zyx zyxwvu zyxw GAS-PHASE ETCHING OF SACRIFICIAL OXIDES USING ANHYDROUS HF AND CHSOH Jong Hyun Lee, Yong II Lee, Won Ick Jang, Chang Seung Lee, and Hyung Joun Yo0 zyxwvuts zyxw zy SemiconductorTechnology Division, Electronics and Telecommunications Research Institute Yusong P.O. Box 106, Taejon 3O5-600, Korea ABSTRACT strength, is necessary to evaluate the release technologies. Pull-in-length can be defined as the beam length of superstructures, beyond which beams are forced into contact with underlying substrate during drying step. On the contrary, the detachment length 4 as shown in Fig. 1, is the beam length, beyond which part remains stuck to the substrate after drying step. One of the major issues in surface micromachining is process-induced failures of overhanged microstructures, after sacrificial removal. This failure process consists of temporary deformation due to capillary force and permanent stiction of the deformed microstructures to substrate due to the residue product. In order to alleviate this failure, some researchers have investigated into the use of low surface-tension liquids, temporary support, sublimation of the final liquid, or supercritical method. These conventional methods, however, are complicated, liable to plasma damage during PR ashing, or require carefbl handling of the samples in rinse liquid. In this paper, we present a newly developed anhydrous HF gas-phase etching (GPE) technology for the removal of sacrificial TEOS (tetraethylorthosilicate)oxide. In order to minimize the capillary force of gas-liquid interface and residue product, methanol of low vapor pressure was employed as a catalyst instead of water vapor. The proposed process features simplicity, virtually no capillary forces and even compatibility with IC process as well. The effectiveness of HF GPE with methanol was verified by successfully fabricating the polysilicon cantilevers up to 1000 um in length with no stiction. The etch rate of HF GPE was 10-15 um/hr for the sacrificial TEOS oxide of 0.1-2 um channel height. k d substrate Fig. 1. Detachment length of a cantilever. From this relation of potential energy versus separation, which has direct relation with pull-in-length, the capillary force is found to be the most signtftcant at the usual dimension of meters in MEMS [2]. We will propose the newly developed HF (hydrogen fluoride) gas-phase etching to reduce the capillary force for longer pull-in-length, and use the detachment length d to evaluate the effectiveness of the proposed method in the fabrication of microstructures. The release technologies have been extensively studied by many researchers. There have been three kind of methods to alleviate this process-induced stiction problem. The first one is just to use the rinse liquid of smaller surface tension, which has linear relation with capillary force [2][3]. However, this could help to reduce the capillary force only to the limited range due to the lack of selection. Another method to alleviate the process-induced stiction is the structure modification or micromechanical temporary support, such as dimples, tethers [4], enhancement of surface roughness [ 5 ] , or photo resist columns [6]. These methods features simplicity, however, also have disadvantages of low productivity, liability of plasma damage, or require additional mask process. The last INTRODUCTION The more sigruficant failure occurs for the higher compliant microstructures which are prerequisite to the fabrication of microsensors with higher sensitivity. The failure process consists of two steps in the conventional wet etching of sacrificial layers. One first one is temporary deformation due to the capillary pulling force between the superstructures and substrate during drylng the rinse liquid after sacrificial etching [1][2]. Next one is the permanent stiction of the deformed superstructures to the substrate due to residue product, which would be generated during sacrificial etching. At this moment, the definition of stiction zyxwvutsrq 0-7803-3744-1/97/$5.000 1997 IEEE 448 zyxwvuts zyxwvu zyxwvutsrqpo zyxwvu zyxwvut zyx zyxwv one is the dry release of sublimation or solidification dry [7][8], supercritical method [9], and HF vapor etch by reducing the gas-liquid interface effect of surface tension [lO][ll]. In the Fig. 2 of phase diagram, we can find there are three ways of alleviating the dotted line of gas-liquid interface, something like a border line of no crossing. Here, the adsorbed HF; is the most significant reactant in the gas-phase etching of sacrificial oxide just as in the liquid-phase etching. Solid Another intermediate species of H2SiF6 can be generated during oxide etching as eq. (3). Critical 6HF(ads) + Si02(s)-+ H2SiF6(ads)+ 2Hzqads) (3) Point At the completion of oxide etching, this adsorbed H2siF6 should be decomposed as HF and SiF4gas, as in eq. (4), and desorbed from the substrate so that it does not leave any residue [151. TeFig. 2. Phase diagram of dry release processes. The first one, flow A going around the fence counterclockwise, replaces rinse liquid with the material of low temperature melting point, such as t-butyl alcohol or p dichlorobenzene, and successively sublimates it at low pressure [rJl[8]. The next one, flow B going around the fence clockwise, also has to replace the rinse liquid with the material of relatively low pressure supercritical state, such as carbon dioxide, and successively evaporate it in vacuum state [9]. These conventional sublimation or supercritical methods show excellent results but still requires complicated high pressure apparatus or carelid handling of the samples. Tbs H2siF6, however, can be decomposed as HF and H2Si4 (silicic acid), as in eq. (9, which is not volatile and tends to leave residue product. zyxwvu From eq. (3,we can figure out that for the release process of less stiction problem, lower water concentration and higher HF pressure are preferable. The severe condensation can OCCUT in HF etch process containing water vapor and may result in shorter detachment length. In order for us to demonstrate the 111 potential of the method, therefore, we intentionally removed water vapor from the etching system by using anhydrous HF and methanol of 1 order higher vapor pressure than that of water, as shown in Fig. 3 [161. The last one, C which we propose to minimize the gasliquid interface effects by crossing over the fence, is HF gas-phase etching (GPE),as a dry release process for sacrificial oxide etching [12]. This process employs anhydrous HF and methanol of low vapor pfessure, instead of water, for the minimization of capillary force and residue product. The proposed process features simplicity, virtually no capillary forces, negligible residue and even compatibility to the future cluster technology, as well. Some researchers have studied on the characteristics of HF gas-phase etching with methanol for native oxide removal in cleaning technology [13][14]. However, these have very different process windows from that of micromachining with respect of condensation, etch rate, and selectivity. HF GAS-PHASE ETCHING D Fig. 3. Vapor pressure of HF, methanol, and water versus temperature. The chemical reaction mechanism of HF etching is shown in eq. (1) and (2). 449 zyxwvutsrq zyxwvutsr zyxwvuts zyxwv zyxw zyxwvuts This methanol would support HF gas-phase etching with smaller water concentration or condensation. From eq (4), we can notice that the smaller water concentration will promotes the generated H2SiF6 to decompose HF gas and SiF4 gas, which would not leave the residue. Fig 4 and Fig. 5 represent the schematic and photograph of the gas-phase etch system using anhydrous HF and methanol. The etch chamber is made of aluminum and coated by Teflon film for anticorrosion. Anhydrous HF and methanol are delivered to the reaction chamber through a shower head for uniform distribution. Those parts of thick lines are heated to prevent condensation. And the flow rates of the anhydrous HF are controlled by this mass flow controller. Flow rates of methanol, controlled by regulated nitrogen carrier gas, are measured by quadruple mass spectrometer. This throttle valve system controls the process pressures in conjunction with a capacitance pressure gauge. Fig. 5 is a photograph of the developed HF phase etch system, whch WW and S / W are very compatible to cluster standard. FABRICATION SEQUENCE Various polysilicon (polycrystalline silicon) microstructures are fabricated to v e e the effectiveness of this HF phase etching. The fabrication sequence of surface micromachined actuator is similar to the typical plysilicon micromachining process, as shown in the figure 6. oxide/nitride /poly-Si Poly-Si =Fig. 4. A schematic of the GPE (gas-phase etch) system using anhydrous HF gas and CH30Hvapor zyx zy Fig. 6. Process sequence for the fabrication of micro superstructures. (a) LPCVD Si@ & Si3N4,(b) LPCVD TEOS / poly-Si / TEOS, (c) RIE TEOS / poly-Si & doping / annealing, (d) HF GPE TEOS After thermal oxidation (0.3 um), LPCVD nitride (0.2 um)and polysilicon (0.2 um)on (100) Si wafer, TEOS (tetraethylorthosilicate) oxide (0.2-2 um) is uniformly deposited as a sacrificial layer. And LPCVD plysilicon (2-5.3 um) and second TEOS (0.8 um) oxides are deposited as a structural and masking layer, respectively. The structure pattern is transferred to TEOS oxide from the mask, and then, to polysilicon by successive RIE (reactive ion etching) process, as shown in figure 7. During the annealing of doped polysilicon for dopant activation, the optimum process condition is found for the minimum residual stress. Fig. 5. A photograph of the developed GPE (gas-phase etch) system using anhydrous €Egas and CH30H. 450 Finally, the sacrificial TEOS is removed by the HF phase etching described above, and the microstructures are released free of stiction. It should be noted that the current process does not employ any structure m a c a t i o n , such as anti-stiction dimples, which are very common for typical surface micromachining processes. zyxwvu zy zyx zyxw EXPERIMENTS AND DISCUSSION Fig 8 and Fig. 9 show the overview of the fabricated polysilicon cantilevers and cross-sectional view for measuring the etch rate in a microchannel, respectively. The thickness of polysilicon is 2 um, width is 10 um, maximum length is 1000 um, and the gap between polysilicon and substrate is 2um. We successfully fabricated cantilevers up to 1000 um in length, the longest one on our mask design, with no stiction, as shown in the Fig. 8. The sacrifkial TEOS oxide was confirmed to be perfectly etched out by using photo thermal radiometry (PTR), which features high resolution and non-destructive testing [ 171. Fig. 7. SEM photograph of the fabricated electrostatic comb structures. Fig. 9. Cross-sectional view of the polysilicon and TEOS sacrificial layer being etched by GPE. Fig, 10 illustrates the detachment lengths as a function of geometric parameter of thickness t and gap h. Compared with conventional wet release for the hydrophobic substrates [2], HF GPE process has an advantage of more than 8 times longer detachment length. This characteristics would considerably improve the productivity of highly compliant microstructures. The etch rate of TEOS oxide is mainly dependent on the partial pressures of HF and methanol, which dominates the ionization reaction between the adsorbed HF and methanol. The etching experiments for bulk TEOS oxide confrmed that the etch rate becomes higher for higher the HF partial pressure, and ratio of methanol and HF partial pressure, as shown in Fig. 1 1. Fig. 8. SEM photograph of the fabricated polysilicon cantilevers with no stiction. (thickness 2jun, width IOjun, maximum length 1000 jun, gap between polysilicon and substrate 2pm ) 45 1 zyxw zyxwvutsrqp zyxwvutsr zyxwvutsr sufficiently lower than the saturation vapor pressures of HF (916.5 torr) and methanol (121.5 torr) at 25 "C. detachment length (um) 1 I 1200 - Other experiments were carried out to characterize the HF GPE for other oxides and analyze the compositions of residue product. And we have found that some residue product in the HF GPE for the TEOS oxide on nitride, and PSG (phosphosilicate glass). The successful sa&cial layer for the HF GPE, developed so far, is only TEOS oxide on crystalline silicon or polysilicon substrate. .:.' (b) 1000 - zyxwvutsrqp 1 !+i 600 800 400 - 2 0 0 t . n ' v 0.5 , 141 (a) , 3/2(Et3h2/y)'/4 , , 1.0 1.5 , 1 I 2.0 2.5 (t3h2)lr4[um]'" Fig. 10. Detachment length of polysilicon on hydrophobic substrate (a) conventional wet release, (b) HF GPE (gas-phase etch). (E : Young's modulus, t : thickness, h : gap, y : surface tension) zyxwvutsrqponm zyxwvutsrq zyxwvu IO2 f L -I =lo1 E Ea Y d 2 .e 0 5 O' sbo ' d o 0 ' l i 0 0 ' 2d00 Etching time (sec) r IO0 ' (a> PMI PHF= 0.66 B PMI PHF= 0.17 PMI PHF= 0.04 0 10-1 IO" 0 2 4 6 8 1 0 1 2 1 4 HF partial pressure (torr) Fig. 11 . The etch rate of GPE for the bulk TEOS oxide. Although we expected that the etch rate for s a d c i a l TEOS oxide might depend on the height of microchannel, not so much Werence could be noticed for the range from 0.1 um to 2 um,as shown in Fig. 12. We believe that there is no severe depletion region of HF etchant in the microchannel at this process condition. 0.0 0.5 1.0 1.5 2.0 TEOS thickness (um) (b) The experimental etch rate of the present HF gas-phase etching is estimated as 10-15 um/hr for the process condition of HF part~alpressure 15 torr, and methanol partial pressure 4.5 torr. These conditions are Fig. 12. Etch rate of HF GPE process in micro channel. (a) according to etching time, and (b) TEOS thickness. 452 zyxwvutsr zyxwvut CONCLUSION We employed newly developed HF gas-phase etch process for the removal of sacrificial TEOS oxide. The proposed process features simplicity, virtually no capillary forces and even compatibility with IC process as well. The effectiveness of HF phase etching with methanol was verified by successllly fabricating the polysilicon cantilevers up to 1000 um in length with no stiction. The etch rate of HF GPE was 10-15 um/hr for the sacrificial TEOS oxide of 0.1-2 um channel height. Further characterizations are under investigation on the process latitude and dynamic properties of fabricated microactuators. ACKNOWLEDGMENTS We would like to thank Dr.H. H. Chung at KIFO, Dr. S . Y. Kang and Dr. J. T. Baek, at ETRI, and C. S . Lee at Genie Tech, for their technical contributions. This research work was supported by the Ministry of Information and Communications, Korea. [7lH. Guckel, J. J. Sniegowski, and T.R. Christenson, 1989, “Fabrication of Micromechanical Devices from Polysilicon Films with Smooth Surfaces”, Sensors and Actuators, Vol. 20, Nos. 1&2, pp. 117122. [8] D. Kobayashi, T. Hirano, T. Furuhata, and H. Fujita, “An Integrated Lateral Tunneling Unit,” E E E Micro Electro Mechanical Systems, Travemiinde, Germany, Feb. 1992, pp. 214-219. [9] G. T. Mulhern, D. S . Soane, and R T. Howe, “Supercritical Carbon Dioxide Drying of Microstructures,” Int. Cod. on Solid-state Sensors and Actuators (Transducers ‘93), Yokohama, Japan, June 1993, pp. 296-299. [lo] M. menberg, F. Larmer, B. Elsner, H. Munzel, and W. Riethmuller, “Novel Process for a Monolithic Integrated Accelerometer,” Int. Conf. on Solid-state Sensors and Actuators (Transducers 9), Stockholm, Sweden, June 1995, pp. 589-592. [ l l ] T. Yoshida, T. Kudo, S . Kato, S. Miyazaki, S . Kiyono, and K. Ikeda, “Strain Sensitive Resonant Gate Transistor,” IEEE Micro Electro Mechanical Systems, Feb. 1995, pp. 317-321. [12] J. H. Lee, H. H. Chung, S. Y. Kang, J. T. Baek, and H. J. Yoo, “Fabrication of Surface Micromachined Polysilicon Actuators Using Dry Release Process of HF Gas-Phase Etching,” IEDM 96, San Francisco,U.S.A, December 1996,pp. 761-764. [13] T. Yoshida, T. Kudo, S . Kato, S . Miyazaki, S . Kiyono, and K. Ikeda, “Strain Sensitive Resonant Gate Transistor,” E E E Micro Electro Mechanical Systems, Feb. 1995, pp. 317-321. [13] K. Torek, J. Ruzyllo, R. Grant, and R. Novak, 1995, “Reduced Pressure Etching of Thermal Oxides in Anhydrous HF/Alcoholic Gas Mixtures,” J. Electrochem. Soc., Vol. 142, No. 4, pp.13221326. [14] Chun Su Lee, Seong IN Woo, Jong Tae Baek, and Hyung Joun Yoo, 1996, “Modeling and Characterization of Gas-Phase Etching of Thermal Oxide and TEOS Using Anhydrous HF Gas and CH3OH Vapor,” J. Electrochem. Soc., Vol. 143, No. 3, p ~1099-1103. . [15] W. Kern, Handbook of Semiconductor Wafer Cleaning Technology, 1993, Noyes Publications, chapter 7. [161 Handbook of Chemistry and Physics, 1990, Edited by D.R Lide, CRC Press. [17l D. H. Kim, and U. Netzelmann, 1993, “Nondestructive Testing of Ceramic Coatings on Steel by Photoacoustic, Photothermal and High Frequency Ultrasound Techniques,” Nondestr. Test. Eval., 1993, vol. 10, pp. 343-350. zyxwvut REFERENCES [l] R Legtenberg, J. Elders, and M. Elwempoek, “Stiction of Surface Micromachined Structures after Rinsing and Drymg: Model and Investigation of Adhesion Mechanisms,” Int. Conf. on Solid-state Sensors and Actuators (Transducers 93), Yokohama, Japan, June 1993, pp. 198-201. [2] C. H. Mastrangelo, and C. H. Hsu, 1993, “Mechanical Stability and Adhesion of Microstructures Under Capillary Forces Part I : Basic Theory,” J. MEMS ,vol. 2, no. 1, pp. 33-43. [3] M. Ohtsu, K. Minami, M. Esashi, “Fabrication of Packaged Thin Beam Structures by an Improved Drying M e t h w IEEE Micro Electro Mechanical Systems, Feb. 1996, pp. 228-233. [4] C.-J. Kun, A. P. Pisano, R. S . Muller, and M. G. Lim, “Polysilicon Microgripper”, IEEE Solid-state Sensor and Actuator Workshop, Hilton Head Island, South Carolina, U.S.A., June 1990, pp. 48-5 1. [5] Y. Yee, K. Chun, and J. D. Lee, “Polysilicon Surface Modification Technique to Reduce Sticking of Microstructures,” Int. Cod. on Solid-state Sensors and Actuators (Transducers 9), Stockholm, Sweden, June 1995, pp. 206-209. [6] D. Kobayashi, C.-J. Kim, and H. Fujita, 1993, “Photoresist-Assisted Release of Movable Microstructures,” Japanese Journal of Applied Physics, Vol. 32, NO. 114 p ~L1642-L1644. . zyxwvuts - zyxwvu zyxwvu 453