Electronic supporting information High Influence of Potassium

Electronic supporting information High Influence of Potassium Bromide on Thermal Decomposition of Ammonia Borane Nikola Biliˇskov∗a , Danijela Vojtaa , László Kótaib , Imre Miklós Szilágyic,d , Dávid Hunyadic , Tibor Pasinszkie , Sandra Flinˇcec Grgacf , Andreas Borgschulteg and Andreas Z¨ uttelg , h September 20, 2016 a

Rudjer Boˇsković Institute, Bijeniˇcka c. 54, Zagreb, Croatia Institute of Materials and Environmental Chemistry, Research Centre for Natural Sciences, Hungarian Academy of Sciences, Budapest, Hungary c Department of Inorganic and Analytical Chemistry, Budapest University of Technology and Economics, M˝ uegyetem rakpart 3, Szent Gellért tér 4, Budapest, Hungary. d MTA-BME Technical Analytical Chemistry Research Group, Szent Gellért tér 4, Budapest, Hungary. e Department of Inorganic Chemistry, Eötvös Lorand University of Sciences, Budapest, Hungary f Faculty of Textile Technology, University of Zagreb, Savska c. 16/9 g ¨ Swiss Fedaral Institute for Materials Science and Technology (EMPA), Uberlandstrasse 129, D¨ ubendorf, Switzerland h Laboratory of Materials for Renewable Energy (LMER), Institute of Chemi´ cal Sciences and Engineering (ISIC), Ecole polytechnique de Lausanne (EPFL), Valais/Wallis Energypolis, Rue de l’industrie 17, CH-1950 Sion, Switzerland b

S1

Contents S1 Experimental S1.1 Chemicals and preparation of mixtures S1.2 IR spectroscopy . . . . . . . . . . . . . S1.3 Raman spectroscopy . . . . . . . . . . S1.3.1 Powder XRD . . . . . . . . . . S1.4 Thermal analysis . . . . . . . . . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

S3 S3 S3 S4 S4 S4

. . . . .

. . . . .

. . . . .

. . . . .

S2 General considerations on IR spectroscopic experiments S2.1 Transmission through KBr pellets . . . . . . . . . . . . . . . S2.2 Matrix influence . . . . . . . . . . . . . . . . . . . . . . . . S2.3 Attenuated total reflection (ATR) . . . . . . . . . . . . . . S2.4 Errors in IR spectra . . . . . . . . . . . . . . . . . . . . . .

. . . .

. . . .

S5 . S6 . S10 . S10 . S13

S3 Reports on IR spectra S14 S3.1 Characteristic IR features. . . . . . . . . . . . . . . . . . . . . . . S18 S3.1.1 Factor group analysis . . . . . . . . . . . . . . . . . . . . S18 S4 Powder XRD patterns

S23

S5 Influence of sample preparation method

S24

S6 Thermal decomposition of ammonia borane

S25

S7 Changes in solid-state decomposition products

S28

S8 Temperature-dependent spectroscopic measurements S31 S8.1 IR spectroscopy . . . . . . . . . . . . . . . . . . . . . . . . . . . . S31 S8.2 TG-IR spectroscopy of gaseous decomposition products . . . . . S55 S8.3 Raman spectroscopy . . . . . . . . . . . . . . . . . . . . . . . . . S55 S9 Mass spectroscopy

S55

S10Thermodynamic measurements

S62

S2

S1 S1.1

Experimental Chemicals and preparation of mixtures

Potassium bromide, KBr, as purchased from Sigma-Aldrich (FT-IR grade, ≥ 99 % trace metals basis) was used. It was stored at 60 ◦C to avoid moistening. Sigma-Aldrich ammonia borane, NH3 BH3 (technical grade, 90 %) was used after recrystallisation from diethyl ether, followed by a wash with ethanol 1 . Ball milling was done by use Spex 8000 SamplePrep mill/shaker. Samples were closed in Ar atmosphere (in MBraun glove box, which ensures ≤ 0.1 ppm O2 and H2 O, respectively) in a hardened steel vessel. Milling time was 30 min. In the present work, we consider thermal decomposition of three systems: neat AB (hereafter denoted as 1), AB in KBr pellet (hereafter denoted as 2) and AB : KBr 1 : 1 mixture (hereafter denoted as 3). Both samples 2 and 3 were prepared by mixing in hand mortar. For 2, KBr and AB were mixed in ∼ 100 : 1 molar ratio, and then the force of 100 kN was applyed to make a pellet. For 3, the ratio of KBr and AB was 1 : 1, and the spectra were taken from powdered sample. In order to characterise products of thermal decomposition of AB, the sample was put into the autoclave. The teflon-line was cooled, and put liquid N2 into the tube. Most of the air was flushed off by the evolving N2 . After that the autoclave was closed. The sample was heated for 24 h in the closed autoclave. The resulting N2 pressure at 150 ◦C was 20 bar. After cooling, the autoclave was opened, and the mass of the foamy product was followed by a precise balance.

S1.2

IR spectroscopy

Temperature-dependent infrared (IR) spectra were obtained in air using a Bomem MB102 Fourier-transform infrared spectrometer with DTGS detector and CsI optics. Each spectrum represents an average of 10 co-added Fourier-transformed interferograms (scans). The nominal resolution is 4 cm−1 , which gives a distance between two points in the resulting spectrum 2 cm−1 . The controlled heating of the samples was allowed by use of Specac’s high-stability temperature controller. The heating rate was 2 ◦C min−1 . The distance between two consecutive spectra is 5 ◦C, and the temperature range r.t. − 200 ◦C. For temperature-dependent transmission IR spectroscopy, a Specac’s GS20730 electrical heating jacket with water cooling system and NiCr/NiAl thermocouple was used. Pellets were prepared by standard procedure, i.e. by grinding ∼ 1 : 100 AB : KBr mixture in an agate mortar and pressing it in a hardened steel die by applying a force of ∼ 100 kN. For temperature-dependent ATR IR spectroscopy, a Specac’s High Temperature Diamond Golden Gate ATR (type GS10642) was used. Its optical beam condensing unit consists of two golden mirrors two ZnSe lenses, and type IIIa diamond prism (high temperature bonded into tungsten carbide using a metal layer) with 2 × 2 mm contact facet, which ensures single reflection at incident angle of 45◦ . First, a standard procedure of ATR IR spectroscopy was applied, i.e. sample of neat AB was pressed (sapphire anvil) on the surface of diamond ATR element. After that, a 1 : 1 AB : KBr mixture, grinded in an agate mortar, was prepared. Temperature-dependent ATR IR spectra of this mixture, powdered and pressed in pellet, respectively, were then recorded in the same

S3

conditions. Influence of grinding time and pressure applied for pelletizing of 1 : 1 AB : KBr mixture was investigated by IR spectroscopy using a Bruker Alpha spectrometer equipped with Platinum ATR accessory (single-reflection IIIa type diamond with 2 × 2 mm facet). The nominal resolution is 4 cm−1 , and each spectrum was taken as an average of 10 co-added Fourier-transformed interferograms (scans). In order to investigate dependence of grinding time, the 1 : 1 AB : KBr mixture was manually ground in agate mortar for 5 − 60 s. Also, one spectrum was taken for the mixture which was not ground, and another was ball milled for 30 min. For dependence of IR spectra on pelletizing pressure, the 1 : 1 AB : KBr mixture was pressed in a hardened steel die, by applying a pressure of 20 − 120 kN. Bruker Alpha spectrometer was also used to monitor, in both transmission and ATR mode, the changes due to adsorption of components from ambient air by 150 ◦C solid decomposition products. The other spectra of solid decomposition products were recorded by Bruker Tensor FTIR spectrometer, also in both transmission and ATR mode.

S1.3

Raman spectroscopy

Temperature-dependent Raman spectra were recorded using a Bruker Senterra Raman microscope of 5 cm−1 spectral resolution (spatial resolution < 5 µm), which uses a 532 nm laser. The 1 : 1 AB : KBr sample was held under sapphire window in a home-made holder, which enables in-situ variable temperature spectroscopy in various atmospheres. For the purposes of this experiment, the sample was closed in holder in MBraun glove box, under Ar atmosphere (≤ 0.1 ppm O2 and H2 O, respectively). S1.3.1

Powder XRD

Powder XRD of ball milled 1 : 1 AB : KBr mixture in a 1 mm capillary was recorded at 30 ◦C using a Bruker D8 Advance diffractometer with Cu anode. The range was 2Θ = 10 − 60◦ with a ∆(2Θ) = 0.0221◦ step and 492.02 ms step time.

S1.4

Thermal analysis

Differential scanning calorimetric (DSC) measurements of neat AB and 1 : 1 AB : KBr mixture (manually grinded in agate mortar and ball milled for 30 min) were performed using a Mettler Toledo high-pressure DSC in 1 bar hydrogen atmosphere. The heating rate was 10 K min−1 over the 25 − 300 ◦C temperature range for all the samples. The hydrogen flow was adapted during measurement to maintain a constant pressure (dynamic mode). The onset and ofset temperatures were determined using the derivative of the DSC peak curve. Thermogravimetric and differential thermal analysis (TG/DTA) was performed simultaneously with a Mettler Toledo TA 4000 system. Measurements were done in air and under N2 flow, in the 25 − 300 ◦C range. Heating rate was 2 ◦C min−1 . Additionally, in order to determine the influence of heating rate on decomposition of AB, TG/DTA were taken at 5 ◦C min−1 and 10 ◦C min−1 .

S4

TG/DTA-MS measurements were done by TA instruments SDT 2960 (which simultaneously records TG and DTA) coupled with ThermoStar Balzers mass spectrometer, which uses a 1 m long sampling capillary heated to 200 ◦C. TG-IR measurements were done by TA instruments TGA 2050 coupled with a BioRad Excalibur Series FTIR spectrometer. These measurements were done in air and N2 atmosphere, respectively, at 2 and 5 ◦C min−1 heating rate in the 30 − 250 ◦C temperature range. Additional TG-IR measurements were carried out using a Perkin Elmer’s Pyris 1 TGA, coupled with Perkin Elmer FTIR spectrometer, equipped with Thermal Analysis Gas Station (TAGS) and DTGS detector. The transfer line, high-temperature flow cell, and TG interface were held at 280 ◦C for the duration of the run to prevent condensation. The evolved gases were transferred through the FT-IR flow cell by a peristaltic pump with a flow rate of 60 mm min−1 .

S2

General considerations on IR spectroscopic experiments

In any spectroscopic experiment, an interaction of electromagnetic radiation with a sample is considered. From a practical point of view, we apply a beam of light, characterised by incident intensity I0 , which passes through a sample. It is then partially transferred, scattered and reflected by sample, and we detect a frequency (or wavenumber or wavelength) dependent resulting intensity I(ν). Ratio between incident and resulting intensity is then called a spectrum: A(ν) =

I(ν) I0

(1)

In this way, spectra contain an information on the sample on molecular level. Concretely, IR spectra represent vibrational modes of the molecular system under consideration, which is subject influences from environment. However, raw experimental IR spectra consist of apparent absorption, caused by optical effects due to experimental setup. Thus, experimental IR spectra generaly need additional workup in order to extract these effects and to ”purify” information on the molecular nature of the system 2,3 . Additionally, one must bear in mind that different geometries of the experiment will provide different information on the same system, depending on what exactly is observed, for example surface of the sample (attenuated total reflection) or bulk (transmission). Thus, to obtain a complete insight into a sample under consideration, it would be often needed to observe it by a combination of IR spectroscopic techniques. In transmission technique with KBr pellet, the incident light beam passes directly through the pellet. Diameter of the beam and pellet are comparable to each other. Thus, the resulting spectrum is a sum of different contributions, which will be discussed later in more details. However, if all other contributions are extracted, the final spectrum represents a sum of the bulk sample, as well as interfaces (and thus interactions) between sample and matrix material. Contact of the powder with ATR element is ensured by pressure screw that pushes the sample onto the diamond surface by applying force. However, one must be aware that in the case of the solid sample, the contact with diamond is achieved through the individual grains. No matter how good the contact is,

S5

Figure S1: IR spectroscopic experiments considered in this study: (a) transmission through KBr pellet; (b) single-reflection attenuated total reflection (ATR). the grains in contact with diamond are always oriented and distributed randomly, as illustrated in Fig. S1(b). Thus, the resulting spectrum is a sum of the spectra of the individual grains, spectra of individual crystal planes, grain/air interfaces, and, in the case of mixture, individual components and their interfaces. This differs to transmission spectra, since in the ATR technique provides an information on the surface layer of the observed system, while transmission technique observe the sample as a bulk.

S2.1

Transmission through KBr pellets

The most often used technique to obtain IR spectra of solid samples is transmission through the KBr pellet. Indeed, from the sampling point of view, this technique is very easy and fast. The sample is mixed with dry, spectroscopic grade KBr in a ∼ 1 : 100 ratio and ground to make mixture as homogeneous as possible. This means that sample particles, small enough not to cause scatter (theoretically < 2 µm), must be homogeneously dispersed in matrix. In place of KBr one can use another alkaline halide, such as NaCl, KCl or CsI, alkaline earth halide, such as CaF2 and BaF2 . Less commonly used matrix materials include Si, Ge, ZnS, SiO2 (Infrasil), SeAsGe glass (AMTIR) and TlBr/TlI (KRS-5). The prepoared mixture is then poured into a mold, and compressed by applying the force of ∼ 100 kN on piston over a period of ∼ 1 min. The mold can be additionally evacuated in order to remove the residual water or other solvents in sample. Following discussion is based on detailed considerations of errors in IR spectroscopy by Jones et al. 4,5 .

S6

The absorption of electromagnetic radiation is best described in terms of the absorption index k(˜ ν ). However, the experimentally measured absorption index contains not only information on the investigated system, but also all contributions due to the imperfections and other sources of error. Thus, we call it apparent absorption index, which is given by: ka (˜ ν ) = 2.302 58

Aa (˜ ν) 4π˜ νd

(2)

where d is pathlength and Aa (˜ ν ) is wavenumber dependent apparent absorbance: Aa (˜ ν ) = log

P0 P (˜ ν)

(3)

where P0 and P are incident and transmitted radiation flux, respectively. Then, transmittance is defined as: P (˜ ν) τ (˜ ν) = P0 and reflectance is given by: ρ(˜ ν) =

R(˜ ν) P0

where R(˜ ν ) is the radiant flux of reflected beam. If scattering and luminiscence could be neglected, then: τ (˜ ν ) + Aa (˜ ν ) + ρ(˜ ν) = 1 In the majority of IR spectroscopic experiments, luminiscence is neglectable, but the nature of pellets does not allow neglection of scattering on particles of matrix and sample. At a boundary between media of refraction indices n1 and n2 , the intensities of the reflected (R) and transmitted (T ) light for unit incident intensity I0 are given by Fresnell equations: r12(s)

=

t12(s)

=

r12(p)

=

t12(p)

=

n1 cos θ1 − n2 cos θ2 n1 cos θ1 + n2 cos θ2 2n1 cos θ1 n1 cos θ1 + n2 cos θ2 n2 cos θ1 − n1 cos θ2 n1 cos θ2 + n2 cos θ1 2n1 cos θ1 n1 cos θ2 + n2 cos θ1

(4)

for assumed 1 → 2 direction of the light beam, where θ1 and θ2 are the angles of incidence and refraction (which can be easily calculated by Snell’s law), r12 and t12 are the amplitudes of the electric vectors of the reflected and transmitted light, while s and p define perpendicular and parallel electric field polarisation vectors (with respect of the plane of incidence on a device). The ratios of the intensities of the reflected and incident components of the beam are: R12(s) (˜ ν)

2 = r12(s) (˜ ν)

R12(p) (˜ ν)

2 = r12(p) (˜ ν)

S7

(5)

Figure S2: Reflection and transmission within a real pellet. and for the refracted beam: T12(s) (˜ ν)

=

T12(p) (˜ ν)

=

n2 (˜ ν) 2 t (˜ ν) n1 n2 (˜ ν) 2 t (˜ ν) n1

(6)

where n2 (˜ ν )/n1 takes into account different cross sectional areas of the incident and refracted beams. These axpressions remain valid for an interface at which one or both of the media are absorbing, provided n(˜ ν ) is replaced by complex refractive index n ˆ (˜ ν ). Fresnell coefficients then become complex and the squared terms in equations (5) and (6) must be replaced by their complex conjugates. As a model of ideal pellet we consider a transparent plate of refractive index n2 (˜ ν ) surrounded by air (or some other gas) of refractive index n1 . The reflectance and transmittance at two interfaces are calculated by equations (4-6). Summation of multiple reflections within the plate gives: Ra (˜ ν)

=

Ta (˜ ν)

=

2R12 (˜ ν) 1 − R12 (˜ ν) T12 (˜ ν )T21 (˜ ν) 2 1 − R12 (˜ ν)

(7)

However, from the physico-chemical point of view, this method comprises a numberous sources of errors. First of all, sample is never homogeneously distributed in matrix. In the other words, pellet contains sample-rich islands, as well as regions which contain only a little or no sample. From the optical point of view, pellet is a thick plate, surrounded by air, with nonparallel bases, whose optical properties are not constant, and which contains various imperfections. This means that optical properties are not constant over the pellet. Also, distribution of particle dimensions is rather broad. Pellet by itself contains many imperfections, such as cracks, holes and bubbles, as well as nonparallel S8

bases. These imperfections introduce additional optical effects, which affect the experimental result (Fig. S2).

Figure S3: Multiple reflections at interfaces of the pellet. First of all, we must take into account the interference between successive internal reflections within the pellet (Fig. S3). Thickness of a typical pellet is d ≈ 0.5 mm, and refraction index n2 (˜ ν ) is not constant, but close to 1.5. Assuming parallel bases of the pellet, this gives rise to interference fringe pattern with a spacing 3 − 5 cm−1 . Taking into account also the nonparallelism between bases, the intensity of reflected rays at interfaces should be taken into account: r12 (˜ ν ) + r22 (˜ ν ) + 2r1 r2 (˜ ν ) cos 2δ 2 2 1 + r1 (˜ ν )r2 (˜ ν ) + 2r1 (˜ ν )r2 (˜ ν ) cos 2δ t21 (˜ ν )t22 (˜ ν) T (˜ ν) = 1 + r12 (˜ ν )r22 (˜ ν ) + 2r1 (˜ ν )r2 (˜ ν ) cos 2δ δ = 2πνd2 n2 (˜ ν ) cos θ2

R(˜ ν)

=

(8)

For a real pellet, relations (7) also become complicated, because all successive interfaces (air/KBr, air/sample, KBr/sample, KBr/air, sample/air etc.) must be taken into account. The resulting intensity relation is as follows: It (˜ ν ) = I0

Ta (˜ ν )Tb (˜ ν) [1 − Rb (˜ ν )]2

(9)

Reflectivity is characterised by two parameters, namely rms deviation from the mean surface level σ, and rms slope of the surface profile m. Generally, σ determines specular reflectance of the rough surface, and m its diffuse reflectance 6,7 . Reflection is then described as: Rr Rm

2

2 2

= R0 e−16π σ ν˜ 32π 2 σ 4 ν 4 θ02 = R0 m2 S9

(10)

where θ0 is the limiting angle of incidence, R0 is reflectance for an ideal surface and the total reflectance is Rr + Rm . General considerations of the effects of surface roughness on transmission 8 give: Tr ≈ T0 [1 − 4π 2 ν˜2 (∆φ)2 ]

(11)

where T0 is transmittance through an ideal surface, and (∆φ)2 is mean square deformation of the wavefront, which is a direct function of surface roughness characterised by σ. It can be evaluated exactly for simple systems, and for a simple plate states: √ (12) ∆φ ≈ σ(n − 1) 2 However, effects of roughness of the pellet surface are negligible in the mid IR region. Additionally, we should not forget to take into consideration scattering on KBr and sample particles, as well as on the imperfections of the pellet. All of these effects make the pellet highly complicated optical system, whose optical properties cannot be controlled, which leads to irreproducibility of the method. However, this method can be good enough if we consider relative changes of the sample due to some perturbation, such as time, pressure, temperature etc.

S2.2

Matrix influence

A common method for preparation of solid samples for transmission IR spectroscopic measurements is the dispersion of the sample in an alkali halide salt (most commonly used is KBr) by grinding. Mixture is then pressed into a pellet. Unfortunately, discussion on possible alteration of the AB sample by matrix is completely absent in literature. However, the effect of alkali halide salts on a number of solid inorganic materials is well documented 9–17 . Among other papers devoted to matrix influence on transmission IR spectra of various systems, of special interest for present study are those dealing with ammonia 9,10 and borohydrides 16,17 . In this context, the study of reaction of AB with MgCl2 and CaCl2 18 is also highly informative. In some cases, the use of KBr as a diluting matrix leads to serious misinterpretations of the IR spectra. Most importantly, exchange reactions between KBr and sample can take place, which causes shifts of existing bands, as well as appearance of new features. Factors responsible for exchange and its rate are numberous, and they are related to type and size of involved ions, applied pressure, time of application of the pressure, size of matrix and sample particles, homogeneity of mixing, surface moisture, solubility of the sample etc.

S2.3

Attenuated total reflection (ATR)

An ATR setup generally consists of precisely polished crystal, that is transparent in a broad region of IR spectrum, and has a relatively big refraction index. An internal optical system, consisting of mirrors and lenses (Fig. S4(a)), directs incident light towards the sample at an specific angle. A detailed description of ATR accessory used in this study is given in section S1, while the following discussion is based on the previous work by Harrick 19 and Biliˇskov 2 .

S10

Figure S4: Single-reflection ATR setup Golden GateTM as used in this study. In the case of nonabsorbing optically rarer medium, Fresnell equations for total reflection on the interface give: q cos β0 − i sin2 β0 − n221 q rˆs = cos β0 + i sin2 β0 − n221 q n221 cos β0 − i sin2 β0 − n221 q rˆp = (13) n221 cos β0 + i sin2 β0 − n221 where rˆs and rˆp are complex amplitudes of parallel and perpendicular component of reflected beam, β0 is angle of reflection (Fig. S1(b)), and n21 a ratio of refraction indices of optically less and more dense medium, i.e. n21 = n2 /n1 . The second quantity of crucial importance for ATR is reflectance R: q n1 2 n1 cos β0 − n2 1 − ( n2 sin β0 ) q Rs = n1 2 n1 cos β0 + n2 1 − ( n2 sin β0 ) q n1 2 n1 1 − ( n2 sin β0 ) − n2 cos β0 q Rp = (14) n1 2 + n cos β n1 1 − ( n2 sin ) 2 0 β0 In an ATR experiment, optically rarer medium (i.e. sample) partially absorbs and attenuates reflected light. In the other words, reflectance will deviates from value 1: R = (1 − a)NR (15) where a is absorption parameter, and NR is effective number of reflections, which always differs from number of reflections obtained from simple geometrical considerations, because it involves also experimental effects, such as scattering of s component of light and beam convergence. Thus, seconds standards should be used to calibrate an ATR accessory. In our case this value is obtained from benzene spectrum and it is NR = 0.78 2 . In the case of absorptive rarer medium, reflectance is again calculated by equations (14), but n2 must be replaced by complex refraction index: n ˆ 2 (˜ ν ) = n2 (˜ ν )[1 − iκ(˜ ν )] S11

(16)

where κ(˜ ν ) is attenuation index, which is related to absorption index k(˜ ν ) by: n2 (˜ ν )κ(˜ ν) =

k(˜ ν )c 4πν

(17)

Figure S5: Electromagnetic wave on the interface between optically denser (n2 , ATR element) and rarer (n1 , sample) medium. λ0 is wavelength of radiation ν )ω in optically denser medium, kz (˜ ν ) = n1 (˜ cos θ and cos φy = √ cos θ , β0 is c 1−1/b (˜ ν)

angle of reflection, θ phase angle and ˆ(˜ ν ) complex dielectric constant 19 . Insight into mechanism of interaction of infrared radiation with sample at the interface of two optical media is derived from Maxwell equations. Frequency remains constant, but amplitude exponentialy decreases with distance from interface z (Fig. S5): E = E0 e−z/dp (˜ν ) (18) where dp (˜ ν ) is depth of penetration, defined as a distance z at which amplitude of electric field vector is E = E0 e−1 , and it is given by: dp (˜ ν) =

λ0 q (˜ ν) 2 2 2π sin β0 − ( nn12 (˜ ν) )

(19)

where λ0 = λ/n1 wavelength of radiation in optically denser medium, and β0 the angle of reflection (or angle of incidence, which are, off course, equal). An S12

ATR spectrum reflects the interaction of the evanescent wave with the sample. Penetrating electric field is continued from sinusoidal field on the interface, as illustrated on Fig. S5. It is important to see in equation (19) that dp (˜ ν ) is directly proportional to λ0 , which results in essential difference between ATR and transmission spectra. Interpolated refraction index spectrum for diamond n1 (˜ ν) 2 −1 and assumed refraction index n2 (˜ ν ) = 1.45 gives dp (4000 cm ) = 1.10 µm and dp (600 cm−1 ) = 7.45 µm. This clearly shows that ATR is technique that gives an information on surface of the sample. In our experiment incident radiation is not polarised. Thus, the ratio of parallel and perpendicular radiation with respect of the plane of interface between ATR element and sample is Is : Ip = 1 : 1. If the angle of incidence is β0 = 45◦ , then Rp = Rs2 and intensity is given by: I(˜ ν) =

1 [Rs (˜ ν )NR + Rs (˜ ν )2NR ]I0 2

(20)

which is a basis for further precise analysis of ATR spectra in the case of homogeneous and isotropic optical media 2 . Evidently, this is the case for ATR element, but, strictly speaking, not for the solid sample, i.e. powder consisting of finely crumbled grains (Fig. S1(b)).

S2.4

Errors in IR spectra

Quality of the spectrum is expressed in terms of signal-to-noise (SN) ratio, as well as contributions from surrounding atmosphere, such as features due to CO2 and H2 O, which are always present if spectra are measured in air. The general rule is that one should record the spectrum as an average of as much as possible individual Fourier-transformed interferograms in order to aquire the optimal SN ratio, defined as: A(˜ ν) SN(˜ ν) = (21) σ(˜ ν) where A(˜ ν ) is average value of the signal (absorbance or reflectance), and σ(˜ ν) is standard deviation. Standard deviation is determined by multiple measurement of the spectra of polystirene as a standard sample. For ATR, SN(˜ ν ) is ∼ 400 (σ(4000 cm−1 = −3 −1 −4 2.4 · 10 ) and σ(1000 cm = 8 · 10 )), while for transmission it is > 6.7 · 104 (σ(4000 cm−1 = 1.8 · 10−5 ), σ(1000 cm−1 = 3.5 · 10−6 )). Concentration of CO2 and H2 O vapour are not constant over time in the atmosphere. This is reflected in spectra as a series of bands in the region 2400 − 2250 cm−1 for CO2 , while water vapour gives rise to 4000 − 3400 cm−1 (νas (OH) and νs (OH)) and 2100 − 1300 cm−1 (δ(HOH)). Variation in concentration of CO2 and H2 O can be compensated by acquisition of background spectrum before every measurement, but it is not possible in the case of in-situ variable temperature measurements. However, these influences are minimised simply due to the higher temperature of the system with respect to the environment.

S13

S3

Reports on IR spectra Table S1: IR spectra of AB and related systems as reported by other authors.

Method Transmission, KBr pellets Transmission, KBr pellets

System AB

Transmission, KBr pellets

AB, AB + AlH3

Transmission, KBr pellets Transmission, KBr pellets

LiAB

Transmission, KBr pellets Transmission, KBr and CaF2 pellets

AB

AB, NH3 BD3 , ND3 BH3 , ND3 BD3 , N(CH3 )3 BH3 , N(CH3 )3 BD3 , NH3 B(CH3 )3 , ND3 B(CH3 )3 , N(CH3 )3 B(CH3 )3 , NH3 BF3 , ND3 BF3 , N(CH3 )3 BF3 , N(CH3 )3 BCl3 , N(CH3 )3 BBr3 AB, LiAB AB

Notes Comparison of AB spectrum with those of Zr(BH4 )4 · 8 NH3 . Comparison of r.t. AB spectrum with products of thermal decomposition after heating to 225 ◦C. Comparison of AB spectrum with those of AB+AlH3 composite at r.t., 120 celsius and 250 ◦C. Comparison of spectrum of neat and confined LiAB. R.t. comparative spectroscopic study.

Ref. 20

21

22

23

24

Ar, pellets were dryed

25

26


AB

Transmission, pellets Transmission, pellets Transmission, pellets Transmission, pellets

KBr

NaAB

Comparative study of transmission IR spectra of AB in different matrices IR spectra of solid dehydrogenation products time-dependent spectra of the samples stored at different temperatures r.t., 55 ◦C and 110 ◦C product

KBr

LiNaAB2

r.t. spectra

30

KBr

AB, NaMgAB3

Solid dehydrogenation products

31

KBr

AB

Products after confined rapid thermolysis.

32

AB, LiAB, LiNaAB2

NaAB,

S14

27

28

29

Transmission, KBr pellets

AB, AB + 0 · 5 Mg

Transmission, pellets Transmission, pellets Transmission, Transmission,

KBr

AB, LiAB

KBr

AB, AB + Mg3 N2

Transmission, pellets Transmission, pellets Transmission, Transmission, pellets Transmission, pellets

R.t. AB and [AB + 0 · 5 Mg]; 2 h milled AB; quench-cooled postdehydrogenated AB (300 ◦C) and [AB + 0 · 5 Mg] (100, 200 and 300 ◦C); [AB + 0 · 5 Mg] after being held at 300 ◦C for 2.5 h. As-prepared samples.

33

35

34

KBr

AB, AB@MOF

Post-milled, aged and postheated samples. As-prepared samples. R.t. spectra of pristine sample and products after different dehydrogenation steps. Samples at r.t.

KBr

AB

Samples at r.t.

39

AB AB confined in C

Samples at r.t. As-prepared and after heating (150 ◦C) samples. As-prepared cycloborazanes.

39

Products of thermal decompositions of PAB. R.t. transmission spectrum (only characteristic bands are listed); DRIFT spectrum is not shown nor commented. Matrix-isolated and gaseous products (B2 H6 and NH3 ). Cryogenic conditions

42

Some spectra were taken from mulls or thin films on KBr. Some variations in peak positions with physical state have been noted. In liquid NH3 , ND3 and (CH3 )2 O solution. The positions of the bands do not differ between the solvents, except in the case of ν(BN). Post-milled, aged and dehydrogenated sample. As-prepared and postdehydrogenated samples.

46

KBr KBr

nujol KBr KBr

Transmission, KBr pellets Transmission, CsI pellets, DRIFT

Transmission, Ar and N2 matrix Transmission, Ar matrix Transmission, KBr pellet

[MgH2 + 2 AB] AB, LiBH4 · AB

(NH2 BH2 )2 , (NH2 BH2 )4 , (NH2 BH2 )5 polyborazine, PAB, PIB, BN AB@BN core-shell

H2 N−BH2 AB, ND3 BD3 AB, NH3 BD3 , ND310 BD3

Transmission, solution

AB, NH3 BD3 , ND310 BD3


MgAB2

ND3 BH3 , ND3 BH3 , ND3 BD3 ,

ND3 BH3 , ND3 BD3 ,

[Li2 (AB)BH4 + LiAB]

S15

36 37

38

40

41

43

44

45

46

47

48

Transmission, pellets Transmission, pellets Transmission, pellets Transmission, pellets Transmission, pellets

r.t. post-milled samples.

49

KBr

LiAB, LiAB · NH3 , [AB + LiNH2 + LiBH4 ] AB, LiAlAB4

R.t., as-prepared samples.

50

KBr

MgBH4 ·6 NH3 + AB

51

KBr

poly-[NH2 BH2 ] x

KBr

[Al(BH4 )3 · 6 NH3 + 4 AB], [Li2 Al(BH4 )5 · 6 NH3 + 3 AB], Na2 (BH4 )(NH2 )

temperature-dependent IR spectra (r.t. - 400 ◦C) Polymeric solid product of first AB dehydrogenation step. As-prepared and dehydrogenation (up to 300 ◦C) products at r.t.

KBr

Transmission, KBr pellets Transmission, KBr pellets Transmission, KBr pellets Transmission, KBr pellets Transmission, synchrotron Transmission, NaCl plate

AB, AB + FeNi, AB + ZrNi AB, AB/MOF-5 AB, MOFs) AB

AB/(various

−[N H2BH2]x −

52

53

α and β phase at r.t.

54

pre- and post-heated (400 ◦C) samples pre- and post-heated (200 ◦C, 1 h) pre- and post-heated (150 ◦C, 1 h) High-pressure IR and Raman study. Solid products of AB decomposition on NaCl plate attached to cold finger

21

55

56

57

58

?

Transmission, gas Transmission, gas

BH(NH2 )2 BH(NH2 )2

ATR

AB, AB+ polyethylene oxide

ATR

AB

ATR

AB

ATR

poly-AB

ATR ATR

AB AB

S16

Isotopomers are also recorded, and the normal coordinate analysis was done Comparison of pristine AB with AB+polyethylene oxide composite, and thermal decomposition product of the composite after heating at 85 ◦C for 10 and 120 min Pristine AB and product ([NH2 BH2 ]x) of catalytic dehydrogenation. R.t. IR and Raman spectrum of as-prepared AB. pressure and temperaturedependent IR and Raman (90 ◦C and 1.38 kPa; 120 ◦C and 2.07 kPa) time-dependence, 80 ◦C Solid products of dehydrogenation r.t. - 1500 ◦C

?

59

60

61

62

63 64

ATR

AB

ATR

AB

ATR

(NH2 BH2 ) n , Na[NH2 (BH3 )2 ] polyborazine AB, CaAB2 · NH3

DRIFT DRIFT DRIFT

DRIFT

N2 H4 BH3 , LiN2 H3 BH3 , NaN2 H3 BH3 , KN2 H3 BH3 AB, MgAB2 · NH3

DRIFT

AB

DRIFT

[AB + LiAlH4 ], [AB + NaAlH4 ] AB

Specular reflection

Gas phase Transmission, synchrotron not specified not specified not specified

AB, Mg(BH4 )2 (AB)2 AB + CoCl2 MgAB2 · 3 NH3

not specified

LiAB

Transmission, phase not specified

gas

AB AB

[AB + T4EGDE]

Pristine and heated samples at 80 ◦C for different times. Spectrum of catalytically dehydrogenated product. R.t.

65

R.t. R.t. and post-dehydrogenated product r.t.

67

R.t. and post-dehydrogenated product R.t. AB and dehydrogenated products. In-situ temperature dependent spectra, r.t. - 260 ◦C AB on rotating disk gold electrode in NaOH solution. 65◦ incident angle AB vapor Temperature-dependent study, 10 − 300 K Spectra taken at r.t. and 220 ◦C at 100 ◦C and 200 ◦C Product after heating up to 300 ◦C R.t., dehydrogenated (250 ◦C) and rehydrogenated solid product. In-situ gaseous products of dehydrogenation. r.t.

70

not specified

AB, (AB + NaAlH4 ), Na(Al(AB)4 ) NaBH4 + 2 AB, KBH4 + AB AB, AB@MOF

not specified

AB, AB@MOF

not specified

cyclo − [NH2 BH2 NH2 BH2 ]NH2 BH3 AB, PAB, PIB, CTB Calculated spectra with experiments

not specified

DFT

S17

66

?

68

69

71

72

73

74 75

76 77 78

79

80

81

As-prepared samples at r.t.

82

Pristine samples and after dehydrogenation at 200 ◦C Pristine samples and after dehydrogenation up to 500 ◦C.

83

84

85

compared

86

S3.1 S3.1.1

Characteristic IR features. Factor group analysis

A detailed factor group analysis for both crystal forms of AB is given by Hess et al. 87 . However, we repeat it here for tetragonal AB (I4mm space group) for the sake of better understanding of given spectroscopic data. AB molecule is of C3v symmetry. Since it contains n = 8 atoms, it posseses 3n − 6 = 18 vibrational degrees of freedom. The irreducible representation is: Γ = 6A1 + 2A2 + 8E

(22)

When considering the vibrational spectrum of the free molecule, the rotational: Γrot = A2 + E

(23)

Γtrans = A1 + E

(24)

and translational: irreducible representations are subtracted from 22, which results in: Γvib = 5A1 + A2 + 6E

(25)

IR active modes are those of A1 and E symmetry, i.e. 11 intramolecular vibrational modes: ΓIR (26) vib = 5A1 + 6E For crystalline phase, contributions from the lattice vibrations and librational motions must be taken into consideration together with intramolecular vibrations. Generally, IR spectra of molecular crystals contain contributions from lattice, acoustic and librations in addition to intramolecular vibrations, which results in an increase of the observed bands. The similarity of here recorded IR spectra at room temperature (Fig. S6) with respect to that observed for the matrix isolated AB 45 , at least in the number of bands, shows that there is no increase of the number of vibrational modes in the solid state. In the other words, factor group analysis provides an accurate description of the vibrational spectrum of solid AB. Custalcean et al. have interpreted this similarity as a result of orientational disorder of the 6 hydrogen atoms of NH3 and BH3 groups about the fourfold axis of the tetragonal I4mm space group 88 . However, factor group analysis of the I4mm phase of AB reveals several together acting factors, which make the IR spectrum of the free and of the solid AB at r.t. to be remarkably similar. Factor group symmetry of I4mm space group is C4v . Two AB molecules occupy the unit cell, i.e. Z = 2. Since there are two lattice points (L) in the I4mm group, there is Z =1 ZB = L AB molecule in Bravais cell. Thus, the number of vibrational degrees of freedom is 3n(Z B ) − 3 = 21, which is greater than the number of intramolecular modes for free molecule. The AB molecule occupies 2a Wyckoff site, which is of 4mm (C4v ) symmetry. The coincidence of site symmetry and factor group symmetry removes one

S18

Figure S6: IR spectra of here considered systems at ambient conditions.

S19

additional vibrational modes, since the number and symmetry of the lattice and acoustic modes are identical, i.e. they cancel each other in summation: Γlattice

= A1 + E

Γacoustic

= A1 + E

Γvib

=

(27)

Γacoustic − Γacoustic = 0

For librational modes, factor group analysis gives three degrees of freedom: Γlibration = A2 + E

(28)

The E mode is degenerate, and modes of A2 symmetry are IR inactive, so only one degenerate band should originate from librational motion. Intramolecular vibrations of I4mm form of AB are somewhat harder to describe, since there is no subgroup-supergroup relationship that connects C3v and C4v point groups. However, Fateley et al. have demonstrated how the correlation between two point groups can be established in a relatively simple way 89,90 . In this method, intramolecular and lattice vibrations are classified under point groups of molecular symmetry, site symmetry and factor group symmetry. Then the correlations can be made by use of correlation tables for corresponding space groups. Comparison of the characters of shared symmetry operators gives the correlation between the two point groups. This can be done by identifying the correlation through the supergroup that contains both point groups and by preserving the alignment of common symmetry operators. For C3v and C4v a common supergroup is C∞v , since the alignment of the rotation and mirror planes are preserved. Insight into correlation tables reveals that characters of representation of translatory modes are of A1 symmetry for Tz , and of E symmetry for Tx and Ty for C3v and C4v , while for C∞v they are of Σ+ and Π symmetry, respectively. On the other hand, characters of representation of rotatory modes are of A2 symmetry for Rz , and of E symmetry for Rx and Ry for C3v and C4v , while for C∞v they are of Σ− and Π symmetry, respectively. Thus, these point groups are correlated by: C3v → (E, C3 = C∞ , σv ) → C∞v → (E, C∞ = C4 , σv ) → C4v

(29)

The irreducible representation of vibrational modes for free AB molecule is given by (26), which equals the irreducible representation of intramolecular vibrational modes for I4mm AB. Finally, the total irreducible representation for I4mm AB is given by: Γtotal = Γintramolecular + Γvib + Γlibration Γtotal = 5A1 + 2A2 + 7E

(30)

which correctly accounts for the 3n − 3 = 21 degrees of freedom. Difference in the irreducible representations between the free and I4mm AB is the addition of librational modes, given by (28). In terms of IR spectrum, only the additional E librational mode is active. Therefore, the similarity of IR spectrum of free and I4mm AB is a result of at least four factors in addition to orientational disorder 88 , namely: 1. identical site and factor group symmetry; 2. direct correlation between irreducible representations for free molecule and crystal site; 3. no net increase in the number of molecules in the Bravais cell; 4. IR inactivity and degeneracy of additional modes in the I4mm phase. S20

Table S2: Positions and relative intensities of symmetric and antisymmentric ν(NH) and ν(BH) bands of pristine AB at ambient conditions. Relative intensities of the bands corresponding to the same functional group are represented by ↑ (more intensive), ↓ (less intensive), and (nr) states for ”not reported”.

νas (NH)

Peak positions in cm−1 νs (NH) νs (BH) νas (BH)

Transmission, KBr pellets 3318 ↑ 3252 ↓ 2280 ↓ 3325 ↑ 3261 ↓ 2293 ↓ 3308 ↑ 3248 ↓ 2258 ↓ 3310 ↑ 3247 ↓ 2268 ↓ 3284 ↑ 3248 ↓ 2263 ↑ 3300 ↑ 3240 ↓ 2250 ↓ 3311 ↑ 3253 ↓ 2347 ↑ 3314 ↑ 3254 ↓ 2347 ↓ 3253 ↑ 3194 ↓ 2348 ↑ 3300 ↑ 3190 ↓ 2351 ↑ 3321 ↑ 3258 ↓ 2337 ↓ 3307 ↑ 3241 ↓ 2351 ↑ (nr) (nr) 2357 ↑ 3320 ↑ 3260 ↓ 2330 ↑ 3304 ↑ 3202 ↓ 2333 ↑ 3309 (nr) 3183 (nr) 2285 (nr) 3323 ↑ 3251 ↓ 2356 ↑ 3315 ↑ 3170 ↓ 2270 ↓ 3312 (nr) 3245 (nr) 2277 (nr) 3318 ↑ 3259 ↓ 2247 ↓ 3298 ↑ 3243 ↓ 2265 ↑ 3319 ↑ 3253 ↓ 2283 ↓ Transmission (other matrices) 3317 ↑ 3185 ↓ 2339 ↑ 3378 ↑ 3256 ↓ 2333 ↑ 3280 ↑ (nr) 2365 ↑ 3284 ↑ 3248 ↓ 2370 ↑ (nr) (nr) (nr) 3451 ↑ (nr) 2509 ↑ 3386 ↑ 3224 ↓ 2415 ↑ (nr) (nr) (nr) (nr) (nr) 2453 ↑ 3323 ↓ 3253 ↑ 2360 ↑ ATR 3308 ↑ 3250 ↓ 2315 ↑ 3400 ↑ 3300 ↓ 2337 ↑ 3306 ↑ 3250 ↓ 2360 ↑ 3400 ↑ 3300 ↓ 2337 ↑ 3301 ↑ 3246 ↓ 2279 ↑ 3312 ↑ 3243 ↓ 2327 ↑ 3302 ↑ 3206 ↓ 2278 ↑ DRIFT 3265 ↑ 3169 ↓ 2343 ↑ 3271 ↑ 3172 ↓ 2350 ↑ 3341 ↑ 3258 ↓ 2423 ↑ Not specified method 3304 ↓ 3248 ↑ 2283 ↓ 3306 ↑ 3246 ↓ 2306 ↑ 3306 ↑ 3246 ↓ 2306 ↑ 3296 ↑ 3178 ↓ 2328 ↑ 3307 ↑ 3230 ↓ 2317 ↑ 3323 ↑ 3250 ↓ 2281 ↑ a KCl

ν(BN)

Ref.

2332 ↑ 2361 ↑ 2311 ↑ 2361 ↑ 2375 ↓ 2350 ↑ 2289 ↓ 2288 ↑ 2288 ↓ 2307 ↓ 2276 ↑ 2285 ↓ 2285 ↓ 2280 ↓ 2260 ↓ 2316 (nr) 2282 ↓ 2315 ↑ 2316 (nr) 2322 ↑ 2313 ↓ 2370 ↑

784 829 769 791 783 784 — 885 798 784 870 780 828 735 782 787 794 790 776 811 770 800

This work

2283 ↓ 2280 ↓ (nr) 2263 ↓ (nr) (nr) 2340 ↓ (nr) 2337 ↓ 2281 ↓

795 779 859 877 755 897 987 754 828 780

This worka This workb

↓ ↓ ↓ ↓ ↓ ↓ ↓

781 825 783 825 781 — 781

This work

2105 ↓ 2112 ↓ 2305 ↓

770 773 776

68

↑ ↓ ↓ ↓ ↓ ↓

756 793 793 783 788 771

76

2214 2100 2283 2100 2205 2204 2124

2313 2214 2214 2211 2211 2215

20 21 22 26 27 28 31 32 33 34 35 38 39 40 46 50 91 24 21 55 56

26c 26 d 39e 43 f 45g 46h 57 i 75 j

60 59 60 60 63 64

70 71

82 82 83 84 81

pellets. pellets. c CaF pellets · 2 d AB on KBr disk e Nujol f CsI pellets. g Cryogenic transmission IR spectroscopy S21 in Ar matrix. h Solution in liquid NH , ND and (CH ) O. 3 3 3 2 i U2A beamline at the National Synchrotron Light Source at BNL. Pressed AB pellets. j AILES beamline at Soleil Synchrotron. Pressed AB pellets. b NaCl

Table S3: Comparison of room-temperature vibrational spectra of AB. Infrared

S22

KBr 3318 3252 3198 2384 2332 2280 2219 2120 1607

s m w,sh sh s m w w m

Transmission KBra KCl 3320 s 3318 3253 m 3250 3196 w,sh 3200 2391 s 2366 2321 s 2342 2282 s 2282 2217 w 2219 2116 w 2117 1606 m 1613 1406

1379 s

1377 s

1163 s

1180 w,sh 1160 s

1066 m

1064 s

797 vw 782 w 727 w

926 888 797 782 727

a Aged

m m w w w

Raman ATR System I System III 3307 s 3309 s 3250 m 3254 m 3194 w 2345 sh 2376 sh 2317 s 2323 s 2278 s 2217 m 2208 m 2117 w 1600 m 1600 m 1444 vw 1385 vw 1373 m 1375 s

System I 3316 w 3252 s 3174 w 2377 s 2330 w 2280 vs

System III 3319 w 3251 s 3174 w 2375 s 2330 w 2281 vs

2121 1598 1445 1385 1369

2120 vw 1597 w overtone ? 1369 w

s m sh sh s m sh w m s

NaCl 3322 s 3251 m 3183 sh 2380 sh 2333 s 2281 m 2216 w 2116 w 1620 m 1435 s

1381 s 1316 s 1258 sh

1376 vs 1360 s 1248 w

1167 m 1100 m 1069 m 1026 sh 927 m 886 m 805 vw 782 w

1165 m

1155 s

1156 s

1180 m 1155 m

1073 m 1028 m

1056 s

1061 s

1063 vw

1180 m 1150 m ? 1063 vw

789 m 774 s 720 w

? ? 789 774 s 720 w

785 w 725 vw

797 vw 782 w 728 w

783 w 727 w

AB.

b According

to 24,46,87,91 to Goubeau and Sawodny, symmetry of this vibration is E 24,91 , while for Taylor it is A1 46 . Hess recognises by Raman spectroscopy two resolved bands, one at 1189 cm−1 of E symmetry, and another at 1155 cm−1 of A1 symmetry 87 . d According to Taylor, symmetry of the band is A 46 . 1 e Hess assigned it to ν( 10BN) 87 . f Hess assigned it to ν( 11BN) 87 . c According

vw w vw vw w

Assignationb ν(NH) ν(NH) overtone ? ν(BH) ν(BH) ? ? δ(NH3 )

δ(NH3 ) ? ? δ(BH3 ) δ(BH3 )

Symmetry E A1

E A1

E

A1

E A1 c

ρ(NBH) ?

Ed

m ν(BN)e ν(BN)f ρ(BH3 )

A1 A1 E

S4

Powder XRD patterns

Figure S7: Powder XRD pattern of 1 : 1 AB : KBr mixture at room temperature, as compared with patterns of KBr 92 and AB, respectively. Powder XRD pattern of system 3, after 30 min ball milling, is compared with those for neat AB (i.e. system 1) and neat KBr 92 . As evident from Fig. S7, the resulting pattern is a linear combination of XRD patterns for individual components, so system 3 is a physical mixture of AB and KBr.

S23

S5

Influence of sample preparation method

A possible influence of preparation method to final AB / KBr mixture is also followed by measuring ATR IR spectra of the samples prepared by applying various pressures for pelletizing, by grinding the mixture in an agate mortar during various times and finally by ball milling the mixture in high-energy conditions for 30 min.

Figure S8: Dependence of pressure applied during the pelletizing of sample 3.

Figure S9: Dependence of grinding time, as well as prolonged (30 min) ball milling on sample 3. It is evident that the samples are not significantly influenced by preparation conditions, particularly when common grinding in an agate mortar for 1 min and pelletizing pressure of 100 kN are applied. However, some sensitivity of BH related modes is evident. Although all the band positions remain intact, the intensities of these bands change with respect of grinding time and pressure, respectively. Specifically, application of high pressure and prolonged grinding, especially high-energy ball milling causes: 1. exchange of relative intensities of νas (BH) and νs (BH) modes; 2. intensification of δ(BH3 ), as well as δ(NH3 );

S24

3. ρ(NBH) band is also intensified, although in a smaller extent with respect of above described bands; 4. all of these bands are sharpened, especially after prolonged ball milling, which indicates an enhanced crystallinity of the sample.

S6

Thermal decomposition of ammonia borane

Thermal decomposition of AB is very complex process which can undergo a wide variety of mechanistic pathways (see, for example, the scheme given by AlKukhun et al. 27 ). A thorough theoretical investigation is given by Matus et al. 93 . The schematic view of these processes is illustrated in the following scheme 94 . 1st step: H H H H B H

B

⊕

H

H

H

N

N

H H

H H

H [NH3 BH2 NH3

] + [BH

4]

H

2H2

–

2

DADB

N

H

B H

2H2 H2 B

NH2

H2 N

BH2

PAB chain polymers CDAB AB AB H2 H2

BH2

BH3 NH2 BH

NH2

NH2

BH2

H2 N

NH2

H2 B

BH2 NH2

ECB

CTB

On the basis of 11B and 15N NMR spectroscopy, Shaw et al. shown that the thermal decomposition of AB in glyme1 at 50 − 95 ◦C proceeds via DADB. This O 1 glyme

=

O

S25

result is interpreted in terms of polarity of the used solvent, which can be also applied for solid-state decomposition of AB. More important, all intermediate species predicted by high-level DFT study by Zimmerman et al. 95 are observed in this experimental study. This strongly enables the use of these computational results as an accurate basis for interpretation of our findings. Thus, we underline here a few of their findings of importance to our study. Since NH2 BH2 is a highly reactive species, it is very hard to detect it experimentally under conditions in which AB decomposition practically occurs. However, this species plays a crucial role in decomposition of AB. This high reactivity of the crucial intermediate also gives rise to very complex network of interconnected mechanistic pathways. Zimmerman et al. organised their discussion as answers to three crucial questions, so we also follow this organisational logic, although for the purposes of this study only the first two questions are important. 1. How do AB and aminoborane (NH2 BH2 ) react? They directly form oligomeric NH3 BH2 NH2 BH3 with a release of 1 H2 . Further oligomerisation is very fast and has a very low barrier. Thus, the production of linear and branched oligo- and polymeric PAB species is prefered at this stage. Additional NH2 BH2 can be produced as a consequence of partial dissociation of AB into NH3 and BH3 . BH3 can dimerise into diborane B2 H6 , but it can also catalyse dehydrogenation of AB into NH2 BH2 . Indeed, as seen in Fig. S37, release of H2 is followed by release of NH3 , while the m/z = 14 signal due to BH3 or m/z = 28 (diborane) are not observed. This indicates a significant dissociation of AB with simultaneous consumption of BH3 for dehydrogenation of AB. At higher temperatures, the autocatalytic formation of NH2 BH2 proceeds to produce even more NH2 BH2 , which is applicable to uncatalysed decomposition of AB. 2. How do aminoborane species oligomerise apart from catalytic centers? The reaction of 2NH2 BH2 immediately leads to cyclic dimer CDAB (see the scheme above) and linear LDEB. Due to very similar formation barriers, these two reactions are competitive. However, CDAB quickly decompose again into 2NH2 BH2 , so the dominant product is LDEB, which is quickly trapped by AB or NH2 BH2 forming LDAB, ECB or ECbB. The last product readily further oligomerise with additional NH2 BH2 . Thus, ECB is the dominant product, which can further lead to slower isomerisation into CTB. In the study by Shaw et al., the most stable intermediates, predicted by Zimmerman et al. 95 , namely CDAB, ECB and CTB are identified, with their well resolved evolution 94 . Expected LDAB was not observed, indicating that its decomposition occurs simultaneously with formation. In the second dehydrogenation step, polyaminoboranes PAB release an equivalent amount of H2 producing polyiminoboranes PIB:

H2

NH2

BH2

NH

n

PAB

BH PIB

S26

n

Table S4: Crucial species formed during the 1st step of thermal decomposition of AB, as predicted by 95 .

Acronim DADB

Structure +

[NH3 BH2 NH3 ] [BH4 ] H2 B NH2

CDAB LDEB

H2 N NH2

NH2

H3 B

58

BH NH3

H3 B BH3 NH2

ECB

Notes

58

BH2

NH2

LDAB

m/z 62

–

60

BH2

BH

NH2

NH2

BH2

87

IR spectrum reported by Pons et al. 85 . Dominant bands: 3260 cm−1 , 2365 cm−1 , 2315 cm−1 , −1 1577 cm

87

Contains bridging H, that should give rise to IR band around 2100 cm−1 .

NH2

ECbB

NH2 BH2 NH2

BH

BH2 H

BH2 H2 N

CTB

NH2

H2 B

87

BH2 NH2 BH

HN

B

NH

HB

81

BH NH

S27

Also, various pathways lead to formation of borazine B and different fused hexagonal rings, which are altogether precursors of boron nitride BN.

S7

Changes in solid-state decomposition products

Published TG curves occasionally show an increase of the mass of the decomposed AB above 150 ◦C (see for example 64,96 )). However, we did not find any satisfactory explanation of this observation, so we decided to cosider this in more details. Our TG experiments show the same behaviour (main text, Fig. 3). Above 120 ◦C, TG for 1 shows a considerably lower loss of the mass when heated in air (10 %) with respect of N2 (25 %). After that, the mass of the 1 slowly, but constantly rise at > 170 ◦C. Heating of somewhat bigger amounts of 1 and 3 in ambient atmosphere in oven up to 150 and 200 ◦C, respectively, resulted in vigorous dehydrogenation followed by ignition of hydrogen and, evidently, boron-containing side-products (green flame). This is recorded and available as supplementary movie. For both 150 and 200 ◦C products, the loss of the mass of AB is between 50 and 60 %. Visually, the products are solid foams (Fig. FOAM!!!!). However, when smaller mass of AB is used, the amount of released hydrogen is not enough for autoignition, so the measurements with lower amounts of AB (as used for all other temperature-dependent measurements in this study) are more reliable with respect of understanding of reaction mechanisms. Transmission IR spectra for 2-125 show that ν(BH) envelope does not change considerably with time. For 1-150 a faster increase of these two spectral features with respect of mass increase is observed. On the other hand, both 2-150 and 3-150 show a faster mass increase with respect of ν(BH). Again, for 3-150, the same behaviour of ν(NH) and ν(BH) is observed, while 2-150 does not show any significant change in ν(BH) region. For X-200, no significant increase of the ν(NH) and ν(BH) region near surface with respect of mass increase is observed. However, system 2-200 shows a well correlated increase of ν(NH) and again no significant change in ν(BH) region. This indicates that near the surface AB is almost completely converted to BN, while AB is still not completely decomposed in the bulk. δ(NH3 ) features significantly change for all the observed products of the 2, which is not observed for 1 and 3. At 125 and 150 ◦C a 1520 cm−1 feature decreases with time with simultaneous growth of the 1620 cm−1 feature, characteristic for pristine AB. Also, the same behaviour is observed for 1325 cm−1 band. However, a drop of this band with time is observed for system 2-200. 1250 cm−1 band in all the three cases shows the same behaviour, i.e. it drops with time. Although the fingerprint region for 2-150 and 2-200 are significantly different, practically equal temporal behaviour of this region for these two systems is observed. The decomposition step for 3 that leads to observed mass drop of 13 % occurs in temperature range 100 − 135 ◦C, continued by a slow drop till 180 ◦C, not observed in N2 . A further drop of the mass is evident above 200 ◦C. When heated in N2 , the mass of the sample remains almost constant, although a small increase of the mass is observed at > 170 ◦C, and its decrease begins at

S28

Figure S10: Time-dependent increase of the mass of products 1-150, 3-150, 1-200 and 3-200. The slope of the linear part of the curves in the initial period (< 10 min) are given along with the legend. temperatures as high as > 230 ◦C. Exposition to ambient conditions result in a rapid time-dependent mass increase, as shown in Fig. S10. Practically the same slope of the linear initial parts of all the curves is observed, i.e. it is (0.0660 ± 0.0053). Thus, all of these curves are practically the same, and this is caused by the fact that the released hydrogen ignites due to the high temperature, causing overheating of the samples. After the individual measurement of the m = f (t) curve, the 1-150 sample was heated and weighed again in five cycles (Fig. S11). After an initial steep ingrease of the mass, sample becomes saturated, which is reflected by plateau. As evident, the process is highly reproducible, and the capacity of the sample increases with cycle number, although the rate of mass increase readily reaches a constant value, as evident from the inlet graph. Additionally, system 1 was heated in TG furnace up to 150 ◦C, and then the heating was stopped for 30 min in air and in N2 atmosphere. In both cases, after some time in which the mass continues to fall, the trend inverts, i.e. an increase of the mass was observed. Surprisingly, this increase is more rapid in N2 with respect to air (Fig. S12). It is interesting to find out what is the molecular background of the above

S29

Figure S11: Time-dependent increase of the mass of product 1-150. Inlet graph shows changes in slope of the initial steep linear (< 10 min) change in mass. The measured values are normalised with respect of the initial mass at t = 0 min. described mass increase. IR spectroscopy is of great help with this regard, so we did a series of dedicated IR monitoring experiments. The variability of the decomposition products of 1 with respect of the atmosphere in which they were stored is illustrated by Fig. S15. This variability should be correlated with the observed increase of the mass with time. These observations motivated the monitoring of temporal evolution of the products X-125, X-150 and X-200 by IR spectroscopy in both transmission (in this case X = 2) or ATR mode (in this case X indicates 1 or 3). Temporal evolutions of the products, as obseved by time-dependent IR spectroscopy, are shown in Figs. S16-S24. Spectra for 1-125 (Fig. S16)and 3-125 product (Fig. S17) show common behaviour of the ν(NH) 3240, ν(BH) 2279, 1565 snd 1390 cm−1 features. However, a considerably different behaviour of all of these features in observed for 2-125 (Fig. S18). The most prominent difference is in ν(BH) region, which remains relatively intact by exposition to air for 2-125. For this sample, the most intense change occurs in the ν(NH) region and for 1325 and 1250 cm−1 bands, that simultaneously grow and drop, respectively. The rest of the spectrum only moderately changes. For 1-125 a slight drop in intensity is observed for 1495 and 1292 cm−1 features. 3-125 shows temporal development of some new features, represented as shoulders, at 2315, 1333 and 808 cm−1 . It should be underlined that fingerprint region is significantly different for 1-125 with respect to 3-125, which indicates that KBr affects some of the processes on surface. Obviously, significantly different behaviour for 2-125 indicates that the majority of the processes for X-125 occur at or near the surface of the sample. All the considered systems X-150 (Figs. S19-S20) show a 3440 cm−1 band, which is assigned to polyimineborane (NHBH)x (PIB), which is not affected S30

Figure S12: Time-dependent mass of the product 1-150 at 150 ◦C in air and N2 . by exposition to air. This band for 1-150 (Fig. S19) is represented only as a weak feature, indicating relatively small amount of PIB at the surface of this sample, but it is very abundant in the presence of KBr. The rest of the spectra are mutually significantly different for these two systems. On the other hand, 1-150 show a complete absence of the ν(BH) region, indicating a complete decomposition of the 1 to BN at this temperature. This indicates lowering of the temperature of PIB production by action of KBr. Additionally, it is important to note a high similarity of the 1-150 to 3-125. For these two systems, all the most important bands are equal. This strongly indicates that ionic character of KBr significantly affects the decomposition of the AB by lowering the dehydrogenation temperature. The growth of the bands in 2500 − 2200 cm−1 region is completely due to absorption of CO2 from air. Differences in 3-200 (Fig. S23) with respect to 2-200 (Fig. S24), and similarity of 3-200 with 1-200 indicate that at 200 ◦C 3 is completely decomposed to BN at the surface, while in the bulk it is still not completely dehydrogenated. The presence of the PIB band at 3440 cm−1 for 2-200 indicate its partial preservation in the bulk of the samples. ν(NH) and ν(BH) region, and the temporal behaviour of these regions, of 2-200 are similar to those for 2-150, indicating a slow decomposition of PIB to BN inside the sample. This is supporte3d by observation of the differences in fingerprint region of these two systems, and comparison with the other systems.

S8 S8.1

Temperature-dependent spectroscopic measurements IR spectroscopy

At room temperature transmission and ATR spectra of all the considered systems are practically equal. The only differences result from optical differences between transmission and ATR experiment. In the temperature range under 100 ◦C no significant spectral changes were observed in all cases. Fig. S26 shows temperature-induced changes in positions of the most important stretching modes, namely ν(NH), ν(BH) and ν(BN). In-situ temperatureS31

Figure S13: Mass increase in time at ambient condition of the X-150 and X-200 systems, with slope of the initial phase. The moment of the end is determined as intercept of the extrapolated lines fitted to plateau and initial phase, respectively.

S32

Figure S14: Comparison of time dependent normalised absorbance of the ν(NH)as and ν(BH)as bands of X-125 systems with mass increase for X-150.

S33

Figure S15: Transmission and ATR IR spectra of decomposition products of 1, cooled to r.t. after heating. The conditions are as follows: X-200 - in air, 2 ◦C min−1 , up to 200 ◦C; X-250-A - in air, 5 ◦C min−1 , up to 250 ◦C; X-250-B - in N2 , 5 ◦C min−1 , up to 250 ◦C; X-150-A - in 20 bar N2 overnight at 150 ◦C; X-150-B - 2 days in N2 flow at 150 ◦C; X-150-C - 5 h in vacuum at 150 ◦C and then 5 days in 1 bar CO2 at r.t. dependent IR spectra, obtained during the course of this study, show first significant change around 100 ◦C. IR bands of both 2 and 3 drop in intensity, while ATR spectrum for 1 remains practically intact. However, the changes are more prominent in the case of 2, especially in the ν(NH) and ν(BN) region (see Table S3). The well resolved bands due to ν(NH) and ν(BN) are significantly broadened giving rise to broad, unresolved envelopes (Figs. S23-S25). Simultaneously, intensity of the ν(BH) region drops without a significant broadening. In the same time, ATR spectra of 1 remain practically intact, while for 3 only a moderate drop in intensity of all spectral bands is not accompanied with any significant broadening of ν(NH) region. Frueh et al. 64 have recorded and assigned solid-phase products of thermal dehydrogenation of AB, as obtained at 120, 210 and 1500 ◦C, respectively. On the other hand, Baumann et al. 52 have followed by combined thermal and spectroscopic measurements thermal decomposition of polyaminoborane (PAB). Additionally, the published IR spectra of cyclic pentameric 41 and polymeric aminoborane 42,58 may be of some additional help. Although both studies are not in situ investigations, the comparison of spectra published in these papers with those obtained in present work is very helpful to understand the changes observed by the present in situ IR spectroscopic measurements. Spectrum of the solid product obtained after release of first equivalent of H2 (120 ◦C product) indicates that it consists mainly of tetrahedral, singly bonded N and B atoms, i.e. it corresponds to polyaminoborane (PAB) family of structures with a general formula −[N H2BH2]x −, where x ≤ 4 52,96 . This solid polymeric product is noncrystalline and does not have a unitary structure. Rather, it should be described as a mixture of various species with common −[N H2BH2]x − composition 52 . It is also noted that branched and cyclised PAB species should have very similar IR spectra, so they cannot be resolved

S34

here. Frueh et al. 64 assign the decrease of original ν(NH) and ν(BH) as a reflection of the loss of high symmetry stretching modes of pristine AB that cannot exist in PAB. In the temperature range between 100 and 125 ◦C, regardless the system, intensity of the ν(NH) related 3310 cm−1 band drops with simultaneous appearance of a broad envelope in the 3400 − 3000 cm−1 region. However, this envelope is somewhat broader for 2 and 3, and in these cases it reaches a maximum at 3280 cm−1 , while for 1 it peaks at 3243 cm−1 . The ν(BH) region also suffers significant changes. For 1, 2315 cm−1 band drops with occurrence of a broad feature centered at 2280 cm−1 . System 2 show a rapid drop of 2340 cm−1 band, while the rest of the region remains relatively intact. The whole ν(BH) region of the system 3 is unchanged over this temperature region. For all the considered systems δ(NH3 ) bands at 1600 and 1370 cm−1 , respectively, drop in intensity. However, 1 shows simultaneous development of the 1560 cm−1 feature, which arises at somewhat higher temperatures for 3. 1370 cm−1 feature suffers a significant broadening in all cases. The feature at 1155 cm−1 , assigned to δ(BH3 ), also drops with simultaneous development of a broad feature centered at 1240 cm−1 . For 1, a broad and very intense band centered at 1140 cm−1 arises, which is observed at higher temperatures for 2 and 3. In all the cases ρ(NBH) band at 1055 cm−1 decreases with temperature, but this drop is somewhat slower for 3. For 2 a band at 1126 cm−1 simultaneously appears and increases. In the case of 2 and 3, only a slight shift of ν(BH) towards the higher wavenumbers, without significant loss of the substructure, is observed. Although intensity of the feature slowly but continuously drops with temperature, for 2 its structure remains almost unchanged even at the temperature as high as 200 ◦C. This indicates formation of extended and stable polyborane (Bn Hm )x polymers in the bulk of the sample. Fingerprint region of 2 and 3 are characterised by good resolved bands, which again differ from dehydrogenation products of 1. However, it should be noted that the 1300 − 800 cm−1 region of ATR spectrum of 3 is considerably more intense with respect of transmission IR spectra of 2. Also, a very good resolved transmission IR band at 1125 cm−1 is represented in ATR spectrum of 3 only as a very weak feature. Another characteristic band of 3 at 859 cm−1 is attributed to ν(BN) mode. However, it is not observed in spectrum of 1125 and 2-125, respectively. The broad and very intense band (dominant at 120 ◦C) is attributed to multiple unresolved δ(BH2 ) and terminal BH3 umbrella modes. 97 Fig. S27 shows comparison of IR spectra at 100 and 120 ◦C. In this temperature range first step of AB dehydrogenation should occur, giving rise to PAB. Together with a dominant ν(NH) feature peaking at 3250 cm−1 , Frueh et al. 64 have reported a weak 3437 cm−1 band, which is assigned to ν(NH) of π bonded segments −[N H = BH]x − of terminal NH2 groups of linear PAB. IR spectrum published by Baumann et al. 52 shows a strong and broad feature peaking at 3450 cm−1 . Spectra obtained during the course of this work does not show any significance of this feature in this temperature region (Fig. S28). Thus, we conclude that in situ monitoring of dehydrogenation process of both neat AB or mixture of AB with KBr shows that exclusively singly bonded PAB species are formed up to 120 ◦C. The region corresponding to ν(BH) vibrations suffer a significant changes in this temperature region (Fig. S29). Above 120 ◦C, the spectra of 2 are qualitatively almost intact, and the S35

changes are limited only to a slow and simultaneous drop of intensity of all bands. The only exceptions include the appearance and increase of the 3430 cm−1 band assigned to PIB above 150 ◦C, as well as disappearance of the 1600 cm−1 band. The rest of the spectra of 2 are characterised by a broad 3553 cm−1 feature due to ν(NH), the envelope due to ν(BH) consisting of at least 4 individual bands. Fingerprint region consists of a series of bands, common for all the three observed systems. However, the 1285 − 1000 cm−1 region is better resolved for 2, while this spectral region of the system 3 becomes equal to 2 only slightly bellow 200 ◦C. For 1, PIB band at 3430 cm−1 appears as low as 130 ◦C and after that it continuously incrases, together with a drop of 1560 cm−1 band. Both ν(NH) and ν(BH) features drop with temperature for 1 and 3 and they practically does not exist at 200 ◦C. Spectra of 1 and 3 behave equally in the 140 − 160 ◦C region. Their fingerprint region consists of δ(NH3 ) band at 1560 cm−1 , which drops in intensity. Very broad and intense envelopes peaking at 1415 and 1130 cm−1 continuously rise and drop, respectively, in intensity. However, for 1 a feature at 1225 cm−1 rises with a drop of 1130 cm−1 band, while the envelope around 880 cm−1 merges into an intense band at 890 cm−1 . Above this temperature region, spectrum of 3 evolves into a spectrum equal to 2, while in the spectrum of the system 1 the band at 1385 cm−1 mainly disappears with simultaneous increase of 1230 cm−1 and a slight drop of 900 cm−1 band and merging of the bands in 1160 − 1000 cm−1 region. For PAB, Jacquemin 97 predicted on the basis of high-level ab-initio calculations, that the umbrella mode of terminal NH3 groups should be located at 1380 cm−1 . This feature, observed for pristine AB, is broadened at 120 ◦C, which is attributed to formation of nonuniform PAB species. Another characteristic band, which appears in ATR spectrum of 3 at 859 cm−1 is attributed to ν(BN) mode. Appearance of this band at 850 cm−1 is also reported by Frueh et al. 64 . However, it is not observed in ATR spectrum of the dehydrogenation product of 1, nor in the transmission spectrum of 1 in KBr pellet, as evident from Fig. S30. The broad and very intensive band (dominant at 120 ◦C) is attributed to multiple unresolved δ(BH2 ) and terminal BH3 umbrella modes 97 . The ν(NH) region remains unchanged with respect to that as observed at 100 ◦C for transmission spectra. However, this region is significantly changed for ATR spectra of both 1 and 3, but these changes are very different, clearly reflecting different processes. Namely, ν(NH) for 3 becomes the same as that observed by transmission. Simultaneously, dehydrogenation of neat AB leads to formation of species which give rise to ν(NH) region peaking at 3245 cm−1 . Similarly, ν(BH) region behaves practically in the same way for 2 and 3, but becomes considerably different for 1. Fingerprint region of 2 and 3 are characterised by rather good resolved bands, which is again dramatically different than that of dehydrogenation products of 1. However, it should be noted that the 1300 − 800 cm−1 region of ATR spectrum of 3 is considerably more intense than that as observed by transmission IR spectroscopy of 2. Also, a very good resolved transmission IR band at 1125 cm−1 is represented in ATR spectrum of 3 only as a very weak feature. Above 120 ◦C, gaseous products, namely borazine (NHBH)3 , together with small quantities of diborane B2 H6 in addition to hydrogen were recognised by combining IR and mass spectrometry 52,96,98 . Although our IR spectroscopic experiments are designed to recognise solid products only, it is important to have in mind also these gaseous products for the sake of interpretation of thermoS36

dynamic measurements, which will be discussed later. At 140 ◦C IR spectrum of system 1 resembles the one published by Baumann et al. 52 . This implies that the formation of PAB species ends in this temperature region, under the present experimental conditions. However, a large difference in intensity of the 1450 cm−1 band (i.e. relatively low intensity of this band as observed in the present work) indicates that in the present conditions highly branched or cyclic PAB species are preferably formed. Above 140 ◦C, the main changes include decrease and, at 200 ◦C almost complete dissappearence of both ν(NH) and ν(BH) features, which is consistent with formation of polyiminoborane (PIB) species of approximative composition −[N HBH]− and finaly NBHy (y ≤ 2). Only a moderate changes are observed in the whole 120 − 200 ◦C region for 2. Qualitatively, both ν(NH) and ν(BH) regions remain almost unchanged, although they continuously decrease in intensity, and the 3430 cm−1 band appears only as a weak feature. Simultaneously, a broad envelope peaking at 1415 cm−1 continuously rise in intensity. The rest of fingerprint region is intact (both qualitatively and quantitatively) in the whole temperature region. Spectra of 3 indicate dramatic changes in the 120 − 140 ◦C region. ν(NH) and ν(BH) regions are broadened, while in the fingerprint region good resolved bands at 1164 cm−1 and 1057 cm−1 merge in a broad envelope peaking at 1164 cm−1 . Although it evidently consists of the features characteristic for both 1 and 2, they are unresolved. Simultaneously, two bands appear, peaking at 1568 cm−1 (which corresponds to the feature observed in the 1 case) and at 1400 cm−1 (which corresponds to the feature observed in both 1 and 2 case). Above 160 ◦C, the broad unresolved feature in the 1320−940 cm−1 region drops with simultaneous increase of the feature peaking at 1400 cm−1 . Finaly, at 200 ◦C the whole fingerprint region of system 2 and 3 become qualitatively equal, while that for 1 is rather different. However, ν(BH) region for 1 and 3 evolve very similarly, ending in rather broad band which peaks at 1980 cm−1 and relatively good defined one at 2147 cm−1 . On the other hand, the evolution of ν(NH) band is similar in the case of system 2 and 3. In all cases, no evidence of the NH and BH terminal groups are found in present IR spectra, since no features around 3700 cm−1 and 2800 cm−1 are observed. Fingerprint region of the present IR spectra should be highly dependent on various conformational differences of the products of thermal decomposition, such as cis-trans isomerism, degree of crystallinity, exact stoichiometry, branching, existence and amount of cyclic structural motives. Thus, it is impossible to give an unequivocal assignment of these spectra. Rather, the three systems, which give the same IR spectra at room temperature, undergo rather different decomposition pathways, which are reflected in significantly different temperature-dependent evolution of their spectra. Additionaly, it is important to note that not only the sample, but also experimental setup must be taken into consideration as an inseparable part of the system under consideration.

S37

Figure S16: Time-resolved ATR IR spectra of the product 1-125. Upper picture shows original spectra, below are the differential spectra, obtained by subtracting the first spectrum (t = 0 min) from the others, and the third picture is 2D representation of differential spectra.

S38


S39

Figure S18: Time-resolved transmission IR spectra of the product 2-125. Upper picture shows original spectra, below are the differential spectra, obtained by subtracting the first spectrum (t = 0 min) from the others, and the third picture is 2D representation of differential spectra.

S40

Figure S19: Time-resolved ATR IR spectra of the product 1-150. Upper picture shows original spectra, below are the differential spectra, obtained by subtracting the first spectrum (t = 0 min) from the others (black lines represent the spectra obtained for > 30 min), and the third picture is 2D representation of differential spectra.

S41


S42


S43


S44


S45


S46

Figure S25: Temperature-dependent (a) transmission (KBr pellets) and (b) single-reflection ATR (neat AB) IR spectra.

S47

Figure S26: Temperature-dependent positions of stretching vibrational bands.

S48

Figure S27: IR spectra at (a) 100 ◦C and (b) 120 ◦C.

S49

Figure S28: ν(NH) region of variable temperature IR spectra in the 100−120 ◦C temperature range: (a) ATR spectra of system 1; (b) transmission spectra of system 2; (c) ATR spectra of system 3.

S50

Figure S29: ν(BH) region of variable temperature IR spectra in the 100−120 ◦C temperature range: (a) ATR spectra of system 1; (b) transmission spectra of system 2; (c) ATR spectra of system 3.

S51

Figure S30: Fingerprint region of variable temperature IR spectra in the 100 − 120 ◦C temperature range: (a) ATR spectra of system 1; (b) transmission spectra of system 2; (c) ATR spectra of system 3. S52

Figure S31: IR spectra at (a) 140 ◦C, (b) 160 ◦C and (c) 200 ◦C.

S53

Figure S32: Variable-temperature transmission spectra of 2. On the right is temperature dependence of baseline absorption measured at 2000 cm−1 .

Figure S33: Temperature dependence of 1380 and 1166 cm−1 band in the case of sample 2.

S54

S8.2

TG-IR spectroscopy of gaseous decomposition products

TG-IR spectra of gaseous products of thermal decomposition of systems 1 and 3 are shown in Figs. S34 and S35, respectively. For system 1, NH3 related bands are well correlated to ν(BH) feature. Altogether, at heating rate 2 ◦C min−1 , they appear in the 135 − 165 ◦C region, peaking at 146 ◦C. After that, a steep decrease in intensity is observed, resulting in disappearance of the bands. On the other hand, intensity of a broad band centered at 1230 cm−1 shows a sigmoid behaviour. It starts to increase in intensity slower but simultaneously with the other observed bands. After a moderate shoulder at 156 ◦C, it continues to grow up to 170 ◦C when it reaches a plateau. TG-IR of system 1, recorded at 5 and 10 ◦C min−1 , and the corresponding intensity profiles are shown in ESI. TG-IR spectra of system 3 show the thermally induced evolution of the NH3 related bands and 1230 cm−1 band. They start to rise at 105 ◦C with equal trend. At 120 ◦C, the intensities of NH3 features reach their maximal value, while the 1230 cm−1 band continues to rise until it reaches a plateau at 130 ◦C. The plateau intensity is 2× as compared to that at 120 ◦C, which corresponds to that as observed for system 1. For 930 cm−1 band, a moderate increase is observed above 155 ◦C.

S8.3

Raman spectroscopy

Equal Raman spectra of 1 and 3 are observed at room temperature, indicating once again that no interaction between AB and KBr is introduced by mechanical mixing of the two components (see Fig. 1 in the main text).

S9

Mass spectroscopy

The results of fitting of EGA-MS curves, given in Tables S5 and S6, show that the release of H2 is accompanied by emission of various byproducts, which is well known. The relative amounts of these products may be different in individual steps (Table S6), but the release of their 2nd equivalents begin imediatelly after the 1st step reaches its peak. The first drop of the mass in TG (see TG presented in the main text) corresponds to hydrogen and ammonia with a minor contribution of borazine. However, borazine significantly contributes to the second drop of the mass for 1. For 3, only a minor amount of borazine is released. In the second decomposition step, the emission of NH3 is in a great extent suppressed for 3 and it is spread over a significantly broader temperature region. Diborane signal at m/z = 28 could not be detected.

S55

Figure S34: TG-IR spectra of system 1.

S56

Figure S35: TG-IR spectra of system 3.

S57

Figure S36: Variable temperature Raman spectra of (a) system 1 and (b) system 3.

S58

Table S5: Temperature ranges (99 % of the area under fitted Lorentzian curve) (in ◦C) in which the gaseous species are released from 1. 1st step m/z

assignation max.

S59

2 17 28 43 58 80

H2 NH3 B 2 H6 NH2 (BH2 −H−BH2 ) cyclo-(NH2 BH2 )2 cyclo-(NHBH)3

116 115 112 106 105 117

N2 range 100 − 132 107 − 123 93 − 129 93 − 124 91 − 124 110 − 124

max. 121 121 121 111 110 123

2nd step air range 105 − 137 108 − 134 104 − 136 102 − 127 94 − 121 114 − 132

max. 148 158 143 — 158 154

N2 range 110 − 186 126 − 190 117 − 165 — 132 − 194 123 − 185

max. 154 156 157 — 162 157

air range 123 − 185 122 − 190 120 − 192 — 135 − 182 127 − 187

Table S6: Ratio of the released gaseous species from sample 1 in N2 with respect of the air atmosphere at 115 and 150 ◦C.

m/z

assignation

1 N2

S60

2 17 28 43 58 80

H2 NH3 B 2 H6 NH2 (BH2 −H−BH2 ) cyclo-(NH2 BH2 )2 cyclo-(NHBH)3

a Area b I(N

118.1 ± 2.9 20.6 ± 1.0 1.1 ± 0.1 0.7 ± 0.04 0.1 ± 0.01 0.6 ± 0.1

st

EGA-MS intensity (I/10−9 )a step 2nd step b air x N2 air

108.5 ± 3.1 13.1 ± 0.3 3.7 ± 0.2 0.7 ± 0.05 0.1 ± 0.01 0.8 ± 0.06

under the fitted Lorentzian bandshape function.

2 )/I(air)

1.08 1.57 0.30 1.00 1.00 0.75

151.8 ± 9.0 31.7 ± 2.8 1.1 ± 0.2 — 0.1 ± 0.01 8.7 ± 0.2

180.9 ± 10.8 15.9 ± 1.0 7.2 ± 0.4 — 0.1 ± 0.01 6.1 ± 0.1

x 0.84 2.0 0.15 — 1.00 1.43

I(1st )/I(2nd ) N2 air 0.78 0.65 1.0 — 1.00 0.07

0.60 0.82 0.51 — 1.00 0.13

Table S7: Intensities and temperature ranges (99 % of the area under fitted Lorentzian curve) (in ◦C) in which the gaseous species are released from 3 in air.

S61

m/z

assignation

2 17 43 80

H2 NH3 NH2 (BH2 −H−BH2 ) cyclo-(NHBH)3

max. 111 111 102 116

1st step range 95 − 128 86 − 130 58 − 130 102 − 125

−9

I · 10

171.8 ± 5.9 24.2 ± 0.6 0.7 ± 0.1 0.2 ± 0.006

max. 168.3 167 167 147

2nd step range 128 − 205 130 − 195 145 − 180 130 − 195

I · 10−9

I(1st )/I(2nd )

242.3 ± 5.4 8.9 ± 1.2 0.07 ± 0.03 0.1 ± 0.03

0.71 2.72 10 2

S10

Thermodynamic measurements

TG/DTA curves for the considered systems are given in Fig. 2 of the main text. They reflect two-step dehydrogenation process. It is expected that AB loose 6.5 % of the initial mass per step if only H2 is released. However, in the case of system 1, in both air and N2 atmosphere, a significantly greater loss of the mass is observed. In both atmospheres, system looses 17 % of the initial mass during the first dehydrogenation step. The second step is in great extent atmosphere dependent. In air, it looses additional 10 % of the mass, while in N2 27 % is lost. The first hypothesis is that in air the loss of the mass is compensated by oxidation of the solid product or by adsorption of some of the species present in the air. In order to explain the observed behaviour, the gaseous products are followed by TG-IR and TG/DTA-MS, while the solid products are followed by ATR and transmission IR spectroscopy. Mass spectroscopy (Fig. S37) reveals significant atmosphere dependent differences in emissions (Table S6, Fig. S38). Namely, H2 release is less efficient in N2 atmosphere in both steps. On the other hand, N2 improves the release of NH3 , especially during the second step. During the first step, only a minor amount of borazine is emitted in both air and N2 . This emission is quite preffered in air. In the second step, a significantly more efficient release of borazine is observed in N2 . EGA-MS gives ion current I as a directly measured quantity, and it is directly related to molar quantity of given species. This allows determination of the quantity µ directly related to mass, by formula: µ=I ·M Comparison of the values of µ for TG/DTA-MS data from air and N2 gives no difference in mass of the gaseous products, released in N2 with respect to air atmosphere: µgas released (N2 ) ≈ 1.0 µgas released (air) . Clearly, the difference in TG curves: mtotal loss (N2 ) = 2.7 mtotal loss (air) cannot be attributed to difference in emissions of gaseous products - decomposition process by itself does not depend on the atmosphere. This observation indicates adsorption of some components of the air by the foamy solid product of 1, which occurs even at at temperatures as high as 150 ◦C. This adsorption partialy compensates the loss of the mass in air. The adsorption becomes a predominant process above 170 ◦C, so above this temperature an increase of the mass is observed in TG. Although observed and reported before (see Fig. 5 in Wolf et al. 96 and Fig. 4 in Frueh et al. 64 ), this increase of the mass was not satisfactorily discussed. TG coupled with EGA-MS is in agreement with results of Frueh et al. 64 , and show that the main species released by thermal decomposition of 1 are hydrogen H2 , ammonia NH3 and borazine N3 B3 H6 . Comparison of the observed mass spectra with borazine spectrum (as taken from 64 ) shows that it is the main boron-containing gaseous species released in both first and second dehydrogenation step (Fig. S39).

S62

When cooled down to room temperature and left exposed to air, the mass of the 150 ◦C product (irrespective if heated in air or N2 ) of 1 continues to rapidly increase, reaching a plateau after ∼ 20 min (Fig. S11). Additionally, it would be important to note a quite unexpected shape of TG curves for all samples, as obtained before recrystallisation of AB (Fig. S40). Namely, after 100 ◦C, decrease of the mass is observed, as expected. However, above 110 ◦C, an unexpected behaviour, i.e. steep increase of the mass, is evident. This happens up to 150 ◦C. After that, mass slowly decreases again. It could be imagined that oxidation occurs in air atmosphere, which would explain the strange shape of TG curve. Thus it is surprising to see that the same happened in N2 atmosphere. The similar behaviour is mentioned by Frueh et al. 64 . They also reported a ”normal” TG when recorded in Ar atmosphere, but the observed behaviour is not further explained in that paper. System 3 shows only a minor drop of mass below 100 ◦C, which is followed by an intense increase of the mass (Fig. S40(b)), comparable to that as evident for system 1 at 10 ◦C min−1 (Fig. S40(a)). This interesting observation is by itself the problem which should be investigated in more details, but this is out of the scope of the present paper. However, since the TG experiments were done in conditions similar to those in which IR spectra were recorded, these results must be taken into account to discuss temperature dependent IR spectral changes. Anomalous behaviour indicated by TG reflects quite a complex processes during the thermal decomposition of AB, which, among others, highly sensitive to atmosphere composition. A comparison of high-temperature spectra (Fig. S41) with those of boric acid B(OH)3 99 , boron trioxide B2 O3 100 and sodium borate 99 excludes extensive formation of boron oxides or, eventualy, potassium borate during the heating process. ATR spectra show some evidence of oxide formation at the surface layer of the samples. However, the observed bands could be also assigned to NH4 Br 99 , but also some other species.

References [1] T. Autrey, A. Gutowska, L. Li, J. Linehan and M. Gutowski, ACS Div. Fuel Chem., Prepr., 2004, 49, 150–151. [2] N. Biliˇskov, Spectrochim Acta A, 2011, 79, 295 – 301. [3] N. Biliˇskov, Kem. Ind., 2010, 59, 11–18. [4] R. Jones, D. Escolar, J. Hawranek, P. Neelakantan and R. Young, J. Mol. Struct., 1973, 19, 21 – 42. [5] J. Hawranek, P. Neelakantan, R. Young and R. Jones, Spectrochim. Acta A, 1976, 32, 75 – 84. [6] H. Bennett and J. Porteus, J. Opt. Soc. Am., 1961, 51, 123–129. [7] I. Ohl´ıdal, K. Navr´ atil and F. Lukeˇs, J. Opt. Soc. Am., 1971, 61, 1630–1639. [8] A. S. Toporets, Opt. Spectrosc., 1973, 34, 559. [9] W. A. Pliskin and R. P. Eischens, J. Phys. Chem., 1955, 59, 1156–1159. [10] O. Y. Ataman and H. B. Mark, Anal. Lett., 1976, 9, 1135–1141.

S63

[11] W.-T. Cheng, S.-L. Wang and S.-Y. Lin, Analyst, 2011, 136, 1036–1040. [12] F. Vr´ atn´ y, J. Inorg. Nucl. Chem., 1959, 10, 328 – 332. [13] V. W. Meloche and G. E. Kalbus, J. Inorg. Nucl. Chem., 1958, 6, 104 – 111. [14] M. Pelletier, L. J. Michot, O. Barres, B. Humbert, S. Petit and J. L. Robert, Clay Minerals, 1999, 34, 439–445. [15] J. Chappell, Forensic Sci. Int., 1995, 75, 1 – 10. [16] C. J. Schutte, Spectrochim. Acta, 1960, 16, 1054 – 1059. [17] E. H. Coker and D. E. Hofer, J. Chem. Phys., 1968, 48, 2713–2719. [18] Y. Li, F. Fang, Y. Song, Y. Li, Q. Zhang, L. Ouyang, M. Zhu and D. Sun, Int. J. Hydrogen En., 2012, 37, 4274 – 4279. [19] N. J. Harrick, Internal Reflection Spectroscopy, Interscience, New York, 1967. [20] J. Huang, Y. Tan, Q. Gu, L. Ouyang, X. Yu and M. Zhu, J. Mater. Chem. A, 2015, 3, 5299–5304. [21] B. Roy, J. Manna and P. Sharma, J. Alloys Compds., 2015, 645, S234–S23. [22] L. Wan, J. Chen, Y. Tan, Q. Gu and X. Yu, Int. J. Hydrogen En., 2015, 40, 1047 – 1053. [23] G. Xia, J. Chen, W. Sun, Y. Tan, Z. Guo, H. Liu and X. Yu, Nanoscale, 2014, 6, 12333–12339. [24] W. Sawodny and J. Goubeau, Z. Phys. Chem., 1965, 44, 227–241. [25] S. M. Lee, X. D. Kang, P. Wang, H. M. Cheng and Y. H. Lee, Chem. Phys. Chem., 2009, 10, 1825–1833. [26] P. Jash, K. Meaux and M. Trenary, J. Undergrad. Res., 2012, 5, 1–5. [27] A. Al-Kukhun, H. T. Hwang and A. Varma, Int. J. Hydrogen En., 2013, 38, 169–179. [28] K. J. Fijalkowski, R. Jurczakowski, W. Kozminski and W. Grochala, Phys. Chem. Chem. Phys., 2012, 14, 5778–5784. [29] K. J. Fijalkowski and W. Grochala, J. Mater. Chem., 2009, 19, 2043–2050. [30] K. J. Fijalkowski, R. V. Genova, Y. Filinchuk, A. Budzianowski, M. Derzsi, T. Jaron, P. J. Leszczynski and W. Grochala, Dalton Trans., 2011, 40, 4407– 4413. [31] X. Kang, J. Luo, Q. Zhang and P. Wang, Dalton Trans., 2011, 40, 3799–3801. [32] M. Weismiller, S. Wang, A. Chowdhury, S. Thynell and R. Yetter, Thermochim. Acta, 2013, 551, 110–117. [33] J.-H. Luo, X.-D. Kang and P. Wang, The Journal of Physical Chemistry C, 2010, 114, 10606–10611. [34] Z. Yang, Y. Wang, J. Liang and J. Chen, Mater. Trans., 2011, 52, 651–653. [35] J. Luo, X. Kang, Z. Fang and P. Wang, Dalton Trans., 2011, 40, 6469–6474.

S64

[36] J. Luo, X. Kang and P. Wang, Chem. Phys. Chem., 2010, 11, 2152–2157. [37] J. Luo, H. Wu, W. Zhou, X. Kang, Z. Fang and P. Wang, Int. J. Hydrogen En., 2012, 37, 10750 – 10757. [38] L. Gao, C.-Y. V. Li, H. Yung and K.-Y. Chan, Chem. Commun., 2013, 49, 10629–10631. [39] M. Hu, J. M. Van Paasschen and R. Geanangel, J. Inorg. Nucl. Chem., 1977, 39, 2147–2150. [40] S. Sepehri, A. Feaver, W. J. Shaw, C. J. Howard, Q. Zhang, T. Autrey and Cao, J. Phys. Chem. B, 2007, 111, 14285–14289. [41] K. W. B¨ oddeker, S. G. Shore and R. K. Bunting, J. Am. Chem. Soc., 1966, 88, 4396–4401. [42] D.-P. Kim, K.-T. Moon, J.-G. Kho, J. Economy, C. Gervais and F. Babonneau, Polym. Adv. Technol., 1999, 10, 702–712. [43] G. Moussa, U. B. Demirci, S. Malo, S. Bernard and P. Miele, J. Mater. Chem. A, 2014, 2, 7717–7722. [44] J. D. Carpenter and B. S. Ault, J. Phys. Chem., 1991, 95, 3502–3506. [45] J. Smith, K. Seshadri and D. White, J. Mol. Spectrosc., 1973, 45, 327–337. [46] R. C. Taylor, Adv. Chem. Ser., 1964, 42, 59–70. [47] J. Luo, X. Kang and P. Wang, Energy Environ. Sci., 2013, 6, 1018–1025. [48] J. Luo, H. Wu, W. Zhou, X. Kang and P. Wang, Int. J. Hydrogen En., 2013, 38, 197–204. [49] J. Luo, X. Kang and P. Wang, Int. J. Hydrogen En., 2013, 38, 4648 – 4653. [50] G. Xia, Y. Tan, X. Chen, Z. Guo, H. Liu and X. Yu, J. Mater. Chem. A, 2013, 1, 1810–1820. [51] J. Chen, T. He, G. Wu, Z. Xiong and P. Chen, Int. J. Hydrogen En., 2013, 38, 16199–16207. [52] J. Baumann, F. Baitalow and G. Wolf, Thermochim. Acta, 2005, 430, 9 – 14. [53] Y. Tan, Q. Gu, J. A. Kimpton, Q. Li, X. Chen, L. Ouyang, M. Zhu, D. Sun and X. Yu, J. Mater. Chem. A, 2013, 1, 10155–10165. [54] M. Somer, S. Acar, C. Koz, I. Kokal, P. H¨ ohn, R. Cardoso-Gil, U. Aydemir and L. Akselrud, J. Alloys Compds., 2010, 491, 98–105. [55] Z. Li, W. Liu, H. Yang, T. Sun, K. Liu, Z. Wang and C. Niu, RSC Adv., 2015, 5, 10746–10750. [56] H. Yang, Z. Li, K. Liu, F. Meng and C. Niu, J. Phys. Chem. C, 2015, 119, 2260–2265. [57] S. Xie, Y. Song and Z. Liu, Can. J. Chem., 2009, 87, 1235–1247. [58] R. Komm, R. A. Geanangel and R. Liepins, Inorg. Chem., 1983, 22, 1684–1686.

S65

[59] A. S. Nathanson, A. R. Ploszajski, M. Billing, J. P. Cook, D. W. K. Jenkins, T. F. Headen, Z. Kurban, A. Lovell and S. M. Bennington, J. Mater. Chem. A, 2015, 3, 3683–3691. ¨ [60] S. Duman and S. Ozkar, Int. J. Hydrogen En., 2013, 38, 180 – 187. ˙ [61] A. K. Figen, M. B. Pi¸skin, B. Co¸skuner and V. Imamo˘ glu, Int. J. Hydrogen En., 2013, 38, 16215 – 16228. [62] R. Xiong, J. Zhang and J. W. Lee, J. Phys. Chem. C, 2013, 117, 3799–3803. [63] J. Zhang, Y. Zhao, D. L. Akins and J. W. Lee, J. Phys. Chem. C, 2010, 114, 19529–19534. [64] S. Frueh, R. Kellett, C. Mallery, T. Molter, W. S. Willis, C. Kingondu and S. L. Suib, Inorg. Chem., 2011, 50, 783–792. [65] J. Zhang, Y. Zhao, D. L. Akins and J. W. Lee, J. Phys. Chem. C, 2011, 115, 8386–8392. [66] P. Bhattacharya, J. A. Krause and H. Guan, J. Amer. Chem. Soc., 2014, 136, 11153–11161. [67] A. D. Sutton, A. K. Burrell, D. A. Dixon, E. B. Garner, J. C. Gordon, T. Nakagawa, K. C. Ott, J. P. Robinson and M. Vasiliu, Science, 2011, 331, 1426–1429. [68] Y. S. Chua, H. Wu, W. Zhou, T. J. Udovic, G. Wu, Z. Xiong, M. W. Wong and P. Chen, Inorg. Chem., 2012, 51, 1599–1603. [69] Y. S. Chua, Q. Pei, X. Ju, W. Zhou, T. J. Udovic, G. Wu, Z. Xiong, P. Chen and H. Wu, J. Phys. Chem. C, 2014, 118, 11244–11251. [70] Y. S. Chua, G. Wu, Z. Xiong, A. Karkamkar, J. Guo, M. Jian, M. W. Wong, T. Autrey and P. Chen, Chem. Commun., 2010, 46, 5752–5754. [71] C. Reller and F. O. R. L. Mertens, Angew. Chem. Int. Ed., 2012, 51, 11731– 11735. [72] Y. Nakagawa, Y. Ikarashi, S. Isobe, S. Hino and S. Ohnuki, RSC Adv., 2014, 4, 20626–20631. [73] M. B. M. Concha, M. Chatenet, F. H. Lima and E. A. Ticianelli, Electrochim. Acta, 2013, 89, 607 – 615. [74] R. L. Sams, S. S. Xantheas and T. A. Blake, J. Phys. Chem. A, 2012, 116, 3124–3136. [75] A. Paolone, F. Teocoli, S. Sanna, O. Palumbo and T. Autrey, J. Phys. Chem. C, 2013, 117, 729–734. [76] L. H. Jepsen, V. Ban, K. T. Møller, Y.-S. Lee, Y. W. Cho, F. Besenbacher, Y. Filinchuk, J. Skibsted and T. R. Jensen, J. Phys. Chem. C, 2014, 118, 12141– 12153. [77] R. Chiriac, F. Toche, U. Demirci, O. Krol and P. Miele, Int. J. Hydrogen En., 2011, 36, 12955–12964. [78] X. Kang, H. Wu, J. Luo, W. Zhou and P. Wang, J. Mater. Chem., 2012, 22, 13174–13179.

S66

[79] Z. Tang, Y. Tan, X. Chen and X. Yu, Chem. Commun., 2012, 48, 9296–9298. [80] J.-E. Seo, Y. Kim, Y. Kim, K. Kim, J. H. Lee, D. H. Lee, Y. Kim, S. J. Shin, D.-M. Kim, S.-Y. Kim, T. Kim, C. W. Yoon and S. W. Nam, J. Power Sources, 2014, 254, 329 – 337. [81] I. Dovgaliuk, L. H. Jepsen, D. A. Safin, Z. Lodziana, V. Dyadkin, T. R. Jensen, M. Devillers and Y. Filinchuk, Chem. Eur. J., 2015, 21, 14562–14570. [82] L. H. Jepsen, J. Skibsted and T. R. Jensen, J. Alloys Compds., 2013, 580, S287 – S291. [83] Z. Li, G. Zhu, G. Lu, S. Qiu and X. Yao, J. Amer. Chem. Soc., 2010, 132, 1490–1491. [84] Y. Peng, T. Ben, Y. Jia, D. Yang, H. Zhao, S. Qiu and X. Yao, J. Phys. Chem. C, 2012, 116, 25694–25700. [85] V. Pons, R. T. Baker, N. K. Szymczak, D. J. Heldebrant, J. C. Linehan, M. H. Matus, D. J. Grant and D. A. Dixon, Chem. Commun., 2008, 6597–6599. [86] C. R. Miranda and G. Ceder, J. Chem. Phys., 2007, 126, 184703. [87] N. J. Hess, M. E. Bowden, V. M. Parvanov, C. Mundy, S. M. Kathmann, G. K. Schenter and T. Autrey, J. Chem. Phys., 2008, 128, 034508. [88] R. Custelcean and Z. A. Dreger, J. Phys. Chem. B, 2003, 107, 9231–9235. [89] W. Fateley, N. McDevitt and F. Bentley, Appl. Spectrosc., 1971, 25, 155–173. [90] K. Nakamoto, Infrared and Raman Spectra of Inorganic and Coordination Compounds, Part A, Theory and Applications in Inorganic Chemistry, J. Willey and Sons, New York, 6th edn, 1997. [91] J. Goubeau and E. Ricker, Z. Anorg. Allg. Chem., 1961, 310, 123–142. [92] V. Meisalo and O. Inkinen, Acta Cryst., 1967, 22, 58–65. [93] M. H. Matus, K. D. Anderson, D. M. Camaioni, S. T. Autrey and D. A. Dixon, J. Phys. Chem. A, 2007, 111, 4411–4421. [94] W. J. Shaw, J. C. Linehan, Szymczak, D. J. Heldebrant, C. Yonker, D. M. Camaioni, R. T. Baker and T. Autrey, Angew. Chem. Int. Ed., 2008, 47, 7493– 7496. [95] P. M. Zimmerman, A. Paul, Z. Zhang and C. B. Musgrave, Inorg. Chem., 2009, 48, 1069–1081. [96] G. Wolf, J. Baumann, F. Baitalow and F. Hoffmann, Thermochim. Acta, 2000, 343, 19 – 25. [97] D. Jacquemin, J. Phys. Chem. A, 2004, 108, 9260–9266. [98] F. Baitalow, J. Baumann, G. Wolf, Jaenicke-R¨ ossler and G. Leitner, Thermochim. Acta, 2002, 391, 159–168. [99] F. A. Miller and C. H. Wilkins, Anal. Chem., 1952, 24, 1253–1294. [100] W. Weltner and J. R. W. Warn, J. Chem. Phys., 1962, 37, 292–303.

S67

Figure S37: EGA-MS profiles of the main gaseous species released during the S68 decomposition of 1. Ionic current is normalised.

Figure S38: Ion current (directly proportional to molar quantity) of the main species observed in gas released during thermal decomposition of 1.

S69

Figure S39: EGA-MS of pure borazine N3 B3 H6 (m/z = 81) with EGA-MS lines observed in this work.

S70

Figure S40: Thermogravimetry (TG) for system 1 and 3 before recrystallisation, as recorded in air and N2 atmosphere at heating rate of (a) 10 ◦C min−1 and (b) 2 ◦C min−1 .

S71

Figure S41: IR spectra at 150 ◦C compared with spectrum of B(OH)3 (taken from NIST webbook, webbook.nist.gov/chemistry).

S72

Figure S42: Differental thermal analysis (DTA) for system 1 and 3 before recrystallisation of AB, as recorded in air and N2 atmosphere at heating rate of (a) 10 ◦C min−1 and (b) 2 ◦C min−1 .

S73

Figure S43: Differential scanning calorymetry (DSC) for (a) system 1 and (b) system 3. Heating rate is 10 ◦C min−1 , H2 atmosphere.

S74