P type transparent conducting CU AL o thin films prepared by PE MOCVD 3

After being normalized to the same thickness of 100nm, the difference in the transmittance shown in Figure 5-9b of films was smaller except for the film grown at the lowest temperature..

It is noted that the films were of different thicknesses After being normalized to the same thickness of 100nm, the difference in the transmittance (shown in Figure 5-9(b))

of films was smaller except for the film grown at the lowest temperature The main reason of the large transmittance of the film grown at 650°C might be the different compositions compared with other films because only this film showed a different XRD spectrum (Figure 5-8)

SEM was used to observe the morphology of the films (shown in Figure 5-10) except the sample grown at 650°C, because it was non-conductive and charged under electron bombardment When the substrate temperature was low, the film was composed of nano-particles of 25-50nm When the substrate temperature increased, the nano-particles gradually linked to each other When the temperature reached 800°C, no obvious particles and intersected chains appeared

AFM was also employed to view the morphology of the films In Figure 5-11, it is clear that the surface of the film grown at 650°C was quite flat and there were some islands of about 100nm diameter In the film grown at 700°C, more and smaller (20-50nm) islands appeared In the film grown at 750°C, rougher and more connected stalagmitic islands were observed However, when temperature increased further, the islands turned into flatter clusters with ridges Here film grown at 750°C appeared to show different information from what was gathered from SEM This is due to the over-enhanced z-scale effect of the AFM image

Figure 5-12 shows growth temperature dependence of the direct optical bandgap of the as-deposited films The bandgap was estimated from a plot of (αhν)2 versus hν which yielded a straight line with an intercept on the hν axis (see appendix B) The film grown at 650°C showed the largest optical bandgap of 3.62eV The bandgap was the smallest at 700°C with the value of 3.45eV and then increased with the growth temperature The bandgap of the films correlated with the result of optical transmittance A film with larger bandgap appeared to have a higher transmittance The electrical properties of the films including sheet resistance, conductivity, mobility and Hall coefficient are listed in Table 5-2 The film grown at 650°C was an insulator showing no conductivity From the table, both the conductivity and carrier concentration increased with the growth temperature The bandgap widening with the growth temperature mentioned above could be due to the increase of the carrier concentration, i.e., the Burstein shift (refer to section 5.4.2) took effect Croitoru and Bannett11 suggested that the contribution of grain boundary scattering potential decreased as the substrate temperature increased With an increase in substrate temperature, the grain size increased12 causing a decrease in grain boundary potential and hence an increase in mobility The decrease in grain boundary potential could also

be responsible for an increase in carrier concentration with substrate temperature, thus

an improvement in conductivity The mobility was of the order of 1.0cm2·V-1·s-1 The Hall coefficient was positive at first, and then decreased with the growth temperature and the sign became negative for the film grown at 800°C The positive sign

undoubtedly indicated p-type conductivity of the films

Table 5-2 Electrical properties of the as-deposited films prepared from dpm precursors (“∞” means out of range, “⎯” means not measurable)

Film A Film B Film C Film D

Bulk carrier concentration (×1018cm-3) ⎯ 5.33 15.7 ⎯

Referring to the expression of Hall coefficient in Appendix C (Eq.C-14), the sign change of Hall coefficient might be due to the existence of two types of charge carriers In the present films, as the crystal size was nanoscale, grain boundary scattering was a very important factor to affect the mobility of the carriers Because the mass of a hole was much larger than that of an electron, the mobility of a hole was

reduced more than that of an electron.13 This resulted in a larger value of b (

where b>0, µn, µp are drift mobilities of electrons and holes, respectively) Although

the hole concentration p was larger than the electron concentration n, the Hall

coefficient R might be negative, which led to inaccurate results and even the wrong sign for the conductivity type In the growth process, a small amount of Cu+ ions were oxidized to Cu2+ ions, which acted as electron donors With the increase of growth temperature more Cu2+ ions existed, causing the inversion of the Hall coefficient From Appendix C, it is known that all the data for conductivity are reliable For the

films grown at 700°C and 750°C, the actual Hall coefficient and mobility for hole conduction should be larger than the experimental data but the actual hole concentration should be smaller However, this does not affect the qualitative analysis given in last chapter and this chapter For the film grown at 800°C, probably because

nb2>p, the Hall coefficient, mobility and carrier concentration changed sign and the experimental values for these parameters were meaningless

To confirm the type of conduction, the Seebeck coefficient was measured A plot of

∆V versus ∆T of the film grown at 800°C is shown in Figure 5-13 Here ∆V is the voltage difference and ∆T is the temperature difference between two ends of the film From the linear fit of the figure, the Seebeck coefficient was found to be +5.45µV/K,

indicating p-type conductivity The Seebeck coefficients of the as-deposited films are

listed in Table 5-3

0.00 0.01 0.02 0.03 0.04 0.05 0.06

Figure 5-13 Seebeck measurement (∆V versus ∆T) of the film grown at 800°C The solid line

is the linear fit curve

Table 5-3 Seebeck coefficients of the as-deposited films prepared at different growth temperatures

Growth temperature 650°C 700°C 750°C 800°C

Seebeck coefficient ── 30.0µV/K 22.5µV/K 5.45µV/K

As reported by Kawazoe et al.5, the Seebeck coefficient of CuAlO2 prepared by laser ablation was +183µV/K with conductivity of 0.095S·cm-1 Nagarajan et al.14 reported that the Seebeck coefficient of CuYO2 prepared by thermal co-evaporation was +284µV/K with conductivity of 0.025S·cm-1 All these values were much larger than those found in the present study The Seebeck coefficient decreased with the increase

of conductivity can be explained by referring to the Seebeck coefficient of degenerate semiconductors given by15

F

B T eE

k2 /32

π

where kB is Boltzmann’s constant For p-type semiconductors, EF is the Fermi energy

measured from the top of the valence band For p-type degenerate semiconductor, the

Fermi level is below the top of the valence band High carrier concentration causes high conductivity and pushes the Fermi level down, leading to a small value of the Seebeck coefficient Compared with other works, the present films had better conductivities and smaller Seebeck coefficients and the Seebeck coefficient decreased when the conductivity increased This conclusion is also applicable to a non-degenerate system as the expression of the Seebeck coefficient is of the form15

)2

5(

T k

E A e

where A is a constant related to the scattering mechanism and here EF is negative With the increase of carrier concentration, the Fermi level becomes closer to the

valence band for a p-type sample, and thus a smaller Seebeck coefficient

To investigate the conduction mechanism of the films, the temperature dependence of resistance was measured in vacuum (2×10-5Torr) from RT to 200°C The resistance was determined by the two-probe method and the temperature was controlled by a temperature controller and measured by a thermalcouple that was in contact with the film surface The natural logarithm of the inverse of resistance as a function of temperature is plotted in Figure 5-14

When the growth substrate temperature was 800°C, ln(1/R) decreased with the inverse

of temperature linearly (see Figure 5-14(a)) However, for the films grown at lower temperatures, the curves were obviously not linear (Figure 5-14(b) and (c)) This suggested different conduction mechanisms for the films grown at different temperatures The details of the conduction mechanisms will be described in section 5.4.1

5.3.3 The effect of oxygen flow rate on the properties of Cu-Al-O films

In this part, the properties of copper aluminum oxide prepared from dpm precursors with the varying oxygen flow rates will be discussed In the work described in section 4.3.3, it was observed that oxygen flow rate had no obvious effect on the transmittance of the films and there was an optimum value for conductivity In this section, the oxygen flow rate was varied from 20 to 35sccm, and other growth parameters were kept constant: the substrate temperature at 750°C, the Ar flow at

2.2 2.4 2.6 2.8 3.0 3.2

-7.83 -7.81 -7.79 -7.77 -7.75 -7.73

20 25 30 35 1.8

2.1 2.4 2.7 3.0

Oxygen Flow Rate (sccm)

Figure 5-15 Growth rate of films prepared from dpm precursors versus oxygen flow rate

The growth rate (Figure 5-15) first decreased with the increase of oxygen flow rate, and then rose slightly when the oxygen flow rate exceeded 30sccm Films grown at 30sccm showed the lowest growth rate

Scattering Angle 2θ (deg.)

XRD spectra of the films are shown in Figure 5-16 The peaks were located at 43.36° (2.088Å), 50.42° (1.810Å), 74.18° (1.279Å) and a big hump existed around 21.30° (4.173Å) All the peaks matched the PDF of β-CuAlO2 and the hump was from the quartz substrate but might also contain the peak from β-CuAlO2

Figure 5-17 shows a high-resolution TEM image of the film grown at 35sccm oxygen flow There were clearly grains of crystals in Figure 5-17 and the sizes of the grains were all less than 10nm The lattice spacings were measured to be around 2.84±0.02Å and 2.46±0.02Å These values could all be attributed to rhombohedral CuAlO2 and the 2.460Å could also be attributed to Cu2O

Figure 5-17 A high-resolution TEM image of the film grown at 35sccm oxygen flow rate

Referring to the discussion in Chapter 4, rhombohedral CuAlO2 existed in the film as nanograins with preferred orientation and could not be observed by XRD β-CuAlO2

detected by XRD decomposed to Cu2O under electron bombardment or upon heating

This conclusion is consistent with all the results from XRD and TEM described in this section

0 10 20 30 40 50 60

(a)

0 10 20 30 40 50 60 70

(b)

D

CBA

Wavelength (nm)

Figure 5-18 Transmittances of the as-deposited films prepared from dpm precursors at different oxygen flow rates: A: 20sccm, B: 25sccm, C: 30sccm and D: 35sccm; (a) original data, (b) after normalization to the thickness of 100nm

The transmittances of these films can be seen in Figure 5-18 Figure 5-18(a) shows the original data of the transmittances of the as-deposited films and Figure 5-18(b) shows

the transmittances after being normalized to the thickness of 100nm In Figure 5-18(b), the sample with the lowest oxygen flow rate had the highest transmittance while the sample with the highest oxygen flow rate showed the lowest transmittance The film at 20sccm showed the transmittance of 33-71% Similar transmittances were observed for the films at 25sccm and 35sccm (13-57%), while the transmittance of the film at 30sccm was about 19-57% The transmittance of the film is affected by the scattering centers in the film Heterogeneities formed in the film will heavily scatter the light.16With the increase of the oxygen flow rate, there were probably more heterogeneities formed in the growth process, which made the film more opaque

Figure 5-19 shows the morphology observed from SEM of the films grown at different oxygen flow rates All the films looked like consisting of intersected chains

or ridges and the size of the ridged clusters of films at 25sccm and 35sccm were significantly bigger than those in the film at 20sccm The film at 30sccm was the most compact one

AFM was also employed to observe the morphology of the films From AFM images (Figure 5-20), it is clear that all the surfaces showed 3-D islands Film A had the smallest particles and film B had the largest particles The particles in film A were more even than other films However, for every film, different sizes of particles and the deep ditches between particles were observed It was probably due to the large roughness of the films, the SEM did not provide accurate information about the surface morphology because it is a surface sensitive technique On the contrary, AFM gave three-dimensional surface image of the film

Hall measurement results are listed in Table 5-4 Most films showed small mobilities

and positive Hall coefficients indicating p-type conductivity The p-type conductivity

was up to 41.0S·cm-1 and the mobility was in the order of 1cm2·V-1·s-1 Again the same problem, as appeared in the previous section, the sign of the Hall coefficient of the most conductive film B was negative Referring to Appendix C and the analysis in section 5.3.2, it is known that for films A, C and D, the actual hole mobilities should

be larger than the experimental data but the actual hole concentrations should be smaller than the experimental data listed in Table 5-4

(a) (b)

Figure 5-20 AFM morphology of as-deposited films prepared from dpm precursors at

different oxygen flow rates of (a) 20sccm, (b) 25sccm, (c) 30sccm and (d) 35sccm The z

scale is 50nm

Table 5-4 Results of Hall effect measurement of the films prepared from dpm precursors at

different oxygen flow rates (“―” means not measurable)

Film A Film B Film C Film D

Sheet carrier concentration (×1013cm-2) 101 ― 21.2 64.2 Bulk carrier concentration (×1018cm-3) 54.3 ― 15.7 40.9

To confirm the type of conductivity, Seebeck technique was employed The Seebeck measurement found that all the films exhibited positive voltage at the cold end, which

meant p-type conductivity of the films The Seebeck coefficients of all the films are

listed in Table 5-5

Table 5-5 Seebeck coefficients of the films prepared at different oxygen flow rates

Seebeck coefficient 17.7 µV/K 15.7 µV/K 22.5 µV/K 11.2 µV/K

The film prepared at 30sccm, which was the least conductive, had the largest Seebeck coefficient while the film at 35sccm had the smallest Seebeck coefficient As discussed in the previous part, the Seebeck coefficient decreased when the conductivity increased Comparing Table 5-5 with Table 5-4, it is seen that the Seebeck coefficients of the films decreased with the increase of conductivity except for the film at 25sccm For the film grown at 25sccm, the Hall coefficient inversion happened, indicating the highest concentration of electrons in the film Thus the abnormal largeness of the Seebeck coefficient was due to the co-existence of the two types of carriers It is also noticed that both the conductivity and the Seebeck coefficient did not have a clear relationship with oxygen flow rate

The bandgap estimated from a plot of (αhν)2 versus hν as a function of oxygen flow rate is shown in Figure 5-21 The bandgap decreased with oxygen flow rate initially and increased after the oxygen flow rate reached 30sccm According to literature,17, 18the bandgap was related to the conductivity On the other hand, Burstein-Moss theory

suggested an increase in bandgap as the carrier concentration increased (see section 5.4.2) Referring back to Table 5-4, the films in the order of conductivity were film B> film D > film A > film C; and the films in the order of carrier concentration were film A > film D > film C From Figure 5-21, the films in the order of bandgap was film A > film B > film D > film C The trend of the carrier concentration was the same

as that of the optical bandgap (Figure 5-21), which meant that the bandgap increased with the increase of carrier concentration and obeyed Burstein-Moss theory

3.45 3.50 3.55 3.60 3.65 3.70

Oxygen Flow Rate (sccm)

Figure 5-21 Optical bandgap versus oxygen flow rate for films grown from dpm precursors

To find out the electrical conduction mechanisms of these samples, the temperature dependence of resistance was studied The natural logarithm of the inverse of resistance as a function of temperature is plotted in Figure 5-22 All the curves were far from linear In section 5.4.1, the conductivity mechanism of films are discussed in

detail and it is found that, in films grown at 750°C and 700°C, which were studied in last section, the grain boundary scattering mechanism dominated

A linear relationship between ln(T/R) and 1/T should exist according to the grain boundary scattering mechanism It is clearly seen that in Figure 5-23 the ln(T/R)~1/T curves for all the films grown at different oxygen flow rates showed very good linearity, which suggested that grain boundary scattering dominated the conduction mechanism in these films

The activation energies of these films were estimated to be 48.3meV, 46.1meV, 42.7meV and 46.5meV, respectively This activation energy was applied to overcome the grain boundary potential, which was much smaller than 0.2eV of the laser-ablation prepared film6 Because for the laser-ablation prepared film, the activation energy was probably used not only to overcome the transport barriers but also to activate the free carriers Once again, it can be noticed that a higher conductivity comes with smaller activation energy

5.3.4 Depth profile

To obtain more information of the elements in the films, depth profiling was employed A depth profile can be accomplished using controlled erosion of the surface by ion sputtering

The films grown at 700°C and 750°C (section 5.3.2) were analyzed by XPS depth profiling The etching current from Ar ions (1keV) was 0.2µA, and the etching period was 2 minutes every time After every etching, XPS spectra were taken until the film was completely etched away A charge compensation current was supplied to eliminate the charge shift

Figure 5-24 shows the depth profiles of the Cu2p3/2 peak of the films B and C discussed in section 5.3.2 Figure 5-24(a) shows 2-D and 3-D depth profile spectra of the film grown at 750°C and Figure 5-24(b) shows 2-D and 3-D depth profile spectra

of the film grown at 700°C In the figures, the long arrow means the direction from the surface to the interior Because the etching rates for different films are uneven, the same etching level for different films does not necessarily correspond to the same depth The factors affecting the etching rate include not only instrumental factors, but also surface roughness, crystalline structure, defects, and sample surface charging All these factors may affect the accuracy of a depth profiling analysis so a rigorous distribution of elements with the depth cannot be obtained However, it is still possible

to analyze some properties of the films qualitatively

Normally the XPS spectra should be plotted after charging calibration using the C1s

peak Now due to ion beam etching, no standard can be used to do this calibration To avoid a large shift of binding energy, a compensation current was used All the peaks

of every level were plotted in their original intensity with an adjusted background level, which made the graph clear and comparable

(b)

Binding Energy (eV)

Figure 5-24 Depth profiles of peak Cu2p3/2 of the films grown at (a) 750°C, (b) 700°C The arrows stand for the direction of depth The left figures are two-dimensional, and the right figures are three-dimensional

Comparing Figure 5-24(a) and (b), it is seen that for the film grown at 750°C, the peak intensity was very weak for the first three levels, then increased sharply until level 6, and decreased after that until level 10 In (b), the peak intensity was not so weak at first, and gradually increased until level 7 That means the copper distribution was more uniform in the film grown at lower temperature Copper appeared less at the surface for both films The low intensity of copper at the surface was also observed by

other workers,19 and was explained by the low surface energy of Al (1200mJ·m-2) compared with Cu (1850mJ·m-2)

It is noted that the peaks at high levels contained more than one component in both films At low levels, all the peaks were mainly at 932.6eV indicating Cu+ and a small shoulder appeared at a higher binding energy (933.8eV) indicating Cu2+ The peak fitting results will be described in a later part (Figure 5-26 and Table 5-6) However,

at high levels, the spectra in Figure 5-24(a) showed one more peak around 935eV from level 7 to level 10; the spectra in (b) only had one peak at around 935eV at levels

7 and 8 The peak shift in (b) can be probably considered as charge shift In (a), above

level 7, a weak peak of Si2p (99eV) appeared which suggested the existence of Si In

the meantime, the peak of copper still had considerable intensity This can be explained by the diffusion The radii of Si4+, Al3+, Cu+ and Cu2+ are 0.039nm, 0.057nm, 0.096nm and 0.072nm,16 in which Si4+ is the smallest With a high growth temperature, the silicon ions might diffuse from the substrate to the film to form a thin interface layer In this layer, copper had two different chemical environments: one was

as usual and another one had silicon ions surrounding This would result in the 935eV

peak of Cu2p3/2 With silicon around, the bond between copper and oxygen would be looser, which caused a higher binding energy of copper core level electrons

As mentioned above, Al had a lower surface energy than that of copper so there was less copper than aluminum at the surface This can be proved by studying Figure 5-25

Figure 5-25 shows the depth profile of Al2p of the film grown at 750°C In the first three levels, unlike copper, peaks of Al2p were clear and intense From level 4 to level

6, the intensity difference was not as large as that of Cu2p3/2 spectra shown in Figure

5-24(a) for the same film On the other hand, from level 7 to level 10, the intensity appeared relatively weaker compared with the peaks of copper at the same levels It is easy to see that aluminum had a more even distribution from surface to interior than copper; but in the region near the interface, less aluminum existed The XPS results suggested that the content of aluminum was more than that of copper at the surface while copper was more than aluminum near the interface Unfortunately the atom ratio

of copper to aluminum cannot be calculated through the quantitative analysis of Cu2p and Al2p because the broad Al2p peak contains Cu3p peak whose binding energy is very close to that of Al2p This problem always exists when making a quantitative

analysis of compounds with both copper and aluminum elements

Binding Energy (eV)

Figure 5-25 Depth profile of peak Al2p for the film grown at 750°C

SIMS gave similar information on the depth distribution of both copper and aluminum elements: (1) copper was concentrated in the interior of the film; (2) aluminum had a more even distribution than copper

To analyze the components of copper, peak fitting was employed to fit the peaks from

level 2 to level 9 of Cu2p3/2 at 932.6eV for the film grown at 750°C (Figure 5-26) Level 1 and 10 were excluded due to too weak signals Each spectrum from level 2 to level 6 was fitted to two peaks The peak at lower binding energy was due to Cu+, and the peak at higher binding energy belonged to Cu2+ At level 2, there was more Cu2+than Cu+ From level 3, Cu+ was more than Cu2+ and the ratio of Cu+/Cu2+ increased sharply until level 6, which proved that the main valence of copper in the film was +1 From level 7, one more peak appeared, which was around 935.4-935.8eV

Table 5-6 XPS peak fitting results for the film shown in Figure 5-24(a)

Level Binding Energy

12

1

peak peak

peak peak

Binding Energy (eV)

Binding Energy (eV)

Binding Energy (eV)

Binding Energy (eV)

Binding Energy (eV)

Binding Energy (eV)

Figure 5-26 Peak fitting of Cu2p3/2 spectra (from level 2 to level 9) in Figure 5-24(a) The green lines are background and the red lines are the sum of all fitted peaks

According to the previous discussion, the third component was attributed to the copper ions in the interface layer The decrease of this component with depth was reasonable due to the diffusion characteristic The fitting parameters including peak positions and area ratios are listed in Table 5-6

100 200 300 400 500 600 700

Figure 5-27 shows the XPS spectrum of the valence band A clear band was resolved

at about 4eV, which was mainly composed of Cu3d, consisting with Yanagi et al.20 As

reported, the tail of the band was attributed to O2p so the upper valence band was primarily composed of admixed states of Cu3d and O2p orbitals

5.4 Further Discussion on Film Properties

Transport phenomenon is the term applied to the motion of charge carriers under the internal or external fields In the absence of an electric field, the electron gas in a

semiconductor is in an equilibrium state, which is established as a result of the interaction of electrons with lattice defects Such defects include lattice imperfections, thermal vibrations of the lattice (phonons) and impurity atoms

In the case of thin films, the surface of a thin film affects the conduction of charge carriers by interrupting carrier transit along their mean free path They may either be diffusely scattered, in which case they emerge from the surface with no memory of their velocity, or they may be reflected so that only their velocity component perpendicular to the surface is reversed, and their energy remains constant Any surface that is not completely specular in its behavior will result in a decrease in the conductivity of the film

There are various scattering mechanisms in crystalline semiconducting materials by which the carriers are scattered Because of the large carrier concentrations and the experimental evidences shown previously, the present films are regarded as degenerate To discuss the conduction mechanism of the present films, different scattering mechanisms are described and analyzed in the following parts

1 Lattice scattering

In addition to the various stationary imperfections, lattice vibrations distort perfect lattice periodicity The degree of distortion is a strong function of temperature Lattice vibrations are categorized into acoustical and optical modes

Acoustic deformation potential scattering occurs when an acoustic wave propagates

in a crystal lattice The expression of mobility in terms of a corpuscular model is21

T k

E N m

c e

B c

ll l

2 / 1 2 2 / 5

*

4 2)2(

−

∆

where c llis the elasticity constant modulus, ∆c is the divergence of strain, and N is the concentration of matrix atoms In a degenerate electron semiconductor, only the electrons near the Fermi level take part in conduction For degenerate semiconductors,

Optical phonon scattering is related to the vibrations in which the neighboring atoms

in a crystal vibrate in opposite phases These vibrations may produce a strain causing

a polarization potential, which results in an electric field traveling through space in the form of a plane wave The interaction of the charge carriers with this wave causes scattering The temperature dependence of the mobility in the case of scattering by optical vibration is of the form21

)1(

a reduction of mobility, thus a reduction in conductivity with the increase of temperature for degenerate semiconductors The lattice vibration scattering is then ruled out because all the present films have better conductivity at elevated temperatures

2 Ionized impurity scattering

Of all the impurities that may be present in the crystal, the greatest effect on the scattering of the carriers is produced by ionized impurities This is because the electrostatic field due to such impurities remains effective even at a great distance According to the Conwell-Weisskoft formula when degenerate charge carriers are scattered by impurity ions, the energy dependence of mobility is21

2 2 3

2 3

2 2 2

)1

ln(

1)

2()(

Ze N

E Z

N e

E m

E m e

i

F i

F F

is the concentration of the scattering centers There is no temperature term in this

formula Shanthi et al.,22 Elich et al.23 and Huang et al.24 also obtained the same result that conductivity was independent of temperature for degenerate semiconductors using different approaches This was different from the case of non-degenerate

semiconductors, which had a temperature dependence25 of the mobility asµi ∝T 2 Thus, the mobility (conductivity) in the case of ionized impurity scattering should be independent of temperature for the degenerate semiconductors and not applicable to the films studied in this project