Journal of Crystal Growth 217 (2000) 201}210
Improvement in wafer temperature uniformity and #ow pattern
in a lamp heated rapid thermal processor
C.P. Yin, C.C. Hsiao, T.F. Lin*
Department of Mechanical Engineering, National Chiao Tung University, Hsinchu, Taiwan, ROC Received 8 July 1999; accepted 31 March 2000
Communicated by K. Nakajima
Abstract
An experimental lamp heated, rapid thermal processor (RTP) for an 8-in single silicon wafer was designed and established to investigate the thermal and #ow characteristics in the processing chamber by transient temperature measurement and #ow visualization. Experiments were carried out to explore the e!ects of placing a highly conducting copper plate right below the wafer on the uniformity of the wafer temperature and e!ects of the showerhead on the resulting #ow distribution in the processing chamber. The measured data indicated that adding the copper plate can e!ectively reduce the nonuniformity of the wafer temperature. Besides, using a showerhead with "ner holes in it results in a better #ow distribution in the processor. 2000 Elsevier Science B.V. All rights reserved.
PACS: 81.15.Gh; 44.25.#f
Keywords: RTP processor; Temperature uniformity; Flow; Convection
1. Introduction
The recent continuing miniaturization of in-tegrated circuit (IC) components and increasing functions of a single IC chip have called for an ultra-high integration of the IC components in a single large wafer. Due to their fast ramp-up and ramp-down rates lamp heated, single wafer proces-sors are preferred in this ultra-high IC integration. However, successful fabrication of the extremely dense submicron circuits on a large wafer still faces
* Corresponding author. Tel.: #886-35712121-55118; fax: #886-35726440.
E-mail address: [email protected] (T.F. Lin).
the problems of the temperature nonuniformity on the wafer and bad #ow distribution in the vicinity of the wafer, among others.
In the design of a lamp heated, rapid thermal reactor, OGzturk et al. [1] proposed that for getting a uniform thin "lm, the thickness of the velocity, temperature, and concentration boundary layers over the wafer must be uniform, the wafer temper-ature should be uniform and the gas #ow in the reactor must be free of any laminar vortices. The higher radiative loss from the wafer edge was found to result in a radial temperature gradient in the wafer [2,3]. This temperature gradient induces a thermal stress which is compressive in central region of the wafer but is tensile towards the edge of the wafer, as noted by Sorrell et al. [4]. They further indicated that it was di$cult to produce the desired
0022-0248/00/$ - see front matter 2000 Elsevier Science B.V. All rights reserved. PII: S 0 0 2 2 - 0 2 4 8 ( 0 0 ) 0 0 4 5 0 - 4
Nomenclature
D processor diameter
QH total air volume #ow rate entering the processor
Re Reynolds number based on the processor diameter ("< D/l)
t time
¹ wafer temperature
< average velocity of the air at the inlet of the processing chamber
Greek letters
l kinematic viscosity of air
injection angle measured from the tangent to the perimeter of the mixing chamber
irradiance distribution in a rapid thermal processor (RTP) by employing only the #at re#ector. Focus-ing re#ectors can be used to improve the temper-ature uniformity. They also proposed to arrange the heating lamps into concentric heating zones. In a numerical study, Campbell et al. [5] gave a two-dimensional solution to the recirculating #ow in a RTP reactor including the convective gas #ow e!ect.
Nenyei et al. [6] found that multiple gas ba%e system, low-temperature guard ring surrounding the wafer, and independent top and bottom heater bank control were important in optimizing the gas #ow in the processing chamber. To improve the wafer temperature uniformity, a cone-shape shield was placed at the edge of the wafer to reduce the heat loss from the edge and to re#ect the radiative energy back into the wafer [7,8]. Cho and Kim [9] used a ring of silicon dioxide formed at the wafer edge to reduce the heat loss from the edge. ZoKllner et al. [10] showed that the peripheral temperature drop could be compensated by a separate focusing lamp ring. Another way to reduce the edge e!ect is to provide a supplementary heat source surround-ing wafer [11,12]. A passive guard rsurround-ing such as an independent heated ring [13] or an annual lamp [14] was noted to improve the wafer temperature uniformity. Moreover, the guard rings can reduce the temperature-gradient-induced wafer deforma-tion by a factor of 10 when compared to a free-standing wafer [15]. Recently, Lee et al. [16] used
concentric Si rings on a planar quartz or Si suscep-tor to improve the wafer temperature distribution. The uniformity of the wafer temperature may also result from the local di!erences in the radi-ation absorption and emissivity of the wafer surface induced by the pattern on the wafer [17,18]. Van-denabeele et al. [19] showed that the patterned oxide layers could cause the wafer temperature nonuniformity up to 80 K.
The above literature review clearly indicates that the #ow and thermal characteristics in a rapid thermal processor for a single wafer remain largely unexplored especially for a large wafer during the transient stage. Moreover, methods to improve the temperature uniformity of the wafer still need to be sought. In the present study an experimental rapid thermal processor for an 8-in silicon wafer is estab-lished to investigate the #ow and thermal charac-teristics associated with the processor through the #ow visualization and transient temperature measurement. Attention is focused on the possible improvement of the wafer temperature uniformity by placing a high-conductivity copper disk below the wafer and on the e!ects of the showerhead on the gas #ow distribution in the processor.
2. Experimental apparatus and procedures
The experimental apparatus established in the present study schematically shown in Fig. 1
CRYS 9564
Mamatha
BG
Giri MV
Fig. 1. Schematic diagram of the experimental apparatus.
consists of six major parts, namely, the heating lamp unit, gas injection unit and mixing chamber, processing chamber, vacuum system, temperature measurement and data acquisition system, and control unit. Note that air is used as the working #uid to replace the inert gases normally used in the real rapid thermal processing.
The lamp heating unit includes 13 410W & 82V OSRAM tungsten}halogen lamps, which are ar-ranged into two zones, an inner ring of four lamps and an outer ring of nine lamps. The input power to each zone can be adjusted for the optimal control of the wafer temperature uniformity.
The gas injection unit consists an air compressor, a #ow meter, four smoke generators, four solenoid #ow control valves, "lters, pressure regulators and injection nozzles. The installations of the smoke generators allows for the visualization of the gas #ow in the processor. Compressed air is injected into the mixing chamber through the four injection nozzles located at the side wall of the mixing
cham-ber. The air is then mixed in the mixing chamber and moves across the showerhead into the process-ing chamber. The showerhead is a thin perforated circular quartz plate of 4 mm thick. It contains 385 small circular holes of having the same diameter of 6 mm. Another showerhead with 2401 "ner holes of having the same diameter of 2 mm is also tested. Moreover, a single vertically downward air jet lead-ing into the processor through a connection pipe and directly impinging onto the wafer has been tested to access the resulting #ow pattern in the processing chamber. In this jet impinging experi-ment no gas is injected into the mixing chamber from the nozzles on its sidewall.
The processing chamber is cylindrical and has an outside diameter of 30 cm. Its side wall is made of 7 mm thick Pyrex glass. The wafer is leveled hori-zontally and "xed centrally. A copper plate of 2 mm thick is placed just below the wafer to provide an additional heat transfer path. Five type-T ther-mocouples are stuck at the back of the wafer at
selected locations and three other thermocouples of the same type are stuck on the copper plate. Note that these thermocouples are positioned at three concentric circles at an equal radial interval of 5 cm. The vacuum system consists of a rotary vacuum pump with a 50 standard liter per minute (SLPM) pumping capacity. The pumping rate can be ad-justed by a manual #ow control valve. The system pressure of the chamber is maintained at the re-quired level by adjusting the #ow rate of the vac-uum pump.
An on/o! control algorithm, which is found to perform better than linear proportional control, is used to control the input power to the lamps. The measured average wafer temperature is calculated by the computer to check whether it is over or under the set temperature. The result is then used to drive a stepping motor which, in turn, connects with a variable resistor and thus can control the required electric current from the power supply. Thus, the input power to the lamps is adjusted.
For clarity, the schematic diagram of the pro-cessing chamber is shown in Fig. 2. The test starts with the air at room temperature ¹ injected into the mixing chamber. Meanwhile, the lamps are turned on until the wafer temperature reaches to the preset level. The air streams mix in the mixing chamber and then move across the showerhead into the processing chamber. It #ows over the wafer and is "nally sucked out of the chamber through the exhaust ports. For a single jet impinging test the air is directly introduced into the processing chamber through an inlet pipe of 8 mm in inner diameter and the showerhead is replaced by a non-perforated quartz plate. After the #ow reaches steady or statistically stable state, we begin the #ow visualization and temperature measurement. The gas #ow pattern in the chamber is illuminated by a vertical plane light sheet from a laser sheet gener-ator and is photographed by a camera from the chamber side.
3. Experimental results and discussion
In this study we "xed the distance between the showerhead or single-tube outlet and the wafer surface at 10 cm. Besides, the chamber pressure is
"xed at 1 atm. The experimental results are present-ed to illustrate the e!ects of the copper plate on the uniformity of the wafer temperature and the e!ects of the showerhead on the resulting #ow distribu-tion.
3.1. Ewects of copper plate on the wafer temperature uniformity
In order to improve the temperature uniformity of the wafer, we have designed a nearly isothermal plate, which is made of high thermal conductivity copper with its upper surface coated with black paint, and placed it beneath the wafer with 1 cm apart (Fig. 2(b)). To compensate the higher energy loss from the wafer edge, another copper plate with an outer thin #ange acting as a guard ring is also tested.
The maximum temperature di!erences across the wafer deduced from the thermocouple data mea-sured at the "ve selected locations on the wafer (Fig. 2(c)) during the entire up and ramp-down processes without placing the copper plate below the wafer for the processor with a single inlet pipe for the limiting case with QH"0 are shown in Fig. 3 for three preset temperature levels for the wafer. Initially at t(0, the whole apparatus is at room temperature ¹ . The time t"0 denotes the instant at which the lamps are simultaneously turned on. The results indicate that the ramp-up rate of the wafer temperature is rather slow at about 2}33C/s. It takes about 50 s for the wafer to reach the preset temperature and then the wafer stays at the preset temperature with certain level of #uctuation. At t"1200 s the lamps are turned o! and the wafer temperature declines gradually. The results show that the maximum temperature di!er-ences at the end of the heating cycle (*¹ ) are 4.8, 8.2 and 10.53C respectively corresponding to the wafer set at 100, 150 and 2003C. When the copper plate without #ange is placed under the wafer with the other parameters "xed at the previous values, the ramp-up rate of the wafer temperature is slight-ly lower. But the maximum temperature di!erences across the wafer reduce respectively to 4.0, 4.8, and 7.23C. For air injected into the processor at con-stant #ow rates of 1}5 SLPM, the results for ¹ set at 1003C suggested that the maximum temperature
CRYS 9564
Mamatha
BG
Giri MV
Fig. 2. Schematic diagrams of the processor (a), copper plate without and with #ange (b) and location on the detection points the wafer (c).
Fig. 3. The wafer temperature and the maximum temperature di!erence across the wafer during the entire heating process for QH"0 without copper plate for the wafer setting temperature of (a) 100, (b) 150 and (c) 2003C.
CRYS 9564
Mamatha
BG
Giri MV
Fig. 4. The wafer and copper plate temperatures at selected points and the maximum temperature di!erence across the wafer during the entire heating process for ¹"2003C for (a) QH"0SLPM, (b) QH"1SLPM, (c) QH"2SLPM, (d) QH"3SLPM and (e) QH"5SLPM.
di!erence is larger for a higher #ow rate and is higher than that for QH"0. Note that at higher wafer setting temperature for ¹"150 and 2003C the maximum temperature di!erences for QH"1}5
SLPM are also all above that for QH"0 except for
QH"1 SLPM at ¹"2003C (Fig. 4). It is noted
from these data that the higher #ow rate produces a larger*¹ . The larger *¹ for a higher gas
Table 1
Maximum temperature di!erences across the wafer at the end of the heating process for wafer set at di!erent temperatures and di!erent inlet gas #ow rates
QH (SLPM) ¹(3C)
100 150 200
(a) Copper plate without yange
0 *¹"4.0 *¹"4.8 *¹"7.2 1 *¹"4.1 *¹"6.0 *¹"7.0 2 *¹"5.0 *¹"8.1 *¹"10.0 3 *¹"6.1 *¹"9.3 *¹"12.0 5 *¹"7.9 *¹"10.9 *¹"15.0 (b) Copper plate with yange
0 *¹"3.2 *¹"4.3 *¹"4.9 1 *¹"2.0 *¹"2.0 *¹"2.8 2 *¹"2.4 *¹"3.2 *¹"5.2 3 *¹"2.6 *¹"3.6 *¹"6.0 5 *¹"4.0 *¹"3.9 *¹ "7.5
#ow rate is considered to result from the larger cooling e!ect associated with a higher QH for the injected gas impinging on the wafer. This is evident by observing that the temperature at center of the wafer is lowest for each setting wafer temperature. The measured data for the copper plate with a #ange indicated that the wafer temperature nonuniformity was reduced signi"cantly by the ad-dition of the radiation shield. For clear comparison the e!ects of the copper plates on*¹ at di!erent gas #ow rates are demonstrated in Table 1. Note that the addition of the radiation shield is more e!ective at higher QH and ¹. These results suggest that the high thermal conductivity of the copper and the addition of the radiation shield do result in better uniformity of the wafer temperature. It should be pointed out that the present method for temperature uniformity improvement is expected to be only slightly a!ected by the patterns on the wafer. Finally, it should be noted from Figs. 3 and 4 that during the ramp down process the wafer temperature is rather close to the copper plate temperature and the uniformity in the wafer tem-perature becomes even better.
It is of interest to compare the present results with those in the literature. For a 4-in wafer heated
from the front and back sides by two banks of lamps, Gyurcsik et al. [20] found that the measured *¹ is slightly above 153C. With an annual guard ring, Load [3] noted that the*¹ was as high as 303C for a 4-in wafer. Through a better multizone lamp control *¹ was reduced to about 12.53C for a 6-in wafer by Moslehi et al. [21]. By using backside heating and a re#ective shower-head, Hebb et al. [18] were able to reduce*¹ to slightly below 103C for an 8-in patterned wafer. Our measured *¹ , when compared with the above data from the literature [3,18,20,21], indi-cates that the method proposed in the present study to improve the wafer temperature uniformity is rather competitive.
3.2. Inyuences of showerheads on yow in processing chamber
In single-wafer rapid thermal processors, various showerheads are normally used to improve the gas #ow distribution over the wafer. In the present study, two di!erent showerheads for "ve di!erent angles of gas injection,"03, 303, 453, 603 and 903, at the injection nozzles on the side wall of the mixing chamber were tested here. Fig. 5 shows the #ow photos taken at steady state for an unheated wafer in the processor for di!erent gas #ow rates with the showerhead having 385 holes and gas injected radially ("903) into the mixing chamber. The result in Fig. 5(a) indicates that for QH" 4 SLPM the #ow moving from the mixing chamber into the processing chamber is dominated by a weak distorted jet around the vertical axis of the processing chamber. Outside this jet the #ow across the showerhead is rather weak. As the jet ap-proaches the wafer, it becomes highly distorted. After hitting the wafer, it spreads somewhat irregu-larly over the wafer. At higher #ow rates of 6 and 8 SLPM several jets form when the gas #ow crosses the showerhead. These jets deform to a larger de-gree and occupy a larger space (Figs. 5(b) and (c)). Weak recirculations are seen outside the jets. It is noted that as the #ow rate is raised to 12 SLPM, the #ow entering the processing chamber is almost evenly distributed (Fig. 5(d)). However, weak #ow recirculations still exist. When the gas is injected tangentially into the mixing chamber ("03), the
CRYS 9564
Mamatha
BG
Giri MV
Fig. 5. Photographs of the #ow pattern in the processing cham-ber with showerhead having 385 holes for "903 for (a) QH"4 SLPM (Re"18.6), (b) QH"6 SLPM (Re"28.0) and (c) QH"8SLPM (Re"37.3) and (d) QH"12 SLPM (Re"55.9).
Fig. 6. Photographs of the #ow pattern in the processing cham-ber with showerhead having 2401 holes for "903 for (a) QH"6 SLPM (Re"28.0), (b) QH"8 SLPM (Re"37.3) and (c) QH"12 SLPM (Re"55.9).
resulting gas #ow is found to be mainly con"ned in the region near the side wall of the processing chamber. Hence, over the wafer the #ow is rather nonuniform. For the injection angle between 03 and 903, the resulting #ow pattern is also between those discussed above for 03 and 903.
When the showerhead contains a much larger number of 2401 smaller holes, the resulting #ow in the processing chamber for "903 shown in Fig. 6 is no longer like jets. But #ow recirculations form near the side wall of the chamber and the main #ow is still restricted to the small zone around the axis of the chamber for QH"6 SLPM (Fig. 6(a)). Note that at a higher QH of 8SLPM the tion is smaller and at QH"12 SLPM the recircula-tion is rather weak (Figs. 6(b) and (c)). Thus, at
QH"12 SLPM the #ow is nearly uniform as it
moves vertically downwards. This suggests that a showerhead with rather small holes is needed to obtain a nearly uniform #ow. It is interesting to note that more resistance is experienced as the #ow crosses the showerhead containing small holes and hence the #ow recirculates more intensely in the
mixing chamber and lasts in a longer period of time. Thus, the downward #ow across the shower-head is more uniform.
4. Concluding remarks
An experimental lamp heated, rapid thermal pro-cessor for processing a single silicon wafer was established here to investigate the #ow in the pro-cessing chamber. A high thermal conductivity cop-per plate was placed beneath the wafer to improve the uniformity of the wafer temperature. Moreover, the use of showerhead to improve the #ow distribu-tion in the processor was examined. The major results can be brie#y summarized in the following. (1) The addition of a high thermal conductivity copper plate beneath the wafer can greatly improve the uniformity of the wafer temper-ature especially at the ramp-down process. (2) A more uniform #ow "eld can be obtained for
the showerhead with small holes at certain inlet #ow rate.
Acknowledgements
The "nancial support of this study by the engin-eering division of National Science Council of Taiwan, ROC through the contract NSC 87-2218-E-009-006 is greatly appreciated.
References
[1] M.C. OGzturk, F.Y. Sorrell, J.J. Wortman, F.S. Johnson, D.T. Grider, IEEE Trans. Semicond. Manuf. 4 (1991) 155.
[2] R. Detaton, H.Z. Massoud, J. Appl. Phys. 70 (1991) 3588.
[3] H.A. Lord, IEEE Trans. Semicond. Manuf. 1 (1988) 105.
[4] F.Y. Sorrell, M.J. Fordham, M.C. OGzturk, J.J. Wortman, IEEE Trans. Electron Devices 36 (1992) 75.
[5] S.A. Campbell, K.L. Kanutson, B.Y.H. Liu, J.D. Leighton, IEEE Trans. Semicond. Manuf. 4 (1991) 14.
[6] E. Nenyei, H. Sommer, J. Gelpey, A. Bauer, Mat. Res. Soc. Symp. Proc. 342 (1994) 401.
[7] P.J. Timans, R.N. Morishige, Y. Wasserman, Mat. Res. Soc. Symp. Proc. 470 (1997) 57.
[8] G.L. Young, K.A. McDonald, IEEE Trans. Semicond. Manuf. 3 (1990) 176.
[9] B.J. Cho, C.K. Kim, J. Appl. Phys. 67 (1990) 7583. [10] J.P. Zollner, I. Patzschke, V. Pietzuch, J. Pezoldt,
G. Eichhorn, Mat. Res. Soc. Symp. Proc. 303 (1993) 177.
[11] R. Kakoschke, E. Bubmann, Mat. Res. Soc. Symp. Proc. 146 (1989) 473.
[12] T.J. Riley, R.S. Gyurcsik, Mat. Res. Soc. Symp. Proc. 303 (1993) 223.
[13] R. Kakoshke, E. Bubmann, H. Foll, Appl. Phys. A 50 (1990) 141.
[14] Y.M. Cho, A. Paulraj, T. Kailath, G.H. Xu, IEEE Trans. Semicond. Manuf. 7 (1991) 34.
[15] J.F. Gongste, T.G.M. Oosterlaken, G.C.J. Bart, G.C.A.M. Janssen, S. Radelaar, J. Appl. Phys. 75 (1994) 2830. [16] K.C. Lee, H.Y. Chang, J.G. Hu, Mat. Res. Soc. Symp. Proc.
470 (1997) 187.
[17] P. Vandenabeele, K. Maex, Mat. Res. Soc. Symp. Proc. 224 (1991) 185.
[18] J.P. Hebb, K.F. Jensen, E.W. Egan, Mat. Res. Soc. Symp. Proc. 387 (1995) 21.
[19] P. Vandenabeele, K. Maex, R. De Keersmaecker, Mat. Res. Soc. Symp. Proc. 146 (1989) 149.
[20] R.S. Gyurcsik, T.J. Riley, F.Y. Sorrell, IEEE Trans. Semicond. Manuf. 4 (1991) 9.
[21] M.M. Moslehi, J. Kuehne, R. Yeakley, L. Velo, H. Najm, B. Dostalik, D. Yin, C.J. Davis, Mat. Res. Soc. Symp. Proc. 224 (1991) 143.
