一种高效的Ge(100)表面清洁方法

Abstract  

We report an efficient method for cleaning Ge substrates similar to the Ishizaka and Shiraki method for cleaning of Si. The  Ge wafers are cleaned in running deionized water and etched with HF. A thin oxide layer was prepared by dipping in a mixture  of H₂O₂ and H₂O for a few seconds and the oxide layer was removed by dipping in HF. This procedure was repeated several  times to ensure the removal of several atomic layers of Ge, Finally the thin oxide layer prepared as described above was thermally  decomposed in ultrahigh vacuum by annealing in the temperature range 300-500°C. The resulting surface gave rise to sharp  emission peaks due to surface states in UPS illustrating a clean surface. Impurities such as carbon and oxygen were below  detection level in XPS and AES. Cross sectional TEM studies showed no defects associated with the cleaning procedure. Ge  buffer layer growth and subsequent SiGe growth showed good morphology and no substrate/buffer defects.

Cleaning of semiconductor substrates is a very  important aspect from the surface science as well as  device fabrication point of view. Within the last decade  or so Ge has emerged as a promising high performance  device material owing to its narrower band gap, high  hole mobility and high solubility limits for p-type dopants. There has been interest in growing heterostructures on Ge substrates as well. Surface  cleaning employing Ar ion sputtering has been used by  various groups. However, a major problem encountered by this procedure is the generation of defects  though it can be minimized to a great extend by  a moderate temperature anneal. Where as an easy  and low temperature method exists for Si cleaning,  cleaning of Ge substrate has been a problem in the  literature. Recently Zhang et al. reported thermal  desorption of ultraviolet-ozone oxidized Ge( 001) for  the preparation of clean substrates. In this Letter we report an efficient and relatively easy method for the  cleaning of Ge substrates.

The samples used in the present study are n-type  wafers with a conductivity of l-5 R. cm. The UPS and  XPS measurements are recorded using a VG CLAM  100 analyzer with He I and Mg Ka sources, respectively, in normal emission geometry. A VG UVL-HI  with higher photon flux was employed for UPS measurements. Ex situ oxidized samples were loaded into  the analysis chamber through a load lock chamber. The  instruments are kept in a clean room. Typical time  required to bring the samples from atmosphere to ultrahigh vacuum are 5-10 min. Typical pressure in the  analysis chamber during XPS me~mements (15 kV,  20 mA) is of the order of 6 X lo- l1 Torr.

Firstly, the wafers are washed in running deionizedwater and rinsed in HF(HF:H₂O in 9:1 ratio;although this ratio is not critical we feel a higher con-centration of HF is desirable) and then washed againin running water.It is to be noted that the wettingcharacteristics are different from that of Si wafers. Athin layer of oxide is prepared by dipping the samplein H₂0₂(H₂0₂:H₂0 in 9:1 ratio)for 10-15 s andthen washed in running water. The oxide layer is etchedoff by dipping the sample in the same solution of HFfor 5-10 s. This procedure is repeated 3-5 times. Theprocedure ensures the removal of several atomic layersof Ge. A final oxide layer is prepared the same way bydipping thc wafcr in H₂O₂ (H₂O₂: H₂O in 9 : 1 ratio)for 10-15 s. The sample is dried by blowing N₂ gas andloaded in to the UHV chamber. All the wet cleaningprocesses are carried out in the same clean room wherethe analysis equipments are kept. The sample isannealed at 300'C for 20 30 min and annealed furtherat around 500°C for 1$ min, The two step annealing isfor the necessary out-gassing and subsequent oxidelayer decomposition.

In Fig. 1 we plot the He I UPS spectra of the oxidized  Ge( 100) surface and that after the anneal. As seen, the  oxidized surface exhibits a prominent peak at around  5.2 eV resulting from 0 2p states. The binding energy  of this peak is different from that of the oxide on Si  surfaces. This difference and related effects associated  with the oxidation process of Ge surfaces will be discussed in detail in a forth coming paper. The surface, after annealing shows a prominent peak due to  the Ge surface states due the presence of Ge dimers on  the surface. This is similar to that of Si surfaces where  a surface state peak is observed at about 0.82 eV below  the valence band maximum. In the case of Si surfaces,  the presence of Si dimers and the resultant surface state  emission in UPS is considered a signature of clean  surfaces. Following similar grounds, we attribute  the existence of Ge dimer states to surface cleanliness.  The possible impurities such as carbon and oxygen  were below detection level in AES and XPS. The  (2 X 1) LEED pattern observed after the annealing was  of good quality. This demonstrates that the procedure  described above is indeed an efficient and easy method  for cleaning of Ge substrates. Furthermore, we  observed clear dimer rows in STM images of Ge( 100)  cleaned using this method. 

Fig. 1. He I UPS of (a) Ge( 100) oxidized by dipping the wafer in  a mixture of HZ02 and H,O (9 : 1) at room temperature and (b)  after annealing to 500°C showing the emergence of surface states.  The wafer was cleaned prior to preparation of oxide as described in  the text.

Cross sectional TEM measurements showed no  defects associated with the cleaning procedure. Ge buffer layer growth and subsequent SiGe growth showed  good morphology and no substrate/buffer defects. Although in cross sectional TEM we observe  only relatively small interfacial areas (perhaps of the  order of 100 pm2 in a typical sample) this observation  establishes that substrate cleaning is effective. We have  recently carried out  misfit dislocation propagation  kinetics in GexSi_1-x/Ge( 100) heterostructures where  the substrate cleaning was achieved using the method  described here.

To conclude, we describe an efficient and easy  method for cleaning of Ge substrates. Spectroscopic  and microscopic evidences suggest that this method is  indeed useful for subsequent growth studies.