@DrChinese
I have started with the Ma paper, and in anticipation of a discussion about coexisting photons, I have modified it so that photons 1 and 4 never coexist, where "never coexist" means in the rest frame of the labs that measure photons 1 and 4, there is no time interval that contains both 1 and 4 a la
Megidish. I also use the time labels of Fig. 1 in Megidish: I, II, III, IV, V. I have taken the liberty of some simplifications that make the maths easier but (hopefully you agree) do not detract from any relevant aspect. The simplifications are:
i) I use a BSM that has a 50/50 chance of occuring, but if it occurs, is perfect, i.e. is a measurement in the basis ##\{\psi^+,\psi^-,\phi^+,\phi^-\}## as opposed to the more accurate ##\{\phi^+,\phi^-,\mathrm{fail}\}##.
ii) I also assume the measurements that occur at I and V by devices A and B are pre-aligned to the aspect ##\omega##, such that a successful BSM will select perfect correlation/anti-correlation between 1 and 4.
Attempting to model the unitary evolution of the experiment: In a conventional swapping experiment described by Ma, the initial state of the 4-photon system is $$|\Psi\rangle = |\psi^-\rangle_{12}|\psi^-\rangle_{34}$$In the Megidish experiment, it is $$|\Psi\rangle = |\psi^-\rangle_{12}|\emptyset\rangle$$where ##|\emptyset\rangle## are the relevant degrees of freedom so that ##U(t_{III},t_I)|\emptyset\rangle = |\psi^-\rangle_{34}##. The initial state including all relevant measurement systems is $$|\psi^-\rangle_{12}|\emptyset\rangle|\Omega\rangle_A|\Omega\rangle_B|\Omega\rangle_\mathrm{BSM}$$The destructive measurement of photon 1 by ##A## is modeled like so:
\begin{eqnarray*}U(t_{II},t_I)|h_\omega\rangle_1|\Omega\rangle_A&=&|\omega,1\rangle_{1,A}\\
U(t_{II},t_I)|v_\omega\rangle_1|\Omega\rangle_A&=&|\omega,0\rangle_{1,A}\\
U(t_{II},t_I)|\psi^-\rangle_{12}|\Omega_A\rangle &=& \frac{1}{\sqrt{2}}(|\omega,1\rangle_{1,A}|v_\omega\rangle_2 - |\omega,0\rangle_{1,A}|h_\omega\rangle_2) = |\chi^{\psi^-}\rangle_{12,A}\end{eqnarray*}and so the full evolution to II is$$U(t_{II},t_I)|\psi^-\rangle_{12}|\emptyset\rangle|\Omega\rangle_A|\Omega\rangle_B|\Omega\rangle_\mathrm{BSM} = |\emptyset\rangle|\Omega\rangle_B|\Omega\rangle_\mathrm{BSM}|\chi^{\psi^-}\rangle_{12,A}$$Next, we evolve to III, the creation of the 34 pair$$U(t_{III},t_{II})|\emptyset\rangle|\Omega\rangle_B|\Omega\rangle_\mathrm{BSM}|\chi^{\psi^-}\rangle_{12,A} = |\psi^-\rangle_{34}|\Omega\rangle_B|\Omega\rangle_\mathrm{BSM}|\chi^{\psi^-}\rangle_{12,A}$$In evolving to IV, let's say there is a 50/50 chance the BSM occurs. A BSM is modelled like so\begin{eqnarray*}U(t_{IV},t_{III})|\Omega\rangle_{BSM}|\chi^{\psi^-}\rangle_{12,A}|\psi^-\rangle_{34}&=&\frac{1}{2}(|\chi^{\psi^+}\rangle_{14,A}|\Psi^+\rangle_\mathrm{23,BSM}\\&&-|\chi^{\psi^-}\rangle_{14,A}|\Psi^-\rangle_\mathrm{23,BSM}\\&&-|\chi^{\phi^+}\rangle_{14,A}|\Phi^+\rangle_\mathrm{23,BSM}\\&&+|\chi^{\phi^-}\rangle_{14,A}|\Phi^-\rangle_\mathrm{23,BSM})\\&=&|\chi^+\rangle_{1234,A,\mathrm{BSM}}\end{eqnarray*}where the ##\chi## terms are of the form, for example $$|\chi^{\phi^+}\rangle_{14,A} = \frac{1}{\sqrt{2}}(|\omega, 1\rangle_{1,A}|h_\omega\rangle_4 + |\omega, 0\rangle_{1,A}|v_\omega\rangle_4)$$If a BSM doesn't occur, the evolution is \begin{eqnarray*}U(t_{IV},t_{III})|\Omega\rangle_{BSM}|\chi^{\psi^-}\rangle_{12,A}|\psi^-\rangle_{34}&=&\frac{1}{2}|\Omega\rangle_\mathrm{BSM}(|\chi^{\psi^+}\rangle_{14,A}|\psi^+\rangle_\mathrm{23}\\&&-|\chi^{\psi^-}\rangle_{14,A}|\psi^-\rangle_\mathrm{23}\\&&-|\chi^{\phi^+}\rangle_{14,A}|\phi^+\rangle_\mathrm{23}\\&&+|\chi^{\phi^-}\rangle_{14,A}|\phi^-\rangle_\mathrm{23})\\&=&|\chi^-\rangle_{1234,A,\mathrm{BSM}}\end{eqnarray*}So the evolution to IV where the BSM may or may not occur is $$U(t_{IV},t_{III})|\Omega\rangle_B|\Omega\rangle_{BSM}|\chi^{\psi^-}\rangle_{12,A}|\psi^-\rangle_{34} = \frac{1}{\sqrt{2}}|\Omega\rangle_B(|\chi^+\rangle_{1234,A,\mathrm{BSM}}+|\chi^-\rangle_{1234,A,\mathrm{BSM}})$$Finally, to model the measurement at V, we use the evolution that acts on states like$$U(t_V,t_{IV})|\Omega\rangle_B|\chi^{\phi^+}\rangle_{14,A} = \frac{1}{\sqrt{2}}(|\omega,0\rangle_{1,A}|\omega,0\rangle_{2,B}+ |\omega,1\rangle_{1,A}|\omega,1\rangle_{2,B}) =|\zeta^{\phi^+}\rangle_{14,AB}$$Like before, if a BSM previously occurred, we have\begin{eqnarray*}U(t_V,t_{IV})|\Omega\rangle_B|\xi^+\rangle_{1234,A,\mathrm{BSM}}&=&\frac{1}{2}(|\zeta^{\psi^+}\rangle_{14,A}|\Psi^+\rangle_\mathrm{23,BSM}\\&&-|\zeta^{\psi^-}\rangle_{14,A}|\Psi^-\rangle_\mathrm{23,BSM}\\&&-|\zeta^{\phi^+}\rangle_{14,A}|\Phi^+\rangle_\mathrm{23,BSM}\\&&+|\zeta^{\phi^-}\rangle_{14,A}|\Phi^-\rangle_\mathrm{23,BSM})\\&=&|\zeta^+\rangle_{1234,AB,\mathrm{BSM}}\end{eqnarray*}If no BSM occurred, we have
\begin{eqnarray*}U(t_V,t_{IV})|\Omega\rangle_B|\xi^-\rangle_{1234,A,\mathrm{BSM}}&=&\frac{1}{2}|\Omega\rangle_\mathrm{BSM}(|\zeta^{\psi^+}\rangle_{14,A}|\psi^+\rangle_\mathrm{23}\\&&-|\zeta^{\psi^-}\rangle_{14,A}|\psi^-\rangle_\mathrm{23}\\&&-|\zeta^{\phi^+}\rangle_{14,A}|\phi^+\rangle_\mathrm{23}\\&&+|\zeta^{\phi^-}\rangle_{14,A}|\phi^-\rangle_\mathrm{23})\\&=&|\zeta^-\rangle_{1234,AB,\mathrm{BSM}}\end{eqnarray*}Putting all of this together, we have the full evolution$$U(t_V,t_I)|\psi^-\rangle_{12}|\emptyset\rangle|\Omega\rangle_A|\Omega\rangle_B|\Omega\rangle_\mathrm{BSM} = \frac{1}{\sqrt{2}}(|\zeta^+\rangle_{1234,AB,\mathrm{BSM}} + |\zeta^-\rangle_{1234,AB,\mathrm{BSM}})$$Before proceeding, maybe it is good to see if you object to this model of the time-evolution of a (simplified) entanglement swapping experiment where photons 1 and 4 never coexist, or want clarification at any stage outlined above.