Variety of steady and excited state interactions in BODIPY aggregates: photophysics in antisolvent systems and ﬂoating layers

Associative behavior of geometrically anisotropic meso-(4-octadecyloxy-phenyl)- boron-dipyrrin (BODIPY) studied spectroscopically in binary solvent mixtures and upon compression in Langmuir ﬂoating layers. Diﬀerent steady and excited state species were found upon monolayer compression and facilitated aggregation in water / acetonitrile systems. This discrepancy points to a big concern in possibility of commonly applied generalizations across diﬀerent aggregating systems. Broad range of decay ﬁtting models were examined to reveal their ben-eﬁts and pitfalls speciﬁc to examination of dye aggregates. Radiative constant gamma-distribution approach and free form ﬂuorescence lifetime distribution with maximum entropy method (MEM) outperformed multiple common techniques for analysis of complex ﬂuorescence decays. MEM could be recommended for analysis of systems where complicated lifetime distributions appear over time or upon external stimuli. Findings and protocols could be utilized as tools in studies of steady and excited-state photophysics of BODIPY aggregates.


Introduction
BODIPY dyes, a boron-difluoride dipyrromethene complexes are "shining" in scope of attention of chemists and biologists for almost two recent decades.
They possess outstanding photophysical properties, such as high molar extinction coefficients, generally high fluorescence quantum yields and narrow absorption and emission bands.Most importantly, desired wavelength shift of both UV-Vis absorption and fluorescence peaks, or sensitivity towards specific parameters could be achieved via straightforward synthetic modifications.[1][2] [3] Biolabels and contrast agents based on BODIPY scaffold are widely commercially available.At the same time, there is also a broad range of publications on sensory BODIPY fluorophores, responsible towards solvent polarity, [4] temperature, [5] viscosity, [6][7] etc., whereas almost none of them are utilized in practice.This seems to be a status quo not just in BODIPY sensing, since fluorescence by itself is a complex process, well known to be influenced by a vast range of microenvironment properties.
One of the most important processes required to be studied for switch from theory to practice is dye association.For example, dye molecules tend to associate heavily in biological media due to their dominant localization at cell interfaces and ever changing solvent content gradients.
Aggregation process was historically of high interest in chemical physics of oligopyrrole dyes.To name a few areas, aggregation effects are significantly important for functioning of photosynthetic systems and for proper mimicking of sophisticated nature-made molecular mechanisms of electron transport.[8][9] It was also found that dye aggregation is one of the most important factors negatively influencing singlet oxygen quantum yield of agents for cancer photodynamic therapy.[10] Aggregation behavior control in literature, unluckily, still often kept in hard brackets of synthetic modification or stated as a byproduct of investigation, even though actual "control" term rather implies certain aggregation properties could be achieved by external stimuli.[11] First investigations of aggregation structure-property relationships for BOD-IPY dyes were carried out by Kim et al., they showed how different mesosubstitution motifs could lead to different aggregative consequences [12] [13].More recent investigation of BODIPY dyes aggregation by Decalzo et al. expanded a range of structural considerations influencing aggregation patterns.[14] For example, aliphatic substitution pattern directly influences possibilities of dense stacking, increase in size of conjugated system leads to increased aggregation tendency and to excimer formation, non-conjugated substituent alters type of formed aggregate, charge transfer induction also leads to different aggregation pathway and so on.Both groups utilized antisolvent systems and examined steady-state spectroscopic evidences.
There were also several attempts of our group and colleagues to explain structure-properties relationships in aggregation in Langmuir-Schaeffer materials.Variation in linear dimensions of bulky aromatic or aliphatic substituents in BODIPY core was found to influence excited state behavior in obtained materials.[15] [16] Concentration control was found to influence emission wavelengths of composite systems in OLED emitters, [17] it was also recently shown how structure alterations actually may change steady-state aggregation patterns in thin films.[18] Even though latter publications are dealing with external aggregation facilitation by compression on an air-water interface, synthetic approach to control of aggregate properties is still in emphasis.Switch from this paradigm is complicated while there is lack of systematic investigations of agggregation effects and known direct means to control them.
Common approach towards studies of organization processes could be characterized as over-generalizing, in particular because of lack of possibilities for in-situ measurements.
Recently, Stuchebryukov et al. developed a technique to measure absorption spectra directly on the air-water interface, which is feasible for real-time observations of aggregation processes, especially interesting for direct comparison to processes in solutions.[19] Since Langmuir trough provides closer model conditions to biological media than solution techniques or bulk investigations, [20] it was especially interesting to investigate spectroscopic consequences of aggregation in confinement.
Here we summarize our findings on associative behavior of low-aggregating dye, both giving our contribution into the field and emphasizing a strong need to employ all of the available aggregation stimuli as a viable part of thorough investigation of fluorophore photophysics.

Solution photophysical investigations
Chosen BODIPY structure was designed for this investigation, such that there were no specific interaction spots, no pronounced charge-transfer or other excited state effects available.In addition, BODIPY-C18 aromatic system stacking is hindered by aliphatic core substitution.It was assumed that BOD-IPY with geometrical anisotropy (having one geometrical axis longer than the others) will demonstrate aggregation-driven changes in electronic absorption and emission properties during concentration in solution.
Investigation of chloroform solution of the dye with concentration of 0.85•10 −2 mol/L in custom 50µm cuvette shows emerging 533 nm absorption band indistinguishable from the main maximum, increase in intensity of vibronic side band and a batochromic shift of a fluorescence maximum (due to the inner filter effect).
No separate long-wavelength emission band was observed.Absorption band batochromic shift and broadening is due to formation of unordered aggregates (dye-dye interactions start to happen in the ensemble, but no spatial relation such as J or H type aggregate could be distinguished).
Compound was further studied in a water / acetonitrile mixtures of various component ratios.It was found that solutions with moderate water content (above 40 Vol.%) are unstable and aggregate with formation of non-absorbing solutions with pink precipitate with green gloss over a course of 5 hours to overnight.Unexpectedly, small rod-or needle-like particles obtained from longer precipitation time have no far order (x-ray diffraction examination for these structures failed in multiple attempts with various setups).
Remarkably, higher water content solutions (above 55 vol.%) were stable over a course of a week (intact both in fluorescence and absorption spectra), indicating stabilizing role of water in association process.Increased solvophobic effect prevents growth of aggregates and precipitation.
Increase of water content in a mixture leads to quenching of fluorescence, and increase in 605 nm band (at 50% water content, peaks at 537 and 605 nm have almost equivalent intensities).Spectra at Figure 1 were normalized to sum intensity of two observed maxima for clarity purposes.Absorption spectra demonstrate no sharp aggregate maximum, and the only actual difference is in vibronic shape due to aggregate formation.[21] Figure 1: Changes in fluorescence spectra, normalized to sum of intensity maxima: left -upon solvent content variation, right -upon BODIPY-C18 concentration variation.
Fluorescence spectra, however, show gradual growth of a separate longwavelength band at 605-610 nm.This band is unlikely to be result of a ground state homomolecular interactions, and should be attributed to BODIPY exciplex (dye-solvent excited state complex or dye-dye excited state dimer).
Increase of BODIPY-C18 concentration did not lead to any distinguishable absorption changes, fluorescence spectra showed batochromic shift of the main band at high BODIPY concentrations Figure 1 (right), probably due to the inner filter effect.

Excited state dynamics
All of the binary mixtures were studied utilizing time-correlated single photon counting (TCSPC) technique to reveal nature of excited state processes.
Typical decay shapes in experiments with varied water / acetonitrile ratio and with varied BODIPY-C18 concentrations are presented in Figure S2.Time-resolved emission scan (TRES) of a concentrated BODIPY-C18 in a midpoint (50 Vol.% water / acetonitrile mixture, Figure 2) yielded two main regions of interest, accordingly being 530-560 nm (main peak) and 600-615 nm (exciplex peak).Exciplex peak appears approximately 500 ps after excitation pulse and decays faster than the main band.Two peaks thus should be considered a distinct fluorophore ensembles.TRES profiles were similar for all of the systems in binary solvent mixtures.Monomer decay yields good regression in a monoexponential model at low water content.Higher water content samples, however, require second component for reconvolution fit, and decays also bear ultrafast component with pronounced anisotropy.This last component does not allow for proper resolution of exponentials, so it was omitted in biexponential analysis.
Instead of direct increase in excited state lifetime expected from latticerelaxing fluorophore, two exponents are distinguishable, meaning that superposition of two ensembles is observed within main peak decay.
Iterative reconvolution towards function [Equation 1] yielded components of 5.47 and 6.6 ns.Relative amplitude of a first exponential falls down starting at 50 vol.%mixture, until relative amplitude inversion (longer exponential domination) happens at 60 vol.%.Overall, change in molecular environment causes 1 ns amplitude-weighted mean lifetime increase over the course of experiment (Figure 3), probably due to aggregation and solvent-related non-radiative processes.
(1) In experiment in a midpoint (with fixed 50% vol.water / acetonitrile ratio and 0.5 to 8 eq.BODIPY-C18 concentrations) global fit for 5 concentrations yields similar lifetime values of 5.3 and 7.4 ns (Figure S3).We still observe separation onto two ensembles, but dye concentration change does not affect their ratio to a significant extent.This indicates that second long component is solely responsible to water content.This also points that aggregate size is concentration-independent since dye within aggregates in a broad concentration range is available to water to the same extent.
Anisotropic nature of a monomer band is increasing proportionally to water content (as found from anisotropy amplitude values).Two sub-nanosecond correlation times could be extracted from anisotropy decay (Figure 3).650 ps correlation time should be attributed to diffusional depolarization and its sigmoid dependence on a water content approves our assumption of particle uniformity.90 ps correlation time rises towards 50-55% water system and then decreases, and should be attributed to resonance energy transfer process between main peak and exciplex peak ensembles.Decrease of amplitude corresponding to the 90 ps depolarization in a high water content system is thus due to decrease in amount of donor species.This is further approved by the fact that in experiment with fixed 50 vol.%water / acetonitrile ratio and varied BODIPY-C18 concentration, anisotropy of both components grows insignificantly after second concentration point (Figure S4).Total of 16-fold concentration increase had no impact on anisotropy, which points out that depolarization could not be dependent on any other system property other than water content.

Exciplex decay analysis
Exciplex band decay has a non-zero rise time, suggesting that exciplexes are acceptor species responsible for fast component in the main peak decay (also responsible for fast anisotropy component in the main peak).Such a behavior is especially interesting because this ensemble may be considered a light-harvesting system, funneling energy towards exciplex species.[22] Exciplexes are more likely to form on the outside of the particles, where water is available.Logically, increasing portion of exciplexes forced by water content growth leads to increased relative amplitude of a fast component in the main peak decays.[23][24] Polyexponential reconvolution fitting does not yield mathematically reasonable results even with discarded head and/or tail of decays.Only three decays (45, 50 and 55 Vol.% H 2 O) yield reasonable data, but lifetimes of two exponential components alter significantly (5.52 to 4.77 ns and 2.07 to 1.79 ns at 45% and 55% Vol.%, respectively).
We observe decrease in lifetimes of excited state species, and there are two major possible reasons for the observed phenomenon: destabilization of a ground state because of significant excited state geometry change or due to further resonance energy transfer with exciplex as a donor.[25] This RET process could be happening from exciplex to water harmonics (more pronounced as water content increases).
Unfortunately, conclusions regarding exciplex decay could not be drawn from such a limited amount of available system properties.
In search for proper reconvolution for description of observed changes we attempted several methods, in details described in the Experimental section.
Two of the approaches involving lifetime distributions yielded feasible results and we were able to justify our aforementioned assumptions.
As seen from Gamma-distribution analysis of exciplex peak in water ratio experiment (Table 1 and Figure S5), as water content in mixture increases, sharp bimodal decay constant distribution broadens until mostly one main peak is found near 0.8 -1.3 ns −1 , similar to value extracted in biexponential analysis.
Constants we recover from low water content solution (modes at 0.18 and 1.87 ns −1 ) should be considered individual radiative constants of main band and exciplex band ensembles (small piece of a broad main peak band is non-zero at 605 nm).Observable changes in position and dispersion of exciplex radiative constant distribution allows monitoring of water-driven resonance quenching.
As shown by Maillard et.al., [26] sensitivity towards water content, realized via overlap of donor luminescence and water absorption band may be effectively utilized in practice (some harmonics of a set of water -OH vibrations are situated in a visible wavelength range).But unlike for fluorophores discussed elsewhere, [26,27] for this molecule we have no need to shift luminescence batochromically since exciplex already possesses sufficient spectral overlap.Moreover, sensing may be considered ratiometric since we both observe concentration-related main / exciplex band ratio and distribution of radiative decay constants for exciplex.Decays at exciplex peak observed for series with varied BODIPY-C18 concentration at 50 Vol.%water / acetonitrile ratio were extracted utilizing the same technique (Figure S6).At lowest concentration we observe pronounced bimodality.Here, 0.43 ns −1 -centered constant is related to ineffective waterdriven exciplex resonant quenching, whereas second distribution is also related to insignificant portion of monomeric dye, indicating incomplete aggregation at this concentration.Subsequent distributions in series are centered around 1.12 ns −1 with lower observed variance as compared to that found in high water content systems.This shows that effect of aggregation induced via antisolvent treatment and concentration increase is different.Fit results are summarized in Table 1(a).
System with 50% water / acetonitrile ratio and 0.4•10 −5 M BODIPY-C18 has highest ambiguities in distributions.Just as found elsewhere for Lorentzian lifetime distributions, [28] systems with highest heterogeneity have broader distributions whereas endpoints provide data which is easier to describe and interpret.
We assume from both experimental series that distribution with mode 1.1 -1.2 ns −1 corresponds to exciplex decay influenced by water electronic-vibrational resonance transfer process.Decreased variance for this component in higher dye concentrations, unachievable with increase in water ratio indicates that particles are more uniform in higher dye concentrations.Interestingly, we clearly see that water-sensitive response is barely dependent on BODIPY-C18 concentration above certain threshold.
Gamma-distribution of lifetimes is a result of direct integration of a stretched exponential function, [29] yet our findings clearly demonstrate that there is no obligatory backward connection.
Namely, distribution accommodating a Gamma-shape not always could be attributed to a stretched exponential function.It was discussed elsewhere that Gamma-distribution indeed satisfies a maximum informational entropy criterion.[30] Moreover, Gamma shape, as stated by the authors, leads to powerlike model, different from stretched exponential function as it bears the so-called heterogeneity parameter, not discussed here in any more details.
Sum gamma-distribution approach allowed us to process all of the exciplex decays in order to extract deep and demonstrative information about excited state processes.Sum of selected distributions, however, fails at the points where radiative constant distribution do not resemble the sum of two gamma functions, suffering from mathematical noise.Due to this reason, approach was unapplicable to monomer band decay and is thus was generally insufficient for investigated system.
More informative results were obtained utilizing free-form distribution with maximum entropy criterion (see Experimental Section), as it requires no prior assumption of a distribution shape or number of modes.
From changes in distributions found in the top group at Figure 4 (and in Table 1(b)) obtained from experiment with water concentration variation, we could directly see advantage of a MEM procedure.As there is no implicit distribution shape, graph consists solely of peaks indicating single lifetimes in low water concentration samples.Then, after 45Vol.% of water, as system becomes more complicated, lifetime distributions are extracted.This has a proper direct explanation: until that point observed response mainly consists of a monomer decay, whereas broad distributions after 45Vol.%are solely defined Interestingly, upon water content variation up from 50%, distribution shape and position remains basically intact.
Bottom group of graphs at Figure 4 demonstrates changes happening upon Reconvolution procedures performed for a main peak decay (Figure S7, S8) are also informative, we see sharp shift towards higher lifetimes in distributions upon increase in water content above 45Vol.%.In BODIPY concentration experiment we observe monotonous shift of peak of a long component distribution.With MEM approach, quantitative analysis of redistribution of relative abundances of short and long lifetimes is finally available.
On a monomer band we could directly see that both water content and concentration increase lead to monotonous increase in a short lifetime relative amplitude.This justifies the nature of the fast component as the one responsible for a rapid resonance transfer from a monomeric dye to excited species.Efficiency of a process is defined both by concentration of water and the dye, logically involving both of them in the observable distribution.
A fluorescence decay analyzed in this way becomes a multidimensional tool and this is especially important when dealing with complicated cases, such as investigations of aggregated fluorophores.
We found that position of a distribution median at exciplex peak could be utilized to assess local BODIPY-C18 concentration up to a value of 0.8•10 −5 M, which is well beyond the required value for biological investigations.At the same time, position of a long lifetime component in a main peak is linearly correlated to water concentration and thus could be utilized to assess solvent polarity.

Langmuir floating layers
Another approach to studies of aggregation processes is organization via external stimulus.Floating layers on an air-water interface, while compressed, serve as an excellent model of gradual organization process, often yielding organizational degree unachievable in any other method, leading to interesting properties.[31] [32] Upon BODIPY-C18 compression, first surface pressure rise happens at trough area corresponding to 140-150 Å2 /molecule.This value is sufficiently small, considering the revolution circle of a bottom projection of the molecule is slightly above 140 Å2 .As indicated by a layer compressibility function [Equation 2], collapse happens at the isotherm bend at a relatively low surface pressure of 3.7 mN/m and area of 120 Å2 /m.
Where π is a surface pressure and A is a mean molecular area.
We have shown that this pressure could be increased by dilution of a layer with a stearic acid.Guest-host floating layers with increasing amount of fatty acid show increase of a surface pressure at the isotherm bend point and increase of the peak layer compressibility (Figure 5).The latter does not fall below zero for 1:3 stearic acid layer.However, changes in luminescence and UV-Vis absorption in host-guest system after collapse are close to a pure layer (discussed later).This emphasizes fact of low co-organization happening between host and guest.Perhaps the acid exists as a separate layer, whereas chromophore units are still prone to interactions with water and between each other (at a corresponding compression).Aggregated floating layer is still elastic after the negative compressibility region, but pronounced hysteresis in multiple compression-relaxation isotherms shows that once aggregates are formed, structure does not relax anymore.This may be described in two different ways -molecules may quickly form 2D rafts at a limiting compression or may form 3D structures.Consistently observed raise in isotherm after 80 Å2 /m corresponding to rectangle of a molecule projection, suggesting the raft formation variant to be more probable.
Floating layers of BODIPY-C18 demonstrate high molecular extinction with absorption shifted by 10 nm as compared to diluted chloroform solution (Figure 6).UV-Vis absorption band shape of a floating layer resemble that of a monomeric dye all the way through compression process (as assumed from the Young modulus).Compression after the bend in isotherm leads to fast emergence of 565 nm high extinction batochromically shifted band, probably devoted to J-aggregate species.Normalization by sum intensity of two peaks transforms graphs to the form with isobestic point, further approving that two bands are related and 1 to 1 conversion is happening.Interestingly, J-aggregate band was never observed in experiments in solution, approving that Langmuir technique is a proper choice for our examinations.
For demonstrative purposes, we also plotted in gray a ratiometric curve of 570 and 535 nm bands to see if monomer / aggregate ratio actually changes all along the compression process.This relative absorption plot saturates at 80 Å2 /m, which mathematically means that absorption intensity of both bands start to grow proportionally from this point.This approves our raft formation assumption, as this is the point where highest density of flat pieces would have been achieved, leading to slipping of rafts on top of each other afterwards.This slipping leads to observed linear proportional growth of both intensities due to layer thickness growth.Also, in solution experiments we were able to see that 3D structures do not produce sharp separate aggregate peaks, thus bulk aggregation may only happen to a small extent.
Layer is weakly luminescent in a gaseous and liquid-expanded state with almost completely quenched main peak at 537nm and 605nm exciplex maximum.Band shape in this case is identical to the one observed in water / acetonitrile system (Figure 7).After bend in isotherm, fluorescence also has two pronounced maxima (Figure S9) completely different from those observed in binary mixture experiment.
Interestingly, ratio between these two maxima upon compression remains unchanged, so they should be attributed to same fluorescent species.Intensity growth is caused by increase in their surface concentration during compression.J-dimer emission is the most probable possibility, and since geometrical constraints during compression do not influence band ratio, they should be attributed to two different electronic transitions, rather than vibronic pecularity.After bend in isotherm J-dimer formation prevents BODIPY-C18 from excited-state interactions with water and leads to a rise in excimer formation processes.Fluorescence intensity rise saturates at 95 Å2 /m, and then starts to grow linearly again after 80 Å2 /molecule, for the ultimate approval of our assumption on aggregated rafts formation.
System of the investigated dye is exceptional in terms of general observations of Langmuir monolayers.Lipophilic luminophores extremely rarely possess bright emission while situated on a water subphase.For BODIPY-C18, we could observe orange emission of a monomolecular layer with a naked eye (Figure S10).It's possible due to equilibrium between two kinds of excited state species.Once molecules are pushed towards each other below certain critical distance, dye molecules do not interact to water molecules in excited state anymore.
We make an important remark here: our current investigations are aimed at further description of the system beyond the notch in isotherm, because even though we have nice spectroscopic evidences to our assumptions, we can't explicitly state that J-aggregates exist in separate rafts rather than 3D structures until we find proper way to examine these species directly.

Conclusion
Utilizing time-resolved techniques we were able to distinguish between homoand hetero-molecular weak interaction effects.We found a suitable model to describe system to the quality, previously hardly achievable for complicated excited state dynamics of aggregates (Gamma distribution sum).Over that, we found out that free form distribution with maximum enthropy criterion is the best out of a broad variety of methods for examination of complex decays.MEM allowed us to recover results, providing multidimensional control possibilities, promising a future development to examinations of fluorophores in a complicated media such as biological systems.
We did a generalization assuming the fluorescence band in a floating layer on water subphase to be caused by exciplex formation (in pseudo-gaseous and liquid-expanded state).We also showed limits of this generallization.System only remains similar to antisolvent system until compression reaches certain limit.Existence of liquid-expanded state of such an aggregation-labile molecule thus mainly possible due to competition between dye-water and homomolecular dye-dye diffusion interactions.Only at the point of compression isotherm where molecules are pushed beyond critical interaction range, aggregation takes place and excimer formation becomes predominant process.This equilibrium is also the reason why aggregation happens almost instantly, and why the Young modulus drops to negative values.Pressure overshoot is caused by lack of time for structural relaxation.Layer compressibility overcomes zero upon further compression at the point where size reaches that of a flat rectangle of a bottom edge of BODIPY-C18 ≈80 Å2 .
We found that generalization between solvent systems and aggregation in interface may possibly lead to erroneous results, even though it is often found in literature.This further emphasizes importance of extended investigations for each self-assembling system to cover.
Interestingly, investigated dye has no pronounced diphilic nature or extended aromatic system as most self-assembling systems do.Via altering our approach we obtained technically desirable J-type aggregates for a dye demonstrating no hallmarks of J-aggregate formation in any solution experiment.
"Soft" motif of geometric anisotropy, in a meantime, may be exact reason for precise tunability of photophysical properties and should be treated accordingly as an approach in design of smart fluorescent materials.

Synthesis
All of the reagents for synthesis were obtained from Sigma-Aldrich ("for synthesis" grade) and used without additonal purification.Solvents for synthesis were obtained from EKOS Russia ("chemically pure" grade) and used without further purification.
4-octadecyloxyphenyl-benzaldehyde was synthesized utilizing Williamson ether synthesis process from octadecyl bromide and 4-hydroxybenzaldehyde (0.1 eq.excess) in DMF with potassium carbonate as a base.Resulting yellowish liquid was diluted with DCM and washed with water to remove any aldehyde residue, and then multiple times triturated with DCM, frozen to -40 • C and evacuated during warm-up to the room temperature to remove any DMF traces.Procedure yields white crystalline powder quantitatively.DPM-C18 ligand was synthesized utilizing the following technique (Figure 8): two equivalents of pyrrole react with aldehyde equivalent in an inert atmosphere in ethanol with HBr aqueous solution to yield dipyrromethene hydrobromide DPM-C18 directly, overnight, avoiding the quinone-driven oxidation stage.DPM-C18 is then isolated on neutral aluminium oxide column with hexane-DCM column with addition of ethanol (10:1:0 gradually to 100:10:

Spectroscopic investigations
Solvents for spectroscopic measurements were obtained from MERCK (HPLC grade) and used without further purification.
UV-Vis absorption was measured on SF-104 spectrometer (Akvilon) and with modular fiber optic system consisting of AvaLight-DHC light source and AvaSpec-ULS2048CL-EVO with qpod 2e sample mount, fluorescence was measured with Varian Cary Eclipse fluorescence spectrometer.
To avoid dynamic range error and inner filter effects, light path length for UV-Vis absorption was decreased from 10 mm to 5 mm and then 1 mm upon concentration increase in a way that absorption never exceeded value of 1 in the scale of optical density, and fluorescence spectra were measured in reflection mode for the same reason.
UV-Vis absorbance and fluorescence of higher BODIPY-C18 concentration in CHCl 3 was measured utilizing custom 50 µm glass cuvette.It was fabricated by placing a 50 µm feeler gauge between two standard 1 mm glass slides, putting cyanoacrylate glue into capillary cavity around the gauge and removal of the gauge after drying.This method does not allow for precise extinction evaluation, since only single-use cuvettes could be fabricated this way, and there exists hardly measurable deviation both in flat capillary thicknesses and an overall cell thicknesses.Though this method was found to suit perfectly for qualitative examination of high concentration solutions to determine the band shapes.

Fluorescence decay measurement and processing
Fluorescence decay profiles were measured using PicoQuant FluoTime 300 setup with PLS500 laser (480 nm excitation mode).For most of the examinations polarizers were set to magic angle conditions, except for anisotropy measurements where required polarization regimes were utilized.
Mono-and biexponential fits were performed utilizing EasyTau 2 software.For custom distribution analysis we used custom python scripts.There, numpy package was utilized for handling mathematics, [34] Matplotlib package for visualization, [35] numpy/scipy fitting procedures as wrapped in lmfit package for the reconvolution process itself (the one providing sophisticated wrapping classes for multiple model trials) and quadpy package for handling numerical integration where required [36] [37] Considering our assumption on particle formation, it was reasonable to attempt to fit decays to stretched exponentials of form [Equation 3].They are often utilized to describe exciton transport distributions in investigations of conjugated polymers and crystalline solids.[38] And serve well to the purpose of complicated decay description in biological media.[29] Our system, however, did not show any good chi-squared values even after we modified the model to accommodate several stretched exponentials and additional monoexponents.
Somewhat universal approach towards transient effects description (including resonance energy transfer) is comprised by models in [Equation 4,Equation 5].
It was possible to obtain reasonable results for our system with variation in parameters, but no fit is possible if parameters were fixed across decays, probably due to the fact that there were no uniform distribution in sample volume.[39] In resonance energy transfer description model [Equation 4], coefficient b describes concentration and availability of donor species, whereas β should be fixed to discrete values depending on the assumption of RET process properties, such as β = 2 in case of FRET β = 6, 3, 0.5 for systems with three-two-or one dimensional transfer process happening.
Equation [Equation 5] is a generalized form for description of systems where exciton recombination is hindered, this also may have been the case to a certain extent, but this equation has degenerate parameters d and s, Hausdorf and spectral dimensionalities, respectively, which are different measures of self-similarity of a material network, this model (percolation model) assumes there are traces which acceptor has to travel towards donor for effective quenching process.[40] Here we refrained from these two quite desirable possibilities for system analysis.Unlike in work of Levitz et.al., [41] where carrier matrix is similar across decays (such that d/s or one of parameters could be fixed) providing rationales for their use in diffusion-hindered systems, systems of pore-adsorbed fluorophores (percolation model) and systems with limited degrees of freedom for RET processes, we assume our systems to be barely organized or spatially constrained.To avoid overinterpretation, we discarded these models until organization could be explicitly stated.Moreover, as seen from careful examination of Equation 3, Equation 4, Equation 5, mentioned models mathematically accommodate each other at certain values of parameters whereas originally serve description of different processes.
We turned to a soft description procedure, the one not really yielding exact parameters for analysis while still providing insight into processes influencing luminescence decay.Free-form distribution analysis proved itself effective in biophysics, where systems imply interaction range distributions.[42][43].Lifetime distributions also are known to handle consequences of aggregation processes.[44] We attempted fitting our systems to different combinations of Lorentz and Gauss distributions of exponential decays with low success, [28][45]to find out that two Gamma-distributions [Equation 6] of radiative constants describe our system in a well-acceptable manner without putting too much restraints or pushing towards restrictive assumptions on organizational properties.
We found it convenient and intuitive to plot radiative constant distribution in form derived elsewhere.[43] This model left us with first insightful data as stated in the Results and Discussion.Though ambiguousity of the results for some of the decays led us further towards more mathematically pure approach.
Maximum informational entropy method (MEM) is a known way to describe extensively complicated decay distributions.[46] Superiority of this method was shown on complicated cases, e.g. for description of changes of tryptophan environment heterogeneity during protein conformational changes and even as a viable tool for fluorescent response-based qualitative analysis of cell cultures.[47][48] Unlike for other approaches, in MEM there is no preliminary assumption on a distribution kind and thus the method produces completely uncorrelated results without "dangling" small intensity components in case two or more distributions are supposed to join.[49] The MEM, despite its outstanding possibilities is not widespread enough, not found in most time-correlated single photon counting (TCSPC) processing software and, to the best of our knowledge, was never implemented in an opensource format, either available as a proprietary software script or in a form of purely mathematical derivations.
We took our effort to develop our own implementation of maximum entropy method using python and the aforementioned mathematical packages.MEM aims towards maximization of Shannon-Jaynes informational entropy [Equation 7] along with normalized amplitudes and chi-squared constrained to unity.Mathematically, the optimization problem was formulated as in [Equation 8], which is a good candidate for a constrained sequential least-squares quadratic programming approach.

S(p) =
where p i is a probability of the i th discrete parameter, which in this specific case is a normalized amplitude of an i th lifetime.

Minimize
Where BKG and β parameters were allowed to be adjusted during optimization to account for background level shift and overall decay scaling, accordingly.Each monoexponential decay was convolved with instrument response function IRF as indicated by the ⊗ sign.
The whole decay was convolved to avoid artifacts at a convolution range borders, but only certain part of decay was included in a χ 2 test, so that a rise time is neglected.Test range beginning was at the highest point of curve for 605 nm decay analysis and lagged for 500ps for 537 nm decay analysis (to discard fast part driving computation unstable).
Test range end was chosen automatically to either be at 30 ns or at the point where less than 0.1% of counts were obtained.Data points were upsampled via linear interpolation with added Poisson noise for improved numerical stability and then subjected to optimization (6) with 100 lifetimes linearly spaced between 0.001 and 10 ns.
Linear spacing was chosen unlike in reference because of the nature of the problem (that way we were able to extract more information).[46] Open-source code for MEM analysis is available from GitHub, sergeyusoltsev/lampMEM/.

Langmuir monolayer investigation
Kibron Microtrough G2 with a surface area of 280 cm 2 was utilized for floating layer studies.Langmuir trough is made of hydrophobic Teflon to maintain subphase meniscus and prevent interaction to deposited specimen.Trough barriers are made of hydrophilic polyacetal to allow for compression process.Surface pressure was calculated from surface tension as measured with Wilhelmy microbalance with platinum plate.Water for subphase was bidistilled and deionized, to achieve (at least) 18MΩcm resistivity.Experiments were performed in the ambient conditions (in air atmosphere at 20±1 • C) BODIPY-C18 was deposited as 130µl of CHCl 3 solution with 4.45•10 −4 mol/L concentration on a deionized water subphase.
UV-Vis absorption spectra of floating layers were measured using an AvaSpec-2048 optic fiber spectrophotometer equipped with deuterium-halogen light source AvaLight-DHC (Avantes) by placing a reflectometry probe (400µm fiber, six irradiating and one measuring fiber combined) orthogonally to a water subphrase over a deep well in the center of a Langmuir trough (to quench mirror reflections from white surface of a trough) at a distance of 2-3 mm from the monolayer.
Reference signal for spectra was taken from an empty subphase (setup scheme in Figure S1).This method allowed us to measure amplified monolayer reflectance as described elsewhere, [19] so graphed absorption spectra for convenient understanding were all multiplied by negative unity.Luminescence of a floating layer was measured by placing the reflectometry probe over a water subphase in the same position as for absorbtion, yet deuterium-halogen irradiation was turned off and fluorescence was measured in a diffuse reflectivity mode as excited with collimated beam from monochromatic light source, again referenced to pure water subphase diffuse reflection signal.
Lightsource was aimed at a projection point of reflectometry probe optical axis and collimated for suitable diffuse spot as described elsewhere.[19] 5. Acknowledgements Authors are grateful to Ilya A. Khodov for NMR investigation of the compound.

Conflict of Interest
Authors declare no conflicts of interests.

Author Contributions
SU synthesized the object and performed experimental procedures, wrote software, analysed, conceptualized and visualized data, and wrote an original draft, OR was managing, providing resources and methodology for the part with Langmuir monolayer investigations, AS developed methodology and supervised part of the project with Langmuir monolayer investigations, YM was managing and providing resources for the part comprising solution and antisolvent system investigations, and supervised original draft writing.

Figure 2 :
Figure 2: Time resolved emission scan: left -plotted as a heatmap, right -plotted as a cut series in selected timespan.

Figure 3 :
Figure 3: Left: Changes in excited state lifetime amplitudes corresponding to lifetimes extracted from each exponential in reconvolution with fixed lifetimes and average amplitudeweighted lifetime τav Amp in series of solvent content variation.Right: Changes in excited state anisotropy amplitudes for extracted correlation times in series of solvent content variation

Figure 4 :
Figure 4: Results of MEM reconvolution procedure: top group -decays measured at 605 nm upon water content variation on the left and recovered distributions of 100 lifetimes for each correspondingly colored decay on the right.bottom group -decays measured at 605 nm upon BODIPY-C18 concentration variation on the left and recovered distributions on the right.Wide graphs (as labeled) indicate residuals and autocorrelation function for each group.Insets are top view of distributions for ease of perception

Figure 5 :
Figure 5: Left (a): dimensions of the molecule in various projections measured according to Van der Waals radii calculated in GFN2-xTB level of theory.[33]Right (b) Characteristic compression isotherms for BODIPY-C18 with stearic acid with compressibility plotted in dashed lines of corresponding color

Figure 6 :
Figure 6: Left graph (a): changes in UV/VIS absorption spectra upon layer compression, with two maxima indicated by colored dashed vertical lines (green for 535 nm and red for 570 nm) normalized to sum of two maxima.Right graph (b): compression isotherm in black solid line with colored points indicating UV/Vis absorption intensities at maxima (green for 535 nm and red for 570 nm).Solid grey line depicts ratio of two peak intensities along the compression process.

Figure 7 :
Figure 7: Characteristic normalized absorption (left) and fluorescence (right) spectra of the dye in various investigated systems.
This work was supported by the Grant of the Russian Foundation for Basic Research (Grant No 20-33-90198).Some of the examinations were carried out using the resources of the Center for Shared Use of Scientific Equipment of the ISUCT (with the support of the Ministry of Science and Higher Education of Russia, grant No. 075-15-2021-671).YM is grateful to the Russian President grant board for funding part of the work comprising steady-state luminescence and absorbance investigations (Project MD-2300.2022.1.37).

Table 1 :
Left (a): Result of two Γ-distributions fit to decays at 605 nm., top -series with varied water ratio, bottom -series with varied BODIPY-C18 concentrations (mol/L) in 50 Vol.%H 2 O mixture.Right (b): Result of MEM analysis for both monomer and exciplex peaks, top -series with varied water ratio, bottom -series with varied BODIPY-C18 concentrations (mol/L) in 50 Vol.%H 2 O mixture