Roland A. - PVD for microelectronics (779636), страница 28
Текст из файла (страница 28)
Electron Bombardment. For a DC magnetron with grounded shields,electron bombardment of the substrate is not significant.5. Heat of Condensation. The latent heat of condensation AH is thechange in enthalpy of the sputtered material associated with its going fromthe gas phase to the solid state. The energy given up upon surface adsorption is on the order of 10 eV, so that the energy flux associated with adsorption is given by J x AH ( ~ 0.2 W / c m 2 for 1 /xm/min of AI).6. Heat of Neutralization. Each Ar ion that is neutralized at the wafergives up the Ar ionization potential of 15 eV, with the total power determined by the flux of Ar + at the wafer. Since the ion current to the waferin conventional PVD is << 1 m A / c m 2, the power associated with the heatof Ar neutralization is << 0.01 W / c m 2.
This is not the case for ionized142R. POWELL AND S. M. ROSSNAGELFIG. 5.25Retlected fraction (R) and maxinmm retlected energy (E,) of reflected neutralized At" during 1 kcV Ar ~ bombardment of various targets. (Reproduced with permission of The McGraw-HillCompanies from D. L. Smith, Thin Film l)el~ositioJt: Prim'ilges & Practice, McGraw-Hill, NY, 1995.)magnetron PVD, in which an appreciable flux of Ar ions can be incident atthe wafer (see Chapter 8).7. Blackbody Irradiation. In general, the temperature of a properlywater-cooled target and the sputter shields will be below 400~ Since theemissivity of metals is very low, the radiated power is in general not a significant contribution to wafer heating, with outgassing of these surfacesduring PVD usually of greater concern.8.
Radiative Heat Transfer. Radiative loss is governed by Stefan's Lawand involves the fourth power of wafer temperature (in K):~,,ut - eo-(T4__T 4)(5.3)where o-is the Stefan-Boltzmann constant (5.67 • 10 -~2 W/cm2-K4), E isthe effective emissivity of the wafer front or back s u r f a c e - which takesinto account both the emissivity of the emitting surface and the environment into which it radiates (e.g., the hot s h i e l d s ) - and T 0 is the temperature of the environment.
In general, the emissivity of a PVD metal iswavelength dependent and a complicated function of surface roughness,SPU7TERING TOOLS143temperature, and film thickness. As a practical matter, the emissivityof thick (> 1000 A) Al, Ti, or TiN films used in PVD is relatively low(< 0.2).Figure 5.26 presents calculated time-temperature profiles for a Si substrate under conditions representative of PVD Ti deposition (1000 A/rnin;25 mW/cm2) and PVD A1 deposition ( I pm/min; 250 mW/cm2).
Given themaximum thickness of Ti/TiN (< 1000 A) and Al (< 1 pm) encounteredin advanced devices, deposition time is < 60 sec. The important points torecognize from Fig. 5.26 are the following: (1) The initial temperature riseof the wafer is linear with deposition time and independent of emissivity.Radiative losses are not significant at low temperature, and Eq. (5.2)(2) Wafer tempershows that the slope of the temperature rise is ll(pCature saturates at a radiation-limited temperature tkit can be quite highunder conditions of high deposition rate.
(3) However. for realistic processtimes below = 60 sec, neither process leads to excessively large wafertemperature increase (AT = 50-100°C). On the other hand, Fig. 5.24 doesnot take into account such things as radiation from a very hot shield or acollimator. or the possibility of a heated wafer chuck. Also. variations inwafer-to-wafer temperature can have a harmful effect on process repeatability. Finally, the incident power density during high-rate, sputter precteanetching (= 1 W/cm2)can be several times larger than during sputter deposition.
resulting i n a proportionately larger temperature increase.While this discussion has focused on temperature rise during processing.wafer cooling is equally important. Figure 5.27 shows the cooldown of aSi wafer after deposition of 1 p m of Al that brought its temperature to225°C [5.27].Depending on your point of view, the cooldown is either unacceptably rapid so that starting wafer temperature at the next processmodule depends strongly on intramodule robotic handling time or unacceptably slow so that wailing for the wafer to stabilize or cool to near roomtemperature will compromise throughput. In view of all the above considerations.
an improved method of controlling wafer temperature duringPVD and related vacuum processing is required.7).The most widely used method of controlling wafer temperature duringPVD is by backside gas conduction i5.29-5.3 11 in which a few Torr ofgas ( v s the few mTorr of process gas used) are introduced into a volumebetween the wafer and the platen (see Fig. 5.28). The gas provides athermally conducting path between the wafer and the platen, whichcould be actively heated or cooled if desired.
Typically a wafer clamp144R. POWELL AND S. M. ROSSNAGEL(a)(b)FIG. 5.26 Calculated time-temperature profiles from ref. 5.27 for a Si substrate under depositionrate and thermal loading conditions representative of (a) PVD Ti (1000 ,Admin; 25 mW/cm2), and (b)PVD A1 (1 ~m/min; 250 mW/cm2).SPUTTERING TOOLSFIG. 5.27145Experimental cooldown of a Si wafer after deposition of 1 /xm of AI (data from ref. 5.27).ring assembly is used to provide sufficient holding force to keep the waferin place against the backside gas pressure (4 Torr results in nearly 4 lb offorce over a 200-mm wafer) with sufficiently low leakage of gas to prevent contamination or variations in process pressure. Because some leakage of gas into the process chamber is inevitable, Ar gas is typically usedFIG. 5.28 Illustration of the use of backside-gas-assisted heat transfer to control wafer temperatureduring vacuum processing.146R.
POWELL AND S. M. ROSSNAGELas the backside gas (e.g., as opposed to inert gases like He, which is costlyand difficult to pump with a cryopump, o r H 2, which is reactive with Ti).As a result, the acronyms BSA (backside argon) and BSG (backside gas)are often used interchangeably in referring to backside-gas-assisted heattransfer.Well-designed wafer chucks with gas-assisted heat transfer can achieveheat transfer coefficients in the range of h = 10-30 mW/cmZ-K, which is100 times greater than without BSA.
The heat transfer coefficient, h, isa measure of the efficiency of the process, and using heat theory we canestimate how well temperature can be controlled. At relatively low wafertemperature, radiative losses can be ignored and the wafer will cool moreor less exponentially (exp(-t/~') with a time constant ~" = pCz/h. Usingz - 0.05 cm for a 200-mm wafer and h = 20 mW/cmZ-K gives ~- ~ 5 sec.So wafers can very quickly be brought to equilibrium. Figure 5.29 showsdata for the rise time of a wafer with and without backside gas admitted between the heater and wafer.
Since h is finite, in general there will be a350..............................~.. . . . . . . .:. . . . . . . . . . . . . .!300 r:..9:i.--~r;of A2o2,_,250 -0n Heater and Wafer200L_~O.150-IO0-EI-,/ i50-......:..eater an, Wa,er9O9w0204060. . . .:,80, a.1 O0.....120...: . .140Time (sec)FIG. 5.29and wafer.Temperature rise time for a wafer with and without backside gas between heater blockSPUTTERING TOOLS147difference or offset between the equilibrium temperature of the wafer andthe heated chuck, which is on the order of AT = ~ / h , where 9 is the energy flux from the heater. For a blackbody heater at 500~ (773 K) withemissivity of 0.5, 9 = e o T 4 ~ 1 W/cm 2.Assuming all this is absorbed by the wafer, a value of h = 20 mW/cmZ-Kleads to the wafer being 50~ cooler than the chuck. Therefore, care mustbe taken when using reported process temperatures, since what is oftenreported is the heater block set point temperature (e.g., measured with athermocouple embedded in the block) and not the actual wafer surfacetemperature, which is extremely difficult to measure accurately withoutphysical wafer contact in the temperature range of 300-500~ where manyPVD processes are carried out.
For example, conventional optical pyrometry is difficult to apply since the spectral emissivity of S i depends onemitted wavelength and is a complex function of wafer temperature, doping, roughness, and the presence of thin dielectric films. Also, the T 4dependence of the radiated power quickly reduces optical signal strengthas wafer temperature falls below 500~Although noncontact methods of wafer temperature based on optical effects such as absorption, interference, scattering, and luminescence havebeen developed, they are not widely deployed in production and not alwayscost-effective to retrofit onto the installed base of PVD tools. Therefore, a toolsupplier will often make controlled measurements using an instrumented testwafer whose surface is in intimate thermal contact with an array of thermocouples of thermal mass sufficiently small compared to the wafer to providean accurate reading and rapid response time (e.g., the wafers provided bySensarray Corporation, Santa Clara, CA).
In this way, the supplier can providedata to a customer ~ albeit not obtained on the customer's actual wafers ormodule during processing ~ of the temperature- and time-dependent correlation between heater block temperature and wafer temperature for a PVD module, preclean module, degas/cooldown station, etc.The physics of backside gas conduction heat transfer is discussed in refs.5.30 and 5.31. Depending on the value of the dimensionless Knudson number Kn given by the ratio of the mean free path in the gas (h) to the gap (z)between the wafer and wafer platen, two regimes need to be considered:the viscous flow or continuum regime where Kn = h/z < 0.01, and themolecular regime, where Kn -- 1. In the viscous regime, the heat conductivity is independent of gas pressure and varies inversely with z.
![](https://s.studizba.com/z.php?f=/templates/si/i/noavatar.png&w=100&h=100&t=1"){kind=avatar}
