Search

Prioritized Experience Replay (PER) enables the model to learn more about relatively important samples by artificially changing their accessed frequencies. However, this non-uniform sampling method shifts the state-action distribution that is originally used to estimate Q-value functions, which brings about the estimation deviation. In this article, an novel off policy reinforcement learning training framework called Directly Attention Loss Adjusted Prioritized Experience Replay (DALAP) is proposed, which can directly quantify the changed extent of the shifted distribution through Parallel Self-Attention network, so as to accurately compensate the error. In addition, a Priority-Encouragement mechanism is designed simultaneously to optimize the sample screening criterion, and further improve the training efficiency. In order to verify the effectiveness and generality of DALAP, we integrate it with the value-function based, the policy-gradient based and multi-agent reinforcement learning algorithm, respectively. The multiple groups of comparative experiments show that DALAP has the significant advantages of both improving the convergence rate and reducing the training variance.

Prioritized Experience Replay (PER) is a technical means of deep reinforcement learning by selecting experience samples with more knowledge quantity to improve the training rate of neural network. However, the non-uniform sampling used in PER inevitably shifts the state-action space distribution and brings the estimation error of Q-value function. In this paper, an Attention Loss Adjusted Prioritized (ALAP) Experience Replay algorithm is proposed, which integrates the improved Self-Attention network with Double-Sampling mechanism to fit the hyperparameter that can regulate the importance sampling weights to eliminate the estimation error caused by PER. In order to verify the effectiveness and generality of the algorithm, the ALAP is tested with value-function based, policy-gradient based and multi-agent reinforcement learning algorithms in OPENAI gym, and comparison studies verify the advantage and efficiency of the proposed training framework.

Lead sulfide (PbS) quantum dot (QD) photovoltaics have reached impressive efficiencies of 12%, making them particularly promising for future applications. Like many other types of emerging photovoltaic devices, their environmental instability remains the Achilles heel of this technology. In this work, we demonstrate that the degradation processes in PbS QDs which are exposed to oxygenated environments are tightly related to the choice of ligands, rather than their intrinsic properties. In particular, we demonstrate that while 1,2-ethanedithiol (EDT) ligands result in significant oxidation of PbS, lead iodide/lead bromide (PbX2) coated PbS QDs show no signs of oxidation or degradation. Consequently, since the former is ubiquitously used as a hole extraction layer in QD solar cells, it is predominantly responsible for the device performance evolution. The oxidation of EDT-PbS QDs results in a significantly reduced effective QD size, which triggers two competing processes: improved energetic alignment that enhances electron blocking, but reduced charge transport through the layer. At early times, the former process dominates, resulting in the commonly reported, but so far not fully explained initial increase in performance, while the latter governs the onset of degradation and deterioration of the photovoltaic performance. Our work highlights that the stability of PbS quantum dot solar cells can be significantly enhanced by an appropriate choice of ligands for all device components.

Perovskite-based solar cells are promising due to their rapidly improving efficiencies, but suffer from instability issues. Recently it has been claimed that one of the key contributors to the instability of perovskite solar cells is ion migration induced electrode degradation, which can be avoided by incorporating inorganic hole blocking layers (HBL) in the device architecture. In this work, we investigate the operational environmental stability of methylammonium lead iodide (MAPbI3) perovskite solar cells that contain either an inorganic or organic HBL, with only the former effectively blocking ions from migrating to the metal electrode. This is confirmed by X-ray photoemission spectroscopy measured on electrodes of degraded devices, where only electrodes of devices with an organic HBL show a significant iodine signal. Despite this, we show that when these devices are degraded under realistic operational conditions (i.e. constant illumination in a variety of atmospheric conditions), both types of devices exhibit nearly identical degradation behavior. These results demonstrate that contrary to prior suggestions, ion-induced electrode degradation is not the dominant factor in perovskite environmental instability under operational conditions.

Hybrid organic-inorganic perovskites (HOIPs) have recently emerged as highly promising solution-processable materials for photovoltaic (PV) and other optoelectronic devices. HOIPs represent a broad family of materials with properties highly tuneable by the ions that make up the perovskite structure as well as their multiple combinations. Interestingly, recent high-efficiency PV devices using HOIPs with substantially improved long-term stability have used combinations of different ionic compositions. The structural dynamics of these systems are unique for semiconducting materials and are currently argued to be central to HOIPs stability and charge-transport properties. Here, we studied the impact of ionic composition on phonon speeds of HOIPs from Brillouin spectroscopy experiments and density functional theory calculations for FAPbBr$_3$, MAPbBr$_3$, MAPbCl$_3$, and the mixed halide MAPbBr$_{1.25}$Cl$_{1.75}$. Our results show that the acoustic phonon speeds can be strongly modified by ionic composition, which we explain by analysing the lead-halide sublattice in detail. The vibrational properties of HOIPs are therefore tuneable by using targeted ionic compositions in the perovskite structure. This tuning can be rationalized with non-trivial effects, for example, considering the influence of the shape and dipole moment of organic cations. This has an important implication to further improvements in the stability and charge-transport properties of these systems.

The efficiency of solution-processed colloidal quantum dot (QD) based solar cells is limited by poor charge transport in the active layer of the device, which originates from multiple trapping sites provided by QD surface defects. We apply a recently developed ultrafast electro-optical technique, pump-push photocurrent spectroscopy, to elucidate the charge trapping dynamics in PbS colloidal-QD photovoltaic devices at working conditions. We show that IR photo-induced absorption of QD in the 0.2-0.5 eV region is partly associated with immobile charges, which can be optically de-trapped in our experiment. Using this absorption as a probe, we observe that the early trapping dynamics strongly depend on the nature of the ligands used for QD passivation while it depends only slightly on the nature of the electron-accepting layer. We find that weakly bound states, with a photon-activation energy of 0.2 eV, are populated instantaneously upon photoexcitation. This indicates that the photogenerated states show an intrinsically bound-state character, arguably similar to charge-transfer states formation in organic photovoltaic materials. Sequential population of deeper traps (activation energy 0.3-0.5 eV) is observed on the ~0.1-10 ns time scales, indicating that most of carrier trapping occurs only after substantial charge relaxation/transport. The reported study disentangles fundamentally different contributions to charge trapping dynamics in the nanocrystal-based optoelectronic devices and can serve as a useful tool for QD solar cell development.