Elucidation of surface-ligand interactions in inorganic lead halide perovskite nanocrystals with vibrational spectroscopy

(1)

Elucidation of surface-ligand interactions in inorganic lead

halide perovskite nanocrystals with vibrational spectroscopy

Benthe Vermaas

Abstract

Colloidally synthesized lead halide perovskite nanocrystals (LHP NCs) are promising materials for the use in electronic and optoelectronic devices. They are characterized by their tunable bandgap energy, high quantum yields and versatile surface chemistry. Due to their large surface-to-volume ratio, the role of the surface increases, eventually becoming dominant. Yet their surface chemistry remains rather unexplored. To elucidate the surface-ligand interactions in inorganic LHP NCs, vibrational spectroscopy was used. More specifically Fourier Transform Infrared Spectroscopy (FTIR) was used to study the vibrational modes of ligands. I studied the FTIR spectra of free ligands, ligand-metal complexes, and LHP NCs during progressive synthesis and purification steps. FTIR spectroscopy has shown to be an effective method to distinguish byproducts, synthesis products and free ligands. The sudden appearance of cesium oleate and oleic acid might indicate the detachment of ligands from the NC surface and the disintegration of the NC. Similarly shifting peaks for different halides in LHP NCs might indicate a halide independent conformational or intermolecular interaction change.

Report Bachelor Project Physics and Astronomy

Size 15 EC

Conducted between 03-2018 - 07-2018

Student number 11026170

Name of institute Hybrid Solar Energy Conversion

Name of University University of Amsterdam and

VU Amsterdam

Name of faculty Faculty of Science

Date of submission August 11, 2018

Supervisor Elizabeth von Hauff

(2)

Dutch popular scientific abstract

(Opto-)elektronische apparaten zijn niet meer weg te denken uit het dagelijks leven. Of het nou gaat om de mobiele telefoon, de computer of de tv: de kans is groot dat jij niet meer

zonder kan. En we willen steeds meer. Groter, sneller, flexibeler, maar ook steeds vaker duurzamer. Helaas wordt de vooruitgang van de ontwikkeling van deze apparaten geremd door problemen met hoge fabricatie kosten en niet effici¨ente en/of flexibele onderdelen. Het gebruik van nanokristallen kan hier de oplossing voor zijn. De productie van nanokristallen

is vaak eenvoudig en goedkoop en levert apparaten op met kwalitatief hoge elektronische eigenschappen. We moeten hiervoor nog wel een probleem oplossen: ze zijn niet erg stabiel en je wilt natuurlijk wel dat je een tijdje door kan met je nieuwe telefoon. Om dit probleem

op te lossen zullen we eerst meer te weten moeten komen over de interacties die er voor zorgen dat het nanokristal stabiel blijft. Daarom zal dit onderzoek gaan over de interacties

(3)

1 Introduction

In the 1990s, lead halide perovskites (LHPs) received significant attention in an optoelec-tronic context. LHPs were investigated as active layers in light-emitting diodes or as channel layers for field-effect transistors [1]. Later, LHPs came into the spotlight in the research field of photovoltaic devices. Since then the power conversion efficiencies of LHPs as thin-film absorber layers has increased from <10% up to >22% [2]. Present day, a promising type of LHP materials is colloidally synthesized nanometer-sized crystals (NCs). LHP NCs have very favorable quantum-size effects which enables size-tunability of their bandgap energies through the entire visible spectral region. Furthermore, they have high quantum yields of 50-90%, photoluminescence (PL) characterized by a narrow emission line width and versatile surface chemistry [3]. Moreover, the synthesis is low-cost and colloidal NCs can easily be processed in solution which makes the use of cheap and fast methods like printing, spraying and doctor blading accessible.

However, with shrinking size comes a greater surface-to-volume ratio and hence the role of the surface increases, eventually becoming dominant. The surface chemistry of LHP NCs remains rather unexplored, causing difficulties in remaining their bright PL, colloidal stability and structural integrity [4]. Surface capping ligands allow control over the colloidal stability, growth and aggregation of the NCs necessary for synthesis, processing and some applications. More generally, they alter their electronic structure by engaging with dangling bonds and thus reducing recombination [5], [6]. Since the raw synthesis of the LHP NCs not only contains ligands attached to the surface, but also unwanted free ligands and synthesis products, it is purified with several purification steps to obtain the best LHP NC solution.

To shed light on the ligand nanocrystal binding mechanisms, this research will focus on the elucidation of surface-ligand interactions in inorganic lead halide perovskite nanocrystals with the use of vibrational spectroscopy. To achieve this, it was necessary to distinguish ligands bound to nanocrystals from synthesis products, including free ligands and ligand-metal complexes. The purification steps of CsPbBr3 and CsPbBr2I will be investigated for

the presence or absence of these free ligands and synthesis products. Furthermore, both LHP NCs with different halides will be compared.

In this research, Fourier Transform Infrared (FTIR) spectroscopy is used to obtain vibra-tional spectra of synthesis products and that of the purification steps of the LHP NCs. As this research is the successor of Simion Beckers An analysis of the purification of CsPbBr3

nanocrystals with FTIR spectroscopy, his data of the ligands is used. For all substances, spectral markers were found to track their presence in the purification steps. The purification steps of CsPbBr3 and CsPbBr2I were compared to try to elucidate how the halide is involved

(5)

2 Theoretical background

2.1 Lead halide perovskite nanocrystals

Nanocrystals are nanometer sized semiconducting materials that have discrete electronic states. The optoelectronic properties of NCs change with their size and shape. The fact that these NCs are smaller that two times the bulk exciton Bohr radius leads to quantum confinement. Hence, the energy levels can be predicted by the particle in a box model where the energy depends on the size of the box. This means that larger NCs have lower electronic energies and thus a redder absorption and photoluminescence (PL) spectrum [7].

More specifically, LHP NCs have the perovskite structure as shown in figure 1. In general, perovskites have the ABX3 structure, where ’A’ and ’B’ are differently sized cations and ’X’

is an anion that binds to both cations. In this research, the focus will be laid upon the perovskites with A = cesium, B = lead and X3 can either be Br3 or Br2I.

Figure 1: Perovskite structure. Perovskites have the ABX3 structure, where ’A’ and ’B’ are

differently sized cations and ’X’ is an anion that binds to both cations.

Due to their size, NCs have a relatively large amount of surface. The crystal lattice terminates at the surface which leads to a high defect concentration. LHPs like CsPbX3

have shown to have a high defect tolerance compared to conventional NCs such as CdSe and InP. Although defects are abundant in LHP NCs because of their low formation energy, they are benign with respect to the optical and electronic properties of the NCs. As an example, in CdSe localized weakly bonding or non-bonding orbitals are present as a result of the displacement or removal of a Cd ion. These orbitals reside within the bandgap and act as

(6)

trap states, as shown in figure 2 [8]. The difference between conventional NCs and LHP NCs is that for conventional NCs the bandgap is formed between the bonding and antibonding states but in LHPs the bandgap is formed between two sets of antibonding orbitals. As a result, the vacancies form states residing within the valance- and conduction band [9]. Hence, these defects are benign towards optical and electronic properties because they do not act as trap states. This enables bright PL and high performance nanoparticle films for optoelectronic devices without surface passivation, which is an absolute necessity for conventional NCs [10].

Figure 2: Defect tolerance in conventional- and LHP NCs. Schematic comparison of con-ventional defect-intolerant NCs (such as CdSe and GaAs) and defect tolerant LHP NCs. Defects in defect-tolerant NCs do not act like trap states and hence are benign towards their optical and electronic characteristics [10].

A major difficulty in this field is the retention of the colloidal state and structural integrity. LHP NCs have certain characteristics which decrease the long-term stability. First, LHPs are soluble in polar solvents. This makes them easily processable into thin films for perovskite PVs. However, this solubility has a negative impact on the long-term stability of the NCs. Second, the internal bonding in LHP NCs and the NC-ligands binding are highly ionic, and the NC-ligand binding has shown to be highly dynamic in solutions [4]. As a result, LHP NCs have a relatively fast ligand desorption in contrast to that of conventional NCs leading to a decreasing colloidal state and structural integrity. Moreover, LHP NCs are vulnerable for a combination of light, moisture and oxygen.

(7)

2.2 Ligands

Surface ligands enable the stabilization of NCs in the form of colloids during synthesis and prevent them against precipitation [11]. Furthermore, they dictate the dispersibility of NCs in different solvents, which plays a great role in the fabrication of multilayered structures [6]. Ligands can also be engaged in a strong chemical bond with dangling bonds at the surface and hence altering the surface energy state from inside to outside the bandgap. For example, to increase the PL and power conversion efficiency of solar cells, surface ligand chemistry can be modified to passivate electronic states, as was done with PbS NCs [12]. In addition, ligands are used for better size control, improved surface passivation and a narrower size distribution of the NC [13].

Figure 3: The covalent bond classification. L-, X- and Z-type ligands contribute respectively 2, 1 or 0 electrons to the metal-ligand bond. Ex-amples of binding motifs of common nanocrystals are shown [4].

The binding of ligands to the surface of the NCs can be described by the Covalent Bond Classification by Green [14],[15]. In this classification system, the number of elec-trons that the neutral ligand contributes to the metal-ligand bond defines the type of lig-and, namely L-, X- or Z-type for respectively 2, 1 or 0 electrons, as shown in figure 3. Al-though the binding mechanisms of ligands to LHP NCs are not fully understood yet, find-ings of De Roo et al. indicate that CsPbX3

NCs are terminated by pairs of X-type lig-ands, yielding a NC(X)2 binding motif [4].

The length of the hydrocarbon chains of the ligands controls the distance between the NCs and hence the ability of electron transfer between NCs. Thus the choice of ligand is dependent on the final use of the NCs. [7]

For the synthesis of CsPbX3, a

combina-tion of oleic acid and oleylamine has been advocated as ligands because of their affinity to metals through their acid and amine anchor groups and their long non-polar carbon chain [16]. Figure 4 shows the chemical structure of oleylamine and oleic acid.

According to J. de Roo et al. there are multiple ways that ligands can attach to the NC surface. Figure 5 shows a schematic drawing of how the ligands attach to the CsPbBr3surface

(8)

Figure 4: Oleic acid and oleylamine. A schematic drawing of the chemical structure of the ligands oleic acid and oleylamine [17].

First, oleic acid can be deprotonated by oleylamine and form oleylammonium oleate. Oleate is binding to the NC surface as an ion pair with oleylammonium resulting in a NC(X)2

binding motif. Second, and also a NC(X)2 binding motif, there is an acid/base equilibrium

of oleylamine with hydrogen bromide which binds to the surface as oleylammonium bromide. Last, in its unprotonated state, oleylamine could bind as a L type ligand to the surface cations. The NC surface ligand interaction of CsPbBr2I is similar to that of CsPbBr3.

Figure 5: CsPbBr3surface ligand binding. Schematic illustration of oleylamine, oleylammonium

oleate and oleylammonium bromide stabilizing the dynamic surface of CsPbBr3. Associated acid/base

(9)

2.3 CsPbX3 nanocrystal synthesis and byproducts

CsPbX3 NCs are synthesized by reacting cesium oleate [Cs(OOCR)] with lead halide (PbX2)

salts in boiling octadecene at 140 - 200 ◦C, depending on the preferred size. Then, a 1:1 mixture of oleic acid (HOOCR) and oleylamine (RNH2) is added to colloidally stabilize the

NCs and to solubilize PbX2. Mixed halides like CsPbBr2I can be synthesized by combining

appropriate ratios of the PbX2 salts [3].

To shed light on this complex reaction mixture and hence its possible byproducts, it is convenient to have a closer look at the underlying reaction mechanism. For simplification, only CsPbBr3 is considered here. The overall perovskite formation reaction can be written as

2Cs(OOCR) + 3PbBr2−→ 2CsPbBr3+ Pb(OOCR)2. (1)

Pb(OOCR)2is an obvious byproduct which one cannot get rid of with these reagents. Cs(OOCR)

is the limiting reagent and thus not expected to be present in the final product. Although this reaction correctly describes the full reaction, octadecene alone cannot solubilize PbBr2

and hence another reaction takes place described by

PbBr2+ xHOOCR + xRNH2 −→ PbBr2−x(OOCR)x+ xRNH3Br, (2)

where x can be one or two. A mixture of HOOCR and RNH2 help colloidaly stabilize the

NCs and solubilize PbBr2 and form a mixture of PbBr2−x(OOCR)xand RNH3Br. Therefore,

after formation the mixture does not only contain CsPbX3 NCs, but also the byproducts

PbBr2−x(OOCR)x and RNH3Br, free ligands RNH2 and HOOCR and solvent octadecene.

Accordingly, the CsPbBr2I NC mixture can also contain PbI2−x(OOCR)_x [4].

2.4 Purification process

After synthesis, the crude solution needs to be purified to get rid of the unwanted byproducts, free ligands and solvent octadecene. The purification of the crude solution is divided into seven steps, as illustrated by figure 6. The unwashed synthesized product is named raw synthesis (RS), which is then centrifuged without adding a solvent. The first supernatant (S1) and precipitate (P1) are separated. Methyl acetate is added to the first precipitate and again centrifuged. Then, the second supernatant (S2) and precipitated (P2) are separated. Once more, methyl acetate is added to the second precipitate and then centrifuged. When the third precipitate is separated from the supernatant(S3), the final product (FP) is obtained.

(10)

Figure 6: Purification process of CsPbX3 NCs. Schematic illustration and naming of the

purification procedure for CsPbX3 NCs. Raw synthesis and precipitates are centrifuged and washed

with methyl acetate and precipitates and supernatants are separated until the final product.

2.5 Vibrational spectroscopy

Infrared and Raman spectroscopy give a fundamental insight of the characteristic vibrational energies of a material, from molecular vibrations to phonon modes. To completely analyze the vibrational energies of a material, often both Raman and IR spectroscopy are necessary since they are complementary techniques. Although some modes are both Raman and IR active, two different physical mechanisms lead to a Raman and IR response. They also differ from a practical point of view. Raman scattering is an inelastic form of scattering and far less probable than the elastic Rayleigh scattering, the intensity is observed to be about 10−6 of the incident light for strong Raman scattering. Hence, Raman signals are very weak, but Raman spectroscopy is easier to perform experimentally than IR spectroscopy. IR signals are stronger, but special substrates and water free environments are required to obtain a reliable spectrum.

2.5.1 Vibrational spectroscopy: Raman and IR

Light incident on a material may be transmitted, reflected, absorbed or scattered. Light that interacts with vibrational modes of a material may be either absorbed or scattered.

(11)

These reactions of radiation with vibrational modes lead to an IR or Raman response. How the photon energy changes the vibrational state is different for IR and Raman. The physical mechanism that leads to an IR response is the transition between molecular vibrational energy levels as a result of the absorption of IR radiation. In order for energy to be absorbed by the molecule, the molecular vibration must change the electric dipole moment of a vibrational mode. Thus if vibrational modes do not have a dipole moment, they are not IR active. The intensity of the IR band that is measured is proportional to the square of the change in electric dipole moment [18].

The physical mechanism that leads to a Raman response is a two-photon inelastic light-scattering event. The energy of the incident photon is larger than the vibrational quantum energy. The incident photon loses part of its energy to the molecular vibration and scatters ei-ther with a reduced frequency which is called Stokes scattering or eiei-ther (but far less probable) with increased frequency which is called anti-Stokes scattering. This interaction between light and matter is an off-resonance condition involving the polarizability of the molecular modes. The polarizability of a vibrational mode is its ability to deform the electron cloud when an external electric field is applied. Incident electromagnetic (EM) radiation induces oscillation dipoles as a result of the polarizability which in turn produces EM radiation. Thus in order for Raman scattering to occur, the molecular vibration must cause a change in polarizability of the vibrational mode.

As a result, Raman spectroscopy can be used best for symmetric or in-phase vibrations and non-polar groups, and asymmetric or out-of-phase vibrations and polar groups are most easily studied by IR [18].

2.5.2 Probing and identifying chemical species with vibrational spectroscopy The use of IR and Raman spectroscopy results in a vibrational spectrum. On the y-axis comes the absorbance or transmittance for IR and the intensity for Raman. On the x-axis comes the wavenumber in inverse centimeter. Each peak corresponds to a vibrational mode with related wavenumber and intensity. Depending on the type of vibration (contracting, bending or stretching), the wavenumber, intensity and shape can differ significantly.

The energy of a vibration is typical for a certain bond and thus the energy (wavenumber) of a peak can be used as a means to identify chemical species. The intensity of the peaks is directly related to the amount of change of the polarizability and dipole moment and thus intensity and absorbance. The absorbance/intensity in turn, is directly related to the concentration of the sample [19].

The width of a peak is related to the lifetime of the vibration τR, which is the time

(12)

According to Heisenbergs uncertainty principle ∆EτR = ~, a long lifetime will result in a

smaller uncertainty in the energy and vice versa. This broadening of the spectral peaks is called homogeneous broadening. The distribution of frequencies that correspond to this shift can be fitted by a Lorentzian distribution with a full width at half maximum of ∆ω = _τ1

R

[20].

A form of inhomogeneous broadening is caused by a Doppler shift that results from the relative movement of the atom with respect to the observer. Inhomogeneous broadening in solids can also arise when vibrational bonds are influenced by their environment which results in a change in frequency. Both shifts have a distribution of frequencies that can be fitted by a Gaussian distribution [20].

Organic compounds, such as oleic acid and oleylamine, can easily be probed with FTIR-and Raman spectroscopy because organic bonds have vibrational energies in the mid-infrared region (600 - 4000 cm−1). Inorganic compounds on the other hand, such as the NCs them-selves, have phonon modes that absorb IR radiation roughly between 10 and 200 cm−1 [21]. The surface-ligand interaction of LHP NCs can as a result be elucidated without actually measuring the NC itself.

3 Methodology

Since many bonds in the substances to be studied are highly active in the mid-infrared region, the Bruker Optics - IFS 66/S FTIR spectrometer (shown in figure 8) is set to probe the mid-infrared region. To obtain reliable results, the deuterated triglycine sulfate (DTGS) detector needs to be cooled with liquid nitrogen at least every 30 minutes to prevent the influence of external infrared radiation. After refilling the detector compartment, the detector needs a minimum of three minutes to stabilize before measuring. After sample placement, the sample chamber is flushed with nitrogen gas to prevent the IR signal of air substances like carbon dioxide and water vapor, who have a strong IR absorption, to be detected. It takes approximately five minutes for the chamber to be completely flushed. Because of fluctuations in time due to a probably unstable IR beam/detector (see appendix A.2) and deviations in substrate thickness, before preparing a sample a background spectrum of the substrate(s) to be used needs to be taken.

Sample thickness and placement is of high importance when using FTIR spectroscopy. When the prepared sample is too thin, no absorption takes place. When the sample is too thick, the IR beam gets completely absorbed by the sample and no light reaches the detector. As a result, not all samples can be prepared with the same approach.

(13)

Figure 7: FTIR spectrometer. Pictures of the Bruker Optics - IFS 66/S FTIR spectrometer. The open chamber on the right picture shows the sample compartment.

The CsPbX3 solutions can be drop casted with a micro pipette onto the CaF2 substrate.

The nanocrystals were received from the Italian Institute of Technology (IIT) in Genova, the group of Prof. Liberato Manna and the nanocrystals were made by Jence Mulder. For the first measurement, 30 µL is drop casted onto the substrate and left for the solvent to evaporate. If after the first measurement the absorbance of the obtained spectrum is too low, more of the CsPbX3 solution can be drop casted onto the substrates as thought necessary. If

the proper absorbance cannot be reached, ZnSe or GaAs substrates can be used as they have a lower absorbance in at least some parts of the spectrum (see appendix A.1).

Figure 8: Salt and CsPbBr3 sample. Pictures of lead oleate salt in an air tight holder (left) and

final product of CsPbBr3 solution drop casted into CaF2 substrate (right).

As samples need to be placed vertically in the FTIR spectrometer, salts cannot simply be placed on top of a substrate. Therefore, salts need to be clammed between two CaF2

(14)

Figure 9: Fit of the CsPbBr3raw synthesis. An example of how peaks are analyzed with Origin

using the Gaussian spectral line profile. The black line represents the crude data points. Green lines represent the obtained peaks from crude data and the red line represents the merge of the obtained peaks.

the holder has a great impact on the absorbance as the density of the salt is not evenly distributed along the surface of the substrates. Trial and error is necessary to get the best holder placement for the finest vibrational spectrum.

Since the CsPbX3 solutions have a low concentration and hence the absorbance is not

always high enough for the peaks to be examined by eye, a fitting procedure may be necessary. For the low absorbance areas of the CsPbBrX3 samples a Gaussian spectral line profile has

been used to fit the peaks using Origin. Figure 9 shows an example of the fitting result of one of the purification steps. The black line represents the crude data points. Green lines represent the obtained peaks from the crude data and the red line represents the merge of the obtained peaks.

(15)

4 Results

4.1 Spectral markers

For peak assignment in the purification steps of the lead halide perovskites nanocrystals, IR spectra were obtained of the possible byproducts lead- and cesium oleate to assign spectral markers. Byproducts oleylammonium bromide and oleylammonium iodide were not available. The spectral markers were chosen in such a manner that the presence of these spectral makers can only be due to the presence of the substance that it belongs to and that the spectral marker is a strong peak in its own vibrational spectrum. Spectral markers were also assigned to the ligands to track their presence. An overview of all assigned spectral markers is shown in table 1.

Substance Spectral marker (cm−1) IR active bond

Oleylamine 1595 δ(NH2)

Oleic acid 1710 v(C=O)

1:1 mix of oleylamine and oleic acid 1436 vs(COO)

Octadecene 1641 v(C=C)

Lead oleate 1509 va(COO)

Cesium oleate 1558 va(COO)

Table 1: Spectral markers. Spectral markers of all substances are shown with corresponding IR active bonds. v = stretching and δ = bending, a = asymmetric and s = symmetric.

4.1.1 Lead oleate

The chemical structure of lead oleate is shown in figure 10.

Figure 10: Lead oleate. Chemical structure of lead oleate.

(16)

Figure 11: Lead oleate. Full vibrational spectrum of lead oleate.

The region between 1000 - 1850 cm−1contains the vibrational modes of the anchor groups and hence is the focus area. Figure 12 shows the 1000 - 1850 cm−1 region and table 2 shows all peaks and some assignments.

Figure 12: Lead oleate. 1000 - 1850 cm−1 region of the IR spectrum of lead oleate with arrows indicating all assigned peaks.

(17)

IR active bonds Lead oleate v(C-C) 1092 v(C-C) 1115 Unassigned 1195 Unassigned 1234 Unassigned 1277 Unassigned 1312 Unassigned 1346 vs(COO) 1405 vs(COO) 1425 δ(C-H in CH2) 1471 va(COO) 1509 Unassigned 1656

Table 2: Lead oleate. All peaks from the lead oleate IR spectrum are noted in this table. v = stretching and δ = bending, a = asymmetric and s = symmetric.

According to L. Robinet and M. Corbeil [22], the small peaks at 1092 and 1115 cm−1 are due to the C-C vibrational modes. The strong peaks located at 1405 and 1425 cm−1 are due to the symmetric vibration of the COO group. At 1471 cm−1 there is a strong peak due to the stretching mode of the CH2. The asymmetric vibration of the COO group gives rise to

the strong peak at 1509 cm−1. Other peaks in this spectrum remain unassigned.

For lead oleate the bending CH2 vibration at 1471 cm−1 and the asymmetric COO

vibra-tion at 1509 cm−1 are good candidates to be a spectral marker. Since the peak at 1471 cm−1 is close to the C-H bending vibration in CH2 at 1466 cm−1due to octadecene, which is greatly

present in the first purification steps of the CsPbX3 solutions, it might be overshadowed by

octadecene and as a result the peak at 1509 cm−1 is the best choice as a spectral marker for lead oleate.

4.1.2 Cesium oleate

(18)

Figure 13: Cesium. Chemical structure of cesium oleate.

Figure 14 shows the IR spectrum of cesium oleate.

Figure 14: Cesium oleate. Full vibrational spectrum of cesium oleate.

Again, the focus area is the 1000 - 1850 cm−1 region, which is shown in figure 15. All peaks and assignments are shown in table 3.

As expected, cesium oleate has many analogous peaks to lead oleate. Again, the small peaks at 1092 and 1115 cm−1 are due to the C-C vibrational modes. In addition, the peak at 1405 cm−1 is also due to the symmetric vibration of the COO group. The CH2 stretching

mode is slightly shifted compared to lead oleate, namely to 1467 cm−1. The asymmetric vibration of the COO group gives rise to the peak at 1558 cm−1. The peak at 1638 cm−1 remains a doubtful one, as cesium oleate was dissolved in octadecene, which has a typical peak at 1641 cm−1 thus might have been due to a residue of octadecene. No literature was found to

(19)

Figure 15: Cesium oleate. IR spectrum of cesium oleate with arrows indicating all assigned peaks.

compare, but Raman measurements on the same sample do not show the typical octadecene peak at 1641 cm−1 and hence indicates that the sample does not contain octadecene [22].

For cesium oleate the only possible spectral marker is the asymmetric vibration of the COO group at 1558 cm−1.

IR active bonds Cesium oleate

v(C-C) 1092 v(C-C) 1115 Unassigned 1196 Unassigned 1227 Unassigned 1271 Unassigned 1307 Unassigned 1343 vs(COO) 1405 δ(C-H in CH2) 1467 va(COO) 1558 Unassigned 1638

Table 3: Cesium oleate. All peaks from the cesium oleate IR spectrum are noted in this table. v = stretching and δ = bending, a = asymmetric and s = symmetric.

(20)

4.1.3 Octadecene, oleylamine and oleic acid

The chemical structure of octadecene is shown in figure 16.

Figure 16: Octadecene. Chemical structure of octadecene.

To obtain spectral markers from solvent octadecene and ligands oleylamine, oleic acid and a mixture of oleic acid and oleylamine, the research An analysis of the purification of CsPbBr3 nanocrystals with FTIR spectroscopy of Simion Becker was used, which was conducted under the same circumstances and with the same apparatus as this research.

Simion Becker identified the stretching C=O vibrational mode at 1710 cm−1 as a spectral marker for oleic acid, the NH2 scissoring mode at 1595 cm−1 as a spectral marker for

oley-lamine and the symmetric COO− vibration at 1436 cm−1 as a spectral marker for a 1:1 mix of oleic acid and oleylamine. For octadecene the C=C vibration at 1641 cm−1 was chosen as a spectral marker [23].

4.2 Purification steps and final product from the CsPbBr3 synthesis

All purification steps of the CsPbBr3 solution were successfully obtained and are shown in

figure 17. The focus area of all purification steps is shown in figure 18. All peaks and assignments are shown in table 4. To verify the presence of byproducts and free ligands, all purification steps were searched for the spectral markers.

Figure 19 shows the purification steps of CsPbBr3 solution with vertical lines which

indi-cate the place of the spectral markers.

The spectral marker at 1436 cm−1 of the 1:1 mix of oleic acid and oleylamine cannot be tracked by eye in figure 19. The peak at 1467 cm−1 seems to have two shoulders. The region 1420 - 1480 cm−1 has been fitted with Origin using a Gaussian spectral line profile for every purification step. After the fitting procedure, the spectral marker of the 1:1 mix of oleic acid and oleylamine appears in none of the purification steps.

(21)

Figure 17: CsPbBr3 purification steps. All purification steps of CsPbBr3 normalized from 0 to

1.

Figure 18: CsPbBr3purification steps. All purification steps of CsPbBr3are shown in the region

(22)

IR active bonds RS S1 P1 S2 P2 S3 FP v(C-C) 1020 1019 1017 1019 v(C-C) 1076 1076 1078 v(C-C) 1092 1094 1093 v(C-C) 1122 1121 1126 Unassigned 1143 1145 1146 Unassigned 1153 Unassigned 1160 1160 Unassigned 1260 1260 1260 1260 Unassigned 1303 1303 1303 1301 1301 1296 1301 Unassigned 1343 SH 1343 SH 1343 SH Unassigned 1353 1353 1353 1353 Unassigned 1369 1369 1369 1369 Unassigned 1378 1378 1378 1378 1378 1378 1378 Unassigned 1390 1390 vs(COO) 1403 1403 1403 1403 Unassigned 1410 1410 δ(C-H in CH2) 1415 1415 1415 1415 δ(C-H in CH2) 1441 1441 1442 1441 δ(C-H in CH2) 1448 1448 1449 δ(C-H in CH2) 1460 1460 1459 1460 1459 1460 1460 δ(C-H in CH2) 1467 1467 1467 1467 1467 1467 1467 Unassigned 1538 1537 1538 Unassigned 1557 1557 Unassigned 1580 v(C=C) 1641 1641 1641 1641 1641 1641 Unassigned 1670 Unassigned 1695 v(C=O) 1711 Unassigned 1743 1743 Unassigned 1821 1821 1821 1821

Table 4: CsPbBr3 IR spectrum peaks. All peaks from the CsPbBr3 IR spectrum of the raw

synthesis (RS), supernatant one (S1), two (S2) and three (S3), precipitate one (P1), two (P2) and the final product (FP) are noted in this table. Peaks that have been fitted with Origin by a Gaussian spectral line profile are marked in bold. v = stretching, δ = bending, SH = shoulder.

(23)

The absorbance of the peaks in the region 1500 - 1600 cm−1 is low and some purification steps have a lot of noise. This region contains the spectral markers of lead oleate, cesium oleate and oleylamine at 1509, 1558 and 1595 cm−1 respectively. P1 and S2 have been fitted in Origin using a Gaussian spectral line profile. S1 had to much noise to be fitted accurately, other peaks could be examined without fitting. The only spectral maker in this region present in at least one of the purification steps was cesium oleate with a spectral marker at 1558cm−1, found in S2 and P2.

Figure 19: Spectral markers. The spectral markers of the 1:1 mix of oleic acid and oleylamine (1436 cm−1), lead oleate (1509 cm−1), cesium oleate (1558 cm−1), oleylamine (1595 cm−1), octadecene (1641 cm−1) and oleic acid (1710 cm−1) are illustrated by two vertical black lines in all purification steps of CsPbBr3. The spectral marker of cesium oleate is found in supernatant and precipitate two.

Oleic acid only appears in precipitate two and octadecene is present in all purification steps except for the final product. Other substances were not present in any of the purification steps.

The spectral markers of octadecene and oleic acid are present in the region of 1600 - 1700 cm−1at 1641 and 1710 cm−1respectively. Octadecene appears to be present in all purification steps except the final product. Oleic acid only appears in P2.

Although the region 1000 - 1400 cm−1 does not contain any spectral markers, it does contain many shifted peaks when comparing different purification steps. The shifting of peaks indicate conformational or intermolecular interaction changes. Figure 20 shows the 1000 - 1400 cm−1 part of the spectrum of CsPbBr3. The C-C, CH2 and CH3 vibrational

modes are active in this region.

4.2.1 Discussion

The presence or absence of the spectral markers has given insight in whether byproducts and/or free ligands are present in the purification steps. For CsPbBr3 no spectral markers

(24)

Figure 20: CsPbBr3purification steps. All purification steps of CsPbBr3are shown in the region

of 1000 - 1400 cm−1. This region does not contain spectral markers, but does show many shifted peaks, indicating conformational or intermolecular interaction changes. The C-C, CH2 and CH3 vibrational

modes are active in this region.

were found in any of the purification steps for lead oleate, oleylamine and the 1:1 mix of oleic acid and oleylamine. This indicates that these substances are not present in free form or that the concentration is too low to measure them. This does not mean that these substances are not present at all in the purification steps. Literature predicts that the vibrational modes belonging to the spectral markers shift when they are attached to the NC surface because of the breakage of the molecular symmetry and/or redistribution of electronic clouds [24] and hence the wavenumber of the vibrational mode that the spectral maker belongs to shifts as well.

The spectral marker of oleic acid, which corresponds to the C=O stretch, only appears in P2. As there was no presence of free ligands before P2, it is possible that the purification procedure changed the relative concentration of other species and hence also the dynamic equilibrium which could result in the oleate oleylammonium ion pair detaching from the NC surface. Because there is no presence of the spectral marker of oleylamine, oleate might be protonated by oleylammonium and the resulting oleylamine binds as a L type ligand to the surface cations. Another explanation for the absence of oleylamine is that it is washed away easier than oleic acid. This leaves us with only oleic acid and thus a peak at 1711 cm−1. This is in agreement with the presence of the spectral marker of cesium oleate in S2 and P2 as the detachment of ligands coordinates with the disintegration of the NC.

(25)

Starting P2, together with the spectral marker of oleic acid, a peak at 1743 cm−1appears. According to Lin-Vien et al., the dimer C=O stretch of oleic acid is in the region of 1680 - 1720 cm−1 and the monomer version at 1740 - 1800 cm−1 [25]. Although uncertain, it is possible that the peak at 1743 cm−1 is due to the monomer appearance of oleic acid. This peak also appears in S3, where the concentration of oleic acid might be lower due to more extensive purification which would increase the fractional concentration of monomers.

The RS, S1 and P1 are almost identical. Differences between these purification steps might also be due to measurement imperfections as these peaks are often very small and the FTIR spectrometer has proven not to be very stable. Starting S2 and P2, significant changes start to appear. Peaks at 1353, 1369 and 1821 cm−1 disappear. As no peaks appear whilst these peaks disappear, this probably indicates the removal of the corresponding molecular bonds. The peaks at 1353 and 1821 cm−1 might come from octadecene, as it has peaks around 1354 and 1833 cm−1 according to the NIST Chemistry WebBook [27] but a real spectrum of octadecene needs to be obtained to confirm this.

Unfortunately, the peak at 1369 cm−1 does not correspond to any of the peaks present in free oleylamine, oleic acid, the 1:1 mix of both, octadecene and lead- and cesium oleate. The peak at 1821 cm−1 is also not present in any of these substances and since it deviates with at least 10 cm−1 from the expected peak position in the octadecene spectrum, it is debatable if it comes from octadecene. This means that either these peaks belong to oley-lammonium bromide/iodide and/or unknown byproducts, they belong to ligands attached to the surface or unknown conformational or intermolecular interaction changes take place. For further research, it would give new insights to obtain vibrational spectra of oleylammonium bromide/iodide.

Furthermore, peaks at 1415 and 1441 cm−1 shifted to 1410 and 1448 cm−1 respectively starting P2 compared to the previous ones. This shifting of these CH2 scissoring vibrations

might be due to liquid/solid phase changes. According to J. Beattie et al., in the solid state the CH2 scissoring modes in the carbon chains arise around 1410 cm−1, around 1440 cm−1

in both solid and liquid phase and around 1460 cm−1 in the liquid phase. The pure liquid vibrational spectrum has a peak at 1440 cm−1 and the solid phase around 1450 cm−1 [26]. Hence both shifts can imply a partial liquid to solid phase transition. As particles are more densely packed in the solid phase, this might mean that we are looking at ligands attached to the surface and not the ones in free form, which would be less densely packed.

The spectral marker of octadecene disappears in the final product. As octadecene is clearly present in all purification steps except the final product, it might overshadow other peaks. Thus it is a possibility that there are more substances present than can be confirmed with FTIR spectroscopy. Furthermore, some regions of all purification steps have a very low

(26)

absorbance. As a result, the reliability of the data is questionable for the noisy parts of the spectra. Because of the low absorbance and noise, some peaks might not be visible.

4.3 Purification steps and final product from the CsPbBr2I synthesis

Not all purification steps of CsPbBr2I were successfully obtained. S2 and S3 had a very low

concentration and evidently the absorbance was too low. Hence, these steps are not included in the analysis. All purification steps that were successfully obtained are shown in figure 21.

Figure 21: CsPbBr2I purification steps. All purification steps of CsPbBr2I normalized from 0

to 1.

The focus area is shown in figure 22 and all peaks and assignments are shown in table 5. Again, all purification steps were searched for spectral markers. Figure 23 shows all purification steps with vertical lines that indicate the place of the spectral markers.

(27)

Figure 22: CsPbBr2I purification steps. All purification steps of CsPbBr2I are shown in the

region of 1000 - 1850 cm−1_{, normalized from 0 to 1.}

IR active bonds RS S1 P1 P2 FP v(C-C) 1016 v(C-C) 1077 1078 1078 v(C-C) 1091 1092 1090 1096 Unassigned 1162 Unassigned 1195 Unassigned 1260 Unassigned 1290 Unassigned 1304 1302 1304 1304 Unassigned 1352 1353 1352 Unassigned 1378 1378 1378 1378 1378 Unassigned 1410 1410 δ(C-H in CH2) 1415 1415 1415 δ(C-H in CH2) 1441 1441 1442 δ(C-H in CH2) 1450 1449 δ(C-H in CH2) 1460 1460 1460 1460 1460 δ(C-H in CH2) 1467 1467 1467 1467 1467 Unassigned 1539 Unassigned 1552 1552 1546 Unassigned 1570 v(C=C) 1641 1641 1641 Unassigned 1660 Unassigned 1821 1821 1821

Table 5: CsPbBr2I IR spectrum peaks. All peaks from the CsPbBr2I IR spectrum of the raw

synthesis (RS), supernatant one (S1), precipitate one (P1) and two (P2) and the final product (FP)

(28)

Figure 23: Spectral markers 2. The spectral markers of the 1:1 mix of oleic acid and oleylamine (1436 cm−1), lead oleate (1509 cm−1), cesium oleate (1558 cm−1), oleylamine (1595 cm−1), octadecene (1641 cm−1) and oleic acid (1710 cm−1) are illustrated by two vertical black lines in all purification steps of CsPbBr2I. The spectral marker of octadecene is found in the raw synthesis, precipitate one

and supernatant one and two, no other spectral markers were found.

As with CsPbBr3, the spectral marker of the 1:1 mix of oleic acid and oleylamine at 1436

cm−1 cannot be tracked by eye. The region 1420 - 1480 cm−1 was fitted for each purification step using Origin with the Gaussian spectral line profile. After the fitting procedure, the spectral marker of the mix at 1436 cm−1 is not present in one of the purification steps.

The spectral markers of lead oleate, cesium oleate, oleylamine and oleic acid at 1509, 1558, 1595 and 1710 cm−1 respectively, are also not present in any of the purification steps.

The spectral marker of octadecene at 1641 cm−1 is found in the RS, S1 and P1.

As for CsPbBr3, the region 1000 - 1400 cm−1 does not contain any spectral markers but

is still an interesting region because the peaks tend to shift when comparing the different purification steps. This region of all purification steps is shown in figure 24. The C-C, CH2

(29)

Figure 24: CsPbBr2I purification steps. All purification steps of CsPbBr2I are shown in the

region of 1000 - 1400 cm−1. This region does not contain spectral markers, but does show many shifted peaks, indicating conformational or intermolecular interaction changes. The C-C, CH2 and

CH3vibrational modes are active in this region.

4.3.1 Discussion

Compared to the CsPbBr3 spectra, the CsPbBr2I spectra has less peaks and the absorbances

are lower. An explanation for this occurrence might be that iodide has a softer basic nature than bromide which results in a weaker acid-base interaction between oleylammonium and the halide in the case of CsPbBr2I compared to CsPbBr3. Therefore, this kind of ligand-NC

binding is lost more easily during the purification process and hence changes the dynamic equilibrium.

For CsPbBr2I no spectral markers were found in any of the purification steps for

oley-lamine, oleic acid, the 1:1 mix of both and lead- and cesium oleate. This indicates that these substances are probably not present in free form, or that the concentration is too low. The spectral marker of octadecene disappears starting P2 instead of the final product, which indicates that there might have been less octadecene to begin with or that the purification procedure has not been completely consistent. Similar to CsPbBr3, the presence of octadecene

could overshadow peaks of other substances. Moreover, the absorbance of the purification steps of CsPbBr2I is even lower than that of CsPbBr3 and hence more peaks might not be

visible and the accuracy decreases even further for the noisy parts of the spectra.

(30)

1441 to 1449 cm−1 and the peaks at 1353 and 1821 cm−1disappear. Similar to CsPbBr3, the

peaks at 1353 and 1821 cm−1 might come from the presence of octadecene [27]. The shifting of the peaks at 1415 and 1441 cm−1 might also be due to a change from liquid to solid phase of the CH2 scissoring vibrations similar to that of CsPbBr3. However, since these shifts are

completely analogous to that of CsPbBr3, they are most likely not halide dependent. This

can also indicate that these shifts might be due to a change in the binding of oleylamine as a L type ligand and/or binding of the ion pair of oleate and oleylammonium to the NC surface. Because the absorbance of P2 and the FP are very low, no significant statements about other peaks in the vibrational spectrum can be made.

5 Conclusion

For the purpose of this research, FTIR spectroscopy has shown to be a fast and effective method to distinguish free ligands, byproducts and synthesis products and to verify if these substances are present in the purification steps of CsPbBr3 and CsPbBr2I. The vibrational

spectra of both LHP NCs have been compared.

The results have shown that no free ligands are present in all purification steps for both CsPbBr3 and CsPbBr2I, except for precipitate two of CsPbBr3 where the spectral marker

of oleic acid appears. The sudden appearance of oleic acid can indicate the detachment of ligands from the NC surface. Furthermore, in this purification step the spectral marker of cesium oleate starts to appear. This might imply that the NCs are disintegrating, which is in correspondence with the detachment of the ligands from the NC surface as this could result in a reduced stability of the NC.

For both halides, but especially CsPbBr3, significant changes start to appear starting

precipitate two. Many peaks disappear or shift compared to the previous purification steps. Most of these peaks do not correspond to any of the peaks present in free oleylamine, oleic acid, the 1:1 mix, octadecene and lead- and cesium oleate. This means that either these peaks belong to oleylammonium bromide/iodide and/or unknown byproducts, they belong to ligands attached to the surface and/or unknown conformational or intermolecular interaction changes take place. For CsPbBr3, the presence of oleic acid, cesium oleate and the non

luminescent character starting precipitate two indicate that these changes might be due to the disintegration of the NCs.

Peaks at 1415 and 1441 cm−1 shift to 1410 and 1449 cm−1 respectively and the peak at 1821 cm−1disappears starting precipitate two for both halides. Thus, these shifts/disappearances are most likely not halide dependent. This indicates that these shifts might be due to a change in the binding of oleylamine as a L type ligand and/or binding of the ion pair of oleate and

(31)

oleylammonium to the NC surface, although it needs to be mentioned that the peak at 1821 cm−1 might also be due to octadecene. Another explanation for the shifting CH2 scissoring

modes is that a partial liquid to solid phase transition occurs. This would be in agreement with looking at a higher fractional amount of ligands attached to the surface compared to free ligands because ligands attached to the surface are more closely packed. To verify these hypotheses, more research is necessary.

It needs to be mentioned that because of the presence of octadecene in almost all purifi-cation steps, some peaks might not have been detectable due to octadecene overshadowing them. Furthermore, the low absorbance and noise in some parts of the spectra make the relia-bility of the data questionable. Future research should focus on obtaining spectra with higher absorbances to increase the reliability. Vibrational spectra of oleylammonium bromide/iodide can also give more insight on the surface-ligand interaction of LHP NCs.

(32)

References

[1] K. D. Karlin, Progress in Inorganic Chemistry (John Wiley Sons, 2007), pp. 1 121. [2] NREL, Best Research-Cell Efficiencies (posted on 04-25-2018). On internet: https://

www.nrel.gov/pv/assets/images/efficiency-chart.png.

[3] M. V. Kovalenko et. al, Nanocrystals of Cesium Lead Halide Perovskites (CsPbX3, X=Cl, Br, and I): Novel Optoelectronic Materials Showing Bright Emission with Wide Color Gamut), Nano Lett., 15, 3692-3696 (2015).

[4] J. De Roo et al., Highly Dynamic Ligand Binding and Light Absorption Coefficient of Cesium Lead Bromide Perovskite Nanocrystals, ACS Nano 10, 2071-2081 (2016).

[5] M. A. Boles, D. Ling, T. Hyeon, D.V. Talapin, The surface science of nanocrystals, Nature materials, 15, 141 - 153 (2016).

[6] C. R. Kagan, E. Lifshitz, E. H. Sargent, D. V. Talapin, Building devices from colloidal quantum dots, Science, 353, 885 - 898 (2016).

[7] A. A. Chistyakov et al., Optoelectronic Properties of Semiconductor Quantum Dot Solids for Photovoltaic Applications, J. Phys. Chem. Lett. 8, 4129 - 4139 (2017).

[8] A. J. Houtepen, Z. Hens, J. S. Owen, I. Infante, Ligand Displacement Exposes Binding Site Heterogeneity on CdSe Nanocrystal Surfaces, Chem. Mater., 29, 752 761 (2017). [9] R. E. Brandt et al., Searching for ”Defect-Tolerant” Photovoltaic MAterials: Combined

Theoretical and Experimental Screening, Chem. Mater., 29, pp 4667 - 4674 (2017). [10] M. Kovalenko, L. Protesescu, M. Bodnarchuk, Properties and potential optoelectronic

applications of lead halide perovskite nanoccrystals, Science 358, 745 - 750 (2017). [11] J. Owen, The coordination chemistry of nanocrystal surfaces, Science 347, 615 - 616

(2015).

[12] X. Lan, et al., Passivation using molecular halides increases quantum dot solar cell per-formance, Adv. Mater., 28, pp 299 - 304 (2016).

[13] M. Klokkenburg, J. Hillhorst, B. H. Ern, Surface analysis of magnetite nanoparticles in cyclohexane solutions of oleic acid and oleylamine, Elsevier, 43, pp 243 - 248 (2007). [14] M. L. H. A. Green, New Approach to the Formal Classification of Covalent Compounds

(33)

[15] M. L. H. A. Green, G. Parkin, Application of the Covalent Bond Classification Method for the Teaching of Inorganic Chemistry. J. Chem. Educ., 91, 807 816 (2014).

[16] R. A. Harris, P. M. Shumbula, H. van der Walt, Analysis of the interaction of surfac-tants oleic acid and oleylamine with iron oxide nanoparticles through molecular mechanics modeling, Langmuir, 31, pp 3934 - 3943 (2015).

[17] S. Becker, An analysis of the purification of CsPbBr3 nanocrystals with FTIR spec-troscopy.

[18] P. Larkin, Infrared and Raman spectroscopy (Elsevier, ed.1, 2011), pp. 1 - 12.

[19] M. Diem, Modern Vibrational Spectroscopy and Micro-Spectroscopy (Wiley, ed. 1, 2015), pp. 1 - 142.

[20] M. Bass, Handbook of optics Vol. I Fundamentals, techniques and design (Van Stryland EW., ed. 2, New York, 1995) pp. 34 - 87.

[21] R. A. Nyquist, R. O. Kagel, Infrared Spectra of Inorganic Compounds, (Elsevier, ed.1, 1971), pp 1 - 3.

[22] L. Robinet, M. Corbeil, The Characterization of Metal Soaps, Studies in Conservation 48, 23 - 40 (2003).

[23] C. Piliego, L. Protesescu, S. Bisri, M. Kovalenko, M. Loi, 5.2% efficient PbS nanocrystal Schottky solar cells, Energy Environ. Sci., 6, 3054- 3059 (2013).

[24] E. Lifshitz, Evidence in Support of Exciton to Ligand Vibrational Coupling in Colloidal Quantum Dots, J. Phys. Chem. Lett., 6, 4336 - 4347 (2015).

[25] D. Lin-Vien, N. Colthup, W. Fateley, J. Grasselli, The Handbook of Infrared and Raman Charecteristic Frequencies of Organic Molecules (Academic Press, ed. 1, 1991), pp 117 -155.

[26] J. R. Beattie, S. E. J. Bell, B. W. Moss, A Critical Evaluation of Raman Spectroscopy for the Analysis of Lipids: Fatty Acid Methyl Esters, Lipids, 39, 407 - 421 (2004).

[27] NIST Chemistry WebBook, 1-Octadecene. On internet https://webbook.nist.gov/ cgi/cbook.cgi?ID=C112889&Mask=80.

(34)

Appendices

A

Comparing substrates

Infrared substrates calcium fluorite (CaF2), zinc selenide (ZnSe), potassium bromide (KBr),

germanium (Ge), gallium arsenide (GaAs) and sapphire were compared to gain insight in which substrates are best to use under certain circumstances while using FTIR spectroscopy. The purpose of these measurements is to know which substrates are useful in the region of interest, which is between at least 1000 - 3750 cm−1, and if there is a quantitative difference between different substrates of the same kind.

A.1 Absorbance in region of interest

A perfect substrate has a constant absorbance that is as close to zero as possible in the region of interest. To compare the absorbance of the substrates, a background spectrum of the empty, nitrogen flushed container was taken. A substrate was placed and when the container was again fully flushed with nitrogen, a spectrum of the substrate was taken and the background spectrum was subtracted. These steps were repeated for all substrates.

In figure 25 the absorbance of CaF2, Ge, ZnSe and KBr are plotted against wavenumber

in inverse centimeter. Data of GaAs and sapphire were lost due to issues with saving data. CaF2 and Ge were measured in the range of 1000 - 3750 cm −1 instead of 750 - 4000 cm−1

due to wrong settings.

Figure 25: IR substrates. The absorbance of calcium fluoride, germanium, zinc selenide and potassium bromide are plotted against wavenumber in inverse centimeter with an N2 background.

(35)

GaAs had a spectrum comparable to ZnSe, but with a lower absorbance, around 0.1 in figure 25. Both substrates are fairly good for measuring in the range of at least 750 - 3750 cm−1 because their absorbance in this range is almost constant and close to zero. Lower wavenumbers might also be measured with these substrates, but in the mid IR mode of the spectrometer no light is detected below 750 cm−1. If this range shows to be interesting, these substrates might be useful but further measurements are necessary to falsify or verify this hypothesis.

Germanium shows not to be a good substrate to measure in the desired wavenumber range. The absorbance is not constant and relatively high.

KBr shows to be an excellent substrate in at least the range of 750 - 3000 cm−1. In the range of 3000 - 3600 cm −1, KBr has a wide shoulder. This might be due to the presence of water. KBr has a high solubility and hygroscopic nature and thus must be kept in a dry environment. This makes KBr less favorable to use.

CaF2 is the best substrate in the range of 1400 - 3000 cm−1. Outside this range the

absorbance goes up and the substrate becomes less useful.

Sapphire is not shown in this figure, but had a high absorbance in the region below 1800 cm−1 and thus showed not to be useful for the desired measurements.

To conclude, sapphire and germanium are never the best choice between the tested sub-strates. The choice of the use of GaAs, ZnSe and CaF2 depends on the absorbance and the

region of interest of the sample to be measured and the amount of samples to be measured as there are more available CaF2 plates due to its low cost. Because cleaning substrates takes

time and time available to measure was limited, CaF2 was mostly used in this research.

A.2 Comparing substrates of the same kind

The container which holds the sample in the FTIR spectrometer takes approximately five minutes to be completely flushed. Due to time limitations, it would be useful to take a single background spectrum of the first substrate to be measured and use this for the following samples. To do so, it needs to be verified that all substrates of the same kind give the same background spectrum. Because sapphire showed not to be useful, those substrates were not compared. ZnSe and KBr were also not compared because of limited samples.

In figure 26, four different samples of CaF2are shown. All samples were compared with the

same CaF2background spectrum. Sample one has been compared with itself. Because it takes

some time to change the samples and it takes at least five minutes to flush the spectrometer with N2, it took approximately eight minutes from one measurement to another.

As expected, sample one gives a straight line with zero absorbance. When comparing the different samples, the shape and maximum absorbance change. If the whole spectrum shifts

(36)

Figure 26: CaF2 substrates. Four different pieces of CaF2 substrates are shown. Measurements

were obtained in numerical order. There seems to be a time dependent decreasing absorbance trend, especially for higher wavenumbers.

with the same amount, it might be due to different sample thickness or that only a part of the IR beam hits the sample. This is not the case.

The change in shape seems to be time dependent. Higher wavenumbers seem to have a higher tendency to decrease. Two hours after the original measurements, sample one was measured again and compared with the original background. This spectrum together with the first measurement is shown in figure 27. If this trend was time independent, we would again expect a straight line for sample one. But the same time dependent trend was spotted again.

This change might be due to an unstable IR beam, or bad flushing/detector cooling. After refilling the N2 container for the detector, sample one gives a higher absorbance. This

indicates that the detector might have been warming up and needed to be refilled earlier than expected. A measurement five minutes after refilling the N2 container already gives a lower

absorbance. This shows that either the detector or the IR laser might not be very stable in time. This trend is also spotted in other substrates.

(37)

Figure 27: IR spectra of CaF2 in time. IR spectrum of the same piece of CaF2 substrate

compared in time.

Although germanium has the highest and an unstable absorbance in the region of interest, the comparison of different Ge samples is still useful to be compared with the trend spotted in CaF2. All samples were compared with the first Ge sample as a background. Sample two

was measured again five minutes after the first measurement. Spectra are shown in figure 28.

Figure 28: Ge substrates. IR spectrum of two different germanium substrates. Substrate two has been remeasured after five minutes.

Germanium seems to follow the same trend as CaF2, which indicates that this effect might

(38)

Three different GaAs samples were compared. All samples were compared to the spectrum of the first GaAs sample as a background and are shown in figure 29 (left graph).

Figure 29: GaAs substrates. Both graphs show the IR spectrum of the same GaAs substrates. In the right graph, substrates two and three were shifted to the absorbance of substrate one to be able to visualize the time dependent absorbance decreasing trend.

As expected, sample one gives a straight zero line. The absorbance of samples two and three are shifted compared to the first sample. This is probably due to the fact that the GaAs samples were small broken pieces. This means that a part of the IR laser beam might have missed the sample and thus no absorption was possible at all for that part. Hence, the absorbance as a whole gets shifted. It seems samples two and three almost completely overlap and the absorbance is close to a straight line, but when samples two and three are shifted up to the absorbance of sample one as in figure 29 (right graph) there is also a deviation at high wavenumbers. However, this deviation is not completely similar to that of CaF2 and Ge.

To conclude, there seems to be a trend that the absorbance tends to decrease in time, especially in the high wavenumber range. This decrease in time is the highest for CaF2 and

less for germanium and GaAs. The decrease in absorbance does not seem to be caused by a difference in thickness or composition of the substrates but rather by an instability of the IR laser/detector or cooling system. For most accurate results, a background spectrum should be taken of every substrate and the sample should be mounted on top of the substrates as fast as possible so changes in the spectrometer are as small as possible.

Elucidation of surface-ligand interactions in inorganic lead halide perovskite nanocrystals with vibrational spectroscopy