Role of solution concentration in formation kinetics of bromide perovskite thin films during spin-coating monitored by optical in situ metrology

Optoelectronic devices based on metal halide perovskites continue to show a improved performance, and solution-based coating techniques pave the way for large-area applications. However, not all parameters influencing the thin film formation process of metal halide perovskites are identified and entirely rationalised over their full compositional range, thus hampering optimised thin film fabrication. Furthermore, while the perovskite deposition via spin-coating and annealing is an easily accessible technique, more profound insights into the chemical formation process are still lacking. Varying the precursor solution concentration is commonly used to vary the resulting thin film thickness. This study shows that varying the precursor solution concentration also affects the thin film morphology and optoelectronic quality. Hence, we herein investigate the influence of the precursor solution concentration on the formation process of a pure bromide-based triple cation perovskite (Cs0.05MA0.10FA0.85PbBr3) by fiber-based optical in situ measurement. During the spin-coating process, in situ UV-vis and PL measurements reveal formation kinetics are strongly dependent on the concentration. Furthermore, we identify delayed nucleation and retarded growth kinetics for more concentrated precursor solutions. In addition, we quantify the shifting chemical equilibrium of colloidal pre-coordination in the precursor solution depending on concentration. Namely, colloids are pre-organised to a higher degree and higher-coordination lead–bromide complexes tend to form in more concentrated precursor solutions. Thus, the modified solution chemistry rationalises retarded perovskite formation kinetics and highlights the precursor concentration as an influential and optimisable parameter for solution-based thin film deposition.


Introduction
Over the last decade, metal halide perovskites (MHPs) rose to prominence in materials and renewable energy research due to their outstanding optical properties, namely a steep absorption onset and a high absorption coefficient, 1 as the photoactive material in optoelectronic devices. Perovskite-based single junction and tandem solar cell devices have today reached record efficiencies of 25.7% and 31.3%, respectively. 2 A unique benet of MHPs is their deposition via solution-based techniques, such as spin-coating in a laboratory and printing or continuous coating techniques on an industrial scale. 3 Thus, production costs are reduced, and MHPs are brought into the game to compete with or complement established silicon-based solar cells. 4 In addition, the unique bandgap tunability of MHPs 5 sparks interest in extending their utilisation into further optoelectronic applications, e.g., light emitting diodes (LEDs), 6 transistors, 7 and detectors. 8 A remaining challenge is the reliable and reproducible fabrication of metal halide perovskite thin lms over their manifold compositional range. High-quality thin lms are essential for manufacturing and improving perovskite-based optoelectronic devices. [9][10][11] Thus, the solution-based perovskite thin lm deposition has been focused on optimising the layer quality in morphology and lm thickness. Such optimisation is achieved by tailoring, e.g., the solvent system, 12-14 the precursor and spectator salts, [15][16][17] setting an anti-solvent drop, 18 and adjusting technical preparation parameters. [19][20][21] The preparation routine and related formation pathways dictate the thin lm quality. Thus, understanding and rationalising perovskite formation processes is key to reliable and reproducible highquality perovskite thin lms. Nevertheless, the formation process's precursor solution chemistry and individual preparation parameters are little explored and understood.
Several intermediate solvate and crystalline phases have been identied, and their occurrence depends on the exact precursor composition and preparation route. [22][23][24][25][26][27] Lately, studies have focused on clarifying the formation process of standard perovskite compositions and recipes, such as MAPbI 3 and (Cs 0.05 MA 0.17 FA 0.83 )Pb(Br 0.17 I 0.83 ) 3 (so-called "triple cation"), by optical and structural in situ measurements. Merdasa et al. 19 describe the overall formation process of the "triple cation" perovskite during spin-coating and annealing, and examine wet-lm thinning. Sutter-Fella and co-workers identify different formation dynamics for MAPbI 3 determined by the lead salt in the precursor solution 28 and an optimised antisolvent dripping time related to the nal MAPbI 3 thin-lm properties. 29 In addition, Taylor et al. 30 classied three types of anti-solvents and their utilisation, explaining differences in solar cell performance.
Such top-down approaches aim to rationalise established preparation routines and reason the device performance. Evaluating various preparation procedures from "The Perovskite Database Project" 31,32 reveals that 97% of metal halide perovskite thin lms are prepared from solution-based processes, 68% of those are fabricated by conventional 1-step spin-coating, and a further 24% include spin-coating as the primary preparation step. Spin-coating preparation depends on parameters such as spin speed, acceleration and time, temperature, atmosphere, annealing temperature and time, solvent system, antisolvent, precursor salts, and solution concentration. 286 solvent and 93 anti-solvent combinations are currently utilised in literature 31 and reect on the herculean task of optimising thin-lm fabrication with empirically determined solvent systems. Thus, a bottom-up approach rationalising the inuence of individual preparation parameters will enable targeted solution-based process development, increase the reproducibility, and perovskite thin lm quality. The rst attempts were made by Merdasa et al. 19 investigating the thinning of the wet lm with increasing spin speed and overall crystallisation processes via optical in situ monitoring. Furthermore, varying the halide ratio to tune the bandgap inuences the formation pathways of MHPs. They form directly from solution, via an intermediate solvate phase, or via both competing pathways depending on the halide ratio. 33 The precursor solution concentration is one modiable parameter for perovskite preparation, used to adjust the thin lm thickness. Usually, a relatively high concentration is used for solar cell application (∼1.2 M) 34 with spin-coating techniques. In comparison, lower solution concentrations are practical in printing and LED manufacturing. 6 Effects of changes in the precursor concentration on the formation kinetic and thin lm quality beyond lm thickness were not yet discussed. Therefore, we report and discuss the inuence of the precursor solution concentration in detail on the formation kinetics of MHPs and underlying solution chemistry, using the example of FA 0.85 MA 0.10 Cs 0.05 PbBr 3 (3CatPbBr 3 ). Kulbak et al. 35 introduced this perovskite composition as the most promising for wide bandgap, stable and high-efficiency solar cell devices. Signicant improvements are postulated upon optimisation of the 3CatPbBr 3 thin-lm quality. However, especially widebandgap bromide-based perovskites are additionally applicable for LEDs. 6 LEDs, as all technologies based on thin lms, require a high thin lm quality. Thus, rationalising underlying lm formation processes is essential to optimise and increase the performance of bromide-based MHPs, not only for solar cells but for a broader range of optoelectronic applications. Also, optimising large-scale deposition will enormously benet from understanding the formation process from solution to the nal lm and make optimisation more efficient and targetoriented.
Bromide-based perovskites form directly from solution during spin-coating as they interact weaker with solvents 36 than their iodide-based counterparts. Iodide-based perovskites form via an intermediate step. 27 For this reason, this fundamental study on the inuence of the solution concentration on the formation kinetics focuses on bromide-based perovskites and deliberately omits the iodide-based materials. Rationalising formation kinetics is most suitable for bromide-based perovskites since additional intermediate and parallel formation steps can be neglected. Thus, data interpretation and conclusion are streamlined since only the formation from solution onto the expected perovskite needs to be considered.
We observe a drastic impact of the precursor solution concentration on the formation kinetics utilising in situ UV-vis and PL spectroscopy. Counterintuitively, an earlier crystallisation onset and a faster crystallite growth during spin-coating are identied for lower precursor solution concentrations. These formation kinetics are unexpected concerning the conventional LaMer model commonly used to describe perovskite crystallisation processes. 37,38 Higher concentrated solutions should reach super-saturation faster due to solvent evaporation and, thus, crystallise earlier.
Small Angle X-ray Scattering (SAXS) and Nuclear Magnetic Resonance (NMR) measurements on the solution concentration series reveal an increased degree of crystallite pre-order for higher concentrated precursor solutions and the formation of colloids. Thus, complex and colloid solution chemistry changes evoked an earlier crystallisation onset and faster growth kinetics for lower concentrated solutions. The higher the solution concentration, the slower the formation kinetics since preformed colloidal structures need to re-structure and re-organise. Thus, the precursor concentration dictates the solution chemistry and predenes critical factors in the formation process. Overall, this study contributes to a systematic characterisation of perovskite preparation parameters and highlights the importance of underlying precursor solution chemistry and characteristics for perovskite formation processes.

Results and discussion
Thin-lm characteristics First, thin lm characteristics are presented. The layers were prepared from a solution concentration series of 0.5 M, 0.8 M, and 1.2 M for preparation. Fig. 1 presents SEM images highlighting the morphology and thickness of the nal FA 0.85 -MA 0.10 Cs 0.05 PbBr 3 (3CatPbBr 3 ) thin lms prepared from the precursor solution concentration series. All thin lms prepared from the concentration series show a low coverage with nonconnected grains. While grains formed from the 0.5 M solution exhibit an undened shape, grains formed from the 0.8 M and especially from the 1.2 M solution are cubic-shaped. For all three samples, the lm thickness is very inhomogeneous and rough. The lm thickness decreases from 900-1000 nm for spin-coating the 1.2 M solution to 300-600 nm for the 0.5 M solution. The limited coverage causes the roughness of the lms. However, with the varying lm thickness, more material is deposited on the substrate using higher concentrated solutions.
Overall, SEM images demonstrate a low-quality morphology of the 3CatPbBr 3 lms of the concentration series. Setting an anti-solvent drop during spin-coating ensures a closed lm with a high coverage and a wrinkled morphology (Fig. S2 †). Although the morphology quality benets from setting an anti-solvent drop, this parameter is not discussed in detail. Setting an anti-solvent drop is a specic technique limited to spin-coating. Herein, we will investigate and explore the solution chemistry and crystallisation kinetics independent of this deposition method. This study aims to rationalise the inuence of the solution concentration. Thus, only an unperturbed formation process without setting an anti-solvent drop is discussed below. Setting an anti-solvent drop would result in higher-quality lms by quenching the process but intervening with the intrinsic formation kinetics. Rationalising the unperturbed process is crucial to rationalise the entire formation of MHP. In a later step, it will help either optimise the anti-solvent drop, replace it with other methods, or develop coating processes where no classical anti-solvent drop is possible. Profound knowledge will be helpful for the inevitable transition to industrial-scale deposition methods like printing and coating techniques. However, the inuence of setting an anti-solvent drop during spin-coating of the 1.2 M 3CatPbBr 3 precursor solution is briey presented in ESI Note 1 and An additional reection at 2q = 12.3 occurs. Due to missing further reections with sufficient intensities, this secondary phase is not identiable. Thus, revising the nal 3CatPbBr 3 lm properties, the solution concentration mainly inuences the thin lm morphology. Overall, we hypothesise that the underlying reason for these differences is the crystallisation process, which determines the optoelectronic properties of the nal samples, especially in high-quality lms. Thus, in the following, we focus on the detailed rationalisation of the formation kinetics depending on the precursor solution concentration.

Formation process and kinetics
To investigate the intrinsic formation process and kinetics of the 3CatPbBr 3 perovskite thin lms by optical in situ metrology, spin-coating is done without an anti-solvent drop. While spincoating parameters are kept constant for comparability, the spin time is increased from 40 s to 120 s compared to the recipe established by Kulbak et al. 35 to follow the entire intrinsic formation process for all three concentrations. Since absorption and scattering inuence the in situ UV-vis signal based on the assembly of the reectance probe (see Fig. S1 †), this measurement mode is referred to as trans-ectance. 19 Although a change in the transectance signal evolves for all three concentrations during prolonged times of spin-coating, no straight absorption edge expected around 500-550 nm for pure bromide-based perovskite is clearly identied. A low change of red tones only minimally indicates the absorption edge since scattering dominates the mean trans-ectance signal (Fig. S4 †) for all solution concentrations over the entire wavelength region. Therefore, only a minor part of the detected light undergoes a transmission process and explains the faintly visible absorption edge around 500-550 nm. Even though the amount of absorbed light increases slightly for a higher concentrated solution, its share is vanishingly small and does not give conclusive insight into the perovskite phase evolution. The implicating evolution of an absorption edge at 500-550 nm suggests the direct formation 33 of the 3CatPbBr 3 , supported by a noticeable absorption edge formation by setting an anti-solvent drop ( Fig. S5 and ESI Note 1 †). The above-described low-quality morphology of the nal 3CatPbBr 3 lms justies the phenomenon of increased scattering accompanied by the low absorption upon crystallisation.
To rationalise the formation process of the 3CatPbBr 3 system comprehensively, in situ PL measurements during spin-coating are presented in Fig. 2(a 2 ), (b 2 ), and (c 2 ) for the 0.5 M, 0.8 M, and 1.2 M solution. In situ PL measurements give complementary insights into the perovskite formation process since the background of the PL signal is less inuenced by scattering due to lter effects than in situ UV-vis measurements in the respective The upper panels (subscript 1) exhibit the UV-vis (transflectance) data, while the lower panels (subscript 2) exhibit in situ PL data. In situ PL is measured with an integration time of 500 ms. While transflectance measurements exhibit a high background signal hampering data analysis, in situ PL measurements display a clear signal evolution that is more straightforward for data analysis. Therefore, in situ UV-vis and PL measurements correlate, supporting and strengthening their results. For higher concentrated solutions, both in situ signals suggest a delayed crystallisation onset.
conguration. In general, light-induced phase segregation 40 is avoided by choosing a purely bromide-based perovskite and, thus, can be neglected in this formation study. All three concentrations display a comparable evolution of the PL emission over time during spin-coating. Due to the direct perovskite crystallisation of pure bromides, a PL signal around 530 nm (corresponding to 2.35 eV, Fig. S6 †) arises with a rapid increase in intensity, decreases fast, and vanishes over the spin-coating process. The higher the solution concentration, the longer the PL signal evolution progress takes. Changes in the in situ PL and UV-vis measurements arise roughly at the same time within one solution concentration. Interestingly, the signal change indicating the start of the formation process is delayed for higher concentrated precursor solutions. From these rst observations, especially for in situ PL metrology, differences in the formation kinetics upon the solution concentration are hypothesised.
In situ PL metrology datasets are analysed and discussed in more detail (Fig. 3) to investigate the concentration-dependent formation kinetics. Fig. S6 † presents the in situ PL measurements for the solution concentration series recalculated by Jacobians transformation 41 in units of energy (eV). Peak analysis in the following uses this recalculated data.
As discussed above, scattering mainly inuences in situ UVvis measurements hiding a clear absorption edge. Thus, in situ UV-vis metrology mainly identies the crystallisation onset by increased averaged transectance (Fig. 3(a)). Increased averaged transectance indicates a change in scattering due to the lm solidication and, thus, translates to the crystallisation onset. For example, the crystallisation of the 0.5 M solution starts at ∼37 s within the spin-coating process, while increasing the concentration to 0.8 M and 1.2 M delays the crystallisation onset to ∼45 s and ∼60 s, respectively. At rst glance, this timedependent behaviour is contrary to an expected crystallisation from the LaMer model commonly utilised to describe perovskite crystallisation from solution. 37,42 Within this model, supersaturation initialises nucleation and subsequent crystal growth. Higher concentrated solutions should reach the saturation limit faster upon solvent evaporation and, thus, crystallise earlier. These assumptions do not appear accurate for the presented 3CatPbBr 3 system. In contrast, the crystallisation onset is delayed for higher concentrated solutions. Setting an antisolvent drop during a relevant process window induces crystallisation ( Fig. S5 and S7 †). Here it must be noted that the dependency of formation kinetics on the solution concentration is inuenced by the experimental design. Deliberately choosing a low boiling point solvent inuences this behaviour and is a way to tune the formation process. 43 Fig. 3(b) presents the evolution of the PL intensity at the leading peak position for the solution concentration series during spin-coating. A PL signal arising around 530 nm is attributed to the presence of the 3CatPbBr 3 perovskite. Thus, the PL signal appearing at this wavelength translates to the formation of the desired perovskite. Furthermore, the rise in PL intensity for all three concentrations correlates well to the averaged transectance (green dashed lines) increase at 38 s, 48 s, and 61 s for 0.5 M, 0.8 M, and 1.2 M solutions. Thus, PL measurements verify a delayed crystallisation onset for higher concentrated solutions, as discussed for in situ UV-vis measurements. Furthermore, slight differences in crystallisation time are attributable to measurements on different samples.
The tted data conrms the considerable signal increase at the onset of crystallisation followed by an immediate decay before being observed in the 2D heatmaps ( Fig. 3(b)). Based on this trend, we hypothesise the following nucleation and growth process: numerous tiny crystal nuclei or nucleation centres precipitate simultaneously during crystallisation induced by solvent evaporation. Similar to perovskite nanocrystals, such nucleation centres would result in the described substantial rise in PL intensity. Very rapid crystal growth of nucleation centres can explain the immediate decrease in PL intensity. During lm formation, the lm condenses more and more due to the growth of crystallites. The growth leads to defects and grain boundaries within the lm and impacts the charge carrier diffusion and non-radiative recombination. These two effects occurring during lm formation quench the PL intensity. A low PL intensity is associated with re-absorption by the already formed crystallites and a lower outcoupling caused by continued solidication 44 during ongoing spin-coating. Optical effects like re-absorption and outcoupling are justied within the refractive index and morphology. A rough and low-quality morphology will increase diffusive reection. Due to continuous growth, the wet lm becomes more comparable to the thin lm properties; thus, optical signatures change. Nevertheless, the measurement setup, e.g. the detector to sample distance, stays the same, and statements are valid during the crystallite growth based on relative changes.
Although the shape of the PL intensity follows the same trend for all concentrations, the maximum PL intensity and the time evolution differ with the concentration. The maximum PL intensity increases from 1.5 × 10 6 counts for the 0.5 M solution to 2.5 × 10 6 counts for the 0.8 M solution. This signicant difference in maximum intensity is hypothesised to point to more nucleation centres forming during initial crystallisation. Interestingly, this trend does not continue for the 1.2 M solution with a reduced maximum PL intensity of 1.3 × 10 6 counts. The reduced maximum intensity indicates the formation of fewer seed crystals or already larger nucleation centres. Seed crystals form more and more from colloidal structures and, thus, are limited in number. This phenomenon will be discussed in more detail below. Surprisingly, the time from the PL onset until reaching the PL maximum, indicated by the broader FWHM, extends with increasing solution concentrations, from 4 s for the 0.5 M, to 6 s for the 0.8 M, and 13 s for the 1.2 M solution. The retarded peak evolution for higher concentrated solutions suggests slower nucleation and crystal growth for higher concentrated solutions. The retarded nucleation and crystal growth accompany the delayed crystallisation onset. The decelerated formation kinetics explain the dened growth of cube-shaped crystallites for lms prepared from higher concentrated solutions (Fig. 1). Fig. 3(c) presents the evolution of the PL peak position for the three established 3CatPbBr 3 solution concentrations. Irrespective of solution concentration, the evolution of the PL peak positions is consistent. It arises at higher energies, drops quickly to lower energies, and then stabilises aer shiing slightly toward lower energies during ongoing spin-coating.
The initial high energy PL emission at 2.38 eV (0.5 M) to 2.36 eV (1.2 M) indicates larger nucleation centres forming from the higher concentrated solutions. Likely, nucleation centres demonstrate PL emission at higher energy, while the PL peak position shis to lower energy during crystal growth. The PL progress supports that small nucleation centres form upon crystallisation and rapidly grow to larger crystallites. Overall, bigger nucleation centres tend to form for the higher concentrated solution. The initial PL peak position at 2.38 eV for the 0.5 M solution shis to 2.36 eV for the 1.2 M one, correlating to crystallite sizes of 7.52 nm and 8.26 nm, extrapolating literature values of MAPbBr 3 nanoparticles. 45 The rapid shi in the PL peak position toward lower energies signals the crystal growth for all three concentrations until a turnover point is reached. Aer this, the PL peak position only shis slightly to lower energies; thus, the crystal growth is slowed. Comparable to the nucleation onset discussed before, the timing of the turnover point is delayed for higher concentrated solutions, from 45 s for the 0.5 M solution within the spin-coating process to 85 s for the 1.2 M solution. Hence, the window for crystallite growth extends from 7 s for the 0.5 M solution to 24 s for the 1.2 M one. This increased growth window indicates a comparable slower crystallite growth within the fast growth regime directly aer nucleation for a higher concentrated solution. The grey dashed lines with different slopes visualise this phenomenon in Fig. 3(c). The atter the slope, the slower the crystallite growth. A slower and, thus, more dened crystallite growth explains the more dened cubic shape of the grains of the lm spin-coated from the 1.2 M solution.
In addition, the turnover point time roughly corresponds to when the PL intensity starts vanishing for all three solution concentrations. This correlation in PL intensity and position emphasises that nucleation and crystal growth mainly occur in the rst seconds aer crystallisation. The same trend in decelerated nucleation and crystal growth process is derived from PL intensity and peak position though absolute timings vary due to different anchor points in both parameters. The fast shi in the PL emission to lower energies is also observed for pure, homogeneous crystallising MAPbBr 3 (Fig. S7 †). Thus, grain growth dominates this shi rather than inhomogeneities caused by the cation distribution.
Overall, re-absorption, acting as a lter effect, appears in perovskites and inuences PL spectra. Due to their high absorption coefficients, the high-energy part of the PL spectrum is re-absorbed and, thus, cut off. The resulting sharp PL onset on the high-energy side becomes more important for ongoing formation since the lter effect increases with perovskite thickness. Additionally, a decay in PL intensity connects to increased re-absorption caused by continued solidication of the perovskite lm. 46 Even in the early stages, re-absorption affects the PL emission peak to a certain extent. Larger particles with a lower bandgap re-absorb the high-energy emission of smaller particles. Thus, the PL emission peak represents the maximum crystallite size at the early nucleation and crystal growth stages.
The PL peak position for all three concentrations is around 2.30 eV when the PL peak position stabilises, indicating a comparable crystal size at this stage of the formation process. However, at the end of the spin-coating process, the stabilised PL peak position lies at higher energy than the PL emission of the nal 3CatPbBr 3 lms (2.23 eV, green dashed line). Slow, ongoing grain growth and residual polar solvent, causing a solvatochromic shi, are accountable for the offset. The annealing step nally removes residual solvents and completes the perovskite formation process. In addition, the wet-lm thinning is also solution concentration-dependent. Overall, lower concentrated solutions result in thinner wet lms (ESI Note 2 and Fig. S8 and S9 †).
In Fig. S10, † we additionally present the in situ UV-vis data for the so-called "triple cation" (Cs 0.05 MA 0.17 FA 0.83 ) Pb(Br 0.17 I 0.83 ) 3 perovskite. Also, for this perovskite composition, the lower concentrated solutions crystalise earlier. While the 0.8 M solution crystallises ∼60 s within the spin-coating process, the 1.2 M solution crystallises at ∼90 s. Exact timings and, thus, detailed kinetics differ between the "triple cation" iodide-based perovskite and the bromide-based archetypes. However, the same trend is demonstrated, and the kinetics discussed above for pure-bromide MHP can be transferred to other perovskite compositions. The delayed crystallisation for higher concentrated solutions is highly relevant for all perovskite compositions and optimising deposition processes for solar cells, LEDs and other optoelectronic applications. Therefore, it can be concluded that the kinetics analysed and the connected reasons in solution chemistry stay comparable for more complex perovskite compositions containing iodide. The formation kinetics of the bromide-based perovskite concentration series is deliberately investigated since those form directly from solution and optical complex to detect intermediate solvate phases, as expected for iodide-based compositions, can be excluded.
In summary, complementary in situ UV-vis and in situ PL measurements reveal a strong dependency of the 3CatPbBr 3 formation kinetics on the precursor solution concentration. Higher concentrated solutions delay the onset of crystallisation and retarded nucleation and growth kinetics.

Solution characteristics and chemistry
The solution concentration has a pronounced inuence on perovskite formation kinetics. However, comprehensive rationalisation of the perovskite formation processes requires further insights into the respective solution chemistry. SAXS (Small Angle X-ray Scattering) measurements were performed to investigate the 3CatPbBr 3 solution concentration series on a nanostructural level. The study by Flatken et al. 47 demonstrates the applicability of SAXS measurements on perovskite solutions and presents the existence of structured colloids and their interaction in solution, resulting in pre-crystalline arrangements. The form factor gives insights into the shape and size of the nano-objects in the colloidal assembly itself. The shape, size and inter-particle interference, reected in the structure factor, directly affect the scattering prole of the solution. Fig. 4(a) presents the scattering curves for the samples with 0.5 M, 0.8 M, and 1.2 M solution concentrations. The black dotted lines demonstrate the actual measured data, while the solid lines show the t from the soware SASt©. 48 A broad maximum evolves for higher concentrated solutions. This specic increase in scattering intensity demonstrates the domination of a structure factor and indicates an interaction, pre-organisation, of the observed particles in the solution. The stronger the maximum, the more particles are involved in forming these pre-organised clusters. The volume fraction derived, assuming a hard-sphere structure factor model, expresses the measure. Fig. 4(b) presents the concentration dependency of the volume fraction, increasing from 0.011 in the 0.5 M solution to 0.056 in the 1.2 M one. Thus, higher concentrated solutions possess a higher structural preorganisation within the colloids.
Using the extended Bragg equation: 49 where q is the position of the peak maximum in q-space, the mean inter-particle distance (d ip ) between the mass centres of the subunits within the colloids is described. Fig. 4(c) shows the concentration-dependent inter-particle distance. The mean inter-particle distance reduces from 3.63 nm in the 0.5 M solution to 1.85 nm in the 1.2 M one. Thus, the individual subunits, on average, come closer together and interact stronger within the higher concentrated 3CatPbBr 3 solutions. The characteristics of the colloidal dispersion strongly inuence the crystallisation process via repulsive interaction between colloids and, by that, the colloidal stability. Nonclassical nucleation and growth mechanisms can explain these phenomena. 50 In addition, 207 Pb NMR measurements on the solution concentration series were performed to investigate the chemical environment of the Pb 2+ ion within the respective solutions. An upeld shi to lower ppm values is observed for increased solution concentration (Fig. 5). In comparison, the peak for the 0.1 M solution is at 1091 ppm, and the peak for the 1.2 M solution shis to 571 ppm. The peak shis linearly to the solution concentration (Fig. 5(b)). Interestingly, the chemical shi of the 1.2 M solution shis close to the peak position for solid-state 207 Pb NMR of FAPbBr 3 at 515 ppm. 51 FAPbBr 3 is chosen as the reference value since FA + is 85% of the cations in the discussed example. Overall, these chemical shis conrm a shied chemical equilibrium. The upeld shi upon the increased solution concentration is connected to an increased electron density around the lead. This effect rstly can indicate an agglomeration of the individual Pb 2+ into a network, which the previously discussed SAXS measurement veried by the described higher pre-coordinated structures. Since the chemical shi of the 1.2 M concentrated solution and the solid-state FAPbBr 3 are comparable, their chemical environments are also comparable in terms of coordination sphere, coordination number, and ligands. This comparison supports the interpretation of a higher pre-ordering of clusters in higher concentrated precursor solutions, comparable to the environment in the nal perovskite, with more electron-donating Br − as ligands coordinating the lead.
In addition, solution-based UV-vis measurements indicate a chemical shi to higher bromide-coordinated lead-halidesolvent complexes or the formation of poly-complexes ( Fig. S11 and ESI Note 3 †). Thus, a third solution-based measurement technique also conrms a shi in the chemical equilibrium.

Conclusion
Changing the precursor solution concentration of MHP precursor solution not only decreases the thin lm thickness but also results in considerable changes in the formation mechanism and kinetics. Lower concentrated solutions exhibit accelerated formation kinetics, namely earlier nucleation and faster crystallite growth, translating into a narrow process window for lower concentrations. Therefore, controlling the lm formation process for lower concentrated solutions gets more complex, potentially leading to lower quality thin lms and a decreased device performance.
We showed that the solution chemistry also depends on the precursor concentration, the chemical equilibrium shis with the increase of the precursor solution concentration. Higher concentrated solutions possess a higher structural preorganisation in colloids interacting more strongly with each other. Thus, the concentration denes the precursor solution fundamentally, such as the chemical interaction and preorganisation of precursor salts and solvents. Overall, the solution chemistry predetermines the formation process and kinetics. Thus, the precursor solution unveils excellent potential to optimise the perovskite thin lm deposition from solution-based techniques, e.g., by new precursor and solvent combinations. This gets especially important for elaborate precursor compositions and possible intermediate phases forming for iodide-based perovskites. Hence, a detailed look at the precursors over the solution chemistry, the individual formation pathways and kinetics to the thin lm properties are necessary to rationalise the formation processes of MHPs fully. Furthermore, the precursor solution concentration should be maintained while optimising the thin lm thickness, not to alternate the underlying solution chemistry and formation kinetics.
The direct insights into the detailed formation kinetics depending on the precursor solution concentration demonstrate the signicance of in situ metrology. Interestingly, small, partly unconscious changes in the preparation process enormously impact the overall formation of MHPs and their thin lm quality. Thus, in situ metrology allows for uncovering crucial preparation parameters, leading to detailed, standardised preparation routines.

Conflicts of interest
There is no conicts of interest to declare.

Acknowledgements
This work was carried out with the support of the Helmholtz Innovation Lab HySPRINT. We like to thank the PTB for the ability to use their facilities at BESSYII to carry out SAXS measurements. C. R. would like to thank Oliver Maus for rheology measurements and Carola Klimm for taking the SEM images. C. R., F. M., E. L. U. acknowledge funding from the German Ministry of Education and Research (BMBF) for the Young Investigator Group Hybrid Materials Formation and Scaling (HyPerFORME) within the program "NanoMatFutur"