REFERENCE: Starov, V. and Lane, L.R., "Use of Polysilicon Deposition in a Cold-Wall LPCVD Reactor To Determine Wafer Temperature Uniformity", Semiconductor Fabrication:
TechnoloaY and Metrology. ASTM STP 990. Dinesh C.
Gupta, editor, American Society of Testing and Materials, 1989.
ABSTRACT: In the Varian 5101 LPCVD reactor operating in a Rapid Thermal Processing (RTF) mode the wafer is heated directly by a radiant heater. The wafer temperature in this case is measured by a pyrometer pointing at its center.
To obtain a temperature map of the entire wafer, we have developed and characterized a polysilicon deposition process. Using the polysilicon thickness as a "chemical thermometer" [1], we have been able to improve the radiant heater design to obtain a temperature uniformity of 0.2%
across a 100 mm wafer and 0.5% on wafers up to 150 mm. As an example of application of our study, the optimized design was used to deposit selective tungsten with thickness uniformity of 4% across 140 mm scan on 150 mm wafer.
KEYWORDS: chemical thermometer, chemical vapor deposition, activation energy, polysilicon growth One of the most important problems associated with cold-wall Chemical Vapor Deposition (CVD) reactors is achieving good coating uniformity. Large temperature differences between the heated wafers and the rest of the reactor create generally non-isothermal environment. Since
Dr. Starov is President of Applied Photonics Research Inc., 421 Grand Street, Redwood City, CA 94062; Mr. Lane is Senior Process Technician at Varian Associates Inc., 48664 Milmont Drive, Fremont, CA 94538. The work presented in this paper was performed at Varian during Dr. Starov's affiliation with the company.
150
thickness of the deposited films depends on the substrate temperature exponentially (for surface-reaction limited processes), variations of several degrees can result in 20-30% thickness non-uniformities. The higher the activation energy for the deposition process, the greater thickness variations will be observed, see below.
Precise temperature control can readily be achieved when the wafer is in intimate contact with a massive heated chuck whose temperature uniformity is assured by design. In newer CVD systems, such as Varian's single- wafer 5101 reactor, the wafer is heated directly by a radiant light source. Fig. 1 (also see Ref. 2). This RTP capability enables the operator to heat and cool the wafer very rapidly, opening the doors to new applications, e.g., in situ annealing of the freshly deposited films.
OUARTZ HAtOalN
RINOmSUIATOR '
=*nnnTmn_
OMMXCHAMMN -
. VACUUM raRT
R I N a W n H A T M
- o A i oerucTOR ' OAt EJECTOR RIMI
FIG. 1 — Varian 5101 Rapid Thermal CVD deposition chamber.
In such reactors, however, temperature control becomes tricky. First, wafer temperature measurement must be accomplished by a remote means since a thermocouple in contact with the back of the wafer would be irradiated by the heater directly. Frontside contact is obviously unacceptable due to interference with the deposition process. Usually, an infrared pyrometer sensing emission from the center portion of the wafer is used [2]. This setup does not allow an easy measurement of the wafer
temperature uniformity and one must resort to other techniques. An additional complication arises from the fact that the surface emissivity of the substrate changes continuously during the initial phase of the deposition to the emissivity of the deposited material.
An alternative technique used in this study is based on "chemical thermometry". Using yields of chemical reactions with known activation energies to estimate system temperature has been suggested previously by Steel et al. [1]. Such calibrated reactions can be used as a
"chemical thermometer" where other methods prove impractical or impossible. For example, in Ref.l, the temperature of a laser heated gas is determined by monitoring decomposition of a "thermometer" compound added to the mixture. Using a thermocouple in that case would be impossible because of its large thermal mass and slow response.
In a similar spirit, to determine the wafer temperature at any point on the wafer, we have developed and characterized a polysilicon deposition process. In this case, the deposited film thickness variations manifest wafer temperature non-uniformities and thus act as a "thermometer".
EXPERIMENTAL
The 5101 test-bed reactor used in our experiments had the configuration shown in Figure 1; it was also described in Ref. 2. Throughout most of the study, we used 100 nun Si wafers with 1000 A oxide. These substrates allowed an easy determination of the deposited polysilicon thickness with standard optical instruments such as a Nanospec automated reflectometer. Several samples were also measured by a Prometrix SpectraMap 200 to get detailed polysilicon thickness maps.
Several heating configurations were tried to optimize the irradiation geometry. One or more samples were deposited using each heating arrangement and the poly-Si thickness was measured at 21 points inside a 90 mm diameter on each 100 mm wafer. Each wafer was thus characterized by the p-Si thickness in the center, the average thickness and the standard deviation.
Most depositions were carried out at 700 C as measured by a calibrated pyrometer at the wafer center.
The wafer temperature was controlled by adjusting the lamp power and was kept constant throughout the deposition.
The reactor pressure was in the range 80-400 mTorr.
Nearly 100% silane was used at the flow rates of up to 100 seem.
To estimate the temperature dependence of the growth rate we carried out several depositions at 650-800 C, keeping the rest of the conditions unchanged.