*3.1. Deposition*

Table 1 shows the deposition conditions for quartz glass samples.

The growth habit of the films on all substrates showed individual nuclei growing in number and size as the number of cycles increased. Figure 1 shows SEM images of the uncapped films on glass for 100, 200, 500, and 1000 ALD cycles (abbreviated to c. hereafter) deposited on soda-lime glass. The films on Si substrates were similar.


**Table 1.** Deposition parameters for films on quartz substrates.

**Figure 1.** SEM images of CuCl deposited on soda-lime glass: (**a**) 100 c.; (**b**) 200 c.; (**c**) 500 c.; and (**d**) 1000 c. Scale bars represent 1 μm.

As the number of cycles increased, the diameter of the crystallites also increased. There was an initial nucleation phase and then the surface was covered in an array of crystallites. The areal density of the crystallites showed a decrease with the number of cycles; some form of agglomeration or ripening was evident. Dimensional change of the crystallites can take place by two processes: (i) Coalescence, where there is poor adhesion between the nanocrystallites and the substrate and they are mobile enough to move across it and to coalesce and (ii) Ostwald ripening, in which there is strong adhesion of the nanocrystallites to the surface and material transfer takes place by the diffusion of atoms across the surface. In this case, there is no significant coalescence of nanocrystallites therefore there is an Ostwald ripening process taking place. Initially, the crystallites had somewhat "liquid-like" shapes whereas the crystallites in the 1000 c. films had a more facetted appearance. This change in habit was accompanied by differences in the chemical composition of the films as discussed later. It is possible that this was caused by the longer deposition time required for the thicker film giving more time for reconfiguration of the crystallites. The size of the crystallites in the CuCl films increases with the number of cycles as shown in Figure 2a, while Figure 2b shows this behaviour in comparison with that reported previously using different precursors [19].

Figure 2 demonstrates that the rate of increase of crystallite diameter is similar with both the current and the previous precursors, once the initial nucleation has taken place, and that the growth on crystalline and noncrystalline substrates is similar. This similarity is not unexpected since the reactant is HCl in both cases; here it is derived by decomposition of pyridine hydrochloride, previously the HCl vapour came from an HCl/butanol solution. The nucleation behaviour is somewhat different: with PyrHCl there is slower nucleation. However, with the HCl/butanol precursor the nucleation behaviour depended on the growth cycle parameters so further exploration of the deposition parameters with PyrHCl may also show differences in nucleation behaviour.

**Figure 2.** (**a**) Crystallite diameter vs. number of atomic layer deposition (ALD) cycles using pyridine hydrochloride (PyrHCl) precursor and (**b**) comparison with previous results using HCl/butanol as precursor (HCl). z PyrHCl–Si, **+** PyrHCl–glass, HCl (short purge), HCl (long purge). In (**b**) error bars have been omitted for clarity.

## *3.2. Crystal Structure*

Films for XRD analysis were deposited on quartz glass both without a capping layer and with a capping layer of ~5 nm of aluminium oxide deposited in situ in order to prevent the films undergoing hydrolysis in atmospheric moisture. The crystal structure was determined by X-ray diffraction (XRD) using GIXRD. The results are shown in Figure 3a,b. The films show the zinc blende crystal structure of γ-CuCl with no evidence of any other crystalline phases. The films with an Al2O3 capping layer also showed the same structure with small additional broad peaks at ~36◦ and 45◦ (marked with asterisks). These do not fit CuCl2 (PDF 9001506), CuO (PDF 1011148), Cu2O (PDF 1010926), or other likely compounds. It is not clear what material they represent. The crystallites showed largely random orientation although with some preferential (220) orientation for thinner films. For randomly oriented films the intensity ratio *I*220/*I*<sup>111</sup> ≈ 0.60 and the ratio here is much higher. However, it must be borne in mind that because of the glancing angle X-ray incidence, any preferential orientation is not parallel to the substrate. From these results we can conclude that the bulk of the film is γ-CuCl.

**Figure 3.** XRD spectra of (**a**) uncapped and (**b**) capped CuCl films on quartz substrates. Unidentified peaks marked with asterisks. Note: The "missing" CuCl peaks ((200), (222), (400), and (420)) have low intensities compared to the (111) peak for randomly oriented material (PDF 1010991).

#### *3.3. Chemical Composition*

The chemical composition was investigated by XPS. The scans were taken before and after Ar sputter-cleaning of the substrate to investigate the extent of surface contamination. Table 2 shows the analysis of survey scans of the deposited material for the unsputtered samples while Table 3 shows the relative concentrations of Cu and Cl after Ar sputter cleaning for 150 s. The XPS analysis was not performed in situ so there will be significant amounts of adventitious carbon from environmental contamination between deposition and analysis for the unsputtered films. The sputtered samples show significantly reduced C 1*s* signal. There is evidence, however, of the presence of carbon as a residue from the organic precursors (see below). There is also some fluorine content in all films which must come from the CuBTMSA precursor. There is significant uncertainty in the Cu/Cl ration due to the dispersed crystallites which allow a large signal from the substrate material to be detected and because the relative sensitivity factor for the XPS quantification for the Cl 2*p* peak is much lower than for the Cu 2*p* peaks by a factor of ~11. Nevertheless, for the unsputtered samples the Cu/Cl ratio is lower than the expected 1/1 ratio for CuCl; however, the sputter-cleaned samples show close to the stoichiometric ratio. The low CuCl ratio for the unsputtered samples arises from the different composition of the surface layer.


**Table 2.** Element concentration obtained from X-ray photoelectron spectroscopy (XPS) survey scan (not Ar-sputtered).


**Table 3.** Relative concentrations of Cu and Cl obtained from XPS (after Ar sputtering).

High-resolution scans of the Cl 2*p* peak from the uncapped samples with peak fittings are shown in Figure 4 for the unsputtered and sputter-cleaned samples. For Q1, Q5, and Q7 the Cl 2*p* peak is well fitted by one 2*p* doublet consisting of 2*p*3/2 and 2*p*1/2 components separated by 1.6 eV. Figure 4a is representative of these. The binding energy values are consistent with those given for copper chlorides although they cannot easily distinguish between Cu(I) and Cu(II) chloride. In general, the CuCl2 should appear at a higher binding energy [24–26]. Taking into account the XRD results which show only CuCl we conclude that peak at ~199.0 eV is CuCl. There is no significant ClO*<sup>x</sup>* formation: this would give rise to 2*p*3/2 peaks at energies approximately 206–208 eV [27], and there is no evidence of these in the Cl 2*p* spectra. The peak details are given in Table 4. For sample Q10, if a single peak fitting is used, the fitting error is significantly higher. However, the measured Q10 signal can be well fitted with two 2*p* components separated by 0.6 eV (Figure 4b). This is consistent with the emergence of two different bonding environments for Cl. The lower energy peak is characteristic of HCl·*n*H2O (198.4 eV) [28] and is also consistent with CuCl bound to an organic ligand [29]. The peak area ratio between this component and the one at slightly lower energy is 1.4 indicating significant surface contamination. Figure 4c shows the Cl 2*p* peak for Q7 after Ar sputter-cleaning. For the Ar-sputtered Q10 sample, again the best fit is with two doublets. In this case, the additional peak appears at higher energy, consistent with the existence of an organic chloride [28,30].

**Figure 4.** XPS Cl 2*p* spectra for uncapped quartz samples showing fitted peaks: (**a**) Unsputtered Q7 500 c.; (**b**) unsputtered Q10 1000 c.; (**c**) Ar-sputtered Q7 500 c.; and (**d**) Ar-sputtered Q10 1000 c.


**Table 4.** Components of Cl 2*p* XPS peaks.

Figure 5 shows the high resolution 2*p*3/2 and 2*p*1/2 raw data scans scan of Cu and the decomposition of the 2*p*3/2 peaks into individual components. Q7 is representative of the other samples with less than 1000 c. deposition. Table 5 shows the peak positions and full width half maxima (FWHM).

It is difficult to separate the Cu0, Cu+, and Cu2+ peaks solely by their binding energies but taking into account the XRD spectra, peak P1 is assigned to CuCl which has been previously seen at 932.4–932.6 eV [31,32]. Peak P2 (~934 eV) could be indicative of CuO, CuCl2, or Cu with an organic or organo-chlorate ligand [33–36]. However, the presence of Cu2+ species should be indicated by an obvious shake-up satellite peak at ~940 eV [37] and there is no evidence of this in Q7 both with and without sputtering (Figure 5a,b). Neither is there evidence for chlorate bonding in the Cl 2*p* spectra. Therefore, Q7 contains only Cu+ bonds and P2 is ascribed to Cu(I) organic bonding. The amount of the Cu organic contamination in both Q7 and Q10 is clearly less after Ar sputtering (Figure 5d,f). Figure 5a,e, Q10 (1000 c., unsputtered), shows a large satellite peak at the expected energy for Cu2+ at ~940 eV together with a peak at 936.7 eV, P3, which may be evidence for some fluoride bound to organic material [38]. However, the Q10 Ar-sputtered sample, Figure 5f, shows the same three peaks as for the Q10 unsputtered sample with no evidence of a satellite peak. Therefore, it is not clear which of the peaks is related to the Cu2+ satellite peak. It may be that the relatively lower intensity of P3 after Ar sputtering gives rise to a much smaller satellite peak which is not visible in Figure 5f.

**Figure 5.** *Cont.*

**Figure 5.** XPS Cu 2*p* peaks, uncapped quartz samples: (**a**) Q7 500 c. and Q10 1000 c., unsputtered; (**b**) Q7 500 c. and Q10 1000 c., Ar-sputtered; (**c**) decomposed 2*p*3/2 peak, Q7, unsputtered; (**d**) decomposed 2*p*3/2 peak, Q7, Ar-sputtered; (**e**) decomposed 2*p*3/2 peak, Q10, unsputtered; and (**f**) decomposed 2*p*3/2 peak, Q10, Ar-sputtered.


**Table 5.** Peak parameters for Cu *2p* XPS.

Observation of the O 1*s* spectra yields further information. The spectra for Q7 and Q10 before and for Q10 after Ar sputtering are shown in Figure 6. The samples are dominated by the peak at 533.3 eV due to SiO2 [39] from the quartz since the substrate coverage is not complete. Both unsputtered samples, Figure 6a,b, also show a peak at ~534 eV which is probably O in an organic ligand indicating some residual CuBTMSA precursor material in the film. In addition, Figure 6b shows a large additional peak at 535.8 eV, which is at an energy which has been ascribed to an O atom linked to a CF containing ligand [30]. This again is likely due to some partially decomposed CuBTMSA precursor. Figure 6c is representative of all the sputtered samples and shows only the SiO2 peak indicating that the oxygen bonded to organic material is only on the surface.

When the Cl 2*p*, Cu 2*p*, and O 1*s* results are taken together with the XRD results, it is clear that the bulk of the films consists of γ-CuCl with a certain amount of organic contamination from the precursor molecules. This is confirmed by the presence of significant carbon content even in the Ar-sputtered samples. The unsputtered samples show, as is to be expected, greater organic contamination, including from F-containing organic fragments and HCl from the CuBTMSA and PyrHCl precursors, respectively. The exact nature of the bonding is the subject of further investigation.

**Figure 6.** Decomposition of O 1*s* peaks for uncapped quartz samples: (**a**) Q7 500 c., unsputtered; (**b**) Q10 1000 c., unsputtered; and (**c**) Q10 1000 c., Ar-sputtered.

Thin uncapped CuCl films will hydrolyse after approximately one day. Figure 7 and Tables 3–5 show measurements on the sputter cleaned sample Q8 (500 c., Al2O3 capped) after exposure to normal atmosphere for approximately 3 weeks. The curves and tabulated data are very similar to those of the uncapped samples with the same number of growth cycles kept in an inert atmosphere (Figures 4c and 5e). They show no evidence of degradation, indicating that 5 nm of oxide capping layer provides an effective barrier to hydrolysis of the films.

**Figure 7.** Decomposition of (**a**) Cl 2*p* and (**b**) Cu 2*p*3/2 peaks of Q8 (500 c. capped, Ar-sputtered) after approximately 3 weeks in normal air.

#### *3.4. Photoluminescence*

The photoluminescence signals obtained from sample Q10 (1000 c., uncapped) were easily measurable at room temperature. Shown in Figure 8 is the experimental spectrum, decomposed into three Gaussian bands located on a linear background. The sum fits the measured data within background noise. The band parameters are listed in Table 6. We use band area as the measure of its strength. The location of the PL maximum is very close to the position of the strongest PL2 band, coinciding with the position of the absorption peak PL2 (3.248 eV) of Table 6 within experimental uncertainty. The weaker but discernible high-energy PL3 band at 3.308 eV lies 0.06 eV above PL2. The lower energy band PL1 is situated at 3.225 eV, approximately 0.10 eV lower than PL2. These results are similar to those observed on CuCl films obtained by thermal evaporation [40] and magnetron sputtering [41]. Those showed that at low temperature the emission peak could be resolved into three components which broadened into one composite peak at room temperature. The main peak was identified as the *Z*<sup>3</sup> free exciton peak observed at energy 3.227 eV at 15 K. This peak shifted to higher energy of 3.243 eV at room temperature, a shift of approximately 0.016 eV [42]. This peak is

almost at the same energy as the main peak reported here (3.248 eV). The peak shown here can also be decomposed into three components. By comparison with these results and allowing for the shift with temperature, the main peak indicating the most probable radiative channel is ascribed to the *Z*<sup>3</sup> exciton. The lower energy peak we see at 3.225 eV (385.4 nm) can be ascribed to the *I*<sup>1</sup> peak due to an exciton bound to an impurity, probably a Cu vacancy [43]. The higher energy peak at 3.308 eV (375.8 nm) can be ascribed to the *<sup>Z</sup>*1,2 exciton. PL mapping across the sample shown in Figure <sup>9</sup> (32 × <sup>32</sup> <sup>μ</sup>m2, step 4 μm) shows intensity varying by a factor of approximately 2, due to local variations of the effective film thickness. No changes in peak position, width or shape were observed.

**Figure 8.** Photoluminescence emission intensity vs. photon energy showing resolved peaks for Q10 (1000 c. uncapped sample). The dashed line is the fitted peak.

**Table 6.** Best-fit parameters of the Gaussian bands used in modelling the PL spectrum of Figure 8.

(**a**) (**b**) **Figure 9.** Uncapped quartz sample Q10, 1000 c. (**a**) White light photography and (**b**) integral PL intensity (arb. units) over the same area of the substrate with 4 <sup>×</sup> <sup>4</sup> <sup>μ</sup>m<sup>2</sup> resolution.
