Spectral sequences without memorizing the machinery first.
Start with a complex hidden inside nested layers. Peel it apart. Compute what each layer says. Then watch the missing cross-layer information show up as diagonal corrections. The notation becomes a label for a motion you have already seen.
A spectral sequence is a controlled way of asking: “What did I miss when I looked at the object one layer at a time?”
In a genuinely graded complex, each layer stays in its lane. The boundary map moves horizontally inside a fixed grade, so we compute homology grade by grade and add the answers.
But Chow’s central move begins when this ideal situation fails: the object is only filtered. Levels are nested, not separate.
We slice the nested filtration into visible quotients. The quotient removes everything already visible downstairs, leaving the new information first appearing at level p.
After peeling, the induced boundary ∂₀ sees only motion confined to a single sheet. We compute homology inside each sheet.
The trap is to think this is the answer. It is only the answer for the peeled model. The real filtered complex may contain boundaries that drop from one layer to another.
Use the page slider as a microscope. On page r, the differential detects information that drops r filtration levels.
The same spectral-sequence event appears in three synchronized representations: a concrete layered object, a page grid, and the formula. Move the cursor over the grid to inspect a cell.
Actual chains live in a nested building. Boundaries may fall from upper floors to lower floors.
The page grid shows surviving classes and diagonal differentials.
The formula names the motion you already saw: dᵣ : Eʳd+1,p+r → Eʳd,p.
If many groups or differentials vanish early, the later pages look the same. Then the spectral sequence has done its work: the surviving data is stable.
This is why spectral sequences are useful. You often do not need to compute forever. You need to know where nothing further can happen.
Do not ask first: “What is the official definition?”
Ask first: “What did my layer-by-layer approximation miss, and how far down can the boundary map leak?”