Since the first organic–inorganic hybrid perovskite solar cell (PSCs) was proposed by Miyasaka’s group^[1], its power conversion efficiency (PCE) has sky-rocketed from 3.8% to 25.2% over the past decade^[2-4]. This quick growth rate has made it as a hotspot in recent years because of its fascinating properties, such as high absorption coefficient^{[5, 6]}, low exciton binding energy^[7-9], tunable band gap^{[10, 11]}, long carrier diffusion length^{[12, 13]}, and superb carrier mobility^{[14, 15]}. Though great progresses have been achieved in term of its PCE, there still exist some problems. Volatility and hygroscopic A-site cations decompose the perovskite structure under extreme environment and destroy the performance of the device^[16-21].

CsPbX₃ is a promising candidate to conquer these problems because Cs⁺ is the most feasible inorganic cation to replay volatility and hygroscopic A-site with suitable tolerance factor^[22-26]. Besides, the structures of organic cation (e.g., CH₃NH₃⁺) have orientation freedom, while Cs⁺ is symmetric without a multiple structure, which make hybrid perovskites show an unstable structure under extreme environments^[27-31]. Among all of the CsPbX₃ materials, CsPbI₃ with a bandgap of ~1.7 eV is a suitable and promising candidate for high performance and stable output photovoltaic material^[32-34].

CsPbI₃ has four different phases (cubic (α), tetragonal (β), orthorhombic (γ) and non-perovskite yellow (δ) phase) and each phase transforms under different temperatures^[35]. At room temperature (RT), CsPbI₃ will finally transfer into non-perovskite phase (δ-phase) with an unsuitable bandgap of 2.75 eV, which limits its practical application^[36-38]. Many researchers have conducted a series of methods to conquer this problem^[39-41]. In the CsPbI₃ PSCs development process, HI hydrolysis-derived intermediate plays an important role^[42-44]: stabilizing the crystal structure, optimizing the fabricated film and improving the device performance.

In this review, we aim to summarize the latest works about CsPbI₃ PSCs based on HI hydrolysis-derived intermediate. First, we briefly review the different crystal and electronic structures of CsPbI₃. We then trace the history and disputes of HI hydrolysis-derived intermediate to make this review more logical. Afterward, we highlight the functions of HI hydrolysis-derived intermediate, and systematically summarize some advanced works about HI hydrolysis-derived intermediate on CsPbI₃ PSCs. Finally, present issues and outlines are discussed to further increase the CsPbI₃ PSCs performance.

Photo-electric properties (e.g., optical transitions, charger transfer) are greatly related to crystal and electronic properties (e.g., phase transition, energy band)^[45-47]. In this section, we mainly discuss the CsPbI₃ perovskite from two aspects: crystal structure and electronic structure.

The CsPbI₃ perovskite structure can be described as: Pb-site and I-site ion form a corner sharing [PbI₆]^4– octahedron, while the Cs cation resides in the cuboctahedral cavities^{[48, 49]}. There are mainly four types of structures: cubic structure (α, Pm m), tetragonal structure (β, P4/mbm), orthorhombic structure (γ, Pbnm), and non-perovskite phase (δ, Pnma)^[46], as shown in Fig. 1(a). CsPbI₃ possesses an unsuitable Goldschmidt tolerance factor (t) in the range of 0.81–0.84, which is smaller than the ideal value (t, 0.9–1) and leads an unstable perovskite structure^[50]. The transitions of each phase in different temperatures are shown in Fig. 1(b)^[35].

The detailed transition temperature between each phase was researched by Even et al., based on density functional theory (DFT) analyzation. The increasing thermal parameters of I^– tended to strength the dynamic motion of the corner-connected [PbI_6/2]^– octahedral, which further induced a change of the unit cell volume and made δ-phase transformed to α-phase at 595 K. Then, with dynamic states gradually relaxing, α-phase transitions to β-phase at 539 K; β-phase transitions to γ-phase at 425 K; finally γ-phase turns to yellow non-perovskite phase (δ-phase) at RT^[46].

The different stability of each phases can be ascribed to the different dissociated energies. The dissociation energy from CsPbI₃ to CsI and PbI₂ for α-, γ-, and δ-phase are 0.04, –0.09 and –0.16 eV, respectively. δ-phase CsPbI₃ shows a small Pb–I–Pb bond angel (95.09° and 91.40°) than α-phase CsPbI₃ (180°) and γ-phase CsPbI₃ (154.74°), which reduces the orbital overlap between Pb and I atoms and also makes δ-phase CsPbI₃ with a deeper defect transition energy level than α-, γ-phase. This indicates that δ-phase is the most stable phase because of its lowest dissociation energy^[51]. Most importantly, different cooling rates change the formation energies for CsPbI₃ phases. α-phase converts to γ-phase when rapidly cooled in dry air, while slowly cooling leads to δ-phase phase because of different formation energies^[52]. Besides, changing the temperature to cause the structure transition, a polar solvent also induces lattice distortion in CsPbI₃. A polar solvent would induce the lattice distortion of CsPbI₃ nanocubes by triggering the dipole moment, which leads to the self-assembly from α-CsPbI₃ to a γ-phase through oriented attachment process^[53]. These studies indicate that metastable phases (contained β, γ-phase CsPbI₃) are more promising than α-CsPbI₃.

The valence band maximum (VBM) of CsPbI₃ perovskite is constituted of antibonding hybridized Pb 6s and X np orbitals, among which X np takes the lead. However, Pb 6p is in the dominant place of conduction band minimum (CBM), as shown in Fig. 1(c)^[54]. Compared with organic A-site cations, Cs⁺ has little effect on CsPbI₃ electronic properties. A-site cations could indirectly influence perovskite electronic properties through Coulombic interactions and steric hindrance to deform the perovskite lattice, which makes perovskite electronic structure close to the band edges and further changes the band gap energetics^[55].

The calculated electronic of α-, β- and γ-CsPbI₃ are depicted in Figs. 1(d) and 1(e)^[46]. According to the tight-binding (TB) and DFT structures, taking band folding into account, we can vividly draw that all the different band gaps of CsPbI₃ phases are direct, and the band gap of α-CsPbI₃ shifts from the R point in the Brillouin zone to Z and Γ for β-CsPbI₃ and γ-CsPbI₃, respectively. This change indicates that the electronic band gap gradually increases with the transition from α-CsPbI₃ to more distorted β- and γ-CsPbI₃ because the [PbI₆]^4– rotations stabilize the top of VBM and destabilize the bottom of CBM^[56].

We summarize the performance of CsPbI₃ PSCs after introducing HI hydrolysis-derived intermediate in Table 1 (sPCE is the stable PCE). Its main functions can be summarized as following:

1) Reducing crystallization energy barrier in low temperature fabrication;

2) Increasing iodide coordination numbers to decrease structural disorder, modifying structure and forming higher-order iodoplumbate complexes;

3) Slowing down the rapid crystalline process and obtaining high-quality CsPbI₃ film;

4) Inducing strain to generate distorted metastable phase (β- and γ- CsPbI₃);

5) Modifying the band gap of perovskites films.

The solution one-step method has advantages of simple, convenience and facile process, and can also be compatible roll-to-roll fabrication technology^[57]. However, the one-step deposition process shows poor morphology and low performance because of the rapid reaction in the solution. The common solution is using intermediates to slow down the quick reaction and make it controllable. However, understanding the composition of the HI hydrolysis-derived intermediate still needs a long time.

In the early stage, researchers focused on using HI additive in CsPbI₃ PSCs fabrication to cause a microstrain and induce a low temperature phase transition process. Meanwhile, extra halides in HI precursor solution tended to fill the vacancies of perovskites, resulting in change of metal–halogen–metal bond connectivity, and consequently cell volumes and optical bandgap^[58]. Besides, PbI₂ first coordinated with DMF in the precursor through Pb–O bonds, but further added HI would eliminate PbI₂–DMF coordination and form higher-order iodoplumbate complexes (e.g., PbI₄²⁻, PbI₅³⁻, and PbI₆⁴⁻), which benefited the formation of high-quality CsPbI₃ film^{[59, 60]}.

In 2015, HI was first used as an additive in CsPbI₃ PSCs. Snaith et al. introduced a small amount of HI in the precursor solution before spin-coating. They found that HI additive could change the solubility of precursor materials and induce a strain to lower the temperature phase transition. Then, strain triggered small crystals appearing and significantly stabilized its structure in RT, as shown in Fig. 2(a)^[42]. Uddin et al. controlled the concentration of HI on purpose to modify CsPbI₃ bandgap. They demonstrated that introducing 36 μL/mL HI would decrease the bandgap from 1.75 eV to optimized value of 1.7 eV and show excellence electronic properties with low charge-transport (12.8 kΩ). From the scanning electron microscopy (SEM) images, optimal concentration HI additive led to the appearance of small grain sizes with a few nanometers, which is beneficial to increase the stability of black phase CsPbI₃^[61]. This conclusion was also proven by Kim et al., who found that HI formed small grains and stabilized the black phase of CsPbI₃ at low temperature^[62]. Furthermore, they investigated the function of NH₄⁺ (e.g., NH₄Cl, NH₄Br and NH₄I) and H⁺ (HCl, HBr and HI) based additive and found that HI additive is the most efficient, and which could reduce the roughness and increase the stability of perovskites films. In particular, when it was exposed in ambient for 3 h, the optimal device with CsPbI₃ perovskite only dropped its PCE from 3.55% to 2.78%^[63].

In addition to HI additive, HI hydrolysis-derived intermediate is more effective because it eliminates water in the HI solution and is an intermediate compound to increase perovskite crystallinity^[64]. It was first proposed in 2015 by Zhao et al. They developed a new precursor compound (named HPbI₃) through a reaction of HI and PbI₂ in DMF solution, and used it to replace PbI₂ in fabrication FAPbI₃-based PSCs, as shown in Fig. 2(b)^[65]. Such HPbI₃ has a pseudo-3D crystal structure, where 1D face-shared [PbI₆]^4– octahedra with intercalated protons (H⁺) for charge balance^[66].

Zhu et al. introduced HPbI₃ into the CsPbI₃ PSCs and assisted with a triple cation NH₃⁺C₂H₄NH₂⁺C₂H₄NH₃⁺ (named as DETA³⁺) to further stabilize the α-CsPbI₃ perovskite phases^[67]. Subsequently, Chen et al. used HPbI₃ to substitute PbI₂ and found that tensile lattice strain appeared in HPbI₃-process CsPbI₃ perovskite^[68]. The tensile lattice strain generated because HPbI₃ crystals serve as a template to guide the nucleation and growth of α-CsPbI₃. A similar work was conducted by Zhao et al., who introduced the I-excess precursor HPbI_3+x and replace PbI₂ to reduce crystallization energy barrier, and fabricated α-CsPbI_3–xBr_x PSCs in low temperature (130 °C). The champion PCE of 13.61% was measured by a reverse scanning, as shown in Fig. 2(c)^[69]. More recently, in order to overcome the low PCE problem of CsPbI₃-based PSC, You et al. used CsI, HPbI_3+x and PbBr₂ as precursor to fabricate high-quality CsPbI_3–xBr_x perovskite films. They developed an inorganic shunt-blocking layer lithium fluoride (LiF) in the ETL/perovskite interface to align the bandgap and suppress the surface defect. Furthermore, a small amount of PbCl₂ were introduced to further suppress the recombination. Finally, they obtained CsPbI_3–xBr_x with the highest PCE of 18.64%, and boosted V_OC to 1.25 V with little loss. After continuous 1 sun equivalent illumination, the best device dropped only 6% of its initial PCE after 1000 h^[70].

Kanatzidis et al. recently claimed that HPbI₃ did not exist and was replaced by a compound of DMAPbI₃, which generated through DMF hydrolysis in HI solution. Importantly, they pointed out that some early reports of inorganic perovskite are actually the hybrid perovskite. They found that DMAPbI₃ possessed a larger tolerance factor and mixing with Cs⁺ could adjust tolerance factor (t) of the compounds (Cs_1−xDMA_xPbI₃) toward an ideal factor (t, 0.9–1). Finally, they achieved a champion PCE of 12.62% in Cs_1−xDMA_xPbI₃-based PSCs, as shown in Fig. 2(d)^[71]. Liu et al. dissolved PbI₂ and HI in DMF to synthesize DMAPbI₃ and further confirmed that no HPbI₃ existed. They used it as precursor to fabricate high-quality Cs_xDMA_1−xPbI₃ perovskite films with 14.3% PCE, and the initial PCE kept more than 85% when exposed in air 20 days without encapsulation, as shown in Fig. 3(a)^[72].

Zhao et al. used PbI₂·xDMAI to fabricate CsPbI₃ PSCs recently, and they concluded that the fabricated perovskites are actually all inorganic composition because the organic ion DMA⁺ are easily lost during the high-temperature (210 °C) annealing process^[73]. Later, they proved that DMAI is a volatile additive, and used it to assist with phenyltrimethylammonium chloride (PTACl) passivation. Finally, they obtained the highest PCE of CsPbI₃ PSC, 19.03%, as shown in Fig. 3(b)^[74]. This conclusion was also confirmed by Pang et al., who found that DMAPbI₃ and Cs₄PbI₆ first formed in annealing 100 °C for 10 min. When the annealing temperature was increased to 180 °C for 15 min, DMAPbI₃ converted into DMA_0.15Cs_0.85PbI₃ with a small amount Cs₄PbI₆ residue. Once the annealing temperature exceeded 200 °C, the γ-phase CsPbI₃ was formed with a small number of DMA_0.15Cs_0.85PbI₃ residue. The number of DMA_0.15Cs_0.85PbI₃ greatly decreased and a little δ-phase appeared, as shown in Fig. 3(c)^[75].

Our groups also confirmed this conclusion. We synthesized a series of intermediate compounds (DMAI and DMAPbI₃) by different ratio of HI/DMF, and used them to fabricate CsPbI₃ PSCs. After detailed analysis, we found that the major component of CsPbI₃ was still inorganic in this reaction route. Most of DMA⁺ organic molecules lost during the annealing process, and only a small amount of DMA⁺ remained to stabilize perovskite structure. Excessive DMA⁺ interacted with Pb²⁺ to further passivate CsPbI₃ surface, as shown in Fig. 3(d)^[76].

In conclusion, the organic molecule DMAI mainly influence the crystallization kinetics and perovskite phase. During the annealing process, DMAI will sublimate quickly, change the rate of crystallization and form metastable (β- and γ-) phase based CsPbI₃. The controllable crystallization kinetics and stable (β- and γ-) phase are beneficial to morphology and stability of perovskite, respectively. Besides, DMA⁺ (2.72 Å) possesses a larger ionic radius than Cs⁺ (1.88 Å)^[77]. Therefore, if the DMA⁺ is residual (non-sublimate) in the CsPbI₃ film: (1) DMA⁺ doped into the lattice of CsPbI₃ can increase its tolerance factor for avoiding crystal structure distortion; (2) DMA⁺ reacted with Pb²⁺ to passivate CsPbI₃ surface, further reducing leakage current generation for increasing the device performance.

As we discussed earlier, HI hydrolysis-derived intermediate showed a lot of advantages in high-quality film fabrication and device performance. Importantly, perovskite films with better crystallinity, morphology, and higher range of absorption are the foundation of efficiency.

The first working α-CsPbI₃ PSCs with a PCE of 2.9% was fabricated in low temperature (100 °C) by Snaith and his cooperators via a small amount HI additive adding^[42]. Many relevant works have been done to boost its performance. Uddin et al.^[61] and Kim et al.^[62] used HI to modify the α-CsPbI₃ film morphology and boost its PCE to 6.44% and 4.88%, respectively. Later, Zhao et al. discovered that PbI₂.xHI could reduce the crystallization energy barrier for α-CsPbI₃ phase. They used PbI₂.xHI assist with two-dimension EDAPbI₄ perovskite to stabilize α-CsPbI₃ phase and avoid lattice distortion. Finally, α-CsPbI₃ phase based on the EDAPbI₄ passivation showed a record PCE of 11.8% together with superior stability, as shown in Fig. 4(a). The α-CsPbI₃ phase perovskite kept its structure after annealing at 100 °C for more than 150 h and stable at RT for months^[78].

Compared with HI, the absence of H₂O molecules in HI hydrolysis-derived intermediate can optimize the perovskite crystallinity and morphology. Chen and his cooperators replaced PbI₂ with HPbI₃ in fabricating stable α-CsPbI₃ film. They found that the bandgap was shifted from 1.72 to 1.68 eV owing to formation of tensile lattice strain. Finally, a HTL free α-CsPbI₃ was obtained with a higher PCE of 9.5%. Besides, the optimal device showed enhanced stability, which maintained 90% of its initial PCE under illumination for more than 3000 h in dry environment, as shown in Fig. 4(b)^[68]. Then, organic terminal groups were widely used to assist HPbI₃ and further improve α-phase CsPbI₃ stability. Zhu et al. introduced DETA³⁺ additive into the HPbI₃ containing precursor to stabilize α-CsPbI₃ perovskite phases. DETA³⁺ has NH₃⁺ or RHN₂⁺ group, which could combine with I^– or [PbI₆]^4– and avoid octahedral tilting. Besides, oil-wet (hydrophobic) hydrocarbon chains of DETA³⁺ increased the CsPbI₃ humidity-resistance. Ultimately, the optimal device kept its structure for more than 6 h in a humid environment (~30 °C, 60%–70% RH), while the reference device changed its color from black to yellow in the same condition in 55 min, indicating a phase transition^[67]. Zhao et al. fabricated HPbI_3+x to form α-CsPbI₃ perovskite film. With the help of PEAI post-treated on the α-CsPbI₃, superior PCE of 13.5% was obtained with improved stability. Because of PEA⁺ terminated on the CsPbI₃ surface and did as a capping layer, PEA⁺-CsPbI₃ remained stable structure after 80 °C annealing for 7 days, while the pure one degraded into non-perovskite in the same condition, as shown in Fig. 4(c)^[79] . Han et al. added organic terminal groups (OTG) into CsI and HPbI₃ precursors to design an inverted planar CsPbI₃ PSCs. They report that OTG induced a steric hindrance and suppressed octahedral [PbI₆]^4– tilting. Moreover, OTG passivated the surface electronic traps states to further increase its performances. Finally, the inverted planar OTG-CsPbI₃ PSCs showed the highest PCE of 13.2% and retained about 85% of its initial PCE for 30 days at RT, while the reference device degraded completely in 3 weeks, as shown in Fig. 4(d)^[80].

Recently, the CsPbI₃ films fabricated by HI hydrolysis-derived intermediate were proved metastable phases (combined β-phase CsPbI₃ with γ-phase CsPbI₃).

The β-phase CsPbI₃ can also be formed at low temperature and show more stable perovskite structure than α-phase one. However, it is difficult to deposit and stabilize its perovskite structure^[81]. Zhao et al. adopted PbI₂·DMAI and CsI as precursor to fabricate stable β-phase CsPbI₃ with stable structure. Furthermore, they used choline iodine (CHI) to passiviate the surface trap states and aligned the energy level in the TiO₂/β-phase interface. β-phase CsPbI₃ PSCs has a particularly high PCE of 18.4% with distinguished ambient stability because of PbI₂·DMAI and CHI. The stability of β-CsPbI₃ PSCs were greatly improved: it retained 92% of its initial PCE after 500 h illumination at the maximum power point, as shown in Fig. 5(a)^[73].

The γ-phase CsPbI₃ is the most stable black phase because of its lowest dissociation energy^[82]. Hence, γ-phase CsPbI₃ was systematically researched to boost its stability and performance. Hu et al. used HI to fabricate high-quality CsPbI₃ film. After that, they introduced a small amount of H₂O into precursor to induce a proton transfer process in CsPbI₃ film, which could manipulate the grains size and improve stability of perovskite. When stored in ambient conditions, the optimal device showed no drop of its performance while the reference’s PCE was greatly degraded, as shown in Fig. 5(b)^[83]. Then, Nazeeruddin et al. developed a soft template-controlled growth method to fabricate pinhole-free γ-phase CsPbI₃ film, where (adamantan-1-yl)methanammonium (ADMA) acted as a template and ionized by HI. They pointed that ADMA absorbed on CsPbI₃ surface and induced a steric effect to further increase the nucleation rate at the initial stage of CsPbI₃ formation. Taking into account of controllable nucleation rate and excellent morphology, superior PCE of 16.04% was obtained with improved stability (drop only 10% after continuous light soaking and heating for 1000 h), as shown in Fig. 5(c)^[84].

In our recent research, we reported the synergistic effect of HI and PEAI additives, where HI transferred to an intermediate (HPbI_3+x) to fabricate distorted black phase-based CsPbI₃ thin films and PEAI induced a steric effects to avoid phase transition. It is noteworthy that the best device maintained 92% of its initial PCE for 60 days storage in ambient (RH ~ 20%–30%, 25 °C), while the reference one degraded to 0.65% in the same condition for 8 days, as shown in Fig. 5(d)^[43]. Then, HPbI₃ was used to fabricate CsPbI₃, assisted with anti-solvent hot substrate spin-coating method, and 15.91% PCE was obtained in a humidity environment (RH ~ 50%)^[85]. Furthermore, we added a small amount of Br^- (5%) to the HPbI₃-CsPbI₃ lattice to increase phase stability by suppressing bulk trap-assisted non-radiative recombination and relaxing lattice strain. The performance was boost up to a record PCE of 17.17% together with excellent stability^[86]. Another work we reported was that 3 mol% Cl^– was added to the γ-CsPbI₃ film with a 16.07% PCE obtained. It not only showed the excepted increase crystalline dynamics for an excellent CsPbI₃ morphology, but also improved crystalline orientation. The Cl-doping CsPbI₃ showed high stability, for optimal device dropped only 0.45% of its initial PCE under continuous light soaking for 200 h. More important, non-encapsulated CsPbI₃ PSCs with Cl^– doping degraded only 6% when stored in RH ~ 30% for 60 days (the fresh one dropped to 85% of its initial PCE in the same conditions)^[87]. We also introduced 2% Pb(SCN)₂ into the DMAPbI₃ and CsI precursor to control the morphology of the CsPbI₃ film. In this case, a PCE of 17.04% with V_OC of 1.09 V was obtained^[88].

One of the notorious problems to limit CsPbI₃ performance is the lower J_SC compared with hybrid one. Thus, we have also developed several strategies to increase its J_SC, such as harvesting short wavelength ultraviolet light (UV-light) or near-infrared (NIR) light, and designing device structure to capture light. First, we developed a downconversion nanoparticles (DCNPs) nitrogen-doped graphene quantum dots (N-GQDs) as an energy-down-shift to harvest the short wavelength (< 350 nm) UV-light. After combining it with HPbI₃-formed γ-CsPbI₃, the optimal device showed an improved short circuit current density (J_SC) from 18.67 to 19.15 mA/cm², with an increase of 2.57%. Meanwhile, its performance was greatly increased 3.15%, from 15.53% to 16.02%^[89]. Furthermore, we developed a core-shell-structured upconversion nanoparticles (UCNPs) to capture the NIR light, making it possible to obtain 15.86% PCE (noted that the improvement of PCE is negligible), J_SC = 19.17 mA/cm^2[90]. We also investigated the influence of different haze glass substrates, and used the optimized one to fabricate CsPbI₃ PSCs with 16.24% PCE. We found that the improvement came from scattering effect of FTO, refractive index and roughness of each layer^[91].

Reducing dimension can further increase the stability of CsPbI₃ PSCs because reducing materials dimension can lead to more symmetric crystal structure and show a smaller surface energy^[92-96].

However, the poor solubility of CsX in the precursor solution would severely limited the thickness of CsPbI₃ film and influence the light absorption. Chen et al. used HPbX₃ and CsAc as new precursor to overcome the poor solubility of Cs⁺ precursor and fabricate α-CsPbX₃ with optimal thickness. They introduced phenylethylammonium iodide (PEAI) to HPbX₃ and CsAc system and further controlled the dimension of CsPbX₃ from three dimension (3D) to two dimension (2D). Finally, a champion PCE of 12.4% in 2D CsPbI₃ perovskite was obtained, and maintained 93% of its initial PCE in ambient for 40 days, as shown in Fig. 6(a)^[97]. Similarly, our group used DMAPbI₃ as a new precursor to fabricated γ-CsPbI₃ perovskite, and then we introduced a judicious amount of PEAI into the DMAPbI₃ contained precursor to convert γ-CsPbI₃ PSCs into a 2D Ruddlesden–Popper (RP) structure (PEA)₂(Cs)_n−1Pb_nI_3n+1 perovskite. And the optimal 2D RP PSCs with highest PCE of 13.65%, showed a similar charge extraction and carrier lifetime compared with the 3D samples. The thermostability was tested in N₂ filled glovebox. The superior 2D RP PSCs kept 88% of its initial PCE when stored at 80 °C for 15 days, while the reference 3D γ-CsPbI₃ degraded to 69%. Besides, 2D RP PSCs maintained its black color whereas 3D γ-CsPbI₃ appeared yellow phase in RH ~ 30% for 12 days, as shown in Fig. 6(b)^[98]. In addition to PEAI, Chen et al. first introduced a novel dual ammonium cation piperazine-1,4-diium (PZD⁺) to generate 2D RP CsPbI₃. It’s noted that one of the ammonium groups coordinated with [PbI₆]^4– octahedral, and another one interacted with I^– to balance charge. Finally, the optimized device achieved a PCE of 9.39% and maintained its performance without decomposition under heating at 100 °C for 24 h^[99].

Pradhan et al. fabricated stable CsPbI₃ nanocrystals (NCs) with superior stability by using a higher temperature (260 °C) than usual (160 °C), and adding olelyamine (OLA) and HI respectively in the reaction process (noted that only OLA or HI are less efficient). Taking the NMR analyzation into account, they found that higher temperature helped the OLA⁺ ligands to occupy the Cs⁺ position on the surface and further stabilized its structure^[100]. To solve adverse shell ligands and unstable core lattices in nanocrystals, Choi et al. developed a rational core-shell design method in CsPbI₃ NCs. They used a novel ligand named 4-fluorophenethylammonium iodide (FPEAI) to enhance the binding force between the ligand and CsPbI₃ NCs and efficient charge coupling between NCs, to increase charge extraction. Besides, H₂PbI₄, which was synthesized with excessive HI, was used to assist doping Mn²⁺ ion into perovskite lattices and to further raise the performance of CsPbI₃ NCs to 13.4% with superb stability (maintained 92% of its PCE in air for 500 h), as shown in Fig. 6(c)^[101].

Although lots of advanced works about HI hydrolysis-derived intermediate have been done to boost PCE of CsPbI₃ PSCs, its PCE still far behind the hybrid ones. Therefore, we need to analyze the urgent problems that remain and develop corresponding strategies to improve the performance of CsPbI₃ perovskites.

Increasing the light absorption.As we show in Table 1, though the performance of CsPbI₃ PSCs has reach 19.03%^[74] , the J_SC (about 20 mA/cm²) is still lower than the hybrid one. This maybe cause by current loss and insufficient utilization of light, resulting from the relatively large bandgap (compared with hybrid perovskite films). Developing high efficiency DCNPs/UCNPs, to capture UV-light or NIR could be an useful solution^[90].

Doping X-site halide to improve CsPbI₃ PSCs stability.The unstable nature of CsPbI₃ is an unsuitable tolerance factor (t), as we mentioned earlier. Partially replacing I^– (2.2 Å) with smaller radius halides, e.g., Cl^– (1.87 Å) and Br^– (1.96 Å) or pseudohalide SCN^– (2.15 Å), could enlarge t value to reach ideal one (0.9–1)^[32].

Ligand assists to stabilize CsPbI₃ crystal structure. Introducing ligand (e.g., OTG^[80], poly-vinylpyrrolidone (PVP)^[102], ADMA^[84], PEAI^[79], and DETA^[67]) into a perovskite precursor and inducing a steric hindrance to further stable CsPbI₃ crystal structure is a useful way to boost its stability. Therefore, it is promising to develop and design new ligands.

Increasing the carrier transport for low dimension perovskite. Low dimension CsPbI₃ tends to increase surface-area-to-volume ratio and raise the Gibbs free energy to make the structure more stable^[103]. It is noted that though low dimension CsPbI₃ can keep CsPbI₃ structure more stable, the PCE of low dimension CsPbI₃ is still far below the 3D CsPbI₃. Hence, we should pay more attention to increase the carrier transport for low dimension CsPbI₃.

Avoiding the disadvantage of intermediate.Liu et al. claimed that intermediates had negative influence during low dimension perovskite deposition process because they slowed down intercalation of ions and increased nucleation barrier, and further caused the byproduct formation^[104]. Although there are no negative effects on HI hydrolysis-derived intermediate so far, we should be more vigilant against HI hydrolysis-derived intermediate.

In conclusion, CsPbI₃ perovskite, particularly the metastable phases (β- and γ-Phase CsPbI₃), is a promising material to replace the unstable hybrid perovskite. Besides, CsPbI₃ PSCs with suitable bandgap make it more suitable to apply in tandem solar cells and commercialization. Based on these advantages, we conclude that CsPbI₃ PSCs maybe the mainstream research direction in the near future, and we should adopt a positive attitude to it.

This work was funded by the National Natural Science Foundation of China (51902148, 61704099 and 51801088), the Fundamental Research Funds for the Central Universities (lzujbky-2020-61, lzujbky-2019-88 and lzujbky-2020-kb06), and the Special Funding for Open and Shared Large-Scale Instruments and Equipments of Lanzhou University (LZU-GXJJ-2019C023 and LZU-GXJJ-2019C019).