Double Category of Bimodules

Let 𝒱 , ⊗ , I 𝒱 tensor-product 𝐼 \mathcal{V},\otimes,I be a Monoidal Category with Coequalisers. The Double Category Bimod ⁢ ( 𝒱 ) Bimod 𝒱 \mathrm{Bimod}(\mathcal{V}) consists of

{A} && {B}

{C} && {D} \end{tikzcd} \] are given by Semilinear Maps; EG, morphisms α : M → N : 𝛼 → 𝑀 𝑁 \alpha:M\to N that satisfy

A⊗Mtensor-product𝐴𝑀{{A\otimes M}}M𝑀{{M}}C⊗Ntensor-product𝐶𝑁{{C\otimes N}}N𝑁{{N}}lMsubscript𝑙𝑀\scriptstyle{l_{M}}f⊗αtensor-product𝑓𝛼\scriptstyle{f\otimes\alpha}α𝛼\scriptstyle{\alpha}lNsubscript𝑙𝑁\scriptstyle{l_{N}}  M⊗Btensor-product𝑀𝐵{{M\otimes B}}M𝑀{{M}}N⊗Dtensor-product𝑁𝐷{{N\otimes D}}N𝑁{{N}}rMsubscript𝑟𝑀\scriptstyle{r_{M}}α⊗gtensor-product𝛼𝑔\scriptstyle{\alpha\otimes g}α𝛼\scriptstyle{\alpha}rNsubscript𝑟𝑁\scriptstyle{r_{N}}

A && B && C

X && Y && Z \end{tikzcd} \]

arise from the universal property of Coequalisers. First, we must give a map M ⊗ B P → N ⊗ Y Q → subscript tensor-product 𝐵 𝑀 𝑃 subscript tensor-product 𝑌 𝑁 𝑄 M\otimes_{B}P\to N\otimes_{Y}Q ; by the universal property of coequalisers, it suffices to give a map γ : M ⊗ P → N ⊗ Y Q : 𝛾 → tensor-product 𝑀 𝑃 subscript tensor-product 𝑌 𝑁 𝑄 \gamma:M\otimes P\to N\otimes_{Y}Q with γ ∘ r M ⊗ m a t h r m i d i f x . . ⋄ l s e i = γ ∘ m a t h r m i d i f x . . ⋄ l s e i ⊗ l N \gamma\circ r_{M}\otimes mathrm{id}{ifx..lse\par i}=\gamma\circ mathrm{id}{% ifx..lse\par i}\otimes l_{N} . Note that ι ∘ ( α ⊗ β ) 𝜄 tensor-product 𝛼 𝛽 \iota\circ(\alpha\otimes\beta) fits the right type, and we have

ι∘α⊗β∘rM⊗mathrmidifx..⋄lsei\displaystyle\iota\circ\alpha\otimes\beta\circ r_{M}\otimes mathrm{id}{ifx..% lse\par i} =ι∘(α∘rM)⊗βabsenttensor-product𝜄𝛼subscript𝑟𝑀𝛽\displaystyle=\iota\circ(\alpha\circ r_{M})\otimes\beta
=ι∘(rN∘α⊗g)⊗βabsenttensor-product𝜄tensor-productsubscript𝑟𝑁𝛼𝑔𝛽\displaystyle=\iota\circ(r_{N}\circ\alpha\otimes g)\otimes\beta
=ι∘rN⊗mathrmidifx..⋄lsei∘(α⊗g)⊗β\displaystyle=\iota\circ r_{N}\otimes mathrm{id}{ifx..lse\par i}\circ(\alpha% \otimes g)\otimes\beta
=ι∘mathrmidifx..⋄lsei⊗lQ∘(α⊗g)⊗β\displaystyle=\iota\circ mathrm{id}{ifx..lse\par i}\otimes l_{Q}\circ(\alpha% \otimes g)\otimes\beta
=ι∘mathrmidifx..⋄lsei⊗lQ∘α⊗(g⊗β)\displaystyle=\iota\circ mathrm{id}{ifx..lse\par i}\otimes l_{Q}\circ\alpha% \otimes(g\otimes\beta)
=ι∘α⊗(lQ∘g⊗β)absenttensor-product𝜄𝛼tensor-productsubscript𝑙𝑄𝑔𝛽\displaystyle=\iota\circ\alpha\otimes(l_{Q}\circ g\otimes\beta)
=ι∘α⊗(β∘lP)absenttensor-product𝜄𝛼𝛽subscript𝑙𝑃\displaystyle=\iota\circ\alpha\otimes(\beta\circ l_{P})
=ι∘α⊗β∘mathrmidifx..⋄lsei⊗lP\displaystyle=\iota\circ\alpha\otimes\beta\circ mathrm{id}{ifx..lse\par i}% \otimes l_{P}

Moreover, the framing constraints on the square still hold up, though this is a bit tedious to check.

When 𝒱 𝒱 \mathcal{V} lacks Coequalisers, we can still form a Virtual Double Category of Bimodules, though the maps are very tricky to define.

References