Characteristic classes turn the global twisting of a vector bundle into cohomology classes that can be calculated, compared and pulled back. They help answer concrete questions: can a continuous nonzero vector be chosen everywhere, can two bundles be isomorphic, and what obstruction survives when local coordinates change? The gold standard is to name the bundle, coefficient system and degree before calculating, then distinguish an obstruction from a complete classification.
This guide develops Stiefel–Whitney, Chern, Euler and Pontryagin classes through original worked cases. You will calculate parity around a loop, combine line bundles, test tangent fields on surfaces, and see why a zero answer sometimes means only that one test has found nothing. The setting is ordinary finite-rank vector bundles over familiar spaces such as manifolds and finite CW complexes; more general bases require additional hypotheses.
Use the Gold Standard master for the collection, the Education parent for learning that transfers, and How X Works for mechanisms across subjects. The existing classifying-spaces guide explains universal bundles; the Thom-space and Euler-class guide develops the geometric obstruction construction.
Choose your characteristic classes route
Chapters 1–3: Name the bundle and its classes
Chapters 4–6: Calculate real and complex examples
Chapters 7–9: Test products and sections
Chapters 10–12: Interpret and repair calculations
Chapter index
Name the bundle and its classes · Chapters 1–3
1. Identify the object before choosing the invariant
Calculate real and complex examples · Chapters 4–6
4. Start with orientation and the first Stiefel–Whitney class
Test products and sections · Chapters 7–9
7. Work a four-dimensional Chern-class example
Interpret and repair calculations · Chapters 10–12
10. Calculate a Pontryagin class and a characteristic number
Transfer and continue · Chapters 13–14
CHAPTER 1 OF 14 · Name the bundle and its classes
1. Identify the object before choosing the invariant
A vector bundle E over X attaches a vector space E_x to every point x of a base space X. Nearby fibres fit together continuously. On a sufficiently small neighbourhood U, the bundle looks like U times a fixed vector space. This local product condition makes familiar linear algebra available in each patch. It does not supply one global basis valid over the whole base.
There are three different dimensions to record. The base has a dimension, the fibre has a rank, and a cohomology class has a degree. For a complex line bundle over a four-dimensional manifold, the complex rank is one, the underlying real rank is two, and its first Chern class has degree two. Confusing these numbers produces incorrect vanishing claims before any calculation begins.
A section selects a vector s(x) in E_x continuously. A nowhere-zero section never selects the zero vector. A frame consists of enough pointwise independent sections to form a basis in every fibre. One nonzero section is therefore much less than a global frame when the rank exceeds one. For a real line bundle, however, a nowhere-zero section supplies a complete basis and trivialises the bundle.
A useful object card
For every problem, write one sentence specifying the base, real or complex fibre, rank and required equivalence. For example: “E is an oriented real plane bundle over the oriented sphere S², and I want an orientation-preserving bundle isomorphism with a product bundle.” That sentence determines which Euler class is meaningful and which triviality claim is being tested.
Also distinguish the tangent bundle TX from an arbitrary bundle E over X. A surface may support many plane bundles. Their Euler classes need not equal the surface’s Euler characteristic. That familiar equality concerns the tangent bundle with its compatible orientations. A calculation about a different bundle cannot borrow it merely because the base is a surface.
CHAPTER 2 OF 14 · Name the bundle and its classes
2. Understand what a cohomology class records
A characteristic class is attached to the bundle but lives on the base. A degree-k class belongs to H^k(X;A), where A specifies coefficients, commonly the integers or the two-element field. It is an algebraic record that can be evaluated on appropriate cycles. It is not generally a single number and should not be reported as one without explaining the evaluation.
For an oriented sphere, H²(S²;Z) is generated by a class u with evaluation one on the oriented fundamental cycle. Writing c₁(L)=5u tells us the whole first Chern class of a line bundle on this base. Writing its Chern number as 5 additionally uses the chosen orientation and generator. Reversing the orientation of the evaluation cycle reverses that number, while the abstract cohomology class remains the same object.
Products of cohomology classes use the cup product. Degrees add: a degree-two class times another degree-two class has degree four. The base’s cohomology ring controls whether that product vanishes. Over S², u²=0 because degree-four cohomology vanishes. Over complex projective two-space, a degree-two generator h has h² nonzero. Identical-looking algebraic expressions can therefore give different answers on different bases.
Keep the coefficient ring visible
Integer arithmetic and arithmetic modulo two are different calculations. The integer 6 is nonzero, but its reduction modulo two is zero. A mod-two obstruction can disappear while an integral obstruction remains. Neither calculation is wrong; the second deliberately retains less information.
A reliable written solution labels the group next to every result. Compare “the answer is zero” with “w₂(E)=0 in H²(X;F₂), but e(E)=6u in H²(X;Z).” The longer sentence tells the reader exactly which obstruction has disappeared and which one still prevents a section. This distinction is central to the worked surface examples below.
CHAPTER 3 OF 14 · Name the bundle and its classes
3. Use naturality to move a problem to another base
If f:Y→X is continuous, the pullback bundle f*E over Y has fibre E_f(y) at y. A characteristic class a satisfies a(f*E)=f*a(E). On the left, calculate after pulling back the bundle. On the right, calculate first and pull back the cohomology class. Agreement is the naturality requirement, not an optional computational shortcut.
Here is an explicit test. Let L be a complex line bundle over an oriented sphere with c₁(L)=4u. Let f:S²→S² have degree −3. Choose a generator v on the domain sphere consistent with its orientation. By the definition of degree, f*u=−3v. Hence c₁(f*L)=4(−3v)=−12v. The sign comes from the map’s degree; it is not evidence that the bundle has a negative rank.
Now compose with g:S²→S² of degree 2. The composite has degree −6. Pulling back directly gives −24 times the domain generator. Pulling back in two stages gives the same answer. This is a small but valuable consistency test: an implementation that gives −12 or 24 has probably lost a factor or reversed a convention.
Restriction is a pullback too
Restricting E to a subspace is pullback along its inclusion. A class can vanish on each chosen test subspace without vanishing on the whole base. Conversely, a nonzero restriction proves that the original class was nonzero. Restriction is therefore a useful obstruction test but an incomplete detector unless the selected subspaces are known to detect all relevant cohomology.
The classifying-spaces specialist article explains how a universal bundle makes this systematic. A classifying map contains more information than its induced characteristic classes. Replacing the map by a few cohomology outputs is useful compression, with a corresponding loss of information.
CHAPTER 4 OF 14 · Calculate real and complex examples
4. Start with orientation and the first Stiefel–Whitney class
Stiefel–Whitney classes apply to real bundles. The class w_i(E) has degree i with mod-two coefficients, and w₀ is the multiplicative identity. For rank r, classes with i greater than r vanish. The first class detects orientability: w₁(E)=0 precisely when the bundle admits a continuous choice of fibre orientation in this standard setting.
Imagine transporting a chosen orientation around a closed loop. Local changes of frame have determinants with positive or negative signs. A negative total sign means that the orientation returns reversed. This cannot be repaired merely by choosing different local frames. The first Stiefel–Whitney class packages this loop test into a cohomology class.
Worked sign audit
Suppose a loop passes through five frame transitions whose determinant signs are −, +, −, −, +. Their product is negative because three reversals occur. In mod-two notation, the contribution is 1+0+1+1+0=1. A second loop with signs −, −, +, −, − has four reversals and gives zero. These tests concern the bundle along the selected loops; they do not establish orientability everywhere unless the loops detect the relevant first cohomology.
For the Möbius real line bundle over S¹, the fundamental loop gives one. Its w₁ is the nonzero generator a of H¹(S¹;F₂). The trivial line bundle gives zero. These are statements about the line fibres, not about an arbitrary physical strip embedded in three-dimensional space.
Changing the cover may increase the number of transitions and change individual signs. The invariant is the resulting cohomology class, not the number of negative entries in a particular notebook. This is a useful distinction in numerical work: presentation data can change while the geometric obstruction remains fixed.
CHAPTER 5 OF 14 · Calculate real and complex examples
5. Combine real bundles without losing the higher classes
The total Stiefel–Whitney class is w(E)=1+w₁(E)+w₂(E)+…. For a direct sum, w(E⊕F)=w(E)w(F). Multiply in the cohomology ring with mod-two coefficients. The direct sum keeps both fibres at each base point; it is different from multiplying two line fibres by a tensor product.
Let L be the Möbius line bundle over S¹. Since w(L)=1+a, the rank-two sum L⊕L has total class (1+a)²=1+2a+a². Modulo two, 2a=0; on S¹, a²=0 for dimensional reasons. Thus the total class is 1. In this particular example the sum is trivial: the doubled transition −I is a rotation of an oriented plane and can be joined continuously to I. The construction, not merely the zero classes, establishes triviality.
Change the base and watch the answer change
Now let L be the tautological real line bundle over real projective two-space. Its mod-two cohomology ring is F₂[a]/(a³), with a in degree one, and w(L)=1+a. The same multiplication gives w(L⊕L)=1+a². The degree-one obstruction cancels, so the plane bundle is orientable. The degree-two class remains nonzero, proving that the plane bundle is not trivial.
The two calculations differ only in the base ring. On the circle, a² vanished. On projective two-space, it survived. A spreadsheet that drops all squares of degree-one variables will get the second case wrong. Over mod-two coefficients, odd-degree classes can have nonzero squares; do not import an inappropriate real differential-form intuition.
For three copies over projective two-space, (1+a)³=1+a+a² after reducing coefficients and discarding a³. For four copies, (1+a)⁴=1 because all surviving binomial coefficients are even. That last result alone is not permission to assert a global frame. It says exactly that these characteristic-class tests have returned zero.
CHAPTER 6 OF 14 · Calculate real and complex examples
6. Build Chern-class calculations from complex lines
For a complex rank-r bundle, c_i lies in H^(2i)(X;Z), and c_i=0 for i greater than r. We use the convention in which the first Chern class of a complex line agrees with the Euler class of its underlying real plane given the complex orientation. Some texts choose different signs for named generators or normalisations; compare definitions before comparing numbers.
Complex line bundles over a finite CW complex are classified up to topological isomorphism by c₁ in H²(X;Z). This is a special completeness result. It must not be extended automatically to bundles of arbitrary rank or to holomorphic isomorphism classes. Additional analytic structure can distinguish objects with the same underlying topological bundle.
Let L_k over S² mean a complex line bundle whose first Chern class is ku. One way to construct it is to glue trivial lines over two hemispheres using a circle-valued transition map of winding number k, with gluing and generator conventions chosen consistently. The integer is global data left after both hemispheres have been made individually trivial.
Tensor product and dual
For complex line bundles, c₁(L⊗M)=c₁(L)+c₁(M), while c₁(L*)=−c₁(L). Thus L₇⊗L₋₄ has first Chern class 3u. The tensor product L₇⊗L₇* has zero first Chern class and is indeed trivial: the evaluation pairing gives its canonical identification with the trivial complex line.
By contrast, L₇⊕L₋₄ has complex rank two. It also has first Chern class 3u, but it is a different kind of object. The identical first Chern output does not erase the rank difference. Always retain the rank alongside the invariant when making a comparison.
On S², the second Chern class of any complex bundle vanishes because H⁴(S²;Z)=0. This is a statement about the base dimension, not evidence that the bundle has rank one. The reason for vanishing belongs in a complete answer.
CHAPTER 7 OF 14 · Test products and sections
7. Work a four-dimensional Chern-class example
Let X=S²×S² with the product orientation. Write a and b for the degree-two classes pulled back from the first and second spheres, normalised so that the evaluation of ab on X is one. The integral cohomology ring satisfies a²=b²=0, while ab is nonzero. Since both classes have even degree, ab=ba.
Choose complex line bundles A and B with c₁(A)=2a+3b and c₁(B)=−a+4b. Such line bundles exist by the topological line-bundle classification. Let E=A⊕B. The direct-sum formula gives c(E)=(1+c₁(A))(1+c₁(B)). Therefore c₁(E)=a+7b.
For the second class, multiply explicitly: c₂(E)=(2a+3b)(−a+4b)=−2a²+8ab−3ba+12b²=5ab. Its characteristic number is 5. The terms disappear because of the cohomology ring relations, not because negative terms or squares are inconvenient.
A comparison that first Chern class misses
Choose A′ and B′ with c₁(A′)=a+2b and c₁(B′)=5b. The sum F=A′⊕B′ has c₁(F)=a+7b and c₂(F)=5ab. These two tests do not distinguish E from F. Without supplying an applicable classification theorem or an explicit isomorphism, the responsible conclusion is that the computed invariants agree.
Now choose C and D with c₁(C)=a and c₁(D)=7b. The sum G=C⊕D still has first class a+7b, but its second class is 7ab. Consequently G cannot be isomorphic to E as a complex bundle over this fixed base: an isomorphism would force equal Chern classes, and 7ab is not 5ab.
This is the practical strength of characteristic classes. A discrepancy gives a decisive obstruction using manageable algebra. An agreement can require more work. Recording these as two different outcomes avoids turning a useful necessary condition into an unsupported equivalence claim.
CHAPTER 8 OF 14 · Test products and sections
8. Use the Euler class to test nonzero sections
An oriented real rank-r bundle has an Euler class e(E) in H^r(X;Z). If it has a nowhere-zero section, its Euler class vanishes. The contrapositive is powerful: e(E) nonzero means every continuous section must vanish somewhere. A zero Euler class is not a universal guarantee that a section exists on an arbitrary high-dimensional base; higher obstructions may matter.
For a closed oriented smooth surface Σ_g of genus g, the Euler number of its tangent bundle is its Euler characteristic 2−2g. The sphere gives 2, the torus gives 0, and a genus-three surface gives −4. The sphere and genus-three surface therefore cannot carry a continuous tangent vector field that is nonzero everywhere.
The torus does carry such a field. View it as a product of circles and choose the tangent direction along the first factor. That explicit construction supplies the existence proof. It is better reasoning than simply pointing to the zero Euler number without naming why the converse is valid here.
Count signed zeros, not just locations
For a smooth tangent field with isolated zeros on a closed oriented surface, the sum of local indices equals the Euler characteristic. Suppose a proposed field on a genus-three surface has three isolated zeros with indices +1, −2 and −3. Their sum is −4, so it passes this necessary index audit. Passing the audit does not itself construct a field with exactly those local behaviours.
A different proposal with four zeros all of index +1 has total +4 and fails the audit. More numerical resolution cannot repair the topological mismatch. One must change the proposed zeros, discover omitted zeros, or identify a false assumption such as a boundary that was ignored.
The existing Thom-space and Euler-class article supplies the construction behind this test. Here the emphasis is choosing the correct bundle and interpreting the result with the appropriate strength.
CHAPTER 9 OF 14 · Test products and sections
9. Compare integral and mod-two information
For an oriented real rank-r bundle, reduction of the Euler class modulo two gives the top Stiefel–Whitney class w_r. This provides a consistency check between two coefficient systems. It also exposes an important limitation: an even nonzero integer class can reduce to zero.
Take an oriented plane bundle over S² with e(E)=6u. Its mod-two class w₂(E) is zero because 6 is even. Its first Stiefel–Whitney class is zero because the bundle is oriented. All positive-degree Stiefel–Whitney classes therefore vanish, but the nonzero Euler class still prevents a nowhere-zero section. The bundle is certainly not trivial.
Compare an oriented plane bundle with e(F)=5u. Its w₂ is the nonzero mod-two generator. In this case both tests detect an obstruction. Reducing 5 and 6 modulo two distinguishes parity; it does not compare their sizes, and it does not say that the six-twist example is less obstructed than the five-twist example.
Underlying real bundles
A complex bundle has a canonical underlying real orientation. Its odd Stiefel–Whitney classes vanish, and its even Stiefel–Whitney classes are reductions of the Chern classes. For the rank-two complex bundle E in Chapter 7, the underlying real rank is four. We obtain w₂(E_R)=a+b modulo two and w₄(E_R)=ab modulo two.
The coefficients follow from 1 and 7 both being odd, and 5 being odd. The subscripts shift: c₁ becomes w₂, while c₂ becomes w₄. Writing w₁=c₁ modulo two would be a degree error as well as an orientation error.
One further caution concerns torsion. Passing from integral to real cohomology kills torsion classes. A differential-form computation can therefore miss integral information even if performed exactly. Changing coefficients is a mathematical operation with information loss, not simply a different notation for the same answer.
CHAPTER 10 OF 14 · Interpret and repair calculations
10. Calculate a Pontryagin class and a characteristic number
Pontryagin classes belong to real bundles and have degrees divisible by four: p_i(E) lies in H^(4i)(X;Z). They can be defined using Chern classes of the complexification, with p_i(E)=(−1)^i c_(2i)(E⊗C). For the real bundle underlying a complex bundle V, a useful first-class identity is p₁(V_R)=c₁(V)²−2c₂(V).
Apply this to Chapter 7’s E over S²×S². We already found c₁(E)=a+7b and c₂(E)=5ab. Squaring the first class gives a²+14ab+49b²=14ab. Subtracting twice the second gives p₁(E_R)=14ab−10ab=4ab. Evaluating on the product fundamental cycle produces the number 4.
Explain what this number does not mean
The number 4 concerns an arbitrary real rank-four bundle on the product of spheres. It is not automatically a Pontryagin number of the manifold’s tangent bundle. One cannot substitute it into a theorem about the tangent bundle and conclude that the manifold has signature 4/3. That false conclusion is caused by changing the object midway through the argument.
For an oriented real plane bundle, the identity p₁=e² is especially convenient. Over S², e=6u has nonzero Euler class but e²=0 because H⁴ is zero. Its first Pontryagin class therefore cannot detect the obstruction already detected by the Euler class. Over a four-dimensional base, the same square may survive.
Characteristic numbers require classes whose total degree matches the manifold dimension and a fundamental class with compatible coefficients. In real dimension eight, p₂ and p₁² can be evaluated; p₁ alone has degree four and is not a top-dimensional number without additional data. Degree bookkeeping is a fast way to reject a formula that otherwise looks plausible.
CHAPTER 11 OF 14 · Interpret and repair calculations
11. Know what stability and curvature forget
Adding a trivial line bundle changes the rank but leaves the positive Stiefel–Whitney and Chern classes unchanged in their respective real and complex settings. This stability is useful when comparing bundles after extra product directions are allowed. It also explains why those classes need not detect every obstruction to a frame at the original rank.
Consider TS². It is not trivial because its Euler number is 2. Nevertheless TS²⊕R is trivial as a real rank-three bundle. At each point x of the unit sphere in R³, a tangent vector v and a normal coefficient t map to v+tx. This gives a fibrewise linear isomorphism to R³ varying continuously with x. The normal direction completes the tangent plane to the ambient space.
The Euler class is sensitive to rank and orientation. After adding a trivial line, the new bundle has an obvious nowhere-zero section in that line, so its Euler class vanishes. It would be a mistake to demand that this new Euler class remain equal to the old degree-two Euler class: they live in different degrees and belong to different rank bundles.
Curvature is another representation, with limits
On smooth bundles, Chern–Weil theory represents real characteristic classes using curvature of a connection. Changing the connection changes the differential forms but not their cohomology classes. This creates a bridge from local geometric calculations to global topological restrictions. It does not imply that all connections on one bundle have the same curvature.
For a computation involving curvature, preserve the connection convention, factors of 2π and i, orientation and coefficient comparison. A sign copied from a source using different conventions can reverse a first Chern number. A real differential form also cannot recover torsion killed by the map from integral to real cohomology. The topological invariant, its chosen differential-form representative and the physical field model are three distinct objects.
The distinction between stable and actual triviality also changes the question a reader should ask. If extra unused directions are allowed, an ambient construction may solve the problem. If the task demands a basis inside the original tangent planes, those extra directions are unavailable. In the sphere example, the radial vector is deliberately outside the tangent plane. It cannot be smuggled back into a proposed pair of tangent basis vectors. Draw the tangent plane and radial line at one point before attempting the global argument; the local diagram makes the difference in admissible vectors visible. Then check continuity as the point moves, because a collection of unrelated fibrewise bases is not a global frame. The requirement is a continuous choice, not merely existence of a basis separately in every vector space.
CHAPTER 12 OF 14 · Interpret and repair calculations
12. Diagnose a calculation before trusting it
Most elementary failures can be repaired by tracing a result back to its input. Begin with the bundle name: tangent, normal, tautological, pullback or direct sum. Next record its rank and structure. Then write the cohomology ring, including coefficients and relations. Only after these steps should a total class be multiplied or evaluated.
If a result has the wrong degree, repair the indexing. Chern index i means degree 2i; Pontryagin index i means degree 4i; Stiefel–Whitney index i means degree i. If a calculation gives a nonzero c₃ for a complex rank-two bundle, the rank bound has been violated. If it produces a nonzero degree-six class on S²×S², the base’s cohomology has been ignored.
Three short repair cases
A learner writes c₂(A⊕B)=c₁(A)+c₁(B). Repair: the sum is c₁ of the direct sum; its c₂ for two line summands is the product c₁(A)c₁(B). Both the algebra and the degrees identify the mistake.
A learner finds w₂(E)=0 and declares the oriented plane bundle over S² trivial. Repair: calculate its integral Euler class. Chapter 9’s e=6u example passes the mod-two test but still cannot have a nonzero section. The proposed implication is false even in a very familiar base space.
A learner obtains equal characteristic classes and declares two arbitrary bundles isomorphic. Repair: either produce a relevant classification theorem with its hypotheses, construct the isomorphism, or weaken the conclusion to “these invariants do not distinguish the bundles.” This repair preserves the useful computation without exaggerating its power.
For computer-assisted work, encode the ring relations explicitly and test the arithmetic independently on a small example. Symbolic expansion alone cannot decide which cohomology generators exist. The geometry supplies the ring; software carries out calculations inside it. Reversing that order turns a formal polynomial into an unsupported geometric claim.
CHAPTER 13 OF 14 · Transfer and continue
13. Transfer the method to a changed example
Work on X=S²×S² again, but change both line bundles. Let c₁(A)=3a−2b and c₁(B)=−2a+5b. Use a²=b²=0 and evaluation of ab equal to one. Predict the ranks and degrees before calculating: A⊕B has complex rank two, c₁ has degree two, c₂ and p₁ of the underlying real bundle have degree four.
First, c₁(A⊕B)=a+3b. Second, c₂=(3a−2b)(−2a+5b)=−6a²+15ab+4ba−10b²=19ab. Third, c₁²=6ab. Thus p₁=6ab−38ab=−32ab. The characteristic numbers from c₂ and p₁ are respectively 19 and −32 with the stated product orientation.
The underlying real bundle has w₂=a+b and w₄=ab modulo two. Compare this with Chapter 7: the same two mod-two outputs occur, but the integral second Chern numbers are 19 and 5. This is an explicit demonstration of information lost by reduction modulo two. It is not a contradiction between theories.
Change the map as well
Let f:X→X act on the first sphere by a degree-two map and on the second by a degree-minus-one map. Then f*a=2a and f*b=−b, so f*(ab)=−2ab. The pulled-back first Chern class is 2a−3b, the second class is −38ab, and p₁ is 64ab.
Check independently from the pulled-back first and second classes: (2a−3b)²=−12ab, and subtracting 2(−38ab) gives 64ab. Both routes agree. If your direct calculation gives −64, inspect the sign from the degree-minus-one map; if it gives 32, inspect the missing degree-two factor.
Finally reverse the orientation used to evaluate the domain manifold. The cohomology classes remain the same, but all their top-dimensional evaluations change sign. Separating a pullback operation from a change of evaluation orientation is the transfer skill this exercise is designed to test.
CHAPTER 14 OF 14 · Transfer and continue
14. Continue with useful questions and an evidence record
Do characteristic classes classify every bundle?
No. They are invariants, and in general distinct bundles can share them. Complex line bundles over the standard spaces considered here are an important exception: their first Chern class gives a topological classification. State the theorem’s scope each time you use completeness. Do not infer a general rule from a successful low-rank example.
Does a zero Euler class guarantee a nonzero section?
Not without conditions on the rank, base and obstruction problem. A nonzero Euler class is a definite obstruction. A zero result removes that obstruction, after which higher obstructions may remain. In familiar low-dimensional cases, additional theorems or explicit constructions can settle existence. The torus construction is a good model of supplying the missing evidence.
Is a characteristic class the same as a characteristic number?
No. A class lives in cohomology. A number results from evaluating an appropriate top-degree class or product on a fundamental cycle. Different evaluation cycles or orientations can change the number. A calculation should preserve the class before compressing it to an evaluation, especially when several independent cycles exist.
Where does this become useful beyond an exercise?
In geometry, characteristic classes obstruct global frames and sections. In topology, they distinguish bundles and contribute to manifold invariants. In mathematical physics, bundles describe global organisation of local field data, but an application needs its own model, regularity and normalisation assumptions. A topological calculation does not by itself validate an experimental interpretation or establish that a particular material realises the model.
A practical learning sequence is to draw line-bundle transitions over a circle, compute tensor powers over a sphere, multiply total classes over a product of spheres, and finally compare a tangent bundle with an arbitrary bundle over the same base. Keep a one-page record containing the object, ring, class formulas, original calculation, obstruction conclusion and remaining unknown. This record makes it possible to check the work after the diagram or coordinate system changes.
For further study, use Allen Hatcher’s Vector Bundles and K-Theory, Chapter 3 for the classical families and obstruction perspective. MIT’s Algebraic Topology II notes, Chapter 5 develop characteristic classes through universal constructions and explain their connection with cobordism. Compare sign conventions explicitly when moving between sources.
Within eduKate, continue to Cobordism Theory to see how suitably chosen characteristic numbers participate in boundary tests. Return to the Euler-class specialist guide when the unresolved question is a section, or the classifying-spaces specialist guide when the unresolved question is classification. Choose the next method according to what the current invariant has left undecided.