For a purified input with reference and channel output , . The reference entropy is fixed through subsequent channels. Thus data-processing inequality for coherent information immediately implies data processing for quantum mutual information between that reference and successive outputs.
Dilate the second quantum channel by an isometry , and set . Its marginal is the final output. Isometry invariance of Von Neumann entropy gives and . The difference between the two values of coherent information is therefore
Nonnegativity follows from Strong subadditivity of Von Neumann entropy. This proves the data-processing inequality for coherent information:
The dilated state here need not be pure. The proof uses only isometry invariance and strong subadditivity, so it applies even when the first channel has already entangled the input with its own discarded environment.
Both output states have the same reference marginal, and . Expand quantum mutual information to obtain the mutual information and coherent information identity
The input-entropy term is identical in the two expressions. Subtract them and use the preceding data-processing inequality for coherent information:
Thus the final quantum channel cannot increase the reference-output quantum mutual information, as required by data processing for quantum mutual information.
Use two results: the isometric Stinespring dilation of a quantum channel, and Strong subadditivity of Von Neumann entropy. Let dilate the given operation, and define
An linear isometry of Hilbert spaces preserves the nonzero eigenvalues, so and . Subtracting the two coherent information expressions gives
The last inequality is Strong subadditivity of Von Neumann entropy, in the form . Thus the data-processing inequality for coherent information is
The lost coherent information is precisely the quantum conditional mutual information between the reference and discarded environment , conditional on the retained output .