Plasma-assisted deposition and crystal growth of thin ...

2 downloads 0 Views 530KB Size Report
1 Institut für Physik, Ernst-Moritz-Arndt-Universität Greifswald, Domstraße 10a, ..... Ziegler J F, Biersack J P, Littmark U, The Stopping and Range of Ions in. Solids ...
Plasma-assisted deposition and crystal growth of thin indium-tin-oxide (ITO) films R. Hippler,1 H. Steffen,1 M. Quaas,2 T. R¨owf,1 T. M. Tun,1 H. Wulff2 1

2

Institut f¨ ur Physik, Ernst-Moritz-Arndt-Universit¨ at Greifswald, Domstraße 10a, 17489 Greifswald, e-mail: [email protected] Institut f¨ ur Chemie und Biochemie, Ernst-Moritz-Arndt-Universit¨ at Greifswald, Soldmannstr. 16, 17489 Greifswald

Advances in Solid State Physics Vol. 44, B. Kramer, Ed., Springer: Berlin Heidelberg (2004), p. 299–312 Summary. Metal oxide films are of significant technological relevance and play an important role, e.g., as protective coatings and in the fabrication of solar cells. Plasma-assisted deposition employing magnetron discharges are frequently employed. Formation and crystal growth of, e.g., indium tin oxide (ITO) films show pronounced dependencies on plasma parameters and on the applied substrate voltage. The influence on chemical composition, crystal structure, density, etc., as well as possible reasons will be discussed.

1 Introduction Metal oxide films are of significant technological relevance and play an important role, e.g., as protective coatings, in the fabrication of solar cells, and in plasma display panels. Thin films of, e.g., tin-doped indium oxide (indium tin oxide, ITO) are characterized by low electrical resistance and high optical transmittance in the visible region. The combination of these properties makes ITO films indispensable in many optoelectronic devices. The macroscopic properties of thin films are directly associated with their chemical composition, microstructure, and defect structure. In plasma-assisted deposition, film properties are to a large extent determined by plasma parameters. Therefore the correlation between deposition conditions and film composition and microstructure is of fundamental interest. Reactive direct current (DC) magnetron sputtering has been reported to yield ITO films with low resistance and good optical transmittance [1]. However, the influence of the deposition conditions on microscopic or structural properties of reactive sputtered ITO films has not been studied in detail, yet. Two deposition procedures are commonly in use; the reactive sputtering from a metallic In/Sn target in the presence of oxygen or from an indium-tin-oxide target. In theses complex processes the energetic and thermal conditions at the substrate surface and the growing film surface play a dominant role. Elementary processes like adsorption, desorption, and diffusion as well as chemical reactions will

2

R. Hippler, H. Steffen, M. Quaas, T. R¨ owf, T. M. Tun, H. Wulff

be affected. Stoichiometry, microstructure and morphology of the deposited films depend on the energetic conditions at the surface. Only recently, more systematic investigations on the role of plasma parameters ion flux and ion energy to the substrate have been reported [2][3]. Many authors use oxygen as reactive gas during the deposition procedure from oxide targets to form stoichiometric ITO; it is well known, that the oxygen partial pressure is a decisive parameter during ITO film deposition by magnetron sputtering [4]. For example, Mergel et al. [5] observed the formation of hyper-stoichiometric ITO films deposited at high oxygen flows. Also, applying a negative substrate voltage during film deposition is rather common. Thus, energy and flow density of charged particles (electrons, ions) to the substrate can be varied. In recent investigations we found that negative substrate voltage influences the composition of the growing film [6][7]. The present study focuses on the relations between the deposition conditions at different oxygen flows and negative substrate voltages and the influence on the chemical composition and the microstructure of ITO films. The plasma in front of the substrate was investigated by energy resolved mass spectrometry. The energy influx to the substrate was estimated by a thermoprobe. Grazing incidence x-ray diffractometry (GIXD) and photoelectron spectroscopy (XPS) were applied to determine the chemical composition and the phase composition of the deposited films. Film thickness and density were measured by grazing incidence x-ray reflectometry (GIXR).

2 Experiment Thin ITO films were deposited on Si(100) wafers by DC planar magnetron sputtering (figure 1). A metallic In/Sn (90/10) target was used. Residual gas pressure was < 10−8 mbar. Deposition pressure (5.6 × 10−3 mbar), gas flow (argon 15 sccm) and discharge power (30 W) were kept constant. The flow of the reactive oxygen gas was varied between 0 and 2 sccm by means of a mass flow controller (MKS Instruments). Substrate voltages Usub between 0 and –100 V were applied. ITO layers were deposited from a metallic and from an oxidized target. To obtain a metallic target, the target was cleaned by sputtering for about 30 minutes with pure argon as working gas. To get an oxidized target we worked for at least 30 minutes with additional 2 sccm oxygen gas flow prior to deposition. X-ray photoelectron spectroscopy (XPS) measurements were performed with a conventional surface analysis system (VG MT 500) to investigate the composition of the deposited ITO films. Samples were transferred under vacuum from the deposition chamber to the surface analysis chamber. Figure 2 shows a typical XPS result for the Sn 3d5/2 peak obtained using the Mg Kα x-ray line (hν = 1253.6 eV), and its decomposition into two, metallic tin and tin oxide, peaks.

Plasma-assisted deposition

3

Langmuir probe substrate

ellipsometer

magnetron

Fig. 1. Deposition chamber with magnetron (schematic).

Grazing incidence x-ray reflectometry (GIXR) investigations were performed on a θ − 2θ diffractometer (Siemens D 5000) with a special thin film attachment [8]. Film thickness and density were determined by fitting the resulting reflectometry curves using the REFSIM1.1 software (Siemens). The deposition time for samples investigated by GIXR was 30 s. Knowing the deposition rate for every deposition condition, a second series of samples was produced with films of 50 nm thickness. These films were investigated by grazing incidence x-ray diffractometry (GIXD) to investigate the phase composition in the films. The angle of the incident x-ray beam was fixed at 0.7◦ /2θ. The scanning region was 25-45◦ /2θ (step width 0.02◦ /2θ, scanning time 2 s per step). The as-deposited samples consist of crystalline metallic In/Sn and/or x-ray amorphous indium-tin-oxide depending on the deposition conditions. Crystalline indium-tin-oxide films were obtained by isothermal postannealing at 200 ◦ C in a high temperature chamber (B¨ uhler HDK2.4) attached to a θ − θ-diffractometer (Seifert XRD 3000). During annealing the films were step-scanned continuously at the measuring conditions mentioned above. For every measurement the integral intensity of the ITO(222) x-ray reflection was determined by a pseudo-Voigt fit procedure. From the time dependence of the integral ITO(222) x-ray intensity during annealing the dif-

4

R. Hippler, H. Steffen, M. Quaas, T. R¨ owf, T. M. Tun, H. Wulff 3400

Sn-3d 3200

0.3 sccm O , 2

intensity (a.u)

3000

V

=0 V

bias

2800

2600

2400

2200

2000

1800

1600

490

488

486

484

482

binding energy (eV)

Fig. 2. Sn-3d5/2 XPS peak of deposited indium-tin-oxide and its decomposition (see text).

fusion coefficients D and the initial amount of amorphous indium-tin-oxide in the as-deposited films were determined. The amount of amorphous ITO in the as-deposited films is characterized by the initial phase composition factor f0 which we define as the concentration ratio of indium oxide relative to the total indium concentration in the film. The mathematical model used here is described in detail in Ref. [9]. The energy influx from the plasma towards the substrate has been measured by a simple thermoprobe [10][11]. The probe consists of a copper substrate (diameter: 3.4 cm, thickness 0.02 cm), which is spot-welded to a thermocouple (type j) and placed within a solid shield. The substrate is only connected to the thermocouple and an additional wire for biasing. The measurement of the total energy influx is based on the determination of the difference between the time derivatives of the substrate temperature during heating (plasma on) and cooling (plasma off). This difference is proportional to the energy influx. The proportionality factor is the specific heat of the substrate. It was determined by a known thermal power source. The energy distribution function IEDF of plasma ions was investigated by energy resolved mass spectrometry employing a so-called plasma monitor (Balzers PPM 421). Mainly positive Ar+ , In+ , atomic O+ and molecular O+ 2 ions and negative O− ions were detected.

3 Results In the following we report results for thin ITO film deposition as function of oxygen gas flow and substrate bias voltage. The deposited films were analyzed by x-ray photoelectron spectroscopy (XPS) and grazing incidence diffractometry (GIXD) and reflectometry (GIXR).

Plasma-assisted deposition

5

Fig. 3. The composition (◦, In; △, Sn; •, O) of deposited indium-tin-oxide films as function of oxygen flow for a metallic target.

3.1 Chemical composition determined by photoelectron spectroscopy (XPS) Oxygen dependence. Fig. 3 shows the composition of thin ITO films deposited in Si(100) as function of oxygen gas flow, with the film composition investigated by XPS. Without oxygen flow, metallic indium/tin is deposited. With increasing oxygen flow, the amount of oxygen in the deposited films increases, while the indium content decreases. Above an oxygen flow of 0.5 sccm a saturation of the oxygen and indium contents occurs. The tin content varies between 11 % and 17 % is, hence, significantly enhanced compared to the stoichiometric composition of the employed In/Sn target. Contrary to indium, the tin contents also shows a pronounced increase with increasing oxygen flow, reaching a maximum around 0.2 sccm followed by a gradual decrease. The above mentioned behavior indicates formation of an oxygen-poor indium-tin-oxide film, well below the stoichiometric oxygen fraction of 0.6 for In2 O3 . Further and more detailed studies reveal the existence of 2 phases, a metallic (oxygen-free) and a completely oxidized ITO phase. Bias voltage dependence. In addition, the composition of the deposited ITO films shows some variation with substrate bias voltage. Fig. 4 shows that the In content increases for negative bias voltages, reaching a maximum composition of about 54 % around –40 V. Similarly, the Sn content increases from 14 % to 16 %, whereas the oxygen content decreases from about 36 % to 30 %. Apparently, a negative

6

R. Hippler, H. Steffen, M. Quaas, T. R¨ owf, T. M. Tun, H. Wulff

Fig. 4. The composition (◦, In; △, Sn; •, O) of deposited indium-tin-oxide films as function of negative bias voltage for a metallic target. Oxygen flow was 0.5 sccm.

bias voltage leads to a more metal-rich and oxygen-poor indium-tin-oxide film. 3.2 Film properties determined by GIXD Figs. 5 and 6 show the x-ray patterns of as-deposited layers. The films were deposited with different oxygen flows and with a metallic target. The layers shown in Fig. 5 were deposited with a substrate voltage Usub = 0 V, the layers of Fig. 6 with Usub = –50 V. Samples deposited at low oxygen flow show reflections attributable to polycrystalline metallic indium [12]. No crystalline indium oxide phases were detected. Crystalline Sn or SnOx phases were also

In(101)

500

Intensity / a.u.

U

400

300

= 0 V

Sub

100 In(110)

In(002)

200

0 sccm O 0.5 sccm O 1.0 sccm O 1.5 sccm O 2.0 sccm O 26

28

30

32

34 2

36

38

2

2

2

2

2

40

/ degree

Fig. 5. GIXD patterns of as-deposited films for various oxygen flows (Usub = 0 V).

Plasma-assisted deposition

7

U

Sub

In(101)

250

= -50 V

200

Intensity / a.u.

150

In(002)

100

50

0 sccm O

2

0.5 sccm O

2

1.0 sccm O

2

1.5 sccm O

2

2.0 sccm O

2

26

28

30

32

34 2

36

38

40

/ degree

Fig. 6. GIXD patterns of as-deposited films for various oxygen flows (Usub = −50 V).

not observed. The latter is not surprising since Sn atoms are believed to substitute In atoms without changing the lattice structure [8]. The amount of crystalline metallic In/Sn decreases with increasing oxygen flow. In case of a grounded substrate (Usub = 0 V) the films apparently become x-ray amorphous for oxygen flows larger than 1 sccm. Applying a negative substrate voltage Usub = −50 V results in partly crystalline metallic films at larger oxygen flows. It appears, hence, that a negative bias voltage has the same effect as reducing the oxygen flow to the substrate. The here observed effect is similar although much more pronounced compared to the bias voltage dependence noted already by XPS investigations (see above

Fig. 7. GIXD patterns of post-annealed films for various annealing conditions. Film deposition was without oxygen flow and without substrate voltage.

8

R. Hippler, H. Steffen, M. Quaas, T. R¨ owf, T. M. Tun, H. Wulff

Fig. 8. Oxygen-to-metal ratio versus oxygen flow. Data are obtained from GIXD (◦, Usub = 0 V; △, Usub = −50 V), and from XPS (•, Usub = 0 V) (see text).

3.3 Post-annealing of deposited indium-tin-oxide films In order to extract further information the as-deposited films have been postannealed for several hours at 200 o C. During annealing the crystal structure changes from an amorphous to a crystalline indium-tin-oxide phase (Fig. 7). The appearance of a crystalline indium-tin-oxide phase is due to a rapid crystallization process of the initially amorphous indium-tin-oxide, and, if not enough oxygen is present in the film, by relatively slow diffusion of oxygen into the film followed by crystallization. A model that separates the two oxygen contributions by their different time dependencies has been applied [9]; it allows to calculate the initial oxygen-to-metal ratio and the oxygen diffusion coefficient D. The model ignores the chemical difference between Sn and In atoms and, hence, treats Sn and In equally. As is expected, the extracted oxygen-to-metal ratio shows a pronounced dependence on oxygen flow. There exists a distinct difference between the GIXD and the XPS measurements. It seems to imply that deposited films consist of two phases: an amorphous metallic In phase not detected by GIXD and a fully oxidized indium-tin-oxide phase. 3.4 Deposition rate and film density determined by GIXR The measured deposition rates and film densities derived from GIXR investigations are displayed in Fig.s 9 and 10. With increasing oxygen flow the deposition rate (in nm/s) decreases by almost a factor of 3. At the same time, the film density increases, reaching the In2 O3 density for oxygen flows larger than about 1 sccm.

Plasma-assisted deposition

9

Fig. 9. Deposition rate of indium-tin-oxide films vs. oxygen flow for a metallic target and for the indicated substrate bias voltages.

3.5 Oxidized target Indium-tin-oxide films deposited with an oxidized target differ significantly from those obtained with a metallic target. Fig. 11 shows the XPS results for as-deposited films (i.e., films that have been kept under vacuum), films that were exposed to ambient air, and post-annealed films. The as-deposited films now show the proper stoichiometric ratios of ITO, i.e., an oxygen content of

Fig. 10. Density of deposited indium-tin-oxide films vs. oxygen gas flow for a metallic target and for the indicated bias voltages.

10

R. Hippler, H. Steffen, M. Quaas, T. R¨ owf, T. M. Tun, H. Wulff

Fig. 11. Composition of as-deposited, air-exposed, and post-annealed indium-tinoxide films (◦, In; •, O; △, Sn).

about 60 %. The tin content is about 5 % and thus in good agreement with the stoichiometric ratio of the target. Air-exposed and post-annealed films show even higher oxygen contents, indicating oxygen-rich indium-tin-oxide. Similar observations have been reported by Mergel et al. [5] and explained by an incorporation of additional oxygen in the bixbyite structure, possibly into constitutional vacancies. 3.6 Ion energy distribution function (IEDF) The observed dependencies of film characteristics on substrate voltage suggest that impinging ions play a significant role during film deposition. In order to investigate this influence we measured the kinetic energy distribution functions of positive and negative ions. Figure 12 shows the energy distribution function IEDF of positive ions measured by energy resolved mass spectrometry. Without oxygen gas flow the positive ion mass spectrum mainly consists of Ar+ ions. The Ar+ signal displays a maximum around 1 eV followed by a pronounced intensity drop. Note that the Ar+ energy distribution displayed in figure 12 was derived from the measured Ar-36 isotope signal and divided by the isotope abundance (0.3365 %) to yield the total Ar+ signal. The energy distribution of In+ ions differs from this behavior as it displays a pronounced tail towards larger energies, presumably reflecting the physical sputtering process by which indium atoms are produced. The measured IEDF also shows that In+ ions only partly thermalize at the gas densities of interest here. The total In+ intensity without oxygen flow is about 2 orders of magnitude smaller than compared to Ar+ . With an oxygen flow of 1 sccm oxygen flow, the In+ signal reduces by a factor of 4, while only a modest reduction is observed for Ar+ ions. The oxygen ion intensity is now reaching a level comparable to the Ar+ signal. Most abundant are molecular O+ 2 ions

Plasma-assisted deposition

11

Fig. 12. Kinetic energy distribution of positive ions in dc magnetron discharge + (argon gas flow 20 sccm, oxygen gas flow 1 sccm). •, Ar+ ; △, O+ , ∇, O+ 2 , ♢, In ; ◦, Sn.

while atomic O+ ions are about one order of magnitude less abundant. Oxygen ions also display a high-energy tail suggesting that a certain fraction of these ions stems from sputtering processes at the magnetron target. Due to the negative biasing of typically –400 V it is unlikely, however, that positive ions directly escape from the target. Opposite to positive ions which become accelerated in the plasma sheath region, negative ions are partly trapped inside the plasma by the positive plasma potential of typically 1-2 Volt. Only negative ions with sufficient kinetic energy can overcome this potential barrier. The most abundant negative ion species is O− . The energy distribution function of negative O− ions consists of several groups. Most intense is a low energy group with near-zero kinetic energies. These ions are believed to be produced inside the plasma. Several groups of high-energy O− exist. Most prominent is a group of ions around 50 eV followed by another small group around 200 eV. The last group has energies of 400-500 eV; its origin is due to the sputtering of oxygen atoms at the target. This is corroborated by its sharp onset which corresponds to the voltage drop at the target. The origin of the first and second groups of the high-energy ions is presumably also due to sputtering of atomic or molecular oxygen atoms at the target. Negative molecular O− 2 ions were also observed, albeit at a significantly reduced total intensity of few percent compared to the total intensity of atomic O− ions. Most of the O− 2 ions have rather small kinetic energies, but few ions with kinetic energies of ≈ 400 eV are also observed.

12

R. Hippler, H. Steffen, M. Quaas, T. R¨ owf, T. M. Tun, H. Wulff

Fig. 13. Kinetic energy distribution of negative oxygen (◦, O− ; •, InO− ).

In addition, small amounts of negative InO− ions were detected. The total intensity is about two orders of magnitude smaller than compared to the O− ions. As shown in figure 13, a significant fraction of InO− ions has large kinetic energies. The energy distribution is somewhat different compared to O− ions, however, suggesting a more complicated formation processes, however. 3.7 Energy influx The energy influx to the substrate was measured for both metallic and oxidized targets. The influence of oxygen flow, substrate bias and oxidation state of the target was investigated. For a metallic target, the measured energy influx increases from 10 to 11.5 mJ cm−2 s−1 with increasing oxygen flow (Fig. 14). For an oxidized target, the energy influx is about 0.9 mJ cm−2 s−1 smaller compared to the metallic target; this difference appears to be independent of the oxygen flow. In order to explain this behavior one has to consider the different contributions to the energy influx. Major contributions to the measured energy influx are due to sputtered neutral particles, electrons, positive and negative ions, and the released heat of condensation [13]. The contribution of sputtered particles to the energy influx is given by the product of flux density and mean kinetic energy. The energy distribution function of sputtered species was calculated with the help of a transport-of-ions-in-matter (TRIM) code [14]. The flux density was estimated from the deposition rate and the mass density (Figs. 9 and 10). Contributions due to positive ions and electrons were calculated with the help of Langmuir probe theory. Finally, the contribution due

Plasma-assisted deposition

13

-2

-1

energy influx (mJ cm s )

11.5 11.0

0.9 mJ cm-2 s-1 10.5 10.0

1.5 mJ cm-2 s-1 9.5

oxidized metallic

9.0

0.0

0.5

1.0

1.5

2.0

oxygen flow (sccm) Fig. 14. Energy influx various oxygen flows (Usub = −11 V).

to the condensation of In, Sn and O particles has been determined from the measured deposition rate and taking the specific heat of condensation into account. In case of a metallic target, no oxygen flow, and for floating potential conditions (Usub ≈ −11 V) the calculation yields a total energy influx of 7.3 mJ cm−2 s−1 . The contribution of sputtered In/Sn atoms amounts to 4.4 mJ cm−2 s−1 , 0.2 mJ cm−2 s−1 are due to the heat of condensation, while positive ions (mainly Ar+ , O+ , and O+ 2 ions) and electrons contribute 2.0 mJ cm−2 s−1 and 0.7 mJ cm−2 s−1 , respectively. The calculated energy influx is somewhat lower than the measured value of 9.8 mJ cm−2 s−1 (Fig. 14). The reason for the difference could be contributions of secondary electrons and backscattered Ar ions, which were not taken into account. In case of an oxidized target and/or with oxygen flow additionally the influence of negative oxygen ions needs to be taken into account. With increasing oxygen flow the deposition rate decreases, which is only partly offset by an increase in film density. Hence, we should expect a decrease of the energy influx with increasing oxygen flow, whereas an increase is observed. The noted difference should be explained by negative ions of relatively high kinetic energy contributing to the energy influx to the substrate.

4 Summary Results for plasma-assisted deposition of indium-tin-oxide films by reactive magnetron sputtering have been reported. Pronounced dependencies of, e.g.,

14

R. Hippler, H. Steffen, M. Quaas, T. R¨ owf, T. M. Tun, H. Wulff

film composition, film density, and deposition rate on oxygen flow, substrate bias voltage, and the oxidation state of the target (metallic or oxidized) have been observed. The observed dependencies are partly attributed to the changing energy-influx to the substrate and to the role played by positive and negative ions.

References 1. Chopra K L, Major S, Pandya D K. Thin Solid Films 102, 1 (1983) 2. Song P K, Shigesato Y, Yasui I, Ow-Yang C W, Paine D C, Jap. J. Appl. Phys. 37, 1870 (1998) 3. Bender M, Trube J, Stollenwerk J, Appl. Phys. A 69 (1999) 397 4. Meng L J, dos Santos M P, Appl. Surf. Sci. 120, 243 (1997) 5. Mergel D, Stass W, Ehl G, Barthel D, J. Appl. Phys. 88, 2437 (2000) 6. Quaas M, Steffen H, Hippler R, Wulff H, Surf. Sci. 454, 790 (2000) 7. Wulff H, Quaas M, Steffen H, Thin Solid Films 355, 395 (1999) 8. e.g., Wulff H, Steffen H, In: Low Temperature Plasma Physics (Hippler R, Pfau S, Schmidt M, Schoenbach K H, Eds.), Berlin: Wiley-VCH, p. 253 (2001), and references therein 9. Wulff H, Quaas M, Steffen H, Hippler R. Thin Solid Films 377-378, 418 (2000) 10. Kersten H, Stoffels E, Stoffels W W, Otte M, Csambal C, Deutsch H, Hippler R, J. Appl. Phys. 87, 3637 (2000) 11. Rossnagel S M, Cuomo J J, Westwood W D (ed.) Handbook of Plasma Processing Technology. NOYES Publications: New Jersey, p. 234 (1990) 12. JCPDS-ICDD 5-642 13. Kersten H, Deutsch H, Steffen H, Kroesen G M W, Hippler R, Vacuum 63, 385 (2001) 14. Ziegler J F, Biersack J P, Littmark U, The Stopping and Range of Ions in Solids, Pergamon Press, New York (2003)