rcap2

Embed Size (px)

Citation preview

  • 8/2/2019 rcap2

    1/3

    Electrochemical and Solid-State Letters, 3 (1) 47-49 (2000) 47S1099-0062(99)07-069-8 CCC: $7.00 The Electrochemical Society, Inc.

    Etching of Silicon by the RCA Standard Clean 1G. K. Celler,*,z D. L. Barr, and J. M. Rosamilia

    Bell Laboratories, Lucent Technologies, Murray Hill, New Jersey 07974, USA

    It is known that the RCA Standard Clean 1, which is used repeatedly during device fabrication, can cause etching of Si. In some

    process flows, such etching can be important when fabricating devices with thin films, for example, in silicon-on-insulator tech-nology. We show that 25-30 of Si is etched away by a modified version of the SC1 clean (1:8:64 parts by weight of NH 4OH,H2O2, and H2O), when it is applied for 10 min to a bare Si layer on top of SiO2. 1999 The Electrochemical Society. S1099-0062(99)07-069-8. All rights reserved.

    Manuscript submitted July 13, 1999; revised manuscript received September 24, 1999. Available electronically November 5, 1999.

    Fabrication of integrated circuits on silicon wafers involves hun-dreds of individual process steps. In the early phase of processing,before dopants are introduced into Si, there is usually little concernabout loss of Si from the surface of the wafers, either by means ofoxidation or etching. The wafer thickness, typically >600 m, is forall practical purposes almost infinite in comparison to the 1000 , any discrepancies betweenthe predicted and measured valued were under 2%, within the rangeof the experimental accuracy for a three-layer structure. (For a mul-tilayer structure, the experimental uncertainty is greater than for asingle film. Not only the optical parameters, n and k, of all the filmsneed to be well known, but also the experimental fit assumes per-fectly abrupt and smooth interfaces.) Therefore, for thicker films itis difficult to measure small losses of Si thickness caused by etching.

    When Si films with a mean thickness of 120 were cleaned andoxidized in dry oxygen at 850C to obtain 100 of pad oxide, itbecame apparent that a drastic loss of Si had occurred. The differ-ence between the mean values of Si before and after 100 oxideformation is shown in Table I. Point by point comparison of the Sifilm thickness, tsi, on one typical wafer is shown in Fig. 1. In bothcases, it is clear that instead of the expected removal of46 of Si,70-75 were lost.

    Since the thickness of oxide grown on the monitor wafer was pre-cisely 100.3 0.5 , there were only two plausible explanations forthe loss of Si

    Downloaded 09 Apr 2012 to 142.58.129.109. Redistribution subject to ECS license or copyright; see http://www.ecsdl.org/terms_use.jsp

  • 8/2/2019 rcap2

    2/3

    48 Electrochemical and Solid-State Letters, 3 (1) 47-49 (2000)S1099-0062(99)07-069-8 CCC: $7.00 The Electrochemical Society, Inc.

    1. Oxidation rate of SOI wafers was significantly higher than thatof bulk Si, either because of defects or because of stresses in the thinSi film;

    2. The SC1 cleaning process consumed 25-30 of Si.Since Si interstitials are injected into Si during thermal oxidation,

    we have always watched for any indication that the buried oxide,which can be either a barrier or a sink for interstitials, affects the oxi-dation rate through its impact on the concentration of interstitials. Ina series of oxidations, with oxides up to 3000 grown on SOIwafers, there was never any measurable difference between bulk Simonitor wafers and SOI wafers. The growth of 100 of SiO2 on 100 of Si also resulted in a correct thickness of SiO2 but a 25-30 shortage of Si. If enhanced oxidation were responsible for the loss ofSi, we would have obtained close to 160 of SiO2.

    To check the effect of the SC1 clean on the film thickness, an SOIwafer with a very thin Si film, 20-30 , was subjected to the stan-dard clean, and the thickness was measured at the same sites imme-diately (5 min) before and within 1 h after the clean. Since theprocess ended with removal of chemical oxide in dilute HF, the Sisurface was hydrophobic and hydrogen terminated, so that the nativeoxide would be

  • 8/2/2019 rcap2

    3/3

    Electrochemical and Solid-State Letters, 3 (1) 47-49 (2000) 49S1099-0062(99)07-069-8 CCC: $7.00 The Electrochemical Society, Inc.

    root-mean-square (rms) before, and no more than 1.2 rms after theSC1. However, surface roughness is not a measure of lost silicon, butan indicator of how conformal the etching is. For conformal etchingthe roughness is not effected.

    ConclusionsSC1 clean procedure removes a finite amount of silicon. Our

    recipe (1:8:64 parts by weight of NH4OH, H2O2, and H2O) etches

    25-30 of Si in 10 min, as compared to 80 for the standard recipe.But even this relatively small loss of Si must be taken into accountin some device structures, and in particular when using SOI wafers,in which the final thickness of Si must be precisely and reproduciblycontrolled. The emergence of SOI technology increases the impor-tance of exploring novel clean chemistries. One advantage of theetching properties of the SC1 chemistry is that it may be intention-ally used for fine tuning the thickness of Si films in SOI applications.

    Lucent Technologies assisted in meeting the publication costs of thisarticle.

    References1. W. Kern,J. Electrochem. Soc., 137, 1887 (1990).2. G. S. Higashi and Y. J. Chabal, inHandbook of Semiconductor Wafer Cleaning

    Technology: Science, Technology, and Applications, W. Kern, Editor, p. 433,Noyes Publishers, Park Ridge, NJ (1993).

    3. J. M. Rosamilia, T. Boone, J. Sapjeta, L. Psota-Kelty, K. Hanson, and G. Higashi,in Proceedings of the Third International Symposium on Ultra Clean Processingof Silicon Surfaces, Acco, Leuven, Belgium (1996).

    4. T. Ohmi, M. Miyashita, and T. Imaoka, in Proceedings of the Microcontamination91 Meeting, p. 491, Canon Communications (1991).

    5. M. Bruel,Electron. Lett., 31, 1201 (1995).6. S. Verhaverbeke and J. W. Parker, in Cleaning Technology in Semiconductor

    Device Manufacturing V, J. Ruzyllo and R. E. Novak, Editors, PV 97-35, p. 184,The Electrochemical Society Proceedings Series, Pennington, NJ (1997).

    7. S. Verhaverbeke, J. W. Parker, and C. F. McConnell,Mater. Res. Soc. Symp. Proc.,477, 47 (1997).

    8. P. Boelen, T. Lardin, B. Sandrier, R. Matthews, I. Kashkoush, R. Novak, andF. Tardif,Mater. Res. Soc. Symp. Proc., 477, 161 (1997).

    Figure 3. Reduction in BOX thickness caused by the SC1 clean followed by90 s immersion in 15:1 H2O:HF.

    Downloaded 09 Apr 2012 to 142 58 129 109 Redistribution subject to ECS license or copyright; see http://www ecsdl org/terms use jsp