Putting PASPT2 on a Firmer Basis
summary
The gist
The paper introduces and analyzes the partial-active-space second-order perturbation theory (PASPT2) method…
In short
The paper analyzes Partial-Active-Space Second-Order Perturbation Theory (PASPT2) variants to improve its reliability for strongly correlated systems. While some variants are strictly size-extensive but suffer from intruder states, the original PASPT2H is near-extensive for target states supported by M0, making it a robust tool for machine learning and solid-state calculations.
Key concepts
- PASPT2
- A method used in quantum chemistry to calculate energy corrections by considering only a subset of orbitals (the active space). It uses second-order perturbation theory to account for electron correlation effects, which are crucial for accurately describing strongly interacting electrons.
- Size-Extensivity
- A property where the calculated energy scales correctly when the system size is doubled. If you double the number of electrons in a system, the calculated energy should scale predictably. This property is important for ensuring that calculations are physically meaningful for large systems like solids.
- Intruder States
- These are states that appear in a calculation but do not correspond to physical states of the actual system. They cause problems because they can introduce unphysical results into the energy calculations, making the method less trustworthy.
Terminology used across episodes
This episode discusses
- Putting PASPT2 on a Firmer Basis · Paper Radio
- PASPT2: a size-extensive and size-consistent partial-active-space multi-state multi-reference second-order perturbation theory for strongly correlated electrons
The paper
Putting PASPT2 on a Firmer Basis · Read on arXiv
Chunzhang Liu, Ning Zhang, Wenjian Liu
Qingdao Institute for Theoretical and Computational Sciences · Center for Optics Research and Engineering, Shandong University
The recently proposed partial-active-space (PAS) multi-state second-order perturbation theory (PASPT2) [Precis. Chem. 4, 997 (2026)] features connected amplitudes and a connected, closed intermediate Hamiltonian. Despite these hallmarks, PASPT2 (denoted as PASPT2H from now on) is not strictly size-extensive, as originally thought (and numerically confirmed), albeit strictly size-consistent. Nevertheless, PASPT2 is near-extensive for the states with major projections on the chosen PAS M 0. This becomes more transparent upon introducing PASPT2X, a strictly size-extensive variant. Compared with the intruder-prone PASPT2X, the intruder-free PASPT2H merely neglects the second-order corrections that are important only for those states with major projections on the orthogonal complement R X of M 0 within the closed space M X (= M 0 plus R X); however, such states are not supported by the chosen finite one-particle basis set. The weak violation of size-extensivity is therefore numerically insignificant for the target states supported by M 0, reinforcing the theoretical basis of PASPT2H.
Transcript
Introduction to the show: ident: Physics. The day in general physics: instruments, methods and results across the field.
Kai: We're looking at the paper: "Putting PASPT2 on a Firmer Basis".
Mira: Here's the gist of the paper. The paper introduces and analyzes the partial-active-space second-order perturbation theory (PASPT2) method, focusing on its properties regarding size-extensivity and intruder states to establish a more robust framework for strongly correlated systems.
Kai: The core finding is that the original variant of PASPT2, denoted as PASPT2H, is near-extensive for the target states supported by M0 and its weak violation of size-extensivity has hardly numerical consequence on those target states Near-extensive for the states with major projections on the chosen PAS M0.
Mira: Methodological Development The authors introduce several variants of PASPT2, including PASPT2M, which is defined by Eq.
Kai: (38), and PASPT2X, which is defined by Eq.
Title and authors: Kai: So we're looking at this paper today, "Putting PASPT2 on a Firmer Basis," which is about making the partial active space second order perturbation theory, or PASPT2, more reliable for strongly correlated systems.
Mira: Right. It’s trying to address some of the fundamental issues with the original version of PASPT2 by adding some new criteria and variants to it.
Lev: From an engineering standpoint, I’m interested in how this affects our ability to run these calculations on actual hardware, because those methods need to be stable for real physical systems.
Kai: Well, what the paper is saying is that they've developed several variants of PASPT2—like PASPT2M and PASPT2X—and they’re showing that while some are strictly size-extensive and size-consistent, they still have problems with intruder states.
Mira: And the main focus here seems to be on the original variant, which they call PASPT2H, because it’s apparently intruder free for all target states supported by a specific choice of the partial active space M0.
Lev: But if it's not strictly size-extensive, how do we know if our results will stay accurate when we try to model something larger than what was in the initial calculation?
Kai: That’s where they introduce a new concept called generalized size-extensivity, which is an operational definition to test the method numerically. They found that for PASPT2H, the size-extensivity errors for the lowest thirteen states are only between one and forty meV when compared against a specific model space M0.
Mira: That’s a relatively small error, which suggests that even though it's not strictly size-extensive in the strict sense, its impact on those target states is minimal.
Lev: So what this means for us running these calculations, it suggests we can use PASPT2H with some confidence when modeling strongly correlated solid states because the energy scaling isn't completely broken.
Kai: Exactly. The paper also introduces several improvements to the original idea, focusing on how to treat static and dynamic correlations in a more adaptive way, which they say is necessary because any fixed reference space is inherently unbalanced.
Mira: That points toward a method that can handle the entanglement between different types of correlation components better than what was there before.
Lev: And another thing they suggest is using hybrid model spaces, where you combine the M0 and the orthogonal complement RX within a larger closed space MX as a buffer to treat states supported by RX more soundly.
Kai: This hybrid approach helps manage states that might not be fully supported by the initial one-particle basis set M0, making them easier to handle in practice.
Mira: It’s also important for machine learning applications because the paper argues that because PASPT2 is state-universal, it becomes a very good model for training machine learning models to find those dressed active space Hamiltonians.
Lev: From an error control perspective, the authors emphasize that PASPT2H can be used to construct an effective Hamiltonian that is "manifestly connected," which helps ensure the energies are connected upon diagonalization, stabilizing calculations for states with major projections on M0.
Kai: So, to wrap up this paper on "Putting PASPT2 on a Firmer Basis," the main conclusion is that PASPT2H is near-extensive for target states strongly projected onto M0, and this weak size-extensivity violation doesn't cause big numerical problems for those states.
Mira: They justify the original claim that this variant is up to date because it’s intruder free for all target states supported by the chosen partial active space M0, which can be generated through methods like sCI.
Lev: And if we consider what this means practically, it suggests we have a method that works well for strongly correlated solid state calculations where size-extensivity is a big deal.
Kai: So, to summarize the core findings of "Putting PASPT2 on a Firmer Basis," they’ve shown that PASPT2H is near-extensive for states with major projections on M0, and this weak violation of size-extensivity has hardly any numerical consequence for those states.
Mira: The paper also lays out some really useful improvements, like the adaptive treatment of static and dynamic correlations, which guides future development because a fixed reference space is just too inflexible to handle everything.
Lev: And that hybrid model space utilization idea is interesting; using the RX space as a buffer means we can get a physically sound approximate treatment for those states with major projections on RX.
Kai: Ultimately, the state-universal nature of PASPT2 makes it an ideal model for training machine learning models to find dressed active space Hamiltonians across different molecular states efficiently.
Mira: It seems like they’ve given us a much more transparent framework for handling the complex entanglement of static and dynamic correlations in these strongly correlated systems.
Lev: So, when we think about putting this on real hardware, PASPT2H offers a stabilized path forward for calculations where size-extensivity isn't perfectly achieved but is at least near enough for our needs.
Kai: We’re wrapping up our discussion on "Putting PASPT2 on a Firmer Basis." It sounds like this work gives us a much more reliable tool for modeling these difficult quantum systems.
The paper's summary: Kai: So, we’ve seen how they tested PASPT2H numerically, and now we’re looking at what they actually conclude about this paper, "Putting PASPT2 on a Firmer Basis."
Mira: Basically, they say that while some other variants of PASPT2 are strictly size-extensive and size-consistent, they still have those pesky intruder state problems.
Kai: But the main focus is on this original version, PASPT2H. And what they find is that for the states that project strongly onto their chosen partial active space M0, it’s near-extensive.
Mira: That means even if it's not perfectly size-extensivity compliant in every single case, the error doesn't really matter numerically for the specific target states they care about. It’s just a small one.
Kai: Right. They use this operational definition of generalized size-extensivity to test it, and they found that for those lowest thirteen states, the error is only around one to forty meV when compared against their chosen model space M0.
Mira: That level of error is what makes them say that it’s practically fine for strongly correlated solid state calculations. It’s not a perfect solution, but it works well enough in practice for those specific states.
Kai: And they point out that this near-extensivity is actually a big plus because the original PASPT2H variant is intruder-free for all target states supported by the partial active space M0, which can be generated using methods like sCI.
Mira: That’s a huge piece of information. If you’re using something like sCI to pick your model space, and PASPT2H doesn't have intruder states in those spaces, that makes it much more reliable for whatever you're trying to calculate.
Kai: So, the paper concludes that the original claim about PASPT2 being the only size-consistent and intruder-free one is actually justified when you look at this variant. It’s a solid foundation.
Mira: And they also highlight how this method has these other advantages, like spin adaptation and being state-universal, which makes it a great starting point for training machine learning models to figure out those dressed active space Hamiltonians.
Kai: It seems like the big implication here is that we can use PASPT2H with more confidence in strongly correlated systems because its error scaling is manageable for the states we are interested in.
Mira: So, it’s not a perfect method yet, but it’s a much firmer basis for moving forward with calculations where size-extensivity isn't perfectly achieved but is at least near enough.
Kai: And this leads us to think about those improvements they suggest, like the adaptive treatment of correlation components and using hybrid model spaces as buffers for states in the orthogonal space.
Mira: Yeah, that hybrid idea is interesting because it lets you treat states supported by the orthogonal space RX more soundly when dealing with complex systems.
Kai: So we’ve seen what they built and where it stands now—it’s a near-extensivity win for PASPT2H, but the real work for hardware is still figuring out how to keep those calculations stable on real superconducting qubits.
The paper's improvements: Kai: So we’ve looked at how they tested PASPT2H numerically, and now we’re going to talk about what improvements this paper suggests for making the method even better.
Mira: Basically, they suggest ways to handle those static and dynamic correlations in a more adaptive way, because a fixed reference space just isn't flexible enough anymore.
Kai: That points toward a method that can adjust how it treats different kinds of correlation components depending on what’s happening in the system right now. It needs to be more dynamic than before.
Mira: And they propose using hybrid model spaces, where you mix the M0 and the orthogonal complement space RX within a larger closed space MX as a buffer to treat those states in RX more soundly.
Kai: That hybrid approach helps manage states that aren't fully supported by the initial one-particle basis set M0, which makes it easier to get physically sound approximate results for those states projecting onto RX.
Mira: And on the machine learning side, they emphasize that because PASPT2 is state-universal, it becomes an ideal model for training machine learning models to find those dressed active space Hamiltonians across different molecular states efficiently.
Kai: That’s a big idea; using the method as a template for AI training means you can capture the physics across many different molecular states without having to re-train everything from scratch.
Mira: It seems like they’re pushing for a more sophisticated framework that allows the theory to evolve with the complexity of strongly correlated systems rather than just being stuck in one fixed approximation.
Kai: So, the implication is that we might see a future where these methods aren't just static calculations but become part of an adaptive system capable of handling those complex entanglement issues better.
Mira: Right. It’s about moving from a set of fixed rules to a more flexible framework that can respond to how correlated the system actually gets.
Kai: That flexibility is exactly what we need when we’re trying to model things that are highly dynamic, like things in condensed matter or even high-energy physics where you have rapid changes.
Mira: And this opens up new avenues for understanding those materials, since they can adapt to the actual correlation structure instead of just being forced into a pre-defined box.
Kai: So we’ve seen the improvements—the adaptive treatment and hybrid spaces—and it looks like the next step is really about building systems that can respond to that adaptability in real time.
Mira: It’s about giving the theory more tools to handle the messiness of real-world correlation structures without just throwing up another complicated, rigid approximation.
Conclusion: Kai: So to wrap up this session on "Putting PASPT2 on a Firmer Basis," the main thing is that they’ve validated their original claim about PASPT2H by showing its near-extensivity for states with major projections on M0.
Mira: It means the method is much more trustworthy for strongly correlated solid state calculations because even though it’s not perfectly size-extensive everywhere, the numerical consequences are minimal for the specific states they are targeting.
Kai: Right. They’ve shown that this variant is intruder-free across all target states supported by their chosen partial active space M0, which is a significant piece of evidence.
Lev: From an error control standpoint, it’s good to see that even with these caveats, the method has a defined scaling for the size-extensivity error, which is helpful when you start thinking about running this on actual hardware where overhead matters.
Mira: I agree. And those other improvements they suggest—like adaptive correlation treatment and hybrid model spaces—show them a clear path for how to evolve this theory beyond just fixing the original version.
Kai: So, the implication is that we can treat these Hamiltonians as more robust starting points for machine learning, which is huge because it allows us to train models that capture physics across different molecular states efficiently.
Lev: And from an engineering view, if you can get a method this stable for theory, it gives us a better target when we try to map those complex interactions onto actual quantum hardware like superconducting circuits.
Mira: So, the paper "Putting PASPT2 on a Firmer Basis" gives us a more transparent and reliable tool for handling the entanglement of static and dynamic correlations in these strongly correlated systems.
Kai: It really does. We’ve seen that this work lays down a much firmer mathematical foundation for using PASPT2H in our research.
Lev: Moving forward, I think we need to see how quickly these adaptive features can be implemented with low overhead so it actually becomes usable on real-world quantum computers.
Mira: That’s the next big question for me—how much flexibility do we actually gain by adding these new components versus just the computational cost of those extra terms?
Kai: Right, so we’ve got this solid update on PASPT2H, but now we need to see how fast we can build systems that can actually utilize these new adaptive features.
More episodes
- 2611.14082-Two atoms with strong light-induced dipolar interactions
- 2611.14040-Characterization of BNLs 20 and 30 micrometer thick LGADs for the Electron Ion Collider
- 2611.13399-Lift-element-based dynamic network of vortex gust-airfoil interactions
- 2611.13476-Hopfion Superlattices in 3D Twisted Moir'e Light Fields
- 2611.13642-Investigation of the timing performance of double-sided p-type 3D silicon sensors produced by CNM
- 2611.13830-SiPM optical readout of a PICOSEC Micromegas precise timing detector
- 2611.14152-Where are the shortest-path nodes in spatial networks?
- 2611.14219-Sub-Ice: semi-automated mapping of basal channels from ice shelf surface elevation data
- 2611.16341-SUMNO: Spatially Unrolled Multigrid Neural Operator for Fluid Flow Surrogates
- 2611.16090-Multi-UTE Myelin Fraction Mapping: Repeatability and Cross-Field-Strength Reproducibility at 1.5T and 3T