Tensors: Index Notation, Transformation Laws and Physical Tensors
CSIR NET Physical Sciences: Cartesian and general tensors, Kronecker delta, Levi-Civita identities, metric, symmetry counting, inertia and field tensors.
- 20 min
On this page · 13 sections
Prerequisites
- Linear Algebra, Matrices and Eigenvalues: orthogonal matrices, similarity transformations and principal axes are the matrix face of rank-2 tensors.
- Dimensional Analysis and Vector Calculus: cross products, curl and the vector identities that index notation proves in two lines.
- Central Forces and Collisions: angular momentum, which the inertia tensor relates to angular velocity.
The Hook: When J Is Not Parallel to E
Index Notation
Einstein summation convention: an index repeated once as a pair in a term is summed over its range (1 to 3 in space, 0 to 3 in spacetime); a repeated index is called a dummy index and can be renamed freely. An index that appears once in each term is a free index; it must be the same in every term of an equation.
\( a_ib_i = \mathbf a\cdot\mathbf b \), \( A_{ij}x_j = (A\mathbf x)_i \), \( A_{ii} = \operatorname{tr}A \).
An index may never appear three times in one term: \( a_ib_ic_i \) is meaningless.
The two special symbols
Kronecker delta: \( \delta_{ij} = 1 \) if \( i = j \), else 0. It “renames” an index: \( \delta_{ij}a_j = a_i \). In \( n \) dimensions, \( \delta_{ii} = n \).
Levi-Civita symbol: \( \epsilon_{ijk} = +1 \) for even permutations of \( (1, 2, 3) \), \( -1 \) for odd permutations, 0 if any two indices are equal. It is totally antisymmetric.
$$(\mathbf a\times\mathbf b)_i = \epsilon_{ijk}a_jb_k, \qquad \det A = \epsilon_{ijk}A_{1i}A_{2j}A_{3k}, \qquad (\nabla\times\mathbf A)_i = \epsilon_{ijk}\partial_jA_k$$The contraction identities
$$\boxed{\epsilon_{ijk}\epsilon_{imn} = \delta_{jm}\delta_{kn} - \delta_{jn}\delta_{km}}$$Contract once more and again:
$$\epsilon_{ijk}\epsilon_{ijn} = 2\delta_{kn}, \qquad \epsilon_{ijk}\epsilon_{ijk} = 6.$$Also: \( \epsilon_{ijk}S_{jk} = 0 \) for any symmetric \( S \) (symmetric times antisymmetric, fully contracted, vanishes). That is why \( \nabla\times\nabla f = 0 \) (since \( \partial_j\partial_k \) is symmetric) and \( \mathbf a\times\mathbf a = 0 \).
Cartesian Tensors
Under a rotation of axes, coordinates transform as \( x'_i = R_{ij}x_j \), where \( R \) is orthogonal: \( R_{ik}R_{jk} = \delta_{ij} \) (that is, \( RR^{T} = I \)).
A Cartesian tensor of rank \( r \) is a set of \( 3^r \) components (in 3D) that transform as
$$\boxed{T'_{ij\ldots} = R_{ia}R_{jb}\cdots T_{ab\ldots}}$$with one factor of \( R \) per index.
| Rank | Components (3D) | Transforms as | Examples |
|---|---|---|---|
| 0 | 1 | \( \phi' = \phi \) | mass, temperature, \( \mathbf a\cdot\mathbf b \) |
| 1 | 3 | \( v'_i = R_{ij}v_j \) | position, force, \( \mathbf E \) |
| 2 | 9 | \( T'_{ij} = R_{ia}R_{jb}T_{ab} \), i.e. \( T' = RTR^{T} \) | inertia, stress, strain, conductivity, dielectric, quadrupole moment |
| 3 | 27 | three factors of \( R \) | piezoelectric tensor |
| 4 | 81 | four factors of \( R \) | elastic stiffness |
For rank 2 the law is a similarity transformation, so everything invariant under similarity (trace, determinant, eigenvalues) is a scalar.
Not every array is a tensor. The components of a matrix that happens to be written in one frame are a tensor only if, by construction, they transform this way in every frame. The test is physical: if \( T_{ij} \) relates two vectors (\( J_i = \sigma_{ij}E_j \)) or is built from vectors (\( a_ib_j \)), it is a tensor.
Algebraic operations
- Addition of tensors of the same rank gives a tensor of that rank.
- Outer product: \( a_ib_j \) (rank 1 times rank 1) is rank 2; in general ranks add.
- Contraction: setting two indices equal and summing reduces the rank by 2. \( T_{ii} \) is a scalar; \( T_{ijj} \) is a vector.
- Inner product: outer product followed by contraction, e.g. \( T_{ij}v_j \).
- Quotient law: if \( X_{ij}a_j \) is a vector for every vector \( a_j \), then \( X_{ij} \) is a rank-2 tensor. More generally, if contracting an unknown array with an arbitrary tensor always gives a tensor, the unknown array is a tensor.
Symmetry and decomposition
Any rank-2 tensor splits uniquely into symmetric and antisymmetric parts, and the symmetric part into its trace and a traceless remainder:
$$T_{ij} = \underbrace{\tfrac{1}{3}T_{kk}\delta_{ij}}_{1\ \text{component}} + \underbrace{\left[\tfrac{1}{2}(T_{ij} + T_{ji}) - \tfrac{1}{3}T_{kk}\delta_{ij}\right]}_{5\ \text{components}} + \underbrace{\tfrac{1}{2}(T_{ij} - T_{ji})}_{3\ \text{components}}$$These three pieces never mix under rotations: \( 9 = 1 + 3 + 5 \). The antisymmetric part is equivalent to an axial vector \( v_i = \tfrac{1}{2}\epsilon_{ijk}T_{jk} \) (only in 3D, because only there does an antisymmetric matrix have 3 independent entries). In the language of Group Theory: SU(2) and O(3), this is the decomposition of the product of two \( l = 1 \) representations into \( l = 0, 1, 2 \).
Symmetry under index exchange is preserved by rotations, so it is a meaningful property of the tensor, not of the frame.
Counting independent components
| Tensor | Independent components in \( n \) dimensions | 3D | 4D |
|---|---|---|---|
| General rank 2 | \( n^2 \) | 9 | 16 |
| Symmetric rank 2 | \( n(n + 1)/2 \) | 6 | 10 |
| Antisymmetric rank 2 | \( n(n - 1)/2 \) | 3 | 6 |
| Totally symmetric rank 3 | \( \binom{n + 2}{3} \) | 10 | 20 |
| Totally antisymmetric rank \( p \) | \( \binom{n}{p} \) | \( p = 3 \): 1 | \( p = 3 \): 4; \( p = 4 \): 1 |
| Riemann curvature tensor | \( n^2(n^2 - 1)/12 \) | 6 | 20 |
| Elastic stiffness \( C_{ijkl} \) (3D, all symmetries) | — | 21 | — |
| Elastic stiffness, isotropic medium | — | 2 (Lamé constants) | — |
For a totally antisymmetric tensor of rank greater than \( n \), every component vanishes: there are not enough distinct index values.
Isotropic tensors
An isotropic tensor has the same components in every rotated frame.
- Rank 0: every scalar.
- Rank 1: none except zero.
- Rank 2: \( \lambda\delta_{ij} \).
- Rank 3: \( \lambda\epsilon_{ijk} \) (under proper rotations).
- Rank 4: \( \alpha\delta_{ij}\delta_{kl} + \beta\delta_{ik}\delta_{jl} + \gamma\delta_{il}\delta_{jk} \).
Physical consequence: an isotropic material has a scalar conductivity (\( \sigma_{ij} = \sigma\delta_{ij} \)) and, because the stiffness tensor is symmetric under exchange of its index pairs (\( C_{ijkl} = C_{klij} \), \( C_{ijkl} = C_{jikl} \)), only two elastic constants (the three coefficients \( \alpha, \beta, \gamma \) above reduce to two because \( \beta = \gamma \)).
Proper and improper transformations: pseudotensors
Orthogonal matrices have \( \det R = \pm1 \). Rotations have \( +1 \); reflections and the inversion \( R = -I \) have \( -1 \). A pseudotensor picks up an extra factor \( \det R \):
$$T'_{ij\ldots} = (\det R)\,R_{ia}R_{jb}\cdots T_{ab\ldots}$$\( \epsilon_{ijk} \) itself is a rank-3 pseudotensor (its components are the same in every frame, which for an improper \( R \) requires the extra sign). Consequently \( \mathbf a\times\mathbf b \) of two polar vectors is an axial (pseudo)vector.
Under inversion \( \mathbf r\to-\mathbf r \):
| Quantity | Type | Under inversion |
|---|---|---|
| \( \mathbf r \), \( \mathbf p \), \( \mathbf E \), \( \mathbf F \) | polar vector | changes sign |
| \( \mathbf L = \mathbf r\times\mathbf p \), \( \mathbf B \), \( \boldsymbol\omega \), spin | axial vector | unchanged |
| \( \mathbf E\cdot\mathbf B \), \( \mathbf S\cdot\mathbf p \) | pseudoscalar | changes sign |
| \( T_{ij} \) (true rank 2) | tensor | unchanged (two factors of \( -1 \)) |
Physical Tensors You Must Handle
Inertia tensor
$$I_{ij} = \sum_\alpha m_\alpha\left(r_\alpha^2\delta_{ij} - x_{\alpha i}x_{\alpha j}\right), \qquad L_i = I_{ij}\omega_j, \qquad T = \tfrac{1}{2}I_{ij}\omega_i\omega_j$$It is real and symmetric, so it can be diagonalised by a rotation: the eigenvectors are the principal axes, the eigenvalues the principal moments. \( \mathbf L \parallel \boldsymbol\omega \) only for rotation about a principal axis.
Stress, strain and response tensors
- Stress \( \sigma_{ij} \): force per unit area in direction \( i \) on a surface with normal \( j \). Symmetric (torque balance on a small cube), so 6 components. Its trace gives the pressure: \( p = -\tfrac{1}{3}\sigma_{ii} \).
- Strain \( \varepsilon_{ij} = \tfrac{1}{2}(\partial_iu_j + \partial_ju_i) \): symmetric part of the displacement gradient; the antisymmetric part is a rigid rotation. \( \varepsilon_{ii} = \nabla\cdot\mathbf u \) is the fractional volume change.
- Response tensors \( \sigma_{ij} \) (conductivity), \( \epsilon_{ij} \) (permittivity), \( \chi_{ij} \) (susceptibility): symmetric for non-magnetic, non-dissipative media, so a crystal has three principal values. Cubic crystals are isotropic for rank-2 properties; uniaxial crystals (tetragonal, hexagonal, trigonal) have two distinct principal values, hence one optic axis.
General Tensors: Curvilinear Coordinates and the Metric
In general coordinates \( x^i \to x'^i(x) \), the transformation matrix is no longer orthogonal, and two kinds of index appear.
Contravariant components (upper index) transform like the differentials \( dx^i \):
$$A'^i = \frac{\partial x'^i}{\partial x^j}A^j$$Covariant components (lower index) transform like the gradient \( \partial f/\partial x^i \):
$$A'_i = \frac{\partial x^j}{\partial x'^i}A_j$$A tensor of type \( (p, q) \) has \( p \) upper and \( q \) lower indices, each transforming with its own factor. Summation is always over one upper and one lower index, which makes \( A^iB_i \) a scalar.
The metric tensor
Distance defines the metric:
$$\boxed{ds^2 = g_{ij}\,dx^i\,dx^j}$$\( g_{ij} \) is symmetric; its inverse \( g^{ij} \) (with \( g^{ik}g_{kj} = \delta^i_j \)) raises indices: \( A_i = g_{ij}A^j \), \( A^i = g^{ij}A_j \).
| Coordinates | \( ds^2 \) | \( \sqrt g \) |
|---|---|---|
| Cartesian | \( dx^2 + dy^2 + dz^2 \) | 1 |
| Plane polar | \( dr^2 + r^2d\theta^2 \) | \( r \) |
| Cylindrical | \( d\rho^2 + \rho^2d\phi^2 + dz^2 \) | \( \rho \) |
| Spherical | \( dr^2 + r^2d\theta^2 + r^2\sin^2\theta\,d\phi^2 \) | \( r^2\sin\theta \) |
| Minkowski (signature \( +,-,-,- \)) | \( c^2dt^2 - dx^2 - dy^2 - dz^2 \) | — |
For orthogonal coordinates, \( g_{ii} = h_i^2 \) (the scale factors), and the “physical” components used in vector calculus are \( \hat A_i = h_iA^i = A_i/h_i \) (no sum). The divergence and Laplacian follow from
$$\nabla\cdot\mathbf A = \frac{1}{\sqrt g}\partial_i\left(\sqrt g\,A^i\right), \qquad \nabla^2f = \frac{1}{\sqrt g}\partial_i\left(\sqrt g\,g^{ij}\partial_jf\right).$$In polar coordinates this gives \( \nabla^2f = \frac{1}{r}\partial_r(r\,\partial_rf) + \frac{1}{r^2}\partial_\theta^2f \), the familiar result.
Covariant derivative
Partial derivatives of vector components are not tensors in curvilinear coordinates, because the basis vectors themselves change from point to point. The fix is the covariant derivative,
$$\nabla_jA^i = \partial_jA^i + \Gamma^i_{jk}A^k, \qquad \Gamma^i_{jk} = \tfrac{1}{2}g^{il}\left(\partial_jg_{lk} + \partial_kg_{lj} - \partial_lg_{jk}\right).$$In plane polar coordinates the only non-zero Christoffel symbols are \( \Gamma^r_{\theta\theta} = -r \) and \( \Gamma^\theta_{r\theta} = \Gamma^\theta_{\theta r} = 1/r \). The \( -r \) term is the centripetal piece: the acceleration \( \ddot r - r\dot\theta^2 \) is \( \ddot x^r + \Gamma^r_{\theta\theta}\dot\theta^2 \).
Four-vectors and the field tensor
In special relativity, with \( x^\mu = (ct, \mathbf x) \) and metric \( \eta = \operatorname{diag}(1, -1, -1, -1) \), Lorentz transformations play the role of rotations. Four-vectors \( p^\mu = (E/c, \mathbf p) \) have invariant length \( p^\mu p_\mu = E^2/c^2 - p^2 = m^2c^2 \).
The electromagnetic field is an antisymmetric rank-2 four-tensor \( F^{\mu\nu} \): 6 independent components, exactly the three components of \( \mathbf E \) and three of \( \mathbf B \). Under a boost \( \mathbf E \) and \( \mathbf B \) mix, and the two Lorentz invariants are
$$E^2 - c^2B^2 \qquad\text{and}\qquad \mathbf E\cdot\mathbf B.$$So a field that is purely magnetic in one frame (\( E = 0 \), \( \mathbf E\cdot\mathbf B = 0 \), \( E^2 - c^2B^2 < 0 \)) can never be purely electric in another.
Memory Tricks
Don’t Confuse
| This | vs | That |
|---|---|---|
| Free index: appears once in every term | Dummy index: appears twice in one term, summed | |
| Polar vector (\( \mathbf r \), \( \mathbf E \)): flips under inversion | Axial vector (\( \mathbf L \), \( \mathbf B \)): does not flip | |
| Contravariant \( A^i \): transforms with \( \partial x'^i/\partial x^j \) | Covariant \( A_i \): transforms with \( \partial x^j/\partial x'^i \) | |
| Cartesian tensors: upper and lower indices identical | Curvilinear: raised and lowered with \( g_{ij} \) | |
| Antisymmetric rank 2 \( \leftrightarrow \) vector: only in 3D | In 4D antisymmetric rank 2 has 6 components (\( \mathbf E \), \( \mathbf B \)) | |
| \( \delta_{ii} = 3 \) (in 3D) | \( \delta_{ij}\delta_{ij} = 3 \) as well; \( \epsilon_{ijk}\epsilon_{ijk} = 6 \) | |
| Principal moments: eigenvalues of \( I_{ij} \) | Diagonal entries of \( I_{ij} \) in an arbitrary frame |
Common Mistakes
Practice
All questions below are original practice questions written for this page. They are not from any CSIR NET paper.
Level 1: Foundation
- Evaluate \( \delta_{ij}\delta_{ji} \) in three dimensions.
- Evaluate \( \epsilon_{ijk}\epsilon_{ijk} \).
- How many independent components does an antisymmetric rank-2 tensor have in four dimensions?
Answers
- 3 (\( \delta_{ij}\delta_{ji} = \delta_{ii} \)).
- 6 (the number of permutations of three objects).
- 6 (\( n(n - 1)/2 \) with \( n = 4 \)): the six components of \( \mathbf E \) and \( \mathbf B \) in \( F^{\mu\nu} \).
Level 2: Part B Style
Practice Q1. The contraction \( \epsilon_{ijk}\epsilon_{ijm} \) equals:
(a) \( \delta_{km} \) (b) \( 2\delta_{km} \) (c) \( 3\delta_{km} \) (d) \( 6\delta_{km} \)
Answer: (b) 2δₖₘ
From \( \epsilon_{ijk}\epsilon_{imn} = \delta_{jm}\delta_{kn} - \delta_{jn}\delta_{km} \) with the second pair set equal: \( \epsilon_{ijk}\epsilon_{ijm} = \delta_{jj}\delta_{km} - \delta_{jm}\delta_{kj} = 3\delta_{km} - \delta_{km} = 2\delta_{km} \). Contracting once more gives 6, option (d)’s number without the delta.
Practice Q2. The number of independent components of a symmetric rank-2 tensor in four-dimensional spacetime is:
(a) 6 (b) 10 (c) 16 (d) 20
Answer: (b) 10
\( n(n + 1)/2 = 4\times5/2 = 10 \). (The energy–momentum tensor and the metric are examples.) (a) is the antisymmetric count; (d) is the Riemann tensor in 4D.
Practice Q3. For position \( \mathbf r = (x_1, x_2, x_3) \), the quantity \( \epsilon_{ijk}\,\partial_j(x_kx_lx_l) \) equals:
(a) \( 2x_i \) (b) \( 0 \) (c) \( x_ix_lx_l \) (d) \( 6x_i \)
Answer: (b) 0
\( \partial_j(x_kr^2) = \delta_{jk}r^2 + 2x_kx_j \). Both terms are symmetric in \( j, k \), and \( \epsilon_{ijk} \) is antisymmetric, so the contraction vanishes. This is \( \nabla\times(r^2\mathbf r) = 0 \): a radial field depending only on \( r \) is curl-free. The other options treat \( \epsilon_{ijk} \) as if it were \( \delta_{jk} \) or ignore its antisymmetry.
Practice Q4. A uniform cube of mass \( M \) and side \( a \) rotates about an axis through one corner. The smallest principal moment of inertia about that corner is:
(a) \( \frac{2}{3}Ma^2 \) (b) \( \frac{1}{6}Ma^2 \) (c) \( \frac{11}{12}Ma^2 \) (d) \( \frac{5}{12}Ma^2 \)
Answer: (b) Ma²/6
The inertia tensor about the corner is \( Ma^2\left[\frac{11}{12}I - \frac{1}{4}J\right] \), with \( J \) the all-ones matrix. Eigenvalues: \( \frac{11}{12} - \frac{3}{4} = \frac{1}{6} \) along the body diagonal; \( \frac{11}{12} \) twice. (a) is a diagonal entry, not a principal moment; (c) is the largest principal moment.
Level 3: Part C Style
Practice Q5. Let \( T_{ij} \) be an arbitrary Cartesian rank-2 tensor in three dimensions. Consider:
- \( T_{ij} \) decomposes into parts with 1, 3 and 5 independent components that do not mix under rotations.
- The antisymmetric part is equivalent to the axial vector \( v_i = \tfrac{1}{2}\epsilon_{ijk}T_{jk} \).
- The symmetric traceless part has 6 independent components.
- If \( S_{ij} \) is symmetric and \( A_{ij} \) antisymmetric, then \( S_{ij}A_{ij} = 0 \).
Which are correct?
(a) 1, 2 and 4 only (b) 1 and 3 only (c) 2, 3 and 4 only (d) 1, 2, 3 and 4
Answer: (a) 1, 2 and 4 only
- Trace (1), antisymmetric (3), symmetric traceless (5): \( 9 = 1 + 3 + 5 \). True.
- \( \tfrac{1}{2}\epsilon_{ijk}T_{jk} \) picks out the antisymmetric part (the symmetric part contracts to zero), and \( \epsilon \) makes it axial. True.
- Symmetric has 6; traceless removes one, leaving 5. False.
- \( S_{ij}A_{ij} = S_{ji}A_{ji} = S_{ij}(-A_{ij}) \) after relabelling, so it equals its own negative. True.
Practice Q6. In plane polar coordinates \( (r, \theta) \), consider:
- The metric is \( g_{ij} = \operatorname{diag}(1, r^2) \).
- For a scalar \( f \), the covariant component of the gradient is \( (\nabla f)_\theta = \partial f/\partial\theta \), while the physical component is \( \frac{1}{r}\partial f/\partial\theta \).
- The Christoffel symbol \( \Gamma^r_{\theta\theta} = -r \).
- The Laplacian is \( \partial_r^2f + \frac{1}{r^2}\partial_\theta^2f \).
Which are correct?
(a) 1 and 4 only (b) 1, 2 and 3 only (c) 2 and 3 only (d) 1, 2, 3 and 4
Answer: (b) 1, 2 and 3 only
- \( ds^2 = dr^2 + r^2d\theta^2 \). True.
- Covariant components of a gradient are partial derivatives; the physical component divides by the scale factor \( h_\theta = r \). True.
- \( \Gamma^r_{\theta\theta} = \tfrac{1}{2}g^{rr}(-\partial_rg_{\theta\theta}) = -\tfrac{1}{2}(2r) = -r \). True.
- With \( \sqrt g = r \), \( \nabla^2f = \frac{1}{r}\partial_r(r\partial_rf) + \frac{1}{r^2}\partial_\theta^2f \); the \( \frac{1}{r}\partial_rf \) term is missing. False.
Practice Q7. A rank-2 tensor has components \( T = \operatorname{diag}(1, 2, 3) \) in a frame \( (x_1, x_2, x_3) \). Consider:
- After a rotation of the axes by \( 90^\circ \) about \( x_3 \), the new component \( T'_{11} = 2 \).
- \( T_{ii} \) and \( \det T \) are the same in every rotated frame.
- Under spatial inversion (\( R = -I \)) all components of \( T_{ij} \) change sign.
- Under spatial inversion, the cross product of two polar vectors changes sign.
Which are correct?
(a) 1 and 2 only (b) 1, 2 and 4 only (c) 2 and 3 only (d) 1, 2 and 3 only
Answer: (a) 1 and 2 only
- A \( 90^\circ \) rotation about \( x_3 \) swaps the roles of the first two axes: \( T' = RTR^{T} = \operatorname{diag}(2, 1, 3) \). True.
- Trace and determinant are similarity invariants. True.
- \( T'_{ij} = (-1)(-1)T_{ij} = T_{ij} \): a true rank-2 tensor is unchanged. False.
- \( \mathbf a\times\mathbf b \to(-\mathbf a)\times(-\mathbf b) = \mathbf a\times\mathbf b \): an axial vector does not change sign. False.
Quick Revision Table
| Idea | Formula or fact |
|---|---|
| Summation | repeated index summed; free indices match across terms |
| \( \delta \) | \( \delta_{ij}a_j = a_i \), \( \delta_{ii} = n \) |
| \( \epsilon\epsilon \) | \( \epsilon_{ijk}\epsilon_{imn} = \delta_{jm}\delta_{kn} - \delta_{jn}\delta_{km} \); \( \epsilon_{ijk}\epsilon_{ijn} = 2\delta_{kn} \); \( \epsilon_{ijk}\epsilon_{ijk} = 6 \) |
| Rank-\( r \) law | \( T'_{ij\ldots} = R_{ia}R_{jb}\cdots T_{ab\ldots} \); rank 2: \( T' = RTR^{T} \) |
| Contraction | rank drops by 2 |
| Quotient law | \( X_{ij}a_j \) a vector for all \( a \) \( \Rightarrow \) \( X \) is a tensor |
| Decomposition | \( 9 = 1 + 3 + 5 \) |
| Component counts | sym \( n(n + 1)/2 \), antisym \( n(n - 1)/2 \), Riemann \( n^2(n^2 - 1)/12 \) |
| Isotropic | \( \delta_{ij} \), \( \epsilon_{ijk} \), three \( \delta\delta \) products |
| Pseudotensor | extra factor \( \det R \); \( \epsilon_{ijk} \), \( \mathbf L \), \( \mathbf B \) |
| Inertia | \( I_{ij} = \sum m(r^2\delta_{ij} - x_ix_j) \); cube corner: \( \frac{1}{6} \), \( \frac{11}{12} \), \( \frac{11}{12} \) (\( Ma^2 \)) |
| Metric | \( ds^2 = g_{ij}dx^idx^j \); spherical \( \sqrt g = r^2\sin\theta \) |
| Divergence | \( \frac{1}{\sqrt g}\partial_i(\sqrt gA^i) \) |
| Polar Christoffels | \( \Gamma^r_{\theta\theta} = -r \), \( \Gamma^\theta_{r\theta} = 1/r \) |
| Field tensor | antisymmetric, 6 components; invariants \( E^2 - c^2B^2 \), \( \mathbf E\cdot\mathbf B \) |
Key Takeaways
What’s Next
- Group Theory: SU(2) and O(3): rotations as a group, and why tensors of rank \( l \) decompose into angular momentum pieces.
- Linear Algebra, Matrices and Eigenvalues: diagonalising symmetric matrices, the engine behind principal axes.
- Quadratic Forms (Mathematical Sciences): the kinetic energy \( \tfrac{1}{2}I_{ij}\omega_i\omega_j \) is a quadratic form, and Sylvester’s law of inertia is the theory behind metric signatures.