Journal of Materials Science: Materials in Electronics

, Volume 23, Issue 4, pp 964–971

Formation of CuInSe2 absorber by rapid thermal processing of electron-beam evaporated stacked elemental layers

Authors

  • Zhao-Hui Li
    • Department of Electronics EngineeringKyungwon University
  • Eou-Sik Cho
    • Department of Electronics EngineeringKyungwon University
    • Department of Electronics EngineeringKyungwon University
  • Mario Dagenais
    • Electrical and Computer EngineeringUniversity of Maryland
Article

DOI: 10.1007/s10854-011-0528-1

Cite this article as:
Li, Z., Cho, E., Kwon, S.J. et al. J Mater Sci: Mater Electron (2012) 23: 964. doi:10.1007/s10854-011-0528-1

Abstract

Copper indium diselenide (CuInSe2) film was formed using the rapid thermal process (RTP) of electron-beam (E-beam) evaporated Cu–In stacked elemental layers as an absorber layer for chalcopyrite solar cells. The E-beam evaporation method could accurately control the thickness of each elemental layer so that the chemical composition of Cu–In precursor could also be controlled by changing the thickness ratio of the Cu/In metal layer. The device-quality Cu-deficient CuInSe2 film was obtained when the thickness ratio of Cu/In was about 1/2.5. Otherwise, the RTP process time was also certified as one of the most important parameters of RTP when the temperature was over 550 °C. A single-phase CuInSe2 was successfully obtained after rapid thermal process under 580 °C for 3 min. Finally, the growth mechanism of CuInSe2 film during RTP process was also summarized.

1 Introduction

Among the polycrystalline materials used as absorbers in thin film solar cells, the chalcopyrite based compounds such as CuInSe2 (CIS), generally alloyed with Ga or S, have achieved the maximum conversion efficiencies over 20%. CIS has one of the highest optical absorption coefficients, so that it can absorb 90% of the incident photons with energies higher than 1.0 eV within 1–2 μm thickness. Otherwise, CIS-based solar cells are very stable, thus allowing long operation lifetimes [17].

Several different approaches have been proposed for the deposition of CuInSe2-based chalcopyrite absorber films. At present, the promising approaches can be divided into two approaches. One is the co-evaporation of each element, but this approach is also being questioned for its suitability for low cost and larger area applications of CuInSe2 based cells [8, 9]. Another approach is a selenization for the preformed Cu–In precursor. The selenization process is usually carried out under an H2Se/Ar mixture ambient. Even though some experimental results revealed that the H2Se gas was the best selenium source, the gas is highly toxic and requires special precaution for its use [10, 11]. An alternative reaction approach is the rapid thermal processing (RTP) of thermally evaporated solid Se on the Cu–In precursor.

The property of the Cu–In precursor directly affects the characteristics of CuInSe2 solar cells. The metal precursor can be formed by a variety of methods, the most common of which is sputtering and evaporation method from alloy or metal sources [1214]. In these methods, the stacked elemental layer (SEL) approach is regarded as the most promising technology because it is an efficient low cost method for the large area commercial mass production. Some papers reported that the sequence of elemental layer affected the surface morphology and structure of CuInSe2 absorber. Usually the first layer of the stacked precursor is a Cu layer in order to avoid the adhesion problems. Furthermore, the Cu/In/Cu…multilayer instead of the Cu/In bilayer can result in the smoother surface morphology and the better crystallinity of CIS films.

On the other hand, the use of the elemental source is more favorable than that of the alloy source, because the latter leads to a change of the chemical composition of the film under different deposition conditions. Otherwise, although the sputtering method is usually used in mass production, unstable discharging always occurs during the sputtering process of the In layer. It could cause the large fluctuation of the thickness and the non-uniformity of the chemical composition of the Cu–In precursor. However, the evaporation method using elemental sources could overcome these weaknesses. Moreover, the selenization of the CIS-based films using solid Se was reported and high efficient devices were also obtained [15, 16]. But there are little reports about the growing mechanism of CIS-based films due to extremely complex reaction process. In the present work, we formed high-quality CIS films by rapid thermal processing of the Se-containing precursors and summarized the growing mechanism of CIS films according to experimental results.

2 Experimental procedure

Molybdenum-coated soda-lime glass (Mo/SLG) plates were used as substrates. These Mo coatings were prepared by in-line DC sputtering on well cleaned soda-lime glass. The sputtered Mo layer was typically 5,000 Å thick with a sheet resistance of ~0.6 Ω/sq. The elemental layers of Cu and In were deposited on the Mo/SLG using the E-beam evaporation method from 99.99% pure source at room temperature. Figure 1a shows the schematic of the E-beam evaporation system. The individual elemental layers were deposited without breaking vacuum from a rotatable crucible. The schematic diagram of the stacked sequence of the elemental layers is also shown in Fig. 1b. The thickness of each metallic layer was controlled by in situ quartz crystal monitoring. The Cu/In thickness ratio was varied from 1/1.8 to 1/3.0 to control the chemical composition of the CuInSe2 layer and the total thickness was kept at about 7,000 Å. Table 1 shows the evaporated thickness of each metallic layer in detail. Here typical rates of evaporation were 1.0 Å/s for the Cu layer and 2.0 Å/s for the In layer. After evaporation of Cu–In multilayer, a Se layer with a thickness of ~2 μm was deposited using thermal evaporation from a tungsten boat at room temperature.
https://static-content.springer.com/image/art%3A10.1007%2Fs10854-011-0528-1/MediaObjects/10854_2011_528_Fig1_HTML.gif
Fig. 1

Schematic of a the E-beam evaporation system and b the stacked elemental layers of the precursor

Table 1

Thickness of each elemental layer by E-beam evaporation

Sub.#

Cu layer (Å)

In layer (Å)

Dep. cycles

Total thickness (Å)

Cu/In thickness ratio

#1

830

1,500

3 cycles

~7,000

1/1.8

#2

780

1,560

3 cycles

~7,000

1/2.0

#3

700

1,630

3 cycles

~7,000

1/2.3

#4

670

1,670

3 cycles

~7,000

1/2.5

#5

580

1,740

3 cycles

~7,000

1/3.0

The Se-containing precursor film was subsequently moved into the quartz reaction tube of a rapid thermal processor, which was provided with 15 tungsten halogen lamps both on the top and at the bottom as shown in Fig. 2. A reproducible thermal budget for each individual run was guaranteed by a computer controlled lamp power supply. The lamp power control was provided by thermocouple measurement. Before the RTP process, the reaction tube was pumped to ~10−2 torr and then the N2 gas (99.999%) was injected and the atmosphere was kept at about 200 torr. The rising rate and falling rate of the annealing temperature were ~60 °C/s and ~30 °C/s respectively. The annealing time and the temperature were varied to form the chalcopyrite CuInSe2 film.
https://static-content.springer.com/image/art%3A10.1007%2Fs10854-011-0528-1/MediaObjects/10854_2011_528_Fig2_HTML.gif
Fig. 2

Schematic of rapid thermal processor for CuInSe2 film

The chemical compositions of CuInSe2 thin films were analyzed by energy dispersive X-ray analysis (EDX). The morphology of these films was observed using a scanning electron microscope (SEM, Hitachi S-4700) and an optical microscope (OLYMPUS, BX51). And the phases and the crystal structure were also analyzed by an X-ray diffractometer (XRD, Rigaku D/MAX-2200) operated at 40 kV and 20 mA using Cu-kα radiation. In addition, the electrical properties were analyzed using Hall-effect measurement (ECOPIA, HMS-3000).

3 Results and discussion

Figure 3 shows the surface morphology of the as-deposited Cu–In multilayer on the Mo/SLG with the different Cu/In thickness ratios, prior to Se film evaporation. We found that the Cu/In thickness ratio affected the surface morphology despite the fact that an indium layer was on top. The morphology of the Cu–In multilayer was characterized by small nodules when the Cu/In thickness ratio varied from 1/1.8 to 1/2.0. These nodules are due to a very In-rich Cu–In alloy [17]. With an increase of the indium proportion in these films, these nodules became gradually larger and formed island-like grains, as shown in Fig. 3c, d, and e. This might come from an agglomeration arising from large amount of deposited indium which forms the CuIn alloy or In cluster on the top layer.
https://static-content.springer.com/image/art%3A10.1007%2Fs10854-011-0528-1/MediaObjects/10854_2011_528_Fig3_HTML.jpg
Fig. 3

Surface morphology of as-deposited Cu–In stacked layer on Mo/SLG with different Cu/In thickness ratios: a 1/1.8, b 1/2.0, c 1/2.3, d 1/2.5, and e 1/3.0

In Fig. 4, the XRD spectrums are shown for the as-deposited Cu–In multilayer with various Cu/In thickness ratios. These peaks correspond to the elemental In, Cu and CuIn alloy. It means that the Cu-, In-, and CuIn-phases co-exist in this metallic layer. For the CuIn-phase, it could be formed via the diffusion and the reaction of the elemental Cu and In during the evaporation process even though at room temperature. From above XRD spectrums, it is clear that the peak intensities of Cu–In precursor varied with different Cu/In thickness ratios. The peak intensities of both CuIn(200) and CuIn(102) decreased with the decrease of the Cu/In thickness ratio due to the relatively smaller Cu amount in the precursor. On the contrary, both In(101) and In(002) peak intensities increased with the decrease of the Cu/In thickness ratio. In other words, the In-phase peak intensity increased with the increase of the In proportion in this metallic layer.
https://static-content.springer.com/image/art%3A10.1007%2Fs10854-011-0528-1/MediaObjects/10854_2011_528_Fig4_HTML.gif
Fig. 4

XRD spectrums of as-deposited Cu–In stacked layer on Mo/SLG with different Cu/In thickness ratios: a 1/1.8, b 1/2.0, c 1/2.3, d 1/2.5, and e 1/3.0

The rapid thermal process was carried out at 550 °C for 2 min after a Se layer with thickness of ~2 μm was deposited on the Cu–In precursor using the thermal evaporation method. Figure 5 shows the top morphology of CuInSe2 films annealed under 550 °C for 2 min. The SEM images of CIS films show that a rough surface was formed after the RTP process. And the grain size also had a large fluctuation range from 1 to 5 μm. It resulted from a rough surface and large grain of the deposited metal precursors. Therefore, in order to form smooth CIS films, an uniform and smooth metal precursor should be deposited before RTP.
https://static-content.springer.com/image/art%3A10.1007%2Fs10854-011-0528-1/MediaObjects/10854_2011_528_Fig5_HTML.jpg
Fig. 5

Surface morphology of the CuInSe2 films annealed at 550 °C for 2 min with different Cu/In thickness ratios: a 1/1.8, b 1/2.0, c 1/2.3, d 1/2.5, and e 1/3.0

In order to measure the electrical properties of CIS films using the Hall-effect measurement system, we attempted to form the CIS films on the bare SLG. Unfortunately this approach proved unworkable since the CIS films were peeled off from substrates during selenization using RTP. Even though, the CIS/Mo/SLG samples were formed and the Hall-effect measurement was carried out, its result was similar to that of the Mo/SLG. Therefore, the Hall-effect measurement of CIS/SLG could not be performed. From this method, the electrical properties cannot be obtained since the Mo layer, which is under CIS layer, has too lower resistivity.

The chemical constituents of RTP-CuInSe2 were also analyzed using energy dispersive X-ray analysis (EDX). Three different regions were measured for each sample and the average result was calculated. We found that the composition had large fluctuations with the different Cu/In thickness ratios. Table 2 just shows the results of the sample with the Cu/In thickness ratio of 1/2.5. The average atomic ratio of Cu/In was about 86%. It revealed that the CuInSe2 layer or its surface layer was a slightly Cu-deficient layer. B. Canawa et al. [18] reported the composition of this Cu-deficient layer corresponds to the stable ordered vacancy compound or ordered defect compound (OVC/ODC) CuIn3Se5. The OVC surface layer was weakly n-type while the bulk of the CuInSe2 absorber was p-type [19]. So the inverted surface could minimize the recombination at the CuInSe2/CdS interface and improve the device performance. To summarize, a high quality CuInSe2 film was obtained when the precursor with a Cu/In thickness ratio of about 1/2.5 was selected.
Table 2

EDX results of CuInSe2 with the Cu/In thickness ratio of 1/2.5 after RTP treatment at 550 °C for 2 min

Measured zone

Atomic concentration (at%)

Cu

Se

In

Cu/In atomic ratio

I

25.15

46.51

28.34

88.74

II

25.17

46.8

28.03

89.80

III

23.95

46.57

29.48

81.24

Average

24.76

46.63

28.62

86.51

The XRD spectrum of RTP-CuInSe2 layer is also shown in Fig. 6. We found that the CuInSe2 peak appeared after RTP for all the samples. During the RTP process, the Cu, In and Cu–In binary alloy diffused and inter-reacted with each other and then formed the CuInSe2 crystalline phase. This reaction has a complex pathway, and some papers reported their reaction process at different annealing temperatures. At the melting point of Se (208–225 °C), the Cu–In metallic phases preformed in the stacked layer begin to decompose and react slowly with Se. At temperatures between 225 and 275 °C the free Cu and In react with the melted Se and forms CuSe, CuSe2, In4Se3, and Cu2−xSe. However, the CuSe, CuSe2 and In4Se3 are always unstable and transform to Cu2−xSe and InSe respectively. CuInSe2 starts to form at temperatures between 375 and 385 °C through the reaction of the Cu2−xSe and InSe under Se vapor atmosphere [2023]. As shown in Fig. 6, the secondary InSe-phase co-exists in the CuInSe2 film. The presence of the secondary InSe-phase may be due to insufficient reaction under such RTP treatment conditions. In order to ensure sufficient reaction, we varied the annealing temperature and time respectively to anneal the stacked precursors. Figure 7 shows the XRD spectrum of the RTP-CuInSe2 under 550–620 °C for 2 min. The XRD results reveal that the InSe peak intensity cannot easily be decreased even at higher temperature than 550 °C in the RTP process. From this analysis it was concluded that increasing the RTP temperature was not an effective way to ensure sufficient selenization reaction.
https://static-content.springer.com/image/art%3A10.1007%2Fs10854-011-0528-1/MediaObjects/10854_2011_528_Fig6_HTML.gif
Fig. 6

XRD spectrums of the CuInSe2 films annealed at 550 °C for 2 min with different Cu/In thickness ratios: a 1/1.8, b 1/2.0, c 1/2.3, d 1/2.5, and e 1/3.0

https://static-content.springer.com/image/art%3A10.1007%2Fs10854-011-0528-1/MediaObjects/10854_2011_528_Fig7_HTML.gif
Fig. 7

XRD spectrum of the CuInSe2 films with the Cu/In thickness ratio of 1/2.5 annealed at different temperatures for 2 min: a 550 °C, b 580 °C, c 600 °C, and d 620 °C

In addition, the Mo back contact layer was broken because the glass substrate became soft when the annealing temperature was over 600 °C. So we varied the RTP time from 1 to 5 min at 580 °C to anneal the precursor and the results are shown in Fig. 8. The InSe peak intensity became smaller with increasing of the RTP time and finally disappeared when the RTP time was about 3 min. Conversely, the CuInSe2 peak intensity increased when the RTP time was increased. For the thin film selenized at 580 °C for 5 min, unfortunately, the film surface appeared gray in color with some dark patches on all sides. This might be due to the oxygenation of the metal element (especially the In element) during the rapid thermal process.
https://static-content.springer.com/image/art%3A10.1007%2Fs10854-011-0528-1/MediaObjects/10854_2011_528_Fig8_HTML.gif
Fig. 8

XRD spectrum of the CuInSe2 films with the Cu/In thickness ratio of 1/2.5 annealed at 580 °C for different time: a 1 min, b 2 min, c 3 min, and d 5 min

According to experimental results and analyses, the reaction mechanism with CuInSe2 film can be summarized as follows:

  1. 1.
    Each elemental layer is deposited on Mo coated glass with the order of Cu/In…Cu/In as shown in Fig. 9a. During and after each elemental layer deposition process, Cu and In atoms interdiffuse and react with each other and form CuIn alloy at the interface between the Cu and the In layer. Prior to RTP treatment, the free-Cu, -In, and CuIn alloy co-exist in the film, as depicted in Fig. 9b.
    https://static-content.springer.com/image/art%3A10.1007%2Fs10854-011-0528-1/MediaObjects/10854_2011_528_Fig9_HTML.gif
    Fig. 9

    Schematic of reaction pathway of precursor during RTP treatment: a Cu, In layer is deposited on Mo coated glass in the order of Cu/In…Cu/In. b Cu and In react each other and form CuIn phase prior to RTP treatment. c Cu, In, and CuIn react with Se and form transient phase Cu2−xSe and InSe. But the Cu2−xSe rapidly reacts with InSe and formed CuInSe2. d Reaction finishes and some Cu2−xSe may co-exist with CuInSe2

     
  2. 2.
    During RTP treatment, the Se reacts with CuIn alloy and each element subsequently forms binary metal selenide. Then these selenides react with each other under Se vapor atmosphere and form CuInSe2. Whether the reaction thoroughly completes or not depends on the reaction temperature and the reaction time. However, the results have shown that the reaction time was more important than the reaction temperature for temperatures higher than 500 °C. Thus, the main reaction can be described by (1)–(4) as
    $$ {\text{Cu}} + {\text{Se}} \to {\text{Cu}}_{{2 - {\text{x}}}} {\text{Se}} $$
    (1)
    $$ {\text{In}} + {\text{Se}} \to {\text{InSe}} $$
    (2)
    $$ {\text{CuIn}} + {\text{Se}} \to {\text{Cu}}_{{2 - {\text{x}}}} {\text{Se}} + {\text{InSe}} $$
    (3)
    $$ {\text{Cu}}_{{2 - {\text{x}}}} {\text{Se}} + {\text{InSe}} + {\text{Se}} \to {\text{CuInSe}}_{2} $$
    (4)
    According to our experimental result, the reaction (4) is relatively slow. It needs a relatively long time to react and form CuInSe2. If the reaction time is insufficient, the residual binary selenide will exist in the film with CuInSe2 (as depicted in Fig. 9c).
     
  3. 3.

    If the reaction time is quite sufficient, the reaction (4) will continue, and finally the reaction will be completed with the depletion of all of the binary selenide. The existence of the Cu2−xSe phase in CuInSe2 film was not verified by the XRD results due to the overlapping between the main peaks for Cu2Se and CuInSe2.

     

4 Conclusions

In this experiment, we prepared the Cu–In stacked elemental layers as a precursor for the formation of CuInSe2 film using the E-beam evaporation method, which could accurately control the thickness of each element and then optimize the chemical composition of the precursor by changing the Cu/In thickness ratios. Device-quality for the Cu-deficient CuInSe2 film was obtained when the Cu/In thickness ratio was about 1/2.5. However, the XRD results showed that the formed film was inhomogeneous since the InSe phase co-existed with the CuInSe2 phase after RTP treatment at 550 °C for 2 min. This might be due to an insufficient reaction during RTP treatment. In order to ensure sufficient reaction and remove the InSe phase, the RTP treatment temperature and time were varied from 550 to 620 °C and from 1 to 5 min, respectively. We found that the higher process temperature could not improve the quality of CuInSe2 film or promote the reaction of selenization. On the other hand, the longer RTP process time can improve the quality of CuInSe2 film and enhance the reaction of selenization. The single phase chalcopyrite CuInSe2 film was formed when the Cu–In precursor was annealed at ~580 °C for 3 min. In addition, the grown mechanism of CIS film was also summarized.

Acknowledgments

This research was supported by Basic Science Research Program through the National Research Foundation of Korea (NRF) funded by the Ministry of Education, Science and Technology (2010-0009454). This research was supported by the Kyungwon University Research Fund in 2011.

Copyright information

© Springer Science+Business Media, LLC 2011