms in Rn.
6.Product of 1-forms as a 2-form. Product of differential forms. Associa-tivity and skew commutativity for k-forms in U ⊂ Rn.
7.The De Rham operator d acting on differential forms, its definition andalgebraic properties. De Rham operator and multipication of forms. DeRham Operator and pull-back operation. Closed k-forms and cohomology(definition for the domains in Rn). Important examples of the closed formsin R2 and R3, angle form in R2; area form for S2 in spherical coordinatesand as a restriction of the closed 2-form in R3.
8.Change of coordinates for k-forms in u ⊂ Rn. Pull-back operation fork-forms for arbitrary smooth maps φ : U → V .
9.Integration of n-forms in Rn. Integration of k-forms along the orientedk-surfaces U → Rn, U ⊂ Rk.
10.The Stokes formula for integration∫· · ·∫
D
dΩk+1
=
∫· · ·∫
∂D
Ωk
,
cases k = 0, 1.11.First pair of Maxwell Equations and De Rham operator d in the 3-
space and in 4-space. Electromegnetic field as a 2-form in R4. Electric andMagnetic fields in R3.
3
12.Minicourse in Linear Algebra. Vectors and covectors. Inner products,Gram Matrices. Nondegenerate inner product. Classification of symmetricInner Products. Euclidean and Minkowski cases. Volume element and GramMatrix. Groups GLn(R) and GLn(C), On, SOn. How many connected piecesdo they have? Same question for group O1,1.
13.Nondegenerate Symplectic Inner Product in R2n and 2-form, its pow-ers and volume element, Pfaffian (square root of determinant) of the skewsymmetric matrix and powers of 2-form.
15. Riemanian and Pseudoriemanian metrics. Euclidean and Minkowskicases, metrics of eulideasn plane, 2-sphere and pseudosphere in the polarcoordinates. Arc length of curves, definition of geodesics as (locally) minimalarc curves. Arc length of timelike curves in Minkowski space. Arc length ofthe ”Fermat metric” in R2. The Fermat principle for the propagation of lightand ”Snells law” in the water/air case.
15. Duality of forms in euclidean metric, the case of 3-space. The op-erators div and curl in terms of differential forms. Product of 1-forms andvector product in R3. Vector product and commutator of skew symmetricmatrices.
16.Differential forms and homotopy of mappings. Poincare’ lemma.17.Curves in the euclidean plane. Curvature and Gauss map. Curvature
as a pull-back of the form dϕ. Integral of curvature along the closed curveand degree of Gauss map.
18.Surfaces in R3. Riemannian metric and curvature form. Mean andGauss curvature. Gauss curvature and Gauss map. Total (integral ) ofGauss curvature for the convex body in R3.
19. Hopf invariant for the maps S3 → S2 and Kelvin-Whitehead integralalong S3.
4
Homeworks
Homework 1
1. Calculate the Jacoby matrix for a shift, rotations, and affine transfor-mations of Rn.
2. Find the restriction of the 1-forms dx and dy to the lines
a) y = λx;
y
x
b) x2 + y2 = 1.
3. Is (ρ(x, y), φ(x, y)) a pair of “Cartesian” type coordinates in R2? Is ittrue for R2 \ 0?
4. In which domain (ρ, φ) may be treated as a pair of “Cartesian” coor-dinates? Is it ok for the domain 0 < φ < 2π?
0
5. Is ρ a “good” Cartesian coordinate for the domain R2 \ 0?
6. Find the restriction of the form
x dy − y dxx2 + y2
to the circle x2 + y2 = 1 (in the coordinates (ρ, φ)).
5
Homework 1 — Solutions
1. Shift: x 7→ x+ b, J = 1.
Rotation: x 7→ Ax, J = A.
Affine transformation: x 7→ Ax+ b, J = A.
2. a) Use x as the parameter. y = λx, x = x, dx→ dx, dy → λ dx.
b) x2 + y2 = 1, x = cos t, y = sin t, use t as the parameter.
dx→ d(cos t) = − sin t dt,
dy → d(sin t) = cos t dt.
3. (ρ, φ) is not a pair of Cartesian coordinates in R2 and R2 \ 0.
4. The domain U ⊂ R2 \ 0 should not contain a closed path around 0.
0
5. No, the level ρ = const > 0 is a circle S1 ⊂ R1.
6.x dy − y dx
ρ2= dφ.
6
Homework 2
1. Calculate the change of coordinates (ρ1, φ1)→ (ρ2, φ2),
ρ1(ρ2, φ2), φ1(ρ2, φ2) = ?
•
•
(0, 0)
(2, 3)
1
2
2. Prove that the 1-form α dφ1+β dφ2 is closed and not exact in a domainV like:
(0, 0)
(2, 3)
V
3. Find a domain in S2 ⊂ R3 in which dθ is a well-defined 1-form.
4. Find a domain in S2 in which dφ is a well-defined 1-form.
5. The cylindrical coordinates (z, ρ, φ) in R3 are defined by z = z, x =ρ cosφ, y = ρ sinφ. In which domain V ⊂ R3 the form dφ is a well-defined 1-form?
7
6. Calculate the integral∮γ
(αz2 + β/z) dz, z = x+ iy, α, β = const
along a closed contour γ.
Homework 2 — solutions
1.x = ρ1 cosφ1,
y = ρ1 sinφ1,
x− 2 = ρ2 cosφ2,
y − 3 = ρ2 sinφ2.
Conclusion:ρ1 cosφ1 = 2 + ρ2 cosφ2,
ρ1 sinφ1 = 3 + ρ2 sinφ2.
Solve these equations:
ρ21 = (2 + ρ2 cosφ2)2 + (3 + ρ2 sinφ2)
2,
cosφ1 =2 + ρ2 cosφ2
ρ1,
sinφ1 =3 + ρ2 sinφ2
ρ1.
2. d(α dφ1 + β dφ2) = 0 (obvious).∮C1
Ω = 2πα,
∮C2
Ω = 2πβ.
3. dθ is OK in the domain S2 \ poles.
4. dφ is OK in S2 \ poles.
5. Cylindrical coordinates in R3. dφ is OK in R3 \ line(x = 0, y = 0).
6. ∮C
(αz2 +
β
z
)dz = 2πiβ
(Cauchy)∮Czn dz = 0, n = −1.
8
Homework 3
1. Prove that the 1-form x dy− y dx is invariant under all rotations of R2
around (0, 0).
2. Prove the same for x dx+ y dy.
3. Introduce “hyperbolic coordinates” (χ, ψ): x = χ coshψ, y = χ sinhψ.In which domain U ⊂ R2 are they defined?
4. Calculate the area 2-form dx∧ dy of R2 in the polar coordinates (ρ, φ):dx ∧ dy = C dρ ∧ dφ, C = ?.
5. Calculate the area 2-form dx ∧ dy in the complex coordinates z, z,dx ∧ dy = C dz ∧ dz, C = ?.
6. Calculate the volume 3-form of R3 in the spherical coordinates (r, φ, θ),
z = r cos θ,
x = r sin θ cosφ,
y = r sin θ sinφ,
dx ∧ dy ∧ dzthe volume 3-form
= (?) dr ∧ dθ ∧ dφ.
Homework 3 — Solutions
1. x dy − y dx = dφ · ρ2 — invariant under rotation.
2. x dx+ y dy = ρ dρ — invariant under rotation: ρ 7→ ρ, φ 7→ φ+ const,dρ 7→ dρ, dφ 7→ dφ.
x = χ coshψ, y = χ sinhψ, x2 − y2 = χ2 > 0
9
x
y
x2−y2>0
3. dx ∧ dy = ρ dρ ∧ dφ.
4.
dx ∧ dy =dz ∧ dz−2i
,
dz = dx+ i dy,
dz = dx− i dy.
5. R3: dx ∧ dy ∧ dz = r2 sin θ dr ∧ dθ ∧ dφ.For the unit sphere S2 (r = 1) we have Area = sin θ dθ∧dφ, 0 6 θ 6 π.
10
Homework 4
1. Prove that d(f dz) =∂f
∂zdz ∧ dz in R2.
2. Calculate the area form in the hyperbolic coordinates x = χ coshψ,y = χ sinhψ.
3. Introduce hyperbolic coordinates in a domain of R3
x = χ sinh θ cosφ,
y = χ sinh θ sinφ,
z = χ cosh θ.
Find a domain where these coordinates are well defined.
4. Find linear transformations of R2 such that ⟨Aη,Aζ⟩ = ⟨η, ζ⟩, where⟨η, ζ⟩ = η1ζ1 − η2ζ2.
5. How many components does the group of such transformations have(see no. 4)?
6. Find the value of the integral∮|z|=1
f(z)
zndz,
where f(z) = a0 + a1z + . . .+ aNzN .
Homework 4 — Solutions
1. df = fx dx+ fy dy = fz dz + fz dz by definition.
Volume element in R2,1 .Coordinates ρ, χ, φ (correction), ρ2 = t2 − x2 − y2 .
t = ρ coshχ , x = ρ sinhχ cosφ , y = ρ sinhχ sinφ
d3σ = ?
dt2 − dx2 − dy2 =
= dρ2[cosh2 χ − sin2 χ cos2 φ − sinh2 χ cos2 φ
]− dχ2 − ρ2sinh2 χ dφ2 =
= dρ2 − ρ2(dχ2 − sinh2 χ dφ2)
G =
1 0 00 ρ2 0
0 0 ρ2sinh2χ
,√detG = ρ2sinhχ
d3σ = ρ2sinhχ dρ ∧ dχ ∧ dφ
3. ds2 = dz dz · g , g = g(|z|2) , dz dz = dx2 + dy2
G =
(g 00 g
),
√detG = g
d2σ = g dx ∧ dy =1
2ig dz ∧ dz
4. ds2 =dz dz
(1 + zz)2(S2 − round sphere)
30
z =1
w, z =
1
z
ds2 =d(1/w) d(1/w)
(1 + 1/ww)2=
dw dw · 1/w2w2
(1 + ww)2/(ww)2=
dw dw
(1 + ww)2
Same!
5. ds2 =dx2 + dy2
c2(x, y), x(t) , y(t) − curve
Arc length:
l(γ) =
∫ b
a
√x2 + y2 dt
c (x(t), y(t))
a b
t
“Geodesics” = i.e. curve with minimal arc length
Example: Fermat Principle Let
c(x, y) =
c2
c1
y=0
a
b
Find geodesic joining a and b.
31
c1a
b
c2
X
y
y=0Aψθ
a = (x1, y1) , b = (x2, y2)
Time from a to A = |a− A|/c1 = T1
Time from A to b = |b− A|/c2 = T2
Total time = T1 + T2 (depends on A), A = (A, 0) .
T1 + T2 =√(x1 − A)2 + y21 / c1 +
√(x2 − A)2 + y22 / c2 = T (A)
minA T (A) = ?
dT
dA= 0 ?
0 =1
c1
(A − x1)√(x1 − A)2 + y21
+1
c2
(A − x2)√(x2 − A)2 + y22
↑sin θ
↑− sinψ
A − x1 < 0 , A − x2 > 0 .
c2c1
=sinψ
sin θ
32
Midterm Tests.
First Midterm Test.
1. Differential 0-forms and 1-forms, define pull-back map, change of coor-dinated, restriction to curve, integration. Newton-Leibnitz formula forintegration. Calculate differential 1-form dz/z in polar coordinates
2. Formulate necessary and sufficient condition for 1-form X = udx+vdyin the planar domain U ⊂ R2 to be closed. Find condition for closedform to be exact in terms of contour integration? Is the form dz/(z2−1)exact in the domain U ⊂ R2 where 2 points ±1 removed?
Define 2-forms and De Rham differential operator d from 1-forms to2-forms for the planar domains.
3. Define exterior (grassmanian) product of any number of 1-forms. For-mulate main properties of this product. Calculate product X1∧X2∧X3
1. Differential 1-forms and change of coordinates. Differential 1- and 2-forms in R2, R3, define the exterior product of 1-forms and De RhamOperator d for 1-forms. What is a closed form? What is an exact form?Calculate 2-form d[(xdy − ydx)/(x2 + y2)s]. Prove that it is closed fors = 1 only in R2minus0.
2. By definition, Electromagnetic Field F is a differential 2-form in R4 orpair (E,B) of 1- and 2-forms in R3 depending on time (i.e. the 4-spaceis presented as R3×R). Formulate the 1st pair of Faraday Laws in R3
(Non-relativistic form) and in R4 = (x0 = ct, x1, x2, x3) (Relativisticform) in terms of differential forms and De Rham Operator. ApplyStokes Formula to express the integral of Electric Field along the closedcontour C in R3 through the magnetic field inside of contour (its timederivative).
3. Prove that for every symmetric indefinite nondegenerate inner productin R2 there exists a basis e, e′ such that Gramm Matrix has a (”light-like”) form
(e, e) = (e′, e′) = 0, (e, e′) = 1
Second Midterm Test. Solutions.
I. First problem is collection of definitions and calculation of dΩ for somespecific 1 - form Ω .
Definitions:
1. 1 - form in R1 , R2 , R3 , . . . :∑i
ui(x) dxi
Change of coordinates x(x′) (pull-back, restriction, . . . )
f(x) → f (x(x′)) ,
34
dxi → ∂xi
∂x′kdx′k ,
matrix form (Jacoby matrix):
dx →(∂x
∂x′
)dx′
2. 2 - form in R2 , R3 , . . . :
Ω =∑ij
aij(x) dxi ∧ dxj
De - Rham operatord : Ω → dΩ
Multiplication of forms ∧ :(f dxi ∧ dxj
)∧(g dxk ∧ dxl
)= fg dx1 ∧ dxj ∧ dxk ∧ dxl
Operator d :
df =∂f
∂xidxi , d
(f dxi ∧ . . . ∧ dxj
)= df ∧ dxi ∧ . . . ∧ dxj
3. Calculate dΩ for
Ω =(x2 + y2
)−s(x dy − y dx)
Exact form: Ω = dω .
Closed form: dΩ = 0 .
Properties
ω − exact ⇒∮C
ω = 0 (closed contour)
ω − closed ⇒∮C
ω = 0 C ⊂ R2
35
D
C
if domain D is simply connected(ω is defined and smooth in D).
II. Second Problem:
Electromagnetic field = 2 - form F in R4 (space-time)
R4 = (x0 = ct, x, y, z)
First part of Maxwell Equations (Faraday Law)
dF = 0
3 − dimensional form in R3 × R :(x, y, z) (t)
E = Ex dx + Ey dy + Ez dz
F = E ∧ dx0 + B , x0 = ct
B =∑
i,j=1,2,3
bij dxi ∧ dxj
d(4)E = 0 ⇔ “Pair” :
a) d(3)B = 0 ,
b) d(3)E + ∂B/∂x0 = 0
(d(4) − in R4 , d(3) − in R3).
D
C
Stokes:∮C
E =1
c
∂
∂t
∫D
B
Second pair of Maxwell Equations:
36
d (∗F ) = J (4) (“4 − current”)
↑ ↑involves non− geometricalinner term
product
This law is nongeometrical(interaction with matter)
First pair dF = 0 is purely geometrical.
III. Third problem: R2 with indefinite inner product (symmetric). Findbasis e, e′ with Gramm Matrix
(e , e′) = 1 , (e , e) = (e′ , e′) = 0 ,
(0 11 0
)= G
Proof. Start with basis
e1 , e2 , G′ =
(1 00 −1
)Put
e = e1 + e2 , e′ = e1 − e2
(e , e) = (e′ , e′) = 0 , (e , e′) = 2
Put
e =e√2, e′ =
e′√2
37
Lectures 1–30
Introductory lecture: History of integration and cal-culus
1. XVII–XVIII centuries:
x∫a
f(y) dy = g(x), g′(x) = f(x), . . .
2. XIX century: idea of a differential 1-form (Green, Cauchy, . . . ) andfirst topological ideas (1820)
0 =
∮C
f(z) dz
C
∂f
∂z= 0 inside C −→ complex analytic function
Integration of differential forms is the best possible type ofintegration:
(i) It is well-defined without use of any other structures (Riemannianmetric, . . . ).
(ii) It is invariant under changes of coordinates both in the space(manifold) and in the body. Admissible changes of coordinatesform a very broad class (even not one-to-one along the body).
(iii) They can be differentiated (d). This differential is invariant.
(iv) They have remarkable “Stokes Property” connecting (body) and(its boundary). It connects them with topology — “De RhamCohomology” known since XIX century.
38
(v) They can be described in the beautiful “differential” language aswell as in the classical vector (tensor) language.
3. Electromagnetism, Faraday laws, Gauss formula and differential forms,topological quantities, Maxwell equations (mid XIX century)
(linking number)
4. Vector (tensor) calculus (in the 3-space), Maxwell, Kelvin, Stokes,Poincare.
Stokes-type formulas:
b∫a
f ′(x) dx = f(b)− f(a)a− +
b
body and boundary
(simplest Newton–Leibnitz formula).
5. Poincare, differential forms and topology (vector (tensor) calculus inmultidimensional spaces) (1895). Hamiltonian systems, Riemann sur-faces.
6. E.Cartan and differential forms (1920s). Discovery of differential formsin the modern notations. Topology.
7. De Rham Theorem (1930s). Proof of the Cartan–Poincare conjecture.
8. Differential forms in modern mathematics. Complex geometry.
Differential forms are special tensors (“contravariant”, skew symmetric).Why are they especially important?
1. They provide a universal integration and differentiation (independentof any other structure like Riemannian metric and so on).
39
2. They present the best bridge between Analysis and Topology. Cauchystarted to use these ideas in early XIX century.
3. They describe the laws of Electromagnetism (Faraday laws and Maxwellequations).
4. In the special case of Complex Geometry everything important canbe written in terms of differential forms including Riemannian metric(“Kahler metric”).
5. The differential notations of Cartan are extremely convenient.
40
Lecture 2. Differential 1-forms
Question: What is a differential 0-form?Answer: Differential 0-forms are functions f(x), 0-bodies are points, 0-chains are collections of points and numbers:
Q =[•
(P1,n1)•
(P2,n2). . . •
(Pk,nk)
], nj ∈ Z,
Q = “0-chain” = (P1, n1;P2, n2; . . . ;Pk, nk).
The “value” (or the “integral”) of the 0-form f along a 0-chain Q is
⟨f,Q⟩ =k∑j=1
njf(Pj).
Example. Q is a single point P with n1 = ±1,
•P1,n1=±1
⟨f,Q⟩ = ±f(Q).
What is a differential 1-form?In the space with coordinates (x1, . . . , xn), R2 = (x, y), R1 = (x),
Remark 1. Change of coordinates for ordinary vectors (like speed of parti-cles). γ(t) : x1(t), . . . , xn(t). Speed: γ(t) = (x1, . . . , xn). New coordinatesx(x′).Speed:
dx
dt=∂x
∂x′dx′
dt
( ∂xi∂x′j
)= J
γ(x)
= Jγ′
(x′)
⇒ γ′
(x′)
= J−1γ(x)
1-form:
η dx = η′ dx′, dx =∂x
∂x′dx′, η
∂x
∂x′dx′ = η′ dx′,
so we have
η∂x
∂x′= η′ or ηJ⊤ = η′
(ηi∂xi
∂x′j= η′j
)It is not the same law: the equality
J−1 = J⊤ ⇔ J ∈ On
(orthogonal)!
49
is true for Orthogonal matrices only.)
Remark 2. For 1-forms this formula makes sense even if the change is noton-to-one
η′ = Jη x = x(x′(t))
x = (x1, . . . , xn), x′ = (x′1, . . . , x′n).May be even det J = 0 in some points (even everywhere). This formula
makes sense even if dim x = dim x′, γ : Rk
(x′)→ Rn
(x)
X =∑
ηi(x) dxi, γ∗X =
∑i,j
ηi(x(t))∂xi
∂tjdtj, t = (t1, . . . , tk).
Example. Let k = 1, n = 2,
γ : R1
t
−→ R2
(x,y)
, x(t), y(t)
1-form η1 dx+ η2 dy = X in R2.“Restriction” of a 1-form to γ = R1 is also a partial case of this formula
γ∗X = η1(x(t), y(t))dx
dtdt+ η2(x(t), y(t))
dy
dtdt.
50
Lecture 4
Coordinates in Rn. Let (x1, . . . , xn) — Cartesian coordinates: −→x = (x1, . . . , xn).−→x 1 = −→x 2 ⇔ the points are distinct. Let U ⊂ Rn be an open domain withthe same cartesian coordinates.
Claim. a) For any closed path γ around 0 in R2 \ 0 we have
∮γ
dφ = 2π. 0γ
b) For S2 \ N,S and γ around N,S we have
∮γ
dφ = 2π
N
S
γ
59
Proof of a).b∫
a
dφ = ∆φ(along γ)
(no matter which path if γ does not cross 0).
0a
bγ
0a
b
γ ∆φ does not dependon the path
Proof of b): Same.
Complex coordinates in R2:
z = x+ iy,
z = x− iy,
∣∣∣ dz = dx+ i dy,
dz = dx− i dy,
∣∣∣ dx =1
2(dz + dz),
dy =1
2i(dz − dz).
Let f = u(x, y) + iv(x, y) and f(x, y) ∈ C.
df = fx dx+ fy dy = Adz +Bdz.
Definition 1. We call A = ∂f/∂z abd B = ∂f/∂z partial derivatives alongthe complex directions.
∂
∂z=
1
2
( ∂∂x− i ∂
∂y
),
∂
∂z=
1
2
( ∂∂x
+ i∂
∂y
).
∣∣∣∣∣∣∣∣ operators ∂z = ∂, ∂ = ∂z
For every 1-form X = U dx+ V dy we can write X = U dz + V dz.
Definition 2. Complex analytic function f = u+ iv is such that:
∂f/∂z = 0 ⇔ ux + vy = 0, vx − uy = 0.
60
Examples: z = ρeiφ.f = 1, z, z2, . . . (any polynomial in z);f = P (z)/Q(z)—a rational function (except “poles”Q = 0), f = 1/z, 1/z2, . . .;f = log z—multivalued, log z = log |z|+ i arg z = log(ρ) + iφ;f = ez, . . .
Important observation (topology):∮γ
f(z) dz = 0 if∂f
∂z≡ 0 inside γ.
γ
the contour γ is “simply con-nected” (no holes inside).
Explanation. Let f(z) = zn. We have
0
γ : |z| = 1ρ = 1
∮γ
dz = 0,
∮γ
z dz = 0, . . . ,
∮γ
zn dz = 0.
z = x+ iy = ρ(cosφ+ i sinφ) = ρeiφ.
Let γ be ρ = 1.∮γ
dz = ?
∮γ
z dz =
∮γ
dw = 0, . . .? dw/dz = z.
The proof will appear later. It follows from the fact that for every smoothfunction f(x, y) we have
∮df ≡ 0.
61
γ
no holes!
← closed path, simply connected inside
What is a closed 1-form?
Definition. a) A 1-form X = u dx+ v dy is closed iff for every point P ∈ R2
there exists a small neighbourhood P ∈ U ⊂ R2 such that X = u dx+v dy =df in U . Same definition for all n > 2.
b) “An exact 1-form” X is such that X = df in the whole domain Vwhere it is defined.
Examples. a) The domain V is R2 \ 0, X = dφ. It is closed but not exactbecause φ is multivalued. The obstruction is
∮γdφ = 2π = 0.
0γ
b) The domain V is S2 \ N,S, X = dφ is closed, but not exact.c) The domain V ⊂ R2 \ 0, where φ is one-valued smooth function, dφ
is exact in V .
62
Lecture 7
Our program and comparison with Bachman:
1. We assume that multivariable calculus is known (partial derivatives,differential of a function).
2. Special coordinates (polar, spherical, complex z = x + iy). Complexcoordinates are missing in Bachman.
3. Theory of differential 1-forms. This theory is briefly done in Bachman.Something appears later only as partial cases of n-forms. This theory(including elementary topology) is fundamental especially complex 1-forms in R2 and Riemann surfaces (Cauchy, Riemann, Poincare).
4. Contravariant property of the category of functions and forms underC∞-maps f : Rm → Rn appears only in the Chapter “Manifold” inBachman, it is called a ”pull-back map” f ∗
From 1-forms to 2-forms:Bachman discusses “product” of 1-forms. We will discuss it later, and startfrom Calculus:Problem. Let a differential 1-form be given in R2 (or Rn): X = u dx+ v dy(in some domain V ).a) Is it locally the differential of some function f :
(x, y) ∈ Uε ⊂ V, X = df, u =∂f
∂x, v =
∂f
∂y?
b) Is it globally differential X = df of any one-valued smooth function in thewhole domain V ?
Examples. 1) X = λ dx+ µ dy, f = λx+ µy, λ, µ = const.
2) X = dφ =x dy − y dxx2 + y2
.
a) V1 = R2 \ 0, dφ is not df for a 1-valued f in V ;
What do we know from Calculus? X = u dx+ v dy ((u, v) = “a covectorfield” in V ⊂ R2),
u =∂f
∂x, v =
∂f
∂y(locally) in Uε ⊂ V.
Uε
Claim. Nearby every point from V (n = 2) the condition
∂u
∂y=∂v
∂x
is necessary and sufficient for the “local” existence of f , ∇f = (u, v), ordf = X.
n > 2: X =∑
i ui dxi:
∂ui∂xj
=∂uj∂xi
.
“Global obstruction” for the existence of a single-valued f = Topology.
Example.
γ V ∮γ
dφ = 2π.
γ1 γ2
dφ1 dφ2 ∮γ1
dφ1 = 2π,
∮γ1
dφ2 = 0,
∮γ2
dφ1 = 0,
∮γ2
dφ2 = 2π.
64
“The cohomology with real coefficients” H1(V,R) is generated by dφ1, dφ2
as a linear space over the field R. Similar answer we have for H1(V,C) usingthe complex-valued 1-forms.
Consider the formal expression
dX =(∂u∂y− ∂v
∂x
)dx ∧ dy (n = 2),
dX =∑i<j
(∂ui∂xj− ∂uj∂xi
)dxi ∧ dxj (n > 2).
Claim. dX = 0 if and only if X is locally exact (i.e. “closed” 1-form).
Formal expressions like a dx ∧ dy,∑i<j
aij(x) dxi ∧ dxj (n > 2)
are called “differential 2-forms”.We extend aij to all pairs (i, j) by the requirement:
aij = −aji, ajj = 0.
The operator d : 1-forms→ 2-forms will be defined.The product of 1-forms X ∧ Y will also be defined by the rule
X = u dx+ v dy,
Y = w dx+ t dy,
∣∣∣ X ∧ Y = (ut− vw) dx ∧ dy.
65
Lecture 8
From 1-forms to 2-forms: let a differential 1-form X = u dx+ v dy be given.It is exact if X = df , i.e. fx = u, fy = v. It is closed if locally there exists afunction f such that fx = u, fy = v.
Example. X = dφ in R2 \ 0.
dφ =x dy − y dxx2 + y2
.
Lemma 1. A differential 1-form
X =n∑i=1
ui dxi
in a domain U ⊂ Rn is closed iff the following equations are satisfied:
∂ui∂xj
=∂uj∂xi
.
(Calculus.)
Introduce a 2-form
dX = Ω =∑i6j
(∂ui∂xj− ∂uj∂xi
)dxi ∧ dxj,
where dxi ∧ dxj = −dxj ∧ dxi by definition.More generally,
3. associativity (extension to k-forms for k > 2):
dx ∧ (dy ∧ dz) = (dx ∧ dy) ∧ dz;
4. multiplications of k-forms is bilinear (“distributive”).
We have “algebra of R(C)-valued differential forms” Λ(U,R(C)) in thedomain U ⊂ Rn with Cartesian coordinates (x1, . . . , xn). A 2-form is∑
aij(x) dxi ∧ dxj, i < j,
a 3-form is ∑i<j<k
aijk(x) dxi ∧ dxj ∧ dxk,
. . .k-form is ∑
i1<...<ik
ai1,...,ik(x) dxi1 ∧ . . . ∧ dxik ,
n-form isa(x) dx1 ∧ . . . ∧ dxn.
Herea(x) = a1,2,...,n(x).
69
Examples.k = 1: X =
∑ui(x) dx
i, n = 2: X = u dx+ v dy;k = 2: Ω =
∑i<j aij dx
i ∧ dxj, n = 2: Ω = a dx ∧ dy;k = 2, n = 3:
2-form = a12 dx ∧ dy + a13 dx ∧ dz + a23 dy ∧ dz.
The “vector” (a23,−a13, a12) = −→η is “an axial vector” in the physical litera-ture.
Operations. d : f → df = fxi dxi.
n = 2. d : X → dX = (uy − vx) dx ∧ dy, a12 = uy − vx.n > 2.
dX =∑i<j
(∂ui∂xj− ∂uj∂xi
)dxi ∧ dxj.
Example. 2-form a dx ∧ dy + b dy ∧ dx = (a− b) dx ∧ dy.
Change of coordinates: what do we already know?Functions x′ → x or x(x′), x1(x′1, . . . , x′n), . . . , xn(x′1, . . . , x′n). f(x(x′))
is the pull-back of the function f for the map x′ → x (map φ : U ′ → U),f(x(x′)) = φ∗f(x′).
1-forms. φ : U ′ → U , x′ → x, x(x′).
φ∗X =∑i
ui dxi φ∗−→
pull-back
∑i,j
ui(x(x′))∂xi
∂x′jdx′
j.
Or u(x)φ∗−→ u(x(x′)) as a function
dxjφ∗−→
∑i
∂xj
∂x′idx′
i.
Partial cases of “pull-back”:
1. Change of coordinates x↔ x′, one-to-one (may be local)
det( ∂x∂x′
)= det J = 0,
J = Jacoby matrix =( ∂xi∂x′j
).
70
2. Restriction to some surface (curve) (number of x′ is less than n).
General definition of the pull-back is given
f(x)→ f(x(x′)) = φ∗f(x′),
φ∗ : df → ∂f
∂xj(x(x′))
∂xj
∂x′ldxl = d(φ∗f) (sum in j, l).
Conclusions. The differential d of a function (0-form) commutes with thepull-back operation: φ∗(df) = d(φ∗f). By definition, the exterior multipli-cation of forms commute also with pull-back map. So we have
Claims. 1. φ∗d = dφ∗ for all k-forms (proof for k = 1 will be given later).2. φ∗(X ∧ Y ) = φ∗(X) ∧ φ∗(Y ). Product of forms is “natural” (commuteswith the pull-back for maps φ : U ′ → U including restrictions and changesof coordinates).
71
Lecture 10
Remarks.
1. Differential of function f is 1-form df =∑fxi dx
i as we write it inAnalysis. Vectors in Analysis we write as fist order differential opera-tors of the form
∑aj(x)∂/∂xj = w.
The scalar product of a 1-form and a vector attached to some point isby definition“the directional derivative” of a function
∇wf =∑i
∂f
∂xiai(x) = ⟨df, w⟩.
2. The homology H1(U) of a planar domains U in fact appear a lot inComplex Analysis as number of “cuts” necessary to make domain sim-ply connected.
Number of cuts =rank of H1(U).
Remaining domain should be simply connected (every closed contour= boundary of a “ball”).
The homology is generated by “cycles”. A cycle c is ∼ 0 iff for everyclosed 1-form X, dX = 0, we have∮
c
X = 0.
3. Cartesian coordinates in U ⊂ Rn (x1, . . . , xn) “inherited” from Rn are(x1, . . . , xn) in U . Other Cartesian coordinates = set (y1, . . . , yn) ofC∞-functions U →
yiR such that −→y 0 = −→y 1 ⇔ the points are distinct.
For t = 0 we have: < Aη, ξ >= − < η,Aξ >. The basis is orthonormal,therefore At = −A.
“Lie Algebra” of SOn consists of constant 2-forms. eAt ∈ SOn.For n = 3 it looks like 3-vectors (Euler).Consider 4-space: x0 = ct, x1, x2, x3. Let F = Fij dx
i ∧ dxj be a 2-form(“Electromagnetic Field”). dF = 0 – Faraday laws.
F = E1 dx1 ∧ dx0 + E2 dx
2 ∧ dx0 + E3 dx3 ∧ dx0︸ ︷︷ ︸
electric field
+∑
α,β=1,2,3α<β
Hαβ dxα ∧ dxβ
︸ ︷︷ ︸magnetic field
.
We assume that (x0, x) = (ct, x) = (time, space).In 3-space R3(x1, x2, x3) we have 1-form E and 2-form H, depending on
t as of parameter.
d(4)F = 0→ d(3)H = 0
and∂Eα∂xβ
dxβ ∧ dxα ∧ dx0 + ∂Hαβ
∂x0dx0 ∧ dxα ∧ dxβ = 0,
therefore∂Eα∂xβ
− ∂Hαβ
∂x0= 0.
Finally in the 3-space R3 we have:
d(3)E − H
c= 0, where x0 = ct,
∂
∂x0=
1
c
∂
∂t.
83
Lecture 14.
Differential forms and physics.
1. Non-relativistic physics. Space R3, (x, y, z) = x, time t is a param-eter.
Electric charge: e.
Electric field: E = (E1, E2, E3) is a 1-form∑α
Eαdxα, α = 1, 2, 3.
In R3: x1 = x, x2 = y, x3 = z, E = E(x, t).
Newtonian equations: mxi = eEi + other forces.
Magnetic field is a 2-form Bαβ, α, β = 1, 2, 3.
B(x, t) =∑α<β
Bαβ dxα ∧ dxβ,
Magnetic (Lorentz) force acting on the charged particle moving withspeed (v1, v2, v3) is fα = e/cBαβv
α where α, β = 1, 2, 3.
Faraday law: d(3)B = 0 (is closed), or
∂B12
∂x3− ∂B13
∂x2+∂B23
∂x1= 0,
B1 = B23, B2 = −B13, B3 = B12, B = (B1, B2, B3).
div B = 0 ↔ ∂B1
∂x1+∂B2
∂x2+∂B3
∂x3= 0.
2. Relativistic physics.
x0 = ct, c ∼= 300000kmsec (speed of light in vacuum).
4-space: (x0, x1, x2, x3) = (ct, x1, x2, x3).
Electromagnetic field:
F =∑j<j
Fij dxi ∧ dxj, i, j = 0, 1, 2, 3,
F = E ∧ dx0 +B = E ∧ cdt+B =
= E1 dx ∧ dx0 + E2 dy ∧ dx0 + E3 dz ∧ dx0++B12 dx ∧ dy +B13 dx ∧ dz +B23 dy ∧ dz.
84
The 1st pair of Maxwell Laws: dF = 0 in the 4-space.
dF = d(3)E ∧ dx0 + d(3)B +∂B
∂x0∧ dx0.
Here d(3) denotes d in the 3-space R3(x, y, z), and d(4) denotes d in R4.
x0 = ct,
therefore
∂B
∂x0=
1
c
∂B
∂t=
1
c
∑α<β
∂Bαβ
∂tdxα ∧ dxβ =
=1
c
∂B12
∂tdx ∧ dy + 1
c
∂B13
∂tdx ∧ dz + 1
c
∂B23
∂tdy ∧ dz.
d(3)E =
(∂E2
∂x− ∂E1
∂y
)dx ∧ dy +
(∂E3
∂x− ∂E1
∂z
)dx ∧ dz+
+
(∂E3
∂y− ∂E2
∂z
)dy ∧ dz, x1 = x, x2 = y, x3 = z.
Conclusion:
(a) d(3)B(x, t) = 0
(b) d(3)E + 1c∂B∂t
= 0.
Corollary:
(a) B = d(3)A (A is vector-potential).
B12 =∂A2
∂x− ∂A1
∂y, B13 =
∂A3
∂x− ∂A1
∂z, B23 =
∂A3
∂y− ∂A2
∂z.
(b) 1cB = −d(3)E in R3.
MagneticField B
metal wire
Electric Field E
C
Current appeaqrs in Crotate it!
(Flux of B) = Circulation of E
85
Lecture 15.
Integration of differential forms. Consider the n-forms in the n-space U ⊂ Rn, x1, . . . , xn.
Ω = f(x) dx1 ∧ . . . ∧ dxn.
By definition,∫· · ·∫
Dn⊂U
f(x) dx1 ∧ . . . ∧ dxn =
∫ (. . .
∫ (∫f(x)dx1
)dx2 . . .
)dxn,
is the ordinary integral.
Let Dn be n-cube (0 ≤ xi ≤ 1).
x1
x2
x11n=10 n=2 n=3
Dn = I1x1× . . .× I1
xn.
What do we know: Let us have a 1 to 1 change of variables: x =x(x′). Change of variables for n-form in Rn: x(x′)
Ω = f(x) dx1 ∧ . . . ∧ dxn =
Ω′
f(x(x′)) det
(∂xi
∂x′j
)dx′1 ∧ . . . ∧ dx′n .
Let D ∼= D′ ⊂ (x′) (i.e. we have a change of coordinates). Then∫D
Ω =
∫D′
Ω′.
Integration of k-forms in n-space along k-surface.
Claim: Consider domain Dk with the boundary ∂Dk. Then:∫Dkx
Domain
dΩ =
∫∂Dkx
Boundary (oriented?)
Ω.
Example (XVII century) – Newton-Leibnitz.
k = 1 :
∫D′df =
∫∂D′
f = f(b)− f(a)
87
Consider a curve in n-space Rn.
A
B
D’D′ :
xi(t), i = 1, . . . , nt = a, x(a) = At = b, x(b) = B
.
∫D′
∂f
∂xidxi =
def
b∫a
∂f
∂xi(x(t))
∂xi
∂tdt =
b∫a
(∂f
∂xi(x(t))
∂xi
∂tdt
)φ∗(df)
=
b∫a
φ∗(df) =
=
b∫a
Φ(t) dt = f(b)− f(a), where Φ(t) =
(∂f
∂xi(x(t))
∂xi
∂t
)x=x(t)
.
What is orientation for k > 1?
For linear space over R orientation is a choice of basis up to linearchange with det > 0.
Orientation in Rn or in U ⊂ Rn is given by Cartesian coordinates(x1, . . . , xn) up to change of coordinates x = x(x′) such that J > 0.
Another point of view: orientation provided at the point P ∈ U ⊂Rn is given by the basis (∂/∂x1, . . . , ∂/∂xn) in the “tangent” spaceTP (Rn) generated by ∂/∂xi.
Consider a submanifold in Rn defined by an equation (globally!):
Mn−1 = f(x1, . . . , xn) = 0, such that (df)Mn−1 = 0.
∆)
f)
∆)
f)
P
Dn
n−1M
(u)externalnormal
“Tangent vectors” to the submanifold M are (∂/∂u1, . . . , ∂/∂un−1) if(u1, . . . , un−1) are local coordinates in Mn−1 near the point P .
88
Take vectors(∇f, ∂
∂u1, . . . ,
∂
∂un−1
), where ∇f =
n∑i=1
(∂f
∂xi
)∂
∂xi.
Compare this basis the chosen oriented basis(
∂∂x1, . . . , ∂
∂xn
).
(∇f, τu) =(∇f, ∂
∂u
)+←−−→
(∂
∂x1, . . . ,
∂
∂xn
)= τx.
Induced orientation at Mn−1 is such that determinant is + (> 0),Mn−1 = ∂Dn.
89
Lecture 16.
Orientation:
Mn−1 − hypersurface in Rn given by equation f (x1, . . . , xn) = 0 ,such that df = 0 in all points x where f(x) = 0 .
n−1M
n−1M
P
Dn
Dn=
∇ f =
(∂f
∂x1, . . . ,
∂f
∂xn
),
∇ f = 0 on M.
Orientation in Rn with given coordinates (x1, . . . , xn) .
∂
∂x1, . . . ,
∂
∂xn− basis in TP (Rn)
Let∂f
∂x1= 0 at P ∈ Mn
(u1, . . . , un−1) = (x2, . . . , xn)
- local coordinates in M near P . “Implicit function theorem”:
x1 = Φ(x2, . . . , xn) if∂f
∂x1= 0
∣∣∣∣P
near P on the surface given by the equation f (x1, . . . , xn) = 0 .
Basis (∇ f
∣∣∣∣P
,∂
∂x2, . . . ,
∂
∂xn
)= A
(∂
∂x1, . . . ,
∂
∂xn
)det A > 0 ↔ x2, . . . , xn given (+)
A
(∂
∂x1, . . . ,
∂
∂xn
)=
(∇ f
∣∣∣∣P
,∂
∂x2, . . . ,
∂
∂xn
)90
Definition:
det A > 0 ↔ (x2, . . . , xn) are oriented coordinates in Mn−1
det A < 0 ↔ (−x2, . . . , xn) are oriented coordinates in Mn−1
“Orientation of ∂ D induced by orientation of Rn ”
Now we prove this theorem for every manifold U of any dimension M ,U ⊂ RM , Ωn−1 - (n-1)-form in U .
φ : In → U
- “singular cube” (smooth map).
We need to proveφ∗ Ω = Ωn−1
(pull-back) in cube In .
By definition ∫Ω =
∫φ∗Ω
φ : In−1 → U In−1
So our theorem follows from the result above.
OK.
117
Lecture 23. Algebraic Boundary.
Stokes formula: ∫D
dΩ =
∫∂D
Ω
We proved it for cube In , 0 ≤ xj ≤ 1 , j = 1, . . . , n . But everyconvex body is isomorphic to cube up to change of coordinates. We calculatenow boundary of cubes and simplices with orientation.
∂ In =n∪i=1
(In−1xi=1 ∪ In−1
xi=0
)x=0 x=1
I
x=0 x=1
y=1
y=0 x
yI
1
2
Orientation of cube is given by coordinates (x1, . . . , xn) , basis e1, . . . , en ,where ej = ∂/∂xj .
Calculate orientation of In−1xi=0 and In−1
xi=1 induced by (x1, . . . , xn) orien-tation of In :
We have (n = xi) for In−1xi=1 and (n = −xi) for In−1
xi=0 . We have τ =(x1, . . . , xi, . . . xn) (−1)? . For (n, τ) we have
(Singularcubes In(s) : Inψs−→ U , andSingularsimplexes σn(s) : σn
ψs−→ U)
Their algebraic boundary can be defined naturally as well as singularchains as finite linear combination of singular simplices (cubes). The cyclesare chains with zero boundary. Singular homology group is defined as factorof space of cycles by the ”exact ” cycles which are algebraic boundaries ofsingular chains. The space of cycles is very big but homology group is not,it is topologically (homotopy) invariant.
121
Lecture 24. Differential Forms and Homotopy PoincareLemma.
Definition. Homotopy process is a map F (C∞ - map)
U ×R F−→ V
x, t → F (x, t) = y
U , V - manifolds (open domains in euclidean spaces U ⊂ RM , V ⊂ RN
or other)F (x, t = const) = ft : U → V
(deformation or homotopy of map ft).Maps f1 = f and f0 = g are called “homotopic maps” (C∞ - homo-
topy).
Theorem. For every closed differential k - form Ω in Λk(V ) we have
f ∗ Ω − g∗Ω = du , u ∈ Λk−1(U) ,
Λk - space of all C∞ differential forms on any manyfold.
Proof. We have by definition dΩ = 0 .Remind.
d f ∗ = f ∗ d , d g∗ = g∗ d
Sod (f ∗Ω) = d (g∗Ω) = 0
Take homotopy process
F : U ×R → V
F (x, t) : F |t=0 = g , F |t=1 = f
Consider k - form F ∗Ω in Λk(U ×R) .Every form A in U ×R has the following form
A = a + b ∧ dt (k − form)
wherea =
∑ai1...ik dx
i1 ∧ · · · ∧ dxik = a(x, t)
122
b =∑
bj1...jk−1dxj1 ∧ · · · ∧ dxjk−1 = b(x, t)
a(x, t) = A|t=const
We have:
dA = da + (−1)k a ∧ dt + db ∧ dtU ×R U U
a =∂a
∂t(x, t) , a = A|t=const (put dt = 0)
Define operator
D : Λk(U ×R) → Λk−1(U)
DA =
∫ 1
0
b ∧ dt , b = b(x, t)
Lemma 1:± (dD ± D d) A = A|t=1 − A|t=0
(sign depends on dimension and plays no role here).
Proof of Lemma.
DA =
∫ 1
0
b ∧ dt
D (dA) = D ( da + (−1)k a ∧ dt + db ∧ dt ) =U ×R U U
= (−1)k (a|t=1 − a|t=0) + dU
(∫ 1
0
b(x, t) ∧ dt)
=
= (−1)k (A|t=1 − A|t=0) + dUDA = D dA
Lemma is proved.
Apply lemma to the form A = F ∗(Ω) in U ×R . We have
A|t=1 = f ∗Ω , A|t=0 = g∗ Ω
dA = dF ∗(Ω) = F ∗(dΩ) = 0
because dΩ = 0 (closed form).
123
So we have
f ∗(Ω) − g∗(Ω) = d u , u = DA
Theorem is proved.
Poincare Lemma: Let U = Ball Dn . Every closed k - form in the Ball isexact.
Proof. Mapf : [0, 1]×Dn → Dn
such that f(x) = x is homotopic to the map g , g(x) = 0 .Obviously we have
g∗(Ω) = 0 , 1 ≤ k ≤ n− 1
Sof ∗(Ω) = Ωk , Ωk − 0 = du
Poincare Lemma is proved.
Cohomology:
Hk(U) = Closed forms /Exact forms
Ω ∼ Ω′ iff dΩ = dΩ′ = 0 , Ω − Ω′ = du
Cohomology Ring is given by Product of forms.
Examples:1. U = pointH0(U) = R , Hk(U) = 0, k = 0.
1′. U = l pointsH0(U) = Rl , Hk(U) = 0, k = 0.
2. U = BallH0(U) = R , Hk(U) = 0, k = 0.
3. U = S1 : H0 = R , H1 = R , Hk = 0 , k = 0, 1.
4. U = Sn : H0 = Hn = R , Hk = 0 , k = 0, n (not provedyet).
5. U = connected domain in R2 , H0 = R , H1 = Rp (p =number of holes), H2 = 0 (!) .
124
Lecture 25. Examples of important closed differentialforms.