Impact of Surface Treatment on the Structural and ... - Springer Link

Journal of ELECTRONIC MATERIALS, Vol. 42, No. 11, 2013

DOI: 10.1007/s11664-013-2649-1 Ó 2013 TMS

Impact of Surface Treatment on the Structural and Electronic Properties of Polished CdZnTe Surfaces for Radiation Detectors SULEYMAN TARI,1,3 F. AQARIDEN,1 Y. CHANG,1 C. GREIN,1 JIN LI,2 and N. KIOUSSIS2 1.—Sivananthan Laboratories, Inc., 590 Territorial Drive, Ste H, Bolingbrook, IL 60440, USA. 2.—W. M. Keck Computational Materials Theory Center, California State University at Northridge, Northridge, CA 91330-8268, USA. 3.—e-mail: [email protected]

We present the effects of surface treatments on the structural and electronic properties of chemomechanically polished Cd0.9Zn0.1Te before contact deposition. Specifically, polished CdZnTe (CZT) samples were treated with four distinct chemical etchants: (1) bromine methanol (BM), (2) bromine in lactic acid, (3) bromine in methanol followed by bromine–20% lactic acid in ethylene glycol, and (4) hydrochloric acid (HCl). The surface structure and surface electronic properties were studied with atomic force microscopy (AFM) and x-ray photoelectron spectroscopy (XPS). AFM images showed that three of the four etchants significantly altered the surface morphology and structure of CZT. All etchants created smoother surfaces; however, all except HCl also introduced high densities of defects. HCl was found to not affect the surface structure. XPS measurements indicated that a thick, 3 nm to 4 nm, TeO2 layer formed about 1 h after etching; hence, it is very important to process devices immediately after etching to prevent oxide formation. Key words: CdZnTe, radiation detector, etching, oxidation, surface, XPS

INTRODUCTION Cadmium zinc telluride (CZT) is a promising material for high-performance semiconductor x-ray and gamma-ray radiation detectors that can operate at room temperature.1–3 CZT bulk materials contain significant numbers of defects such as Cd vacancies, Te precipitates, impurities, and stoichiometric imbalances resulting in microcrystalline material. There is a comprehensive international effort to improve the crystalline quality of CZT.4 However, the performance of a radiation detector depends not only on the quality of its bulk material but also on its surfaces and metal/CZT interfaces. Surfaces and metal/ semiconductor interfaces can be the dominant factors influencing detector performance, especially for relatively soft x-ray photons and large, pixelated arrays, since they can significantly contribute to leakage currents, which in turn affect a detector’s (Received December 14, 2012; accepted May 23, 2013; published online June 18, 2013)

3252

signal-to-noise ratio and energy resolution. Atomically smooth and defect-free surfaces may be needed for high-performance CdZnTe-based detectors. However, surface structure, oxidation, defects, and contamination can alter the surface electronic structure to a large extent and may diminish detector performance. There are a number of reports addressing surface passivation5–8 and contact electrode9–11 improvements to decrease detector leakage currents. The effect of the various chemical solutions on the relation between the surface roughness and the leakage current of CZT detectors has also been studied.12 Chemomechanical polishing (CMP) is typically performed on CZT to obtain a smooth surface. After the CMP process, CZT is chemically etched in order to remove scratches that result from abrasives used in the polishing slurry. However, the chemical etch may significantly alter the CZT surface and introduce defects. It is also not easy to create reproducible surfaces with chemical processes. The effects of polishing and etching of CZT surfaces have been recently reported.13

Impact of Surface Treatment on the Structural and Electronic Properties of Polished CdZnTe Surfaces for Radiation Detectors

We present in this work the effects of surface treatment on the structural and electronic properties of CMP Cd0.9Zn0.1Te before contact deposition. CZT samples were treated with various chemical etchants. The surface morphology and structure were studied with bright-field optical microscopy (OM) and atomic force microscopy (AFM), and surface electronic properties were studied with x-ray photoelectron spectroscopy (XPS). EXPERIMENTAL PROCEDURES We have a significant history with polishing CZT substrates for subsequent molecular beam epitaxy (MBE) epilayer crystal growth. We achieved, through successive CMP and chemical polishing (CP) processes, a surface root-mean-square (rms) roughness value, as determined by AFM, of 0.8 nm on a Cd0.96Zn0.04Te wafer. The state-of-the-art rms roughness value is 0.5 nm for Cd0.96Zn0.04Te wafers from Nikko Materials, as routinely measured by us via AFM. We have transferred our expertise in Cd0.96Zn0.04Te wafer polishing to bulk Cd0.9Zn0.1Te samples for radiation detection applications.14 Single-crystal Cd0.9Zn0.1Te samples of size 20 mm 9 10 mm 9 5 mm were purchased from Creative Electron. The surface morphology of as-received samples was characterized with OM and AFM. The samples were then lapped using a Logitech PM4 system followed by two steps of CMP using a Logitech CP 3000 system. The details of the polishing process are reported elsewhere.14 Alumina abrasives of 1 lm size in a slurry were used for CMP. CP is in general applied as the last step to eliminate scratches from alumina abrasives. We did not apply CP in this study so as to clearly see the effects of the etching on surface morphology and structures. After CMP, four samples were etched in different solutions: (1) 4% bromine in methanol (BM), (2) 4% bromine–20% lactic acid in ethylene glycol (BLE), (3) BM followed by BLE (BMBLE), and (4) 4% hydrochloric acid (HCl) in deionized (DI) water. HCl is not an etchant, but we used it here to compare its effect on the CZT surface with the etchants. HCl, in general, is used to remove

3253

the thin oxide layers from CZT surfaces prior to the deposition of contacts. The etched samples were rinsed and dried as follows: BM etching was followed by two methanol rinses and then N2 drying; BLE and BMBLE were followed by ethylene glycol rinses, a DI water rinse, and then N2 drying; and HCl was followed by a DI water rinse and then N2 drying. Structural and electronic characterization techniques were employed before and after polishing and after surface treatments. AFM is extensively used for CZT surface and defect morphology studies. In this work, a Veeco D3100 scanning probe microscopy (SPM) system configured for tappingmode operation was used for AFM measurements. The measurements were done on CZT samples after polishing and after surface treatment steps with various scan areas. The surface chemical composition and the oxidation state of the samples were analyzed with XPS using a SSX-100 spectrometer with a monochromatic and focused Al Ka source (E = 1486.6 eV). Survey spectra were recorded with a 1000 lm spot diameter, and core-level spectra were recorded with a 600 lm spot diameter and 50 eV pass energy. The corresponding energy resolution for core-level scans is 0.9 eV, as measured by the full-width at half-maximum of the Au 4f core level. RESULTS The surface morphologies of CZT samples as received, after CMP, and after etching were studied with OM. As seen in Fig. 1a, there are visible defects on the surface of as-received CZT. These large defects, 30 lm to 50 lm in diameter, are likely Te inclusions and appear on each sample with a density of about 10 cm 2. The samples were first mechanically polished using 1-lm alumina abrasive in a slurry to eliminate these defects. The samples were then polished with a CMP process. Figure 1b shows the sample after CMP; the defect has been removed, but the alumina abrasive used for polishing created scratches. This is expected because CZT is a soft material.

Fig. 1. Optical microscopy images of (a) an as-received bulk CZT sample and (b) the same sample after CMP.

3254

Tari, Aqariden, Chang, Grein, Li, Kioussis

Fig. 2. Optical microscopy images of CZT samples after etching with (a) BM, (b) BLE, (c) BMBLE, and (d) HCl.

Figure 2 shows OM images of four CZT samples after surface treatment via the four etching processes. As seen in Fig. 2a, BM etches away the scratches and the surface looks smoother; however, a high density of etch pits is present. BLE etching (Fig. 2b) results in a relatively lower number of etch pits, but the surface has an orange-peel-like structure. The BMBLE etchant resulted in a smooth surface and a low density of etch pits along with a trace of scratches, as shown in Fig. 2c. As seen in Fig. 2d, the scratches are still present and the surface shows some roughness, which means HCl does not affect the surface morphology. All three employed etchants, except HCl, significantly affect the surface morphologies by smoothing the surface and introducing etch pits. It is not straightforward to obtain reproducible surfaces after etching. AFM scans were performed over various scan areas on the samples after the etching process. Before measurements, a frequency scan was performed to find the resonant frequency of the cantilever and the reference phase point or zero phase was set. It is very important to optimize the parameters and confirm that the tapping-mode AFM measurements do not damage the CZT surface. Figure 3 shows an AFM image of a CZT sample after CMP over a scan area of 5 lm 9 5 lm.

Fig. 3. AFM image of a CZT sample after CMP with a 1-lm grit size alumina abrasive (before etching).

Scratches resulting from the alumina abrasive are clearly seen, and the surface roughness is found to be 4 nm. A typical roughness found after CMP with a 1-lm grit size alumina abrasive is about 3 nm to 4 nm.


3255

Fig. 4. AFM images of CZT samples after etching with (a) BM, (b) BLE, (c) BMBLE, and (d) HCl.

AFM images of CZT samples after etching are shown in Fig. 4. The images all have a vertical height scale of 200 nm. The etchants significantly change the surface structure by smoothing the surface and introducing bump-like defects (Fig. 4a–c). Although the surfaces appear smoother than the unetched surface, the rms roughness of the BM-etched sample increases to 8 nm due to the formation of surface defects. For the BLE- and BMBLE-etched samples the surface roughness changes almost by half. There is no indication of any structural change or defect formation for the HCl-etched CZT. The scratches seen before etching are still present, as seen in Fig. 4d. Table I summarizes the rms roughness of CZT samples after CMP and after the various etchants. XPS is a nondestructive technique to study surface electronic properties and contamination. Relative concentrations of atomic constituents can be easily and quantitatively measured. Chemical state information in the surface region can be assessed by determining the exact positions of the constituent peaks, which can be done via fitting the high-resolution peaks. XPS survey and high-resolution peak

Table I. AFM rms surface roughness of CZT samples after CMP and after various etchants Sample

Roughness, rms (nm)

CMP BM BLE BMBLE HCl

5.7 8.2 2.2 2.9 5.5

scans for Te 3d, Cd 3d, and Zn 3d were recorded from a CZT sample after CMP and each of the four etching techniques. XPS was employed on etched surfaces: (1) immediately after etching, (2) 1 h after etching, and (3) 24 h after etching, in order to investigate the oxidation states of the CZT surfaces. The time elapsed after etching until preparation of a CZT sample for contact deposition is important to eliminate oxidation of the freshly etched surface. After etching, the CZT surface is Te rich and is very vulnerable to oxidation. Therefore, it is crucial to keep the time short between etching the surface to

3256


Fig. 5. XPS high-resolution spectra of Te 3d peaks (a) before and (b) immediately after etching for the various etchants.

remove oxygen layers and contact deposition for detector fabrication. Because contacts need to be deposited on two sides of CZT detectors and the presence of sample-handling issues for deposition, especially for electron beam and sputtering deposition, the time elapsed to the start of contact deposition may be 10 min to 30 min. All expected photoelectron peaks were clearly seen in a survey scan (not shown). Figure 5 shows the XPS spectra of Te 3d peaks before etching (after CMP), and immediately after each of the four etchants. The time elapsed after etching until loading into the XPS system was about 3 min. The Te 3d5/2 and Cd 3d5/2 peaks were fitted with a single mixed Gaussian plus Lorentzian, and the background was subtracted with a Shirley peak. All spectra were shifted with respect to the C 1s peak, which is at 284.6 eV. In the high-resolution scan for CMP, shown in Fig. 5a, Te and TeO2 peaks are seen clearly at 572.2 eV and 575.5 eV, respectively. The TeO2 thickness calculated from XPS compositional analysis was found to be about 1 nm. Figure 5b shows the high-resolution spectra of the Te peaks for the case immediately after etching. The positions of the Te 3d5/2 and Te 3d3/2 peaks at the low-energy side were found to be 572.2 eV and 582.2 eV, respectively, with spin–orbit splitting of 10 eV. The spectra shown in Fig. 5b are normalized for comparison purposes and to clearly observe the chemical shifts. There is no peak shape change after any etchant, indicating that the peaks correspond to only Te 3d. However, there is a very low-intensity peak for the BM-etched sample at about 575.6 eV, corresponding to TeO2. This oxidation shows that, immediately after etching, namely within a few minutes, oxide layers start forming on CZT when the BM etchant is employed. The reason for this very thin (couple of atomic layers) oxide layer may be either that oxidation starts immediately after BM etching, which makes it more vulnerable to oxidation, or that the oxide layer which formed after

the CMP process was not totally removed. Aspects of the etching process, such as the time elapsed between the preparation of the BM solution and starting the etch, and very precise control of agent concentrations, need to be optimized to verify the former assumption. We have some earlier results (not published) supporting the latter one. Full-width at half-maximum (FWHM) values of the Te 3d5/2 and Cd 3d5/2 peaks were found to be 1.46 ± 0.02 eV and 1.15 ± 0.05 eV, respectively, for all etchants immediately after etching, which corresponds to oxide-free surfaces, comparable to the values found for Te and Cd on clean CdTe surfaces. The intensity ratios of Cd 3d5/2 to Te 3d5/2 were found be 0.5 for BM, BLE, and BMBLE, and 0.61 for HCl etching, which indicates that the CZT surface is not stoichiometric but Te rich after etching. The sensitivity factors of Te 3d and Cd 3d are 5.70 and 3.97, respectively, and were taken into account in the calculations of the ratios. Note that different surface roughness observed after etching may impact the sensitivity factors of each species. The XPS spectra 1 h and 24 h after etching are shown in Fig. 6a and b, respectively. In both cases, there is a high-intensity peak at 572.2 eV along with a Te peak at 575.6 eV. The chemical shift between the Te and TeO2 peaks is 3.3 eV, leading us to conclude that the peak at 575.6 eV is associated with the presence of TeO2. The intensity ratio of TeO2 to Te is close to 1 for all etchants except BM, for which the ratio is about 0.6, indicating that the oxidation of a BM-etched surface is slower than for the other etchants. Assuming that a very thin layer (couple of atomic layers) of oxide is left on the CZT after BM etching, we can explain the fact that the BM-etched sample oxidized more slowly than the other samples for the 1 h case. We have seen in our study that oxide layer formation is slower and the thickness of the oxide layer is less than for a freshly prepared and etched CZT surface (not published). There is no significant Te 3d5/2 peak shape change,


3257

Fig. 6. XPS high-resolution spectra of Te 3d peaks at 1 h (left) and 24 h (right) after etching.

and the FWHM is found to be 1.5 ± 0.01 eV, which is very close to that immediately after etching. The difference is about 0.04 eV, which is not large enough to be associated to any other oxidation states of Te. The thickness of oxide was calculated from XPS compositional analysis of the Te and TeO2 peaks and found to be between 2 nm and 3 nm. For the case of 24 h after etching, which is shown in Fig. 6, the TeO2 to Te ratio is close to 1 for BM, BMBLE, and HCl, being almost saturated. This ratio is close to 2 with formation of the thickest oxide layers for BLE. The TeO2 layer was also found to be thicker for the case of 1 h after BLE etching, which indicates that it forms a surface structure that is prone to oxidation. It is possible that local surface roughness variations may have an impact on the oxidation peaks detected for BLE. However the HCl-etched sample does not show structural modifications, in contrast to the other etchants. Therefore, we believe that surface chemistry may affect the oxidation of CZT, although we did not observe significant changes in surface chemistry within the limit of our XPS instrument’s sensitivity. Figure 7 shows the spectra of the Cd 3d3/2 peaks of etched CZT samples after 24 h. The peak position and FWHM were found to be 404.6 eV and 1.25 ± 0.02 eV, respectively. The Cd spectra are normalized to clearly discern peak shape changes. Any peak shape change indicates the presence of a chemical shift; For example, the presence of Cd oxides is indicated by an increase in the Cd-peak FWHM. However, analysis of only one Cd spectrum is insufficient to conclude that oxidation has taken place because the CdO chemical shift is about 0.2 eV, which cannot be easily measured in raw spectra. The Cd spectrum needs to be fitted and any peak widening associated with the formation of Cd oxides. As seen in Fig. 6, Te on the CZT surface is already oxidized to a large extent 24 h after etching. The FWHM of the Cd peak immediately after etching was found to be 1.15 ± 0.05 eV, namely

Fig. 7. XPS high-resolution spectra of the Cd 3d peaks 24 h after etching.

when there is no oxidation of any kind. The difference in FWHM between immediately after and 24 h after etching is 0.1 eV, which is less than the CdO shift. The FWHM values of Cd peaks after each treatment versus time are plotted in Fig. 8. We take the immediately after etching case as the reference because the surface is oxide free and the Te and Cd peaks are not expected to have any chemical shifts. There is no significant change in the FWHM values after BM and BMBLE. The largest FWHM change is seen for HCl, where the change is 0.16 eV. We cannot associate this small peak width change with the formation of CdO. More work and detailed analysis of the Cd peak may be needed to clarify this observation. According to OM and AFM analysis, the BMBLE etch process appears to produce a better surface morphology and structure, in agreement with previous work.12 More work may be needed to optimize etching of CZT, possibly using various

3258


elapsed before contact deposition should not be more than a few minutes to realize oxide-free interfaces. BMBLE appears to be a better etchant as far as surface structure is concerned. However, we propose to improve the surface smoothness by further improving the CMP process and possibly CP to remove very fine scratches, instead of etching. HCl appears to be appropriate to remove the thin, 1 nm, oxide layer without altering the surface morphology and structure. Also, XPS analysis suggests that the HCl etch forms a more stoichiometric CZT surface than other etchants and does not appear to cause a chemical reaction. ACKNOWLEDGEMENTS The research at Sivananthan Laboratories and California State University at Northridge was supported by DTRA Grant No. HDTRA1-10-1-0113. Fig. 8. FWHM values of the Cd 3d5/2 peak for various etchants: immediately, 1 h, and 24 h after etching.

concentrations of etchant agents, especially for BMBLE. CONCLUSIONS CZT surfaces were polished by CMP and etched with various etchants. OM and AFM images clearly indicate that the BM, BLE, and BMBLE etchants significantly alter the surface morphology and structure. HCl does not appear to have a significant effect on the CZT surface. XPS analysis shows that all etchants are effective at removing oxide layers formed after CMP, and the surfaces are not stoichiometric but rather Te rich. The CZT surface is oxidized to a large extent 1 h after etching. A slight increase in the FWHM of the Cd peaks was observed after exposing BLE- or HCl-treated CZT to the atmosphere for 24 h. We showed that the time elapsed between etching CZT and contact deposition is crucial to prevent formation of TeO2. Oxide-free surfaces are desirable for detector fabrication to prevent formation of insulating layers at the interface of CZT and a contact. We suggest that the time

REFERENCES 1. O. Limousine, Nucl. Instrum. Method A 504, 24 (2003). 2. C. Szeles, IEEE Trans. Nucl. Sci. 51, 1242 (2004). 3. G. Yang, W. Jie, Q. Li, T. Wang, G. Li, and H. Hua, J. Cryst. Growth 283, 431 (2005). 4. Y.J. Li, G.L. Ma, X.N. Zhan, and W.Q. Jie, J. Electron. Mater. 31, 840 (2002). 5. S. Wenbin, W. Kunshu, M. Jiahua, T. Jianyong, Z. Qi, and Q. Yongbiao, Semicond. Sci. Technol. 20, 343 (2005). 6. G. Zha, J. Wang, T. Tan, and P. Li, Appl. Surf. Sci. 253, 3476 (2007). 7. M. Amman, J.S. Lee, and P.N. Luke, IEEE Trans. Nucl. Sci. 53, 3035 (2006). 8. M.C. Duff, D.B. Hunter, A. Burger, M. Groza, V. Buliga, and D.R. Black, Appl. Surf. Sci. 254, 2889 (2008). 9. Q. Li, W. Jie, Li Fu, X. Wang, and X. Zhang, Appl. Surf. Sci. 253, 1190 (2006). 10. E.J. Morton, M.A. Hosain, P. De Antonis, and A.M.D. Ede, Nucl. Instrum. Method A 458, 558 (2001). 11. S.H. Park, J.H. Ha, Y.H. Cho, H.S. Kim, S.M. Kang, Y.K. Kim, and J.K. Kim, IEEE Trans. Nucl. Sci. 55, 1547 (2008). 12. H. Chen, J. Tong, Z. Hu, D.T. Shi, G.H. Wu, K.-T. Chen, M.A. George, W.E. Collins, A. Burger, R.B. James, C.M. Stahle, and L.M. Bartlett, J. Appl. Phys. 80, 3509 (1996). 13. A. Bensouici, V. Carcelen, J.L. Plaza, S. De Dios, N. Vijayan, J. Crocco, H. Bensalah, E. Dieguez, and M. Elaatmani, J. Cryst. Growth 312, 2098 (2010). 14. F. Aqariden, S. Tari, K. Nissanka, Jin Li, N. Kioussis, R.E. Pimpinella, and M. Dobrowolska, J. Electron. Mater. 41, 2893 (2012).