Understanding the virtual states referenced in multiphoton absorption studies The Heisenberg energy-time uncertainty tells us that we can have so-called virtual states between eigenstates as long as the lifetime of these states is at most:
$\tau = (\frac{h}{4 \pi E_v})$
Where $h$ is the Planck constant ($6.62607 * 10^{-34} J*s$, or $4.135668 * 10^{-15} eV*s$), and $E_v$ is energy difference between our virtual state and the nearest eigenstate.  If we assume $E_v \approx 1 eV$ for a virtual state, we can estimate that this state can only exist for at most $\tau_2 \approx 3.29106 * 10^{-16}$ seconds.
Now, in the literature and wikipedia, I've repeatedly seen the claim that two-photon absorption (https://en.wikipedia.org/wiki/Two-photon_absorption) involves a single photon event promoting a molecule to a virtual state, and because the energy-time uncertainty allows this state to exist for some time $\tau$ (as estimated above for a $\approx 1 eV$ transition), if another photon strikes within this time window the energy of the two photons can be summed to allow for a transition with double the energy of a single photon.  For example, two $\lambda \approx 2.48 \mu m$ photons with energies of $\approx 0.5 eV$ each can, through a virtual state, sum to allow for a $\approx 1 eV$ transition correspond to the absorbance of a single $\lambda \approx 1.24 \mu m$ photon.
However, I'm having trouble understanding where the "memory" for the first photon absorbance is stored, since a virtual state is not a real eigenstate with distinct observables.  What mechanism allows for the energies of the two photons to be summed?  And are these energies being summed, or is there some energy loss associated with transitioning through a virtual state?  Please note that I fully understand there may not be an intuitive "memory mechanism", I just mean "memory" in the sense of being able to write down something like a descriptive state vector for a virtual state that accounts for the state's energy.
Finally, what does it mean to say that a virtual state can exist for "at most a time $\tau$" in the context of the energy-time uncertainty principle?  Is this just the expected lifetime for a state decaying exponentially with decay parameter $\lambda = \frac{1}{\tau}$?

A good soundbite for this question would be: how can we write down a quantum mechanical expression for a virtual state (say, in the context of raman spectroscopy) in Hilbert space?  And if it's ultimately correct to describe the virtual state using an interaction Hamiltonian between some particle's dipole and the EM field of an incoming photon (of some wavelength), how do we write down a description of this Hamiltonian in a rigorous manner?  Does this Hamiltonian give us any insight into the lifetime of the virtual state?
Also, why do I often see references to a Rigged Hilbert space when doing literature searches on this topic?  Take for example these two papers by Gadella, M.:
Gadella, M. A description of virtual scattering states in the rigged Hilbert space formulation of quantum mechanics. J. Math. Phys. 24, 2142 (1983).
https://doi.org/10.1063/1.525966
Gadella, M. Construction of rigged Hilbert spaces to describe resonances and virtual states. Physica A: Statistical Mechanics and its Applications 124(1–3), March 1984, 317 (1984).
https://doi.org/10.1016/0378-4371(84)90249-8

From (https://arxiv.org/abs/1209.5620): Leon-Montiel, R. J. et. al. Role of the spectral shape of quantum correlations in two-photon virtual-state spectroscopy. (2013).
"The quantum mechanical calculation of the TPA [Two-Photon Absorption] transition probability shows that its value can be understood as a weighted sum of many energy non-conserving atomic transitions (virtual-state transitions) [22, 23] between energy levels."
 A: I know that this question is old like more than full seven years, but I would like to answer this question.
First, absorption occurs only when photons interact with matter or something. It does not have to do with memory as you said, and it has to do with the matter that the photon is interacting with, for example an electron bound to the matter. The wavefunction of a bounded electron (bounded by the attractive force of a nucleus or nuclei of an atom or a molecule) is deformed due to the interaction with the photon. The electron is interacting with the binding potential of the matter. If the intensity (i.e., flux) of the photon is not enough, then the wavefunction of the electron cannot escape enough from the center of the binding potential. On the other hand, if the intensity of the photon is enough, then
the electron can oscillating with the frequency component not allowed by the single photon frequency because of the binding potential. Mathematically, Fourier trasformation of the dipole acceleration of the matter has frequency component corresponding to the two photon. Therefore, the two photon absorption all has to do with the deformation of electron wavefunction in matter.
Regarding the Heisenberg uncertainty principle,

The Heisenberg energy-time uncertainty tells us that we can have so-called virtual states between eigenstates as long as the lifetime of these states is at most:
\tau=(\frac{h}{4πE_v})
Where h is the Planck constant (6.62607∗10−34J∗s, or 4.135668∗10−15eV∗s), and Ev is energy difference between our virtual state and the nearest eigenstate. If we assume Ev≈1eV for a virtual state, we can estimate that this state can only exist for at most τ2≈3.29106∗10−16 seconds.

The Heisenberg uncertainty principle is overused or misused here and there maybe due to its popularity and concise but powerful meaning in physics. It does not have to do with Heisenberg uncertainty principle.
