Synthesis, characterization and DFT studies of zinc-doped copper oxide nanocrystals for gas sensing applications †

,


Introduction
Hydrogen is widely used in industries, as a power source in aerospace, for metal sintering and annealing, in research laboratories, biomedical systems, automotive and transportation equipment, etc. [1][2][3][4][5][6][7][8] Therefore, the reliable, selective and fast detection of hydrogen gas leaks appears to be the strict necessity for preventing the accumulation of hydrogen in air (4% H 2 ). 60][11][12] Due to their unique properties and specic gas sensing mechanism, p-type copper oxide nanostructures have demonstrated promising potentials in various gas sensing applications, especially in detection of ethanol vapour and other volatile organic compounds (VOCs), like benzene, acetone and toluene, etc. [13][14][15][16][17] Copper oxide semiconductors and their derivatives are excellent candidates for the fabrication of low-cost p-type nanosensors due to high abundance of copper in nature, 18 simple synthesis and nanointegration methods. 193][24] In this context, the change of chemical and physical properties by doping of foreign elements is an important and efficient way to further control the gas sensing characteristics of semiconducting oxides such as sensitivity, response and recovery times, as well as selectivity. 16rom the existing literature, it is difficult to nd any study which demonstrates the p-type nanomaterials based hydrogen gas sensing with high response/selectivity characteristics.
In this work, the synthesis of zinc-doped copper oxides, Zn x Cu 1Àx O y , nano-crystalline lms by a synthesis from chemical solutions (SCS) method and rapid thermal annealing (RTA) process has been developed.We report on the details of the growth process, phase analysis, chemical composition, vibrational and sensorial performances of the functional copper oxide treated by conventional thermal annealing (TA) in electrical furnace and by RTA in air.A mixed phase of CuO:Zn/ Cu 2 O:Zn in samples TA-treated at 400-450 C for 30 min and RTA-treated at 525 C for 60 s was detected by XRD, EDX and micro-Raman measurements.An enhanced H 2 gas response and improved stability to humidity of RTA-treated samples with high Zn content was demonstrated by gas sensing measurements.Based on the ionosorption model and density functional theory (DFT) calculations, the gas sensing mechanism was proposed.

Experimental
Pure copper oxides, Cu x O y , and zinc-doped copper oxides, Zn x -Cu 1Àx O y , nano-crystalline lms were grown on glass substrates (microscope slides 76 mm Â 25 mm Â 1 mm) via SCS.The procedure of substrate cleaning was described in our previous work. 25The substrate surface was sensitized with SnCl 2 /HCl solution. 26,27For SCS deposition process the substrates were immersed vertically in the solutions using a microcontroller based system to grow Cu 2 O nanostructured lms with desired thickness, by controlling the number of cycles and time of substrate immersion in the solutions.The deposition system includes a home-made robot arm, controlled by a microcontroller through a graphical user interface (GUI), with the possibility to send commands and set parameters through a wireless module.A typical schematic for our set-up with related description is represented in Fig. S1.† The main advantage of the presented system is the exclusion of errors due to human factors which actually leads to high reproducibility of the samples and commodity in synthesis process.Wireless communication allows operation control out of fume hood in case of work with toxic chemicals or in the case of emanation of harmful gases aer chemical reactions in the deposition process.
The aqueous copper-complex solution, as cation precursor, comprises a mixture of 1 M copper sulfate (CuSO 4 $5H 2 O) and 1 M sodium thiosulfate (Na 2 S 2 O 3 $5H 2 O).In order to prepare the Zn-doped nano-crystalline lms, two new baths, one containing Cu + ions and the other lled with only deionized (DI) water, were included in the growth system.The Zn doping of Cu 2 O nanostructured lms was achieved by adding ZnSO 4 $7H 2 O in the copper complex solution corresponding to the necessary concentration of Zn in nanostructured lms.For doping concentrations (#1) -1.5 mM, (#2) -4 mM and (#3) -6 mM of zinc ions were added in complex solutions, thus the Zn doping level from 0 to 3 wt% Zn was varied according to EDX results.Deionized water (resistivity 18.2 MU cm) was used as solvent.NH 4 OH was added to the initial solution to control the complex solution and the value of pH was 10.The copper complex solution was diluted with DI water to obtain 0.1 M copper concentration and was maintained at room temperature during the lm deposition.The anionic precursor solution comprises sodium hydroxide (NaOH) diluted with DI water to obtain 2 M and was maintained at 80 C during the lm deposition.All chemicals were of reagent grade.Aer deposition, the nanocrystalline lms were rinsed well with DI water, and then dried for 5 min at 150 C in air ux.The global reaction of copper complex solution preparation for Cu 2 O nano-crystalline lms deposition was described in previous works. 11,28A single deposition cycle of a Cu 2 O nano-crystalline lm is described by the following two steps: (1) immersion of substrate in the hot (80 C) alkali solution for the adsorption of the (OH) À ions on the glass substrate; (2) dipping this substrate in the thiosulfatocuprate(I) for reaction of the copper ions and zinc ions with adsorbed (OH) À .This is one cycle of the ion layers adsorption and reaction processes in our method.The copper ions are formed by the dissociation equilibrium. 11,28quation of Cu 2 O forming is represented by: 11,28 Post-growth treatment is an important factor to increase the crystallinity of nanostructured oxide lms aer chemical synthesis. 11,28In this work, the as-grown lms were subjected to two types of post-growth treatments in air, namely conventional thermal annealing (TA) in an electrical furnace and rapid thermal annealing (RTA). 29Typical schematic temperature proles for TA and RTAtreatments at 450 C and 525 C are represented in Fig. S2a and b, † respectively.A typical schematic diagram of our RTA set-up is represented in Fig. S2c.† The samples morphology was investigated by using scanning electron microscopy (SEM) REM-ZEISS (at 7 kV).The chemical composition of the SCS samples was examined by EDX spectroscopy.Concentrations of Zn dopants in copper oxide layers were noted and found as (#1 or c1) -1.9 wt% Zn, (#2 or c2) -2.7 wt% Zn; (#3 or c3) -3.0 wt% Zn in the case of RTA treated samples and (#1 or c1) -1.9 wt% Zn, (#2 or c2) -2.4 wt% Zn; (#3 or c3) -2.8 wt% Zn in the case of TA treated specimens.Crystallographic data were obtained from X-ray powder diffraction (XRD) investigations using a Seifert XRD3000 PTS instrument operating at 40 kV and 40 mA, with CuKa 1 radiation, l ¼ 1.54184 Å. Secondary ion mass spectroscopy (SIMS) studies were performed with a Physical Electronics ADEPT 1010 quadrupole analyzer.The primary beam (25 nA) was applied to a 300 mm by 300 mm area, with the detection of negative secondary ions from an area of 100 mm by 100 mm at the centre.Technical information of the experiment can be found in our earlier reports. 7,30,31Micro-Raman studies were performed at room temperature with a WITec alpha 300 RA system in a backscattering conguration.The Nd-YAG laser power was less than 4 mW at the sample. 7The investigated sample sizes were 5 mm in width and few cm in length.Each measurement was performed on an individual sample and re-checked on several ones from the same set.Electrical studies were carried out using the two-point probe method at room temperature.Gas sensing properties were evaluated according to the previous reports. 8,32,33ifferent values of relative humidity (RH) were generated using a bubbling system.Air was passed through water and then was continuously injected into the chamber to create the necessary RH value.The humidity was continuously monitored by a standard hygrometer.More details can be found in previous works. 8,32,335][36][37] The projector augmented wave method was certied to describe the interaction between electrons and ions, 38 and the non-local exchange correlation energy was considered using the Perdew-Burke-Ernzerhof functional. 39,40he Hubbard model was involved to treat strong correlations within the DFT + U method in the Dudarev formalism. 41ecently, we have determined a value of U that can efficiently describe both copper I and II oxides 42 and this effective U value of 7 eV was chosen for the localized 3d electrons of Cu, whereas U ¼ 6 eV was taken for the Zn-3d electrons. 43The atomic structures were relaxed using the tetrahedron method with Bloch correction as implemented in VASP and the positions of all atoms in the cell were fully relaxed until the atomic forces on each ion were less than 0.01 eV ÅÀ1 .To model Zn doping, we have used our earlier bulk and surface models of CuO and Cu 2 O. 42,44 We have further used the implementation of the DFT-D2 approach described by Grimme 45 to account for long-range dispersion forces.
The binding energy of H 2 molecule was calculated from the relation: where molecule in the centre of a broken symmetry cell with lattice constants of 20 Å, sampling only the gamma-point of the Brillouin zone with the same accuracy parameters as described for the surfaces.A Bader analysis was performed with the code developed by Henkelman and co-workers, 47,48 to quantify the charge transfer between H 2 molecule and the surfaces.

Morphological SEM and compositional EDX studies
To investigate the surface morphology of SCS grown Zn x -Cu 1Àx O y nano-crystalline lms, SEM was used.The corresponding images are shown in Fig. 1 and S3.† In general, the nano-crystalline layers are homogeneously and continuously covering the glass substrates (see Fig. 1 and S3 †).
Large islands layers are difficult to see in the sintered lms formed from well-packed nanocrystallites of copper oxides covering completely the whole substrate.Typical SEM images of the as-grown lms are presented in Fig. 1a-c for Zn-doped copper oxide layers with concentrations (#1 or c1), (#2 or c2) and (#3 or c3), respectively.SEM of undoped CuO-Cu 2 O can be found in our recent work. 13The lm thickness was measured by cross-sectional SEM and the typical thickness was $1 mm.For samples aer RTA at 725 C (noted as RTA725 in Fig. 1g-i) or aer TA at 650 C (noted as TA650, see Fig. 1m-o and S3 and S4 †), relatively large particle agglomerates can be observed in contrast to lms treated at lower temperatures of RTA at 525 C (RTA525, see Fig. 1d-f) or TA at 450 C (TA450, see Fig. 1j-l).In these samples, marked as RTA525 and TA450, the lms morphology appeared most uniform and the nanoparticle size was smaller (of about 40-60 nm).The lms are formed from densely packed grains of approximately pyramidal form (Fig. 1d-f and j-l).
With further rise in sintering temperature up to 650 C (for TA treated samples), and up to 725 C for RTA, it appears that some fusions of grain boundaries have occurred (Fig. S5 †) leading to a marginal increase in particle size, of about 120-200 nm with a columnar morphology (Fig. 1g-i and m-o). 49Fig.S5 † shows a low magnication TEM image of a Zn-doped copper oxide lm.The lm consists of accumulated nanocrystals and its thickness amounts to 1 mm.From the TEM image in Fig. S5, † it is conrmed that the CuO:Zn lms are quite porous.We observed several pinholes on the CuO:Zn sample.Typical width and length of the Zn-doped CuO nanocrystals are 50 nm and 200 nm, respectively.Such columnar growth in copper oxide has been shown by Richthofen et al., 50 but Chauhan et al. 49 did not observe any preferred grain elongation.The random size and projection distributions of grains, suggest the random nucleation and growth mechanism, and hence an isotropic grain growth. 51

X-ray diffraction structural analysis
In all crystallized solid materials, divalent Cu surroundings are always distorted by a large Jahn-Teller effect which generally leads to square planar groups, which are more stable. 52When the Cu 2+ ions are substituted by metal (Zn 2+ ) ions in the Zn 2+ doped copper oxide sample, the difference in ionic radii 53,54 between Cu 2+ (r ¼ 0.071 nm) and Zn 2+ (r ¼ 0.074 nm) leads to deformation of the CuO lattice and the strain energy may cause dislocations and imperfections in the monoclinic crystal lattice.The formula (Zn x Cu 1Àx O 1Àı , for example x ¼ 0.03) indicates that 3 wt% of Zn atoms replace the copper (Cu) atoms to maintain charge neutrality, some of the oxygen vacancies ðV cc o Þ (ı ¼ 0.03) are created in the system.Therefore, it is reasonable to suggest that due to these reasons, the doped samples could have a higher electrical conductivity.
Fig. 2a shows X-ray powder diffractograms of the as-grown Cu 2 O layers (curve 1) and of the RTA treated ones (RTA525, curves 2-4), where two coexisting phases (cubic Cu 2 O and the monoclinic CuO) are detected.6][57] The comparison of the respective peak intensity of specimens aer RTA at 525 C shows that the Cu 2 O amount is higher when compared to CuO layers (Fig. 2a, curves 2-4).In the XRD pattern of TA650 and RTA725 specimens, all XRD peaks are indexed completely to the CuO monoclinic symmetry 54,55 (JCPDS 45-0937, as presented in Fig. 2b, curves 2-4).X-ray powder diffractograms of all samples showed the same features.The peak at 2q ¼ 36.4 is strongest and was indexed to the Cu 2 O(111) crystal plane, and other peaks at 2q ¼ 29.6 , 42.3 , 61.4 , 73.6 and 77.4 are also in good agreement with the diffraction pattern of the Cu 2 O cubic phase 54,55 (JCPDS 05-0667, as presented in Fig. 2).Even if CuO peaks are weaker compared to Cu 2 O peaks on the sides of the strongest peak two bands at 2q ¼ 35.5 and 38.7 indexing to the (À111)/(002) and ( 111) planes (see Fig. 2a, curves 2-4) of the monoclinic CuO can be observed, respectively.The spectra modications should be due to a removal process of the Cu 2 O phase on the specimens. 55The effect of rapid thermal annealing at 725 C for 60 s on the structure of CuO layers doped with different concentrations of zinc is presented in Fig. 2b.
The oxide average crystallite sizes are computed from Debye-Scherer's equation from the broadening of diffraction lines (À111) and (111): 54,55,58 where l is the X-ray wavelength, bthe full width at half maximum (FWHM) of the diffraction peak, D hklthe average crystallite size, and qthe Bragg angle.The grain sizes are calculated by using the peaks corresponding to the (À111) plane of CuO:Zn, which are 3.23 nm, 3.46, and 3.63 nm for samples RTA annealed at 725 C and doped with concentrations #1, #2, and #3, respectively.The mid-crystallite size increases slightly with increasing concentration of Zn in the materials.It can be caused by higher ionic mobility and lower activation energy of zinc ions. 56As a result it leads to a larger crystallite size due to easier transfer of Zn ions from trapping sites to nucleation sites in the growth of crystallites. 56The grain sizes calculated by using the peak corresponding to the (111) plane of CuO:Zn are 3.27 nm, 3.15 nm, and 3.14 nm for samples doped with concentrations #1, #2, and #3, respectively, aer RTA at 725 C. Note that the FWHM of the peak is obtained by tting the XRD patterns using soware Rayex. 7For samples marked as RTA525, the sizes of Cu 2 O aggregates D 111 are 2.43, 2.29 and 1.09 nm, for Zn-doping concentrations 1, 2, 3, respectively.This is consistent with the corresponding SEM images of the CuO nanocrystals (see Fig. S3 †).It was observed that the Cu 2 O aggregates increase in size with the treatment temperatures, but decrease with the doping level.For CuO, the aggregate sizes D 111 are 2.43, 2.39, and 1.47 nm, for samples doped with concentrations #1, #2, and #3, respectively (sample sets TA650-treated).
The information on the structural characteristics of the CuO nano-crystalline layers were obtained by computing the strain (3), the average crystallite size (D), as well as the dislocation density (d) according to the relations: 59,60 where b is the FWHM and q is the diffraction angle.
Lattice parameters (a s b s c, a ¼ g ¼ 90 s b for monoclinic structure) and the unit cell volume for the CuO layers were found by following the equations: 61 The calculated structural parameters of the Zn-doped copper oxide thin lms indicate no signicant changes in values.According to the XRD results, the Zn-doped CuO samples contain no detectable impurities, but the lattice parameters slightly differ from those of pure CuO (the a parameter is increased and the b parameter is decreased) which can be interpreted as an indicator that Zn is incorporated in the CuO lattice as discussed before. 62,63There are no additional peaks of Zn impurities, suggesting that the Zn-doping did not modify the original CuO lattice. 62,63Therefore, it could be also suggested that the Zn 2+ ions are spread in the CuO matrix. 64In addition, the lattice constants are determined using the least square renement method in the unit cell program.The obtained parameters of the doped sample are: a As annealing temperature increases (450-650 C for TA and 525-725 C for RTA), the crystalline nature of CuO has been improved indicated by enhanced intensity of the corresponding XRD peaks. 65However, with further increase in temperature up to 650 C for TA and 725 C for RTA samples, a slight decrease in crystallinity was observed (not shown here).The strain values were calculated with eqn (4) by using the peak corresponding to the (111) plane of Zn-doped CuO lms, and the resultant are À0.077,À0.094, and À0.112, respectively, for samples doped with concentrations #1, #2, and #3 aer RTA annealing at 725 C. The strain values are À0.149,À0.265, and À1.499, respectively, for samples doped with concentrations #1, #2, and #3 aer RTA at 525 C. The negative sign indicates that it is compressive by nature.The dislocation density for SCS samples is determined to be decreased with increasing Zn concentration.As well as, structural studies of the SCS/RTA thin lms reveal that the diffraction angle and the corresponding d value of the lattice spacing of the layers changes slightly with variation in Zn content. 56The decrease in the b value of the (111) peak indicates on the crystallite size growth RTA at 725 C and with zinc doping with concentrations #1, #2, and #3, respectively.The main motive of crystallite size increase is associated with strain and indicates a reduction in concentration of lattice imperfections. 56ccording to XRD experimental results, a negligible lattice deformation was found for Zn-doped CuO due to ionic radii difference.It may be a result of shorter Cu Zn -O bonds and smaller [Cu Zn O 4 ] units in nano-crystals. 54In Fig. 2b, there is a very small shi ($0.026 ) to a higher 2q angle value of the (0 0 2) XRD peaks for low Zn-doped CuO with respect to those of pure CuO.No essential change in the crystalline structure was observed, which indicates that most Cu atoms are in the CuO lattice.

Micro-Raman studies of doped samples
Copper(II)oxide (CuO) has a monoclinic structure and affiliate with the C 6 2h space group (two molecules for every primitive cell). 66It has twelve phonon branches since there are four atoms in the primitive cell. 67The zone-centre optical-phonon modes are G ¼ A g + 2B g + 4A u + 5B u . 68Among them, there are nine optical modes, from which three modes (A g + 2B g ) are Raman active, 68,69 other six ones 3A u + 3B u are infrared active. 67Furthermore, there are normally three acoustic modes, which are of A u + 2B u symmetry.In the A g and B g Raman modes the oxygen atoms move in b-direction for A g and perpendicular to the b-axis for B g modes. 67However, the IR modes implicate motion of both O and Cu atoms and the dipole moment is along the b-axis for the A u modes, but perpendicular to it in case of the B u modes. 67ig. 3 shows a micro-Raman spectra of various samples (RTA-annealed and TA-annealed) with CuO:Zn nanomaterials.Three micro-Raman peaks were observed in sample c1, at about 288, 330 and 621 cm À1 .With an increase in grain size (samples aer TA treatment at 650 C, see Fig. 1m-o and 3b), these micro-Raman peaks become higher, and move slightly to larger wavenumbers. 69A shi to 292 cm À1 and 295 cm À1 was found in samples c2 and c3, and the FWHM of the peak was decreased from 21 to 11 and 7 cm À1 , respectively.The broadening and red shi of the micro-Raman spectra happens with a decrease in the grain size. 69The micro-Raman curves show three main phonon modes in the crystallite-like CuO nanomaterials, at 288, 335 and 624 cm À1 , interrelated to the A g , B 1g and B 2g symmetries, 70 respectively.
Fig. 3 conrms continuous red-shi and the broadening of the A g phonon mode when the Zn concentration was increased.Based on reports, 69,71,72 the phonon modes in CuO single crystal shi to a lower frequency mode and the linewidth is found to broadened as a result of the nanodimensions effect. 73oexistence of mixed phases, namely cuprite Cu 2 O:Zn and tenorite CuO:Zn were observed in our samples.It is considered that the peak shis as the nanoparticle diameter changes, 70 also the size effects of phonon-electron coupling 74 and the polarization dependence of Raman phonon modes play an important role in the Raman shis and lead to unconventional properties. 70Comparing the Raman spectra of CuO powder 68 and single crystal, 71 we note the micro-Raman peak at 288 cm À1 as A g and 330 and 621 cm À1 as B g modes.These wavenumbers are slightly different in comparison with those in the scientic literature (298, 345 and 632 cm À1 ) due to size effects. 69However, in sample c3 when the grain size is larger, micro-Raman peaks have been observed at 296, 342, and 628 cm À1 , which corresponds to bulk CuO.No Cu 2 O modes are present in the samples aer TA650-treatment, demonstrating single phase characteristics of our CuO nanomaterials.The bands at about 301, 347 and 633 cm À1 , may be assigned to the A g (296 cm À1 ), B g(1) (346 cm À1 ) and B g(2) (631 cm À1 ) modes and are in agreement with previous data. 71,75The absence of Cu 2 O modes 76 in the micro-Raman studies of the specimens conrms the phase purity of the nanocrystals aer annealing at 650 C (TA650).
Secondary ion mass spectroscopy (SIMS) shows (see Fig. S6 †) that while the concentrations of Cu and O remain constant as a function of depth, the Zn concentration varies rapidly as a function of depth and its behaviour seems to track that of the Si substrate.It is possible that the SIMS signal of Zn maybe contaminated by the Si signal.

Gas sensing response
For gas sensing investigations the fabricated gas sensor structures are mounted in a test chamber at the required operating temperature for 30 min before exposing to the designated gases in order to stabilize the electrical baseline.Reported data in literature demonstrates good selectivity and gas response of the copper oxide based sensors to ethanol vapours. 14,20Thus, we select hydrogen gas (100 ppm) and ethanol vapour (100 ppm) as our test gases to demonstrate changes in selectivity of the Zn-doped copper oxide based nano-crystalline lms.The gas response is presented as S ¼ DR/R air Â 100%, assuming DR ¼ R gas À R air , R gas and R air are the resistances of the sensor specimen under gas exposure and in ambient air, respectively.
In this work, gas sensing properties are investigated mainly in dependence of Zn concentration in copper oxide nano-crystallite lms (1 mm thick).In order to determine the inuences of Zn dopant content on gas sensing performances of copper oxide lms, three concentrations are studied as presented above.More detailed investigations on the effect of treatment type and thickness of nano-crystalline lms, as well as detailed gas sensing mechanism are reported partly in paper, 81 as well as will be reported in future work.
Fig. 4a shows the gas response of the copper oxide samples with 2.8-3.0 wt% Zn to H 2 gas and C 2 H 5 OH versus type of thermal annealing: as-grown Cu 2 O samples, TA450 and TA650-treated with 2.8 wt% Zn at operating temperature (OPT) of 400 C, and RTA525 and RTA725-annealed with 3.0 wt% Zn at OPT of 300 C. As-grown Zn-doped copper oxide samples demonstrated poor gas response and selectivity to hydrogen gas, S H 2 $ 42% and S EtOH $ 23%, while thermal annealing improved the gas sensing characteristics of the fabricated sensor specimens.The highest gas response of the RTA sample set annealed at 525 C with S H 2 $ 756% and S EtOH $ 79% was found, having a good selectivity to hydrogen gas versus ethanol (S H 2 /S EtOH z 9.5).Also, we investigated thermal annealing at 650 C (samples TA650) and rapid treatment at 525 C (samples RTA525).Fig. 4b shows the response of sensor structure to hydrogen gas versus concentration of Zn dopant: Zn(1) -1.9 wt% Zn; Zn(2) -2.7 wt% Zn; Zn(3) -3.0 wt% Zn in the case of RTA treated samples, as well as of Zn(1) -1.9 wt% Zn, Zn(2) -2.4 wt% Zn; Zn(3) -2.8 wt% Zn in the case of TA treated samples.As expected, the gas response of the Zn-doped copper oxide based sensors increases as Zn doping concentration increases for both types of treatments, which indicates that the main factor in hydrogen gas response change is the dopant in the nanomaterial.
The optimal operating temperature (OPT) of the fabricated sensors was also investigated.Fig. 4c shows the gas response to hydrogen gas vs. OPT for the samples doped with concentration Zn(3).As observable, the OPT is higher for TA650-treated specimens when compared to the case of RTA525, 400 C versus 300 C, respectively.The lowest OPT is preferable for lower power consumption of the sensor heaters, thus it is a very important parameter from the industrial point of view.However, the gas response and rapidity of the sensor is strongly dependent on the OPT value. 6,14,20At low operating temperatures of below 200 C, the hydrogen gas molecules do not possess sufficient thermal energy to react with the adsorbed oxygen molecules (e.g.O 2 À ) on the surface of copper oxide, which lead to low reaction rate and lower gas response. 77At OPT higher than 300 C the oxygen atoms mainly adsorb at the surface of copper oxide nano-crystallites. 14,20,77In this case, the thermal energy is sufficient to overcome the activation energy barrier of the reaction of the hydrogen molecules with ionosorbed oxygen species. 20,77ig. 4d shows the dynamic response of sensor structures on copper oxide nano-crystalline lms treated TA650 and RTA525 toward hydrogen gas at OPT of 400 C and 300 C, respectively.The calculated response s r and recovery s d times (as time to accomplish 90% of the full gas response and time to take 10% of full recovery to initial baseline value, respectively) are s r $ 1.9 s and s d $ 8.5 s for TA650-treated sample at OPT of 400 C, while for the RTA525 treated one at lower OPT value was s r $ 2.4 s and s d $ 6.9 s, respectively.However, at the same OPT the RTA treated samples are faster, for example at 400 C the s r $ 1.4 s and s d $ 5.4 s.All data on rapidity of the sensors are included in Table S1.† The best thermal annealing regime to obtain higher gas response and faster response/recovery from CuO:Zn lms is RTA at 525 C for 60 s.
Fig. 4e shows the results on sensor resistance and gas response versus concentration of hydrogen gas (p H 2 ) of the sample treated RTA525 and measured at OPT of 300 C with Zn(3) doping level and it is observed that the response follows a power law relationship on gas concentration , where b is the slope of log R vs. log p H 2 , and b ¼ 0.398 (see Fig. 4f), 14 which is in concordance with different results on copper oxide based sensors. 6Fig. 4f shows the estimation of lowest detection limit (LDL) using the method reported by Volanti et al. 6 R min is dened as R min ¼ R air + 3s 0 , where R air is the resistance in air before exposure to hydrogen gas at OPT of 300 C, and s 0 is the standard deviation of the air response. 6For calculation of s 0 , the sensor was placed in the test chamber at OPT of 300 C for 60 min, in order to stabilize the sensor resistance, and then the next 60 min before exposing to hydrogen gas are used to calculate s 0 .In this work we found that s 0 ¼ 5.43 U and R min ¼ 5471 U.
The parameters obtained by linear tting of the log vs. log plot are presented in Fig. 4f.Estimated LDL was $0.5 ppm of hydrogen gas which was comparable to one of the best values for CuO hydrogen gas based sensor reported in literature. 6We note the relatively low value of R min (in the order of kU), which is important during lm integration in electronic devices for gas detection applications.Usually, to increase the gas response to reducing gases in case of sensors based on n-type oxide metals, one uses acceptor doping which decreases the concentration of the electrons in the semiconductor 16,78 and leads to an increase in total resistance of the sensor (to MU range). 79As a result, the sensing measurements in laboratory conditions are possible only with high precision test measurements tools in the majority of cases.Thus, the integration in a sensor module is a problem due to the necessity of ampliers with very high input impedance as result of the high resistance of the sensors.Therefore the total cost of the device increases.In our case, a kU range of the sensor resistance value can solve those problems.Also, the fast response and recovery times associated with high gas response and selectivity to hydrogen gas, demonstrated good potential for fabrication of stable, robust, low-cost and fast hydrogen gas sensor applications.
According to literature analysis, sensors based on copper oxide or other p-type metal oxide nanostructures with such high response and selectivity to hydrogen have not been reported until now by other research groups.In most cases, p-type semiconductor based sensors demonstrated good gas response to ethanol vapour and other VOCs, such as benzene acetone and toluene. 16For comparison, in Table S2 † the most relevant data on the hydrogen gas sensors based on p-type and n-type oxide metal nanostructures which indicate response to hydrogen gas and ethanol vapour ratio have been presented in detail.

The effect of humidity on the gas sensitivity
To check the inuence of the humidity on gas sensing performances of the copper oxide based nanostructured lms, gas sensing measurements were performed in 90% RH and compared with those in 30% RH for samples treated RTA at 525 C and TA at 650 C with 3.0 wt% Zn (Fig. 5a and b).As can be observed, the gas response decreased in humid atmosphere by z35% and z65% for RTA and TA treated samples, respectively.Also a decrease in rapidity of the sensors can be observed (see Table S1 †).Thus, the improved stability to humidity of the RTA treated samples is indicated.One of the main possible reasons could be a lower response to water vapour of the RTA treated samples (see Fig. S7 †), while the decrease in gas response can be attributed to competition of adsorbed oxygen species on the surface of the CuO nanocrystals between water vapour and H 2 molecules. 80More details on the inuence of humidity on gas sensing performances of the Zn-doped copper oxide nanostructured lms versus concentration of Zn and related mechanism have been presented in previous work. 81

Gas sensing mechanism proposed
The reactivity of active sites on the surface of Zn-doped copper oxide nano-crystalline lms is strongly dependent on its doping concentration, type of defects and impurities. 78Thus, surface phenomena are important to understand the sensing mechanism of the sensor based on p-type semiconductor nanomaterials.As we mentioned above, adsorbed oxygen can be ionized into several forms, namely atomic O À or O 2À and molecular O 2 À . 82At temperatures below 200 C, predominant surface oxygen species are in molecular form O 2 While at temperatures above 200 C predominant surface oxygen species are atomic species O À and O 2À : Ionosorbed oxygen creates the surface acceptor states and traps electrons from the valence band, which results in increased concentration of holes and resistance decreases (for p-type semiconductors). 80A schematic illustration of the proposed gas sensing mechanism is presented in Fig. 6a and b, where the transparent blue zone shows HAL with lower resistance in comparison with core region (represented by dark yellow region).At exposure to hydrogen gas the following reaction occurs (taking in account that h + + e À ¼ null): 14 Equation ( 12) is represented schematically in Fig. 6b.With decrease in the holes concentration, HAL width becomes thinner and the resistance of the sensor structure increases respectively.
Electron sensitization is known to be an efficient tool to control/tune the gas response and selectivity of the oxide metals with both types of electrical conductivity. 78In our case it is performed by control of Zn concentration in copper oxide nanocrystalline lms.It has been shown that addition of zinc in copper oxide lms decreases the electrical resistivity of the sensing material. 13,81For example, in the case of RTA at 525 C the electrical resistivity of the sample was decreased from 3.1 Â 10 4 U cm to 1.95 Â 10 4 U cm by increasing Zn concentration from (#1) to (#3).The created acceptor centres are ionized at room temperature by capturing electrons from the valence band and generating holes.In general, it is assumed that Zn-doping of Cu 2 O results in donor doping.Thus, further investigations are needed to determine the acceptor centres origin related to zinc ion doping of the Cu 2 O and CuO.Due to different oxidation states of copper ions in Cu 2 O (Cu 1+ ) and CuO (Cu 2+ ), incorporation of Zn 2+ should occur by different mechanisms.In both cases, the concentration of holes is increased, which leads to a decrease of the electrical resistance value. 13,81Taking into account that the value of the sensor resistance under exposure to gas R gas can be theoretically very large, as a result sensor sensitivity (S ¼ R gas /R air ) can be improved by decreasing the R air value. 85However, this can't explain the high selectivity to hydrogen gas of the Zn-doped copper oxide based sensors.
Since peaks of ZnO or Zn were not observed in the XRD data, it is reasonable to assume that most of Zn dopant is dispersed well on the surface of CuO nano-crystallites as amorphous ZnO, 86 which was demonstrated by XPS measurements in previous work. 81Addition of ZnO to metal oxides is known to enhance the surface oxidation by H 2 gas molecules, which lead to a higher selectivity to hydrogen gas. 87Yu et al. 87 demonstrated an increase in hydrogen gas selectivity by addition of ZnO to pellet-type SnO 2 sensors.Also, the lower response of copper oxide based sensors more likely is induced by an inhibition of ethanol vapour molecules catalytic oxidation due to Zn doping of nano-material. 13,16,20,86

DFT calculations
In order to gain further insights into the H 2 sensing mechanism, we performed density functional theory based calculations on the interaction of H 2 with Zn-doped CuO(111) and Cu 2 O(111) surfaces.
Zn-Doped CuO surface.As already mentioned, we have used our earlier slab model for CuO(111), where at the base of the surface simulation cell two layers of atoms were xed at their relaxed bulk positions to simulate the bulk phase of the crystal, while above these two layers, the surface was represented by two layers of atoms whose positions were allowed to change freely in the optimization process.The top layer consists of 3-and 4-coordinated Cu and O atoms.3-Coordinated O atoms (O SUF ) are the most exposed atoms, each connected to one 3-coordinated (coordinatively unsaturated -Cu CUS ) and two 4-coordinated (coordinatively saturated -Cu CSA ) Cu atoms (Fig. S8 †).We rst needed to identify the most energetically favourable dopant site for Zn ion on a (1 Â 1) surface slab, which we achieved by substituting Zn in place of distinct surface Cu atoms in the top layer (Fig. S8 and S9 †).We found that Cu CUS is the preferential site for Zn doping, as the Cu CUS-Zn doped structure is more stable by 0.20 eV compared to the Cu CSA-Zn substituted structure.In order to comply with the experimental doping of $3%, we next performed our calculations in a (2 Â 1) supercell of CuO(111) surface (consisting of 31-Cu, 32-O and 1-Zn atom), corresponding to 3.125% doping of Zn.No reconstruction takes place as a result of Zn substitution and Zn binds to nearby surface oxygens, O SUB and O SUF , with bond lengths of 1.928 and 1.855 Å, respectively, as shown in Fig. S9.† As a result of doping, the band-gap decreases slightly by $0.1 eV, decreasing its value from 0.900 eV in the undoped CuO(111) to 0.890 eV in the Zn-doped CuO(111) surface.
Next, we modelled the interaction of the H 2 molecule with this Zn-doped CuO(111) surface.In a series of calculations, we rst placed the H 2 molecule close to all surface atoms in different orientations, but we found that the molecule desorbs from the surface in all the initial input congurations, except when it was placed near the most exposed O SUF atoms.We observed that the bonds between the surface oxygen O SUF and the Cu CUS surface atoms are broken, with O SUF connecting to the H 2 molecule to form an adsorbed water molecule, as shown in Fig. 7.The same O SUF is bonded to the Zn atom on the surface with a bond length of 2.022 Å.Moreover, among all the surface O SUF atoms, the most stable conguration was found by placing the H 2 molecule near the O SUF atom, which is bonded to the Zn dopant atom.Our results are similar to recent work on H 2 adsorption on an undoped CuO(111) surface, 88 where the bond between the surface oxygen O SUF to the Cu CSA atoms is broken and O SUF connects to a nearby Cu CUS surface copper atom aer forming a H 2 O molecule.Our calculated binding energy is À80.1 kJ mol À1 , which is $8 kJ mol À1 larger than that achieved on the CuO(111) surface without Zn-doping.The charge density difference, together with Bader charge analysis (Fig. 7), shows that there is signicant charge transfer between the newly formed H 2 O molecule and surface atoms Cu CSA and O SUF .As a result of the H 2 molecule interaction with the surface, the Fermi energy increases by 0.210 eV, from À1.563 to À1.353 eV, while the calculated band-gap decreases by only $1.3% to 0.877 eV.The calculated densities of states are shown in Fig. S10a.† Zn-Doped Cu 2 O surface.Here, we have again used our earlier slab model, consisting of ve atomic layers, in which the three uppermost layers were free to relax during the optimization, while the bottom two layers were kept frozen at the optimised bulk positions.The Cu 2 O(111) surface consists of four different Fig. 6 Schematic representation of H 2 gas sensing mechanism for gas sensor based on Zn-doped copper oxide nano-crystalline films.(a) At exposure to ambient air, the oxygen atoms (illustrated by red spheres) are adsorbed at the surface of nano-crystallites.By trapping free electrons from the bulk, the hole accumulation layer region (HAL) with lower resistance is created, that lead to a current flow (illustrated by dark red arrows).HAL is illustrated by the blue transparent region, while core region is illustrated by dark yellow.(b) At exposure to hydrogen gas, the width of HAL region will decrease due to injected electrons from hydrogen gas molecules, which are oxidized to H 2 O molecules.As a result, the current flow through the HAL region will decrease, that is illustrated by smaller dark red arrows.
atomic sites in the top layer, namely a coordinatively unsaturated copper Cu CUS , an outermost oxygen O SUF , a coordinatively saturated copper Cu CSA and a sub-surface oxygen atom O SUB , which is 4-fold coordinated as shown in Fig. S11.† Similar to the CuO(111) surface, we have doped Zn into a (2 Â 1) supercell slab (consisting of 40Cu and 20O atoms in ve layer slab), corresponding to 2.5% doping (Fig. S12 †).However, in order to maintain charge neutrality, as a result of substitution of Zn 2+ ion in place of Cu 1+ in Cu 2 O, we also created a Cu vacancy by removing one Cu atom near the Zn dopant atom.We investigated two congurations by substituting two distinct surface coppers atoms, Cu CUS and Cu CSA , and creating nearby Cu vacancies.We found that the Cu CSA -substituted structure, shown in Fig. S12, † is 0.21 eV more stable.As a result of the Cu vacancy, surface reconstruction takes place and in the relaxed surface structure, the bond is broken between the Cu CUS atom close to the vacancy and the oxygen atom in the second layer; the Cu CUS then bonds to one top surface oxygen atom, O SUF , and one sublayer oxygen atom, O SUB , with bond lengths of 1.827 and 1.886 Å, respectively.The Zn atom binds to O SUF , O SUB and an oxygen atom in the second layer with bond lengths of 1.833 1.886 Å and 2.016 Å, respectively.We have shown side and top views with annotated bond lengths in Fig. S12.† We note that with this 2.5% doping of Zn, the bandgap of the slab does not change, remaining at 0.824 eV.
We next investigated the interaction of the H 2 molecule with the surface and found that the H 2 molecule dissociates on the surface with a small binding energy of À24.6 kJ mol À1 .This binding is weaker compared to that on the pristine Cu 2 O(111) surface without Zn-doping, 42 where the calculated binding energy for the dissociative adsorption of the H 2 molecule was found to be À78.6 kJ mol À1 .As shown in Fig. 8, one of the H atom binds to the Zn atom with a bond length of 1.533 Å, while the other H atom binds to the nearby surface O SUF atom with a short 0.978 Å bond length.Bader charge analysis shows that one of the H atoms, bonded to Zn, accepts a charge of À0.309 e À , while the second H atom, bonded to O SUF , transfers charge to the surface O SUF atom, becoming H + with a calculated positive charge of 0.647.As such, the hydrogen molecule has dissociated into a proton and hydride species, where one hydrogen atom transfers charge to a comparatively more electronegative surface oxygen atom, while the other hydrogen atom gains charge from a comparatively less electronegative surface Zn atom (Fig. 8).The Fermi energy of the system increases by 0.350 eV, from À1.274 to À1.239 eV, while the calculated band gap decreases slightly by $5% as a result of the interaction of the H 2 molecule with the surface, increasing to 0.865 eV (Fig. S10 †).
Our calculations reveal that the Zn doping inuences the binding energies of the H 2 molecule on both CuO and Cu 2 O surfaces and charge re-distribution takes place.While the Fermi energy changes as a result of the interaction of the H 2 molecule with the surface, the changes in the band-gaps are only small.Our results are similar to H 2 sensing on ZnO surfaces, where the H 2 molecule acts as a charge acceptor, decreasing the Fermi level by only a small change of 0.01 eV in the band gap value. 89

Conclusions
Structural, morphological, chemical and vibrational properties of as-grown and post-growth thermal treated copper oxide samples prepared via SCS were investigated in details.High crystallinity, mixed Zn-doped CuO/Cu 2 O phase and slight changes in the crystal constants of the samples have been demonstrated.XRD and micro-Raman studies of the as-grown Cu 2 O layers and of the RTA treated ones (RTA525 at 525 C for 60 s) detected two coexisting phases (cubic Cu 2 O and the monoclinic CuO).The two phases ratio can be controlled by the Zn-doping level in SCS and by annealing regimes of TA or RTA.
Gas sensing investigations showed higher gas response and selectivity to hydrogen gas of the Zn x Cu 1Àx O y nano-crystalline lms with 3.0 wt% Zn versus pure copper oxides.By Zn doping in copper oxide and formation of a nano-heterojunction aer RTA at 525 C for 60 s it was possible to change the sensing properties from ethanol vapour (pure copper oxide) to hydrogen gas (zinc-doped copper oxide with mixed phases).For samples RTA-treated at 525 C for 60 s at operating temperature of 300 C the gas response for 100 ppm H 2 was S H 2 $ 765%, which corresponds to an increase of sensor resistance by one order, with fast response times (s r $ 2.4 s) and recovery times (s d $ 6.9 s).Based on literature analysis, no reports on such high hydrogen gas response and selectivity of sensors based on p-type nanomaterials, have been published to-date by other groups.High crystallinity of the samples synthesized at relatively low temperatures (#80 C) and the existence of mixed Zn-doped CuO/Cu 2 O phases resulted in better sensing properties of the sensors and demonstrated good potential for fabrication of reliable, selective and fast hydrogen sensors with high gas response.A stabile sensor performances in humid air was found for CuO:Zn by exposure of samples to rapid thermal annealing process for the rst time.
Dispersion-corrected DFT + U calculations were employed to investigate the hydrogen gas sensing mechanism on both the CuO and Cu 2 O surfaces, which revealed that Zn doping facilitates the binding of H 2 where charge transfer between the molecule and the surface is responsible for the sensing behaviour.
Such nanomaterials are of great interest for other applications where a p-type oxide is requested.
¼ 4.6871 Å, b ¼ 3.4252 Å, and c ¼ 5.1333 Å, and the unit cell volume is 81.3110Å3 .The lattice constants of the doped sample are the monoclinic data of undoped CuO (a ¼ 4.6877 Å, b ¼ 3.4266 Å, c ¼ 5.1328 Å, unit cell volume ¼ 81.2979 Å3 ) obtained under the same conditions.It can be seen that the length of both a-axis and b-axis of doped samples shrink, while c-axis expand with the Zn 2+ ion doping in CuO because the ionic radius of Cu 2+ ion (0.71 Å) is smaller than that of Zn 2+ ion (0.74 Å).However, it is speculated that Zn 2+ ions are indeed incorporated into the CuO lattice and replaced some of the Cu ions to form solid solutions of Zn x Cu 1Àx O with low doping amount of Zn 2+ ions (x ¼ 0.03).

Fig. 3
Fig. 3 Series of micro-Raman spectra scanned of copper oxide nanocrystallites doped with different concentrations (Zn(1), Zn(2), Zn(3)) of zinc presenting the phonon modes in the top of the nanostructures corresponding to the A g , B 1g and B 2g symmetries, respectively.Typical micro-Raman spectra of the: (a) CuO/Cu 2 O nanocrystallites RTA-annealed at 525 C for 60 s; (b) CuO nanocrystallites TA-annealed at 650 C for 30 min.

Fig. 4
Fig. 4 (a) Gas response of the copper oxide samples with 2.8-3.0 wt% Zn doping level versus type of thermal annealing: as-grown Cu 2 O samples (As-gr), TA-treated at 450 C (TA450) and at 650 C (TA650) at operating temperature (OPT) of 400 C, RTA-treated at 525 C (RTA525) and at 725 C (RTA725) at OPT 300 C. (b) Gas response of the sensor to H 2 versus content of Zn, for TA650 and RTA525.(c) Gas response to H 2 versus OPT of the samples with Zn(3) doping level.(d) The dynamic response of TA650 (curve 1) and RTA525 (curve 2) toward H 2 at OPT of 400 C and 300 C, respectively.(e) Resistance and gas response of sensor toward concentrations of H 2 for RTA525 at OPT 300 C with Zn(3) doping level; (f) estimation of the lowest detection limit for H 2 detection using log vs. log plot.

Fig. 5
Fig. 5 Dynamic hydrogen gas response of the samples treated RTA at 525 C (operating temperature 300 C) and TA-annealed at 650 C (operating temperature 400 C) in 30% and 90% RH.

Fig. 7
Fig. 7 (a) Side view of H 2 on the Zn-doped (2 Â 1) slab of the CuO(111) surface, (b) electronic density difference plot of the H 2 bonded structures on the Zn-doped CuO(111) surface, showing charge transfer in the regions between the H 2 and the surface atoms upon interaction.Blue contours indicate electron density increases by 0.01 electrons ÅÀ3 and yellow contours indicate electron density decreases by 0.005 electrons ÅÀ3 .Bader charge values are in e À .Cu, O, Zn and H atoms are represented by blue, red, grey and salmon pink colours respectively.

CuO/Cu 2 O-Zn+H 2 is the total energy of the H 2 molecule adsorbed on the Zn-doped CuO or Cu
2 O surface, E CuO/Cu 2 O-Zn is the energy of the naked surface slab of Zn-doped CuO or Cu 2 O, and E H 2 is the energy of the isolated H 2 molecule.Within this denition, negative adsorption energy indicates an exothermic process.While calculating, E H 2 , we modelled the isolated H 2