Filtering Lithium Out of Perovskite Cells: KIER's NiOx Route to 1,000-Hour Damp-Heat Stability

Researchers at the Korea Institute of Energy Research (KIER) and UNIST have found a way to use a lithium salt to dope the hole-transport layer of a perovskite solar cell and then take the lithium back out before the layer is deposited. According to the team's paper in Advanced Science, the cells reached a power conversion efficiency of 25.24%. They also retained over 95% of their initial efficiency after 1,000 h under damp-heat stress. The tool is a micro-sized nickel oxide (NiOx) powder that acts as a redox catalyst and is then filtered out of the solution.

The result matters because durability, not peak efficiency, is the main barrier for perovskite top cells in tandems. Efficiency in the regular architecture is already high. What fails is the doped organic layer that carries holes out of the device.

01Why the standard dopant package is the weak point

In the regular (n-i-p) architecture the hole-transport layer is usually spiro-OMeTAD. Undoped, it conducts poorly. The standard fix is lithium bis(trifluoromethanesulfonyl)imide (LiTFSI) plus 4-tert-butylpyridine (tBP), which helps dissolve the salt and improves film formation. Li-TFSI doping is a widely adopted strategy for improving power conversion efficiency. The same package is also a well-documented source of instability, for three reasons.

• tBP loss. KIER's release describes the volatile additive evaporating under real operating conditions. This creates unwanted gaps between layers and so interfacial defects. The paper names tBP as the additive in question. The KIER summary does not name it.
• Lithium migration. Li⁺ is small. Earlier literature notes that Li-TFSI can easily migrate through the octahedral halide cage of the perovskite. Migrating ions can act as a random dopant and shift the charge balance of the whole device. Lithium can also accumulate at the perovskite/electron-transport-layer interface, where it hurts efficiency and increases hysteresis.
• Hygroscopicity. LiTFSI absorbs moisture and aggregates in the film, which degrades band alignment, adhesion and morphology. That is a direct problem at 85% relative humidity.

The paper argues that tBP has taken most of the blame, and that Li⁺ migration is an equally pivotal yet frequently overlooked factor. Many tBP-free formulations have been tried. The harder problem is removing tBP and keeping the lithium from moving, without giving up doping efficiency.

02The catalytic capture-and-filter process

The process works on the precursor solution, not on the finished device. The steps below follow the paper's abstract and the KIER release.

1. Add NiOx powder to the spiro-OMeTAD/LiTFSI solution. The powder is micro-sized on purpose, so a filter can catch it. Per the abstract, it induces ultrafast oxidation of spiro-OMeTAD through a redox cascade while promoting LiTFSI solvation and Li⁺ sequestration. The authors report that oxidative doping takes only seconds.
2. Let NiOx take the place of tBP. In a tBP-free formulation LiTFSI dissolves poorly. The paper says the negatively charged NiOx improves LiTFSI solvation. Lithium ions bind to the oxide surface, so the powder does the solvation job tBP normally does.
3. Filter. The reacted NiOx particles along with sequestered Li⁺ and Li-related byproducts are removed during filtration. The remaining solution is spiro-OMeTAD that has already been oxidised. It contains no tBP, Li⁺ or NiOx.

The result is a film that is p-doped without carrying mobile lithium. The oxidised spiro-OMeTAD radical cations still need a counter-anion. The sources I reviewed do not detail the final composition, but the likely partner is the TFSI⁻ anion left behind. A related UNIST-led paper that used ethylene carbonate in place of tBP described liberated TFSI⁻ ions stabilising a larger population of spiro-OMeTAD radicals. That is my inference from the related work, not a statement from the NiOx paper.

Earlier tBP-free work mostly kept lithium in the film and tried to hold it still, using agents such as 12-crown-4, functionalised NiOx nanoparticles and ethylene carbonate. KIER's route does the opposite and removes the lithium. The NiOx also does a second job, as a catalyst for oxidation, so the process needs no additional doping agent.

03What the numbers show, and what they don't

The 0.03-point efficiency gap is not meaningful. The paper's own wording is that the result is commensurate with conventionally doped devices. The claim is parity in efficiency with much better stability. It is not an efficiency gain.

Damp heat at 85°C/85% RH for 1,000 h is a standard harsh benchmark, and it is the right one for a lithium-and-moisture problem. Still, for the wider field, several points are worth keeping in mind.

• Others have reached similar figures. The UNIST ethylene-carbonate approach reported a maximum PCE of 25.56% (certified 25.51%). It also reported resilience through 1,000 hours of 85°C/85% RH damp heat, and it demonstrated mini-modules of 23.22% and 22.14% at 25 and 100 cm² aperture. The KIER result is therefore not the only route to damp-heat stability in this architecture. Its distinguishing feature is that lithium is removed instead of anchored.
• Damp heat is one stress among several. The summaries I reviewed describe the test only as a high-temperature, high-humidity trial. I did not see operational data, meaning continuous illumination at maximum power point. I also did not see UV exposure or thermal cycling in these sources.
• Cell size is not stated. Small lab devices commonly outperform modules, and the summaries do not give the active area. I found no module demonstration for the NiOx route.
• Removal is described, not quantified, in the sources I saw. The abstract says the particulates and lithium are removed completely. I did not see lithium concentration measurements in the material I reviewed. The paper's full text and supporting information are the place to check this.
• No tandem device yet. KIER frames the work as a step toward tandem commercialisation, and the lead researcher says it addresses a chronic lifespan problem. The sources describe single-junction perovskite cells, and the tandem link is a projection.

04Why this bears on commercialisation

Perovskites are most useful as the top cell on silicon, where the tandem has to survive as long as the silicon module. A secondary report on the work cites perovskite lifetimes of two to 10 years versus 25 to 30 for silicon. That is a general industry figure and not a result from this study. Closing that gap means removing each known degradation driver from the stack. Doped spiro-OMeTAD's tBP and lithium are among the best-characterised of those drivers.

The process also looks workable in manufacturing terms. It adds a mixing step and a filtration step to the precursor solution preparation, and uses a commodity oxide powder. It does not change the deposition process or the device layout. A filtration step is routine in solution processing, although handling spent catalyst and lithium-bearing waste at scale would need a plan, and the sources do not discuss one. Whether results hold for batch-to-batch reproducibility, larger areas and slot-die or blade-coated films is not addressed in the material I reviewed.

Spiro-OMeTAD itself also remains a cost and thermal-stability concern. The same review literature points to the material's glass-transition temperature, which in tBP-free layers was reported as high as 125 °C. This work does not change that fundamental limit.

Next steps are clear from the gaps: certified efficiency, module-scale devices, tests under illumination at elevated temperature, and then integration into perovskite/silicon tandems. KIER and UNIST have shown that mobile lithium and tBP can be removed from the hole-transport layer while efficiency holds. Whether that is enough to meet a 25-year warranty remains unproven.