Proposition 4.2.5. Let \(F\colon C \to D\) be a functor between stable \(\infty \)-categories. Then the following are equivalent:

(1)

The functor \(F\) is exact;

(2)

The functor \(F\) is right exact, i.e.Β preserves initial objects and pushouts;

(3)

The functor \(F\) is left exact, i.e.Β preserves terminal objects and pullbacks.

Proof of claim: To construct an inverse, consider a diagram \(z \leftarrow x \rightarrow y\) in \(C\), and consider the following pullback diagram (ignoring the dashed morphisms):

Commutative diagram involving Omega x, Omega z, 0, Omega y, a, b, c, x, y, z, and further objects.

Note that each of the pullbacks that appear here may be obtained as a fiber, hence is assumed to exist in \(C\). More explicitly, starting from the span \(z \leftarrow x \to y\), set \[ b := \fib (x \to y), \qquad c := \fib (x \to z). \] Then define \(a\) as the fiber of the composite \(b \to x \to z\). The nullhomotopy of \(a \to b \to x \to z\) induces a map \(a \to c\), and the square

Commutative square with maps a to b, c to x, a to c, b to x.

is a pullback by the pasting law. The other displayed pullback squares are obtained in the same way from the fiber sequences \(b \to x \to y\), \(c \to x \to z\), and \(a \to b \to z\). The construction is functorial in the input span because each step is obtained by applying the fiber construction to a functorially specified morphism, together with the universal maps in its defining pullback square. Thus a morphism of spans induces compatible morphisms of all the objects in the displayed diagram. We now claim that an inverse to the restriction functor is given by

Commutative triangle with vertices x, y, z.
Commutative square with maps Omega x to Omega y, Omega z to a, Omega x to Omega z, Omega y to a.

Note that this functor does indeed land in \(P\) since the equivalence \(\Omega ^{-1}\colon C \iso C\) preserves pullback squares. Furthermore, it provides an extension of the original diagram, hence defines a right inverse. Finally, if we start with any pullback square, we may rewrite it in the form

Commutative square with maps x to y, z to Omega -1 a, x to z, y to Omega -1 a.

and we see that the above procedure recovers the original pullback square, showing that it is also a left inverse. This finishes the proof of the claim. β–‘