Experiments on gas cooling of wafers

Experiments on gas cooling of wafers

Nuclear Instruments and Methods 189 (1981) 169--173 North-l-IoUand Publishing Company 169 EXPERIMENTS ON GAS COOLING OF WAFERS Monty KING Eaton Corp...

315KB Sizes 19 Downloads 85 Views

Nuclear Instruments and Methods 189 (1981) 169--173 North-l-IoUand Publishing Company

169

EXPERIMENTS ON GAS COOLING OF WAFERS Monty KING Eaton Corporation, 2433 Rutland Drive, Austin, Texas 78758, U.S.A. and

Peter H. ROSE Nova Associates, Inc., 133 Brimbal Avenue, Beverly, Massachusetts 01915, U.S.A.

The temperatures reached by wafers during implantation can be high enough to destroy resist materials and higher temperatures can cause partial annealing in silicon. This paper examines the applicability of gas cooling of the wafer backside and presents data showing that very effective cooling can be provided with a heat transfer coefficient of between 20-50 mW cm -2 C-1 depending on the gas used.

1. Introduction In medium current ion implanters, which generally employ electrostatic scanning and a target chamber which processes a single wafer at a time, the final or equilibrium temperature, if controlled by radiative heat loss alone, is often above that desired by the process engineer. For this reason, various methods have been employed to introduce a conductive component to reduce the temperature rise. A very effective method o f cooling wafers is to use a layer o f vacuum grease to bond the wafer to a cooled substrate. This technique is unacceptable except for research purposes and a less effective, adequate alternative is to press wafers lightly against a spongy material bonded to a cooled backing. There are various room temperature vulcanizing silicones than can be used for this purpose, and commercially designed system are available, such as Waycool TM, which make use o f this method. In high current ion implanters beam powers are more than an order o f magnitude greater and for thermal reasons a batch process is used to average the beam over a number o f wafers and thus reduce the temperature rise to an acceptable level. As the performance o f high current implanters continues to improve, increasing the batch size eventually becomes a cumbersome way o f reducing the temperature rise and, as in the medium current machine, some form o f conductive cooling is desirable. Parry [1,2] has analyzed the process o f wafer heating during implantation in typical situations and a comparison o f his calculations with different experi0 0 2 9 - 5 5 4 X / 8 1 / 0 0 0 0 - 0 0 0 0 / $ 0 2 . 5 0 © 1981 North-Holland

mental conditions shows a good agreement e.g., Glawischnig [3]. Parry's equation for temperature rise expressed in a form suitable for computation is, A T = ( A t / p A I C ) [ P o - ~(T4w - Ts4)

- / 3 ( T ~ - T~h ) - H ( T w -- Twh)] ,

(1)

where T = temperature rise in °C, t = time in s, p = density o f silicon g c m -s, A = total radiation area, l = thickness o f wafer cm, C = specific heat o f silicon j g-1 C-1, Pb = beam power W, Tw = instantaneous wafer temperature, T s = temperature o f surroundings, T w h = t e m p e r a t u r e o f wafer holder, a = o A e w e s / [e s + ew(1 - es)], b = o A e w e w h / [ e w h + ew(1 - ewh)], o = Stefans constant, ew, es, ewh = are the emissivities o f wafer, surroundings, and wafer holder respectively, H = the conductive heat loss coefficient from the wafer to wafer holder W/°C. It is with the term H that this paper is primarily concerned. Before proceeding it is worth noting that Parry demonstrated in his papers that, under the conditions normally occurring in an ion implanter and at the power densities possible, the front to back temperature drop in the wafer is small (<5°C), (see also Glawischnig [3]. He also showed that the sideays diffusion o f heat in the wafer is not a large effect and one which, in any case, only reduces the temperature rise in the area heated b y the beam, and concluded, on examining the effect o f an instantaneous deposition o f power into the small volume o f material where the ions are adsorbed, that is produces a negligible thermal spike o f < I ° C at the available energy Ill. TEMPERATURE CONTROLLED IMPLANTS SPECIAL TECHNIQUES

M. King, P.H. Rose / Gas cooling of wafers

170

densities. Under these conditions the conductive heat loss can play a dominant role in reducing the temperature rise, provided some conductive medium, solid liquid or gas is inserted between the wafer and a cooled platen.

2. Conductive cooling The most simple and obvious way to provide conductive cooling is to clamp a wafer firmly to a very flat cooled platen (fig. l a). tt/A is generally in the range of 0 to a few mW cm -2 C -1 depending on the number of points and total area over which the wafers touch. The wafer can only be clamped around the rim and generally most of the area of contact is close to the edge of the wafer. It is also observed that the cooling is sensitive to the clamping force in a somewhat erratic manner depending on the flatness of the wafer and the smootlmess and flatness of the platen. This variability may be concealed in a production situation because wafers introduced into the vacuum are hnplanted before they have time to outgas. Gas trapped between the wafer and platen and gas evolved by implantation will be responsible for some of the heat conduction and improve the consistency of wafer to wafer performance. During the steps of processing the wafer becomes less flat and the back rougher which reduces the effectiveness of clamping as a cooling teclmique. Clamping the wafer over a thin layer of conformable material as shown in fig. lb provides a value of H/A ~ 1 8 - 2 4 mW cm -2 C -1 and provides temperature control with quite good consistancy (-+20%) over batches of wafers. There are a large number of suitable

RTV materials such as for example Eccosil No. 4952 which have a thermal conductivity in the range 0 . 0 1 0.015 W cm -x C -x wlfich can be compared with the much higher values 2.3 and 1.5 W cm -1 C -1 for aluminum and silicon. The force applied to the edge of the wafer compresses the conformable film against the backside of a wafer and at the best it can be estimated that the effective area of contact is only 1 0 20% of the wafer area. The effectiveness of the method depends on choosing the best balance between the elasticity, thermal conductivity and thickness of the material. Bowing the wafer over a spherical or cylindrical surface improves the uniformity of the compressive force. Instead of using a simple layer of material between the platen and the wafer, gas filled foams, liquids or a complex of soft metal sprhags can be employed and it is possible to increase the value of H over that for simple material. Perhaps it is because ion implantation is a vacuum process and a poor vacuum can spoil the uniformity of an i_mp!ant that gas (or liquid) cooling has been ignored until recently. Gas cooling has been applied by one of us [4] in a commercially available medium current implanter with very successful result using the configuration shown in fig. 2. To provide a fairly uniform pressure distribution over most of the back of the wafer the platen is recessed in the central region by 0.0005 (2/am). Fig. 3 shows data obtained on an Eaton implanter, the value for 11/,4 is about 18 W cm -2 C -t using nitrogen gas, the direct contact caused by the clamping in this instance contributing 10% to the conductivity term. If the value of H can be varied by changing the applied gas pressure and if hydrogen is used instead of nitrogen a seven fold increase in H can be obtained which is a higher value

FlatSurface

'

(a)

~ e r

Wafer

Sprin~Loaded

i

t co I

Piaten Wafer

1

Platen = INl

• T.C.Gauoe / ~ S o [ e n 7

(b)

Curve'~dPlaten

Fig. 1. (a) Shows wafer damped to a very flat surface by a clamp ring. (b) An example of wafer cooling by a therm',dly conductive RTV. The wafer is clampcd over a cylindrical surface shown much exaggregated in the illustration.

Leak

Valve

GAS

Operated Valve

Fig. 2. Schematic of cooling technique successfully employed in the Eaton implanter. l]ae gas ted into the center is constrained from leaking into the vacuum system by clamping the wafer with a spring loaded ring against a very flat surface.

M. King, P.H. Rose / Gas cooling o f wafers 7-'"

I

i

I

6OL

i



I

WAFER COOLING

|

4

t~o~

Wafer

Temperature

*C

Fig. 3. F_~luflJbrium temperature of wafers in the Eaton implanter as a function of pressure for different power inputs. The waters were heated by a scanned 200 keY ion beam. The measured temperature gradient between the center and the edge of the wafer was less than 1 * C.

171

3. D i s c u s s i o n o f gas c o o l i n g

There are many detailed treatments of gas conductivity, see for example Dushman [5], the main features can be seen in fig. 3. As the pressure of the coolhag gas !s increased the conductivity increases eventually reaching a value which is independent of the pressure. Knudsen (see ref. 5) has calculated the conductivity of a gas at low pressures, that is the region in which L > g , where L is the mean free path of gas molecules in the gas and g is the gap between the wafer and the platen. For convenience, table 1 lists the mean free path of a number of gasses at 25°C and a pressure of 1 Tort, the value at other temperatures and pressures varies as x/T/P. Knudsen obtained for the energy transfer from a hot to an adjacent cold surface the expression H 273 -~ = a.A ~wh (Tw - rwh)

than that reported by any other method. With no vacuum seal around the edge of the wafer and with a clamping force of 2 X 10 -2 N cm -2 the leakage of nitrogen, or argon, which are the cooling gases usually employed, into the system is about 10 -4 Torr 1 s-1. This modest gas flow and consequent small increase ha pressure in the electrostatic beam scanner line does not spoil the performance of the ion implanter and wafers can be processed with excellent uniformity at a consistent and controllable temperature. Fig. 4 shows time response and compares the temperature of an uncooled and gas cooled wafer.

Chart

IOC

I

Speed 2era/rain without Gas Coolin I I I I l 1

i Chart Speed 5cm/mm. ~= with Gas Cooling I;I

I

I

I

I

I

L

- BEkM CFF I

W cm -2 C -1 ,

(2)

where A is the free molecule heat conductivity with the values shown in table 2 and a is a constant called the accommodation coefficient. The accommodation coefficient takes into account the molectfle-surface interaction. In most engineering situations the surfaces are covered with layers of oxide and adsorbed gases and a has a value close to

Table 1 The mean free path L in um for different gases at 25 ° and at 1 Torr.

Gas

L

H2 He N2 H20

93 147 46 30

t -rE oo

E E i~°?"1'*"?"?"~"1

i

111 I

POWER OENSIT Y- LO w.~

Table 2 Values of the molecular heat conductivity A at 0°C in mW cm -2 Tort -1 C -1 . For reference, values of thermal conductivity k, in mW -] C -1 are also given in the table.

i i t t i L

Cm

Fig. 4. Left: A portion of a strip cart showing the temperature rise of a wafer without gas cooling at a beam power of 0.5 W/cm 2. Right: A portion of the strip chart showing the reduced temperature when gas cooling is used. 4 Torr pressure of argon was used with a beam power of 1.0 W/cm 2.

Gas

A

k

H2 He N2 H20

60.72 29.36 16.63 26.49

1.62 1.35 0.22 0.14

III. TEMP. CONTROLLED IMPLANTS/SPECIAL TECHNIQUES

M. King, P.H. Rose / Gas coolhzg of wafers

172

Table 3 Thermal accommodation coefficients a for: (a) metals having absorbed molecules and oxide layers on the surface and (b) a clean aluminum surface shown for comparison [6]. Gas

Surface

~

a (a)

Aluminum

0.95-0.97

-

H2 Ne

Platinum

0.3

-

Aluminum Aluminum

0.89

-

iii

0.073

unity (see table 3). The table also gives one example of the accommodation coefficient for a molecularly clean surface and as can be seen, H would be reduced by more than an order of magnitude in these circmnstances. The authors were unable to fred a value for the accommodation coefficient o f single crystal silicon. An interesting feature of gas cooling in the molecular region is that H is independent of the gap for g < L. According to eq. (2) and at a nitrogen pressure 1 Torr, H = 16a mW cm -2 C -I, which should correspond to the linear region of cooling shown by the data of fig. 3. The measured values of H taken from these curves indicate that a is significantly less than unity. At much higher pressures where the conductivity o f a gas has become constant, the heat transfer is given by the conventional formula,

H = k A (Tw - Twh). g

Wafer

(b)

Air N2

RTVS e a t ~ R i n g

(3)

The minimum value o f g is limited by the bowing of a wafer having pressure on one side and vacuum on another. For example the displacement of the center of a 75 mm wafer clamped around the edge was measured to be 5 ~an at 5 Torr and 380/am at 1 atm. Substituting reasonable values for g again leads, as in the molecular region, to values of H considerably higher than the measured values. Investigations are continuing to fired reasons for these discrepancies.

4. Conclusions

Gas cooling offers the hnplanter designer a system with a heat transfer capacbility o f 2 0 - 5 0 mW cm -2

Fig. 5. Schematic of a gas cooled rotating disk target system. The inset shows a possible technique for sealing the edge of the wafer using an elastomer.

C -1 depending on the gas used. The method is clean, consistent and simple to apply. For single wafer processing in medium current implanters the system shown in fig. 2 using gas as the conductive medium appears to be close to an optimum from a practical viewpoint. Hydrogen or helium provide the best cooling provided the vacuum system can pump these gases. In the case o f a batch type target chamber such as a disk the amount of gas leaking into the vacuum system is increased proportionally to the number of wafers in the chamber. In some systems this gas can significantly affect current measurement and hence dose and the dose uniformity. To avoid these effects the edge of the wafer may be sealed as shown in fig. 5. The same figure shows a partial cross-section of a disk target chamber modified so as to include gas cooling. It is worth pointing out that a high current system which employs only mechanical scanning and in which the beam is stationary has an advantage over other systems in that accurately dosed and inform implants may be made at high operating pressures in the region of the target chamber. In such machines the path length after analysis can be short which reduces the effect of charge exchange and the actual beam line pressures may be as high as 10 -4 Torr without affecting the beam current reading of the Faraday cage, and in any case no uniformity error is introduced. This has the additional advantage that the heavy outgassing which occurs at the beginning of an implant is not a limitation.

M. King, P.H. Rose / Gas cooling o f wafers

References [1] P.D. Parry, J. Vac. Sci. Technol. 13 (1976) 622. [2] P.D. Parry, J. Vae. Sci. Technol. 15 (1978) 111. [3] tt. Glawischnig, Conf. on Low energy ion beams, Bath (1980).

173

[4] M. King, Eaton Corporation, Patent applied for. [5] S. Dushman, Scientific foundations of vacuum technique Wiley, New York, 1949). [6] M.L. Wiedman and P.R. Trumpler, Trans. ASME 68 (1946) 57.

III. TEMP. CONTROLLED IMPLANTS/SPECIAL TECHNIQUES