Roland A. - PVD for microelectronics (779636), страница 71
Текст из файла (страница 71)
The relatively inefficient use of electrical energy in DC magnetron sputtering is due to two effects" (1) the nature of physical sputtering and (2) the nature of the diodeplasma system.As described in Chapter 2, physical sputtering is a momentum transferprocess from the incident, energetic ion to the atomic lattice of the target.Under somewhat random collisional processes, one or more of these targetatoms is ejected due to the bombarding particles. A relevant example is thecase of Ar § sputtering of A1Cu.
In this case the operating or discharge voltage of the cathode is perhaps 500 V, which imparts 500 eV of kinetic energy to the incident Ar ion. The sputter yield of 500 eV Ar § on AICu isabout 1.0, and the average kinetic energy of the ejected AI atom might be10 eV. Therefore, from a simply particle point of view, the emissionprocess is only (1.0 • 10)/500 = 2% efficient in terms of returned energyto the discharge. However, there are also additional processes to consider.The secondary electron yield for Ar + on A1 might be ~ 5%, and each secondary electron picks up the full discharge potential as it returns back intothe plasma. There are also other minor sources of energy from the cathode:SPUTTERING TARGETS379t h e r m a l b l a c k b o d y radiation f r o m the slightly h e a t e d surface, r e f l e c t e dneutrals (which are a very small effect for Ar + on A1), and s o m e opticalemission.
T h e net result is that only 10% or so o f the incident e n e r g y returns from the c a t h o d e in the f o r m o f energetic particles or p h o t o n s ; ther e m a i n i n g 90% is a b s o r b e d as heat by the c a t h o d e and must be takena w a y by w a t e r c o o l i n g of the c a t h o d e (air c o o l i n g is not sufficient to dealwith the heat load on p r o d u c t i o n D C m a g n e t r o n s but can be used insmaller, r e s e a r c h - o r i e n t e d sources).Target cooling, then, is a key e n g i n e e r i n g p r o b l e m for m a g n e t r o n sputtering.
A standard 12- to 13-inch-diameter target is generally rated at2 0 - 2 5 kW, which m e a n s that it must have sufficient water cooling to absorb nearly 20 k W of thermal energy. The water flow r e q u i r e m e n t s are onthe order o f 5 gallons or more per minute, requiring water lines o f about1-inch d i a m e t e r ( c o m p a r a b l e to the flow when filling a car's 15-gallon gastank in 3 minutes from a service station pump). As a practical matter, thep o w e r capability of a m a g n e t r o n sputtering cathode is limited primarily bycooling and not by plasma issues.The water flow can be calculated as follows. Taking the heat capacity ofwater as 1.0 cal/gm-~ using chilled water near 0~ and using the maximumoutput water temperature of 100~the incoming power P in watts(joules/sec) can be written asP=gm-~xcalx (100~xarTwhere dM/dT is the mass flow rate of the water ( 1 gal/min = 65 gm/sec).
At an applied power of P = 1 kW, we calculate a minimum water flow of ~ 3 gm/sec =1/20 gal/min. This is an absolute minimum, however, since it is preferable forsafety reasons not to have an output water temperature exceeding 35~ This consideration leads to a practical value about 5 times higher, or an effective flow rateof about 1 gal/min per 4 kW of applied power.The p o w e r density on the cathode of a swept-field m a g n e t r o n varieswith time as the m o v i n g etch track r e v o l v e s across the cathode space. Eventhough the instantaneous p o w e r density in the etch track might be 10 timesgreater than the average p o w e r density ( t i m e - a v e r a g e d over many rotationsof the m a g n e t array), thermal calculations typically use the average p o w e rdensity. For example, a 12-inch d i a m e t e r AICu cathode operating at 20 k Whas an average p o w e r density of 20 W / c m 2.
A s s u m i n g a backing plate o f380R. POWELL AND S. M. ROSSNAGEL0.25 inch and a target thickness of 0.5 inch, this leads to the surface temperature at the cathode being a few tens of degrees higher than the backside water temperature. However, if the magnet rotation is stopped and theetch track remains stationary, the power density would be > 200 W/cm 2,resulting in potential local melting of the A1Cu target. Needless to say, production sputtering tools have interlocks that detect both adequate waterflow as well as magnet rotation.
This situation is not unique to PVD. Forexample, in a high-current batch ion implanter, each wafer in the batchmay be rotated through the high-power density ion beam with only a fewmsec spent under the beam per rotation. Should the beam be allowed todwell too long on a given wafer, excessive photoresist heating or even catastrophic melting of the Si wafer could occur.At the PVD shield and wafer locations there are also concerns aboutthermal status during deposition. For example, a conventional-diameterA1Cu cathode (12- to 13-inch diameter) might be operated at 20 kW to obtain a deposition rate of 1 /~m/min. Approximately 10% of this 20 kW, or2 kW, is delivered to the discharge chamber, where it eventually reachesthe wafer, chamber walls, and fixturing.
At the wafer, the high depositionrate can cause significant heating (see Section 5.3.4). Assuming an approximate atom size of 2.5 A, a deposition rate of 1 /~m/min is equivalentto about 67 atomic layers/sec. Each arriving atom brings along its kineticenergy plus its heat of condensation, whichalong with other minor contributions from the plasma ~ might amount to 12-15 eV per adatom. Integrating this over a 200-mm wafer leads to a deposited energy flux of50-70 W.
The power deposited on the shields is significantly higher thanthis because they are located closer to the cathode and also function electrically as the de facto anodes in the plasma circuit. Shields can easilyreach temperatures of 200-300~ during continuous operation. Collimators too, as described in detail in Chapter 6, have been measured to exceed450~ during extended, high-power operation. The thermal cycling of allof these chamber parts is important in that the resulting stress in the filmsinevitably deposited on them can result in subsequent delamination andflaking.Target heating can affect the PVD process in a variety of ways. For example, excessive target heating can cause undesired outgassing of impurities or induce thermal stress resulting in particle emission and possiblycracking. Thermomechanical damage to magnetron parts, harmful effectsto permanent magnets, or even loss of a critical dimension by thermal expansion are possible.
It is even possible for solder-bonded targets to physically fall off of their backing plates due to thermal-stress-induced delamination. More subtle effects of heating might include changes in targetSPUTTERlNG TARGETS38 1microstructure - such as increased grain size - that can affect sputteredfilm properties. For example, A1 films are observed to have the best sheetresistance uniformity when the A1 target has both a preferred (100) orientation and fine grains (< 100 pm), so recrystallization of a fine grain target is to be avoided.
As a rule of thumb, recrystallization is a concern whenthe operating temperature of the target exceeds one-half of its meltingpoint. For a pure A1 target (Tmp= 660°C = 933 K), this means that temperature should be kept below = 194°C. In cases of reactive sputtering ofTIN, where the degree of target nitridation is of direct concern, increasesin Ti target temperature can cause the process to shift from a non-nitrided,metallic mode to a nitrided mode.
Radiant heat from a hot sputtering target can also increase substrate temperature, affecting film microstructureand composition. Therefore, active cooling of the target is critical.A variety of methods for controtling cathode temperature have been employed. most of which rely on water cooling with particular attention to reducing the thermal resistance between the target and a cooled backingplate. For example, an e.rpirtzsion contuci method has been successfullyused with conical magnetrons in which the thermal expansion of the targetbrings it into physical contact with a water-cooled ring [I 1.21.
As the target expands. the contact pressure can become large enough for effectiveconduction cooling. When sputtering power is removed. the target coolsand contracts so that i t is no longer i n contact with the ring, which makesevcntual replacement very easy.Planar magnetrons typically employ a bonded contact between the target and water-cooled hacking plate. The trend is away from solder orepoxy and toward more reliable diffusion bonding i n which solid state diffusion at elevated temperature and/or pressure is used to "cement" the target and backing plate. For example, concerns over debonding have prevented very high power operation of Ti targets solder bonded to Cubacking plates.
However, Ti targets that have been diffusion bonded to1.8 kpsi) canhigh-strength. A1 backing plates (tensile strength of bondreliably operate at powers up to 30 kW 11 1.3, 11.41. Whatever method ofbonding is used. verification of bonding integrity is often provided by thetarget supplier using nondestructive ultrasonic imaging (= I0 MHz) of thecompleted target-backing plate assembly. Finally, we note that bondingcan be avoided totally if the target and backing plate are machined fromthe same monolithic block of material: however, this can lead to significantly increased costs.To cool the backing plate and other cathode components, a bathtub-likearrangement can be used as shown i n Fig. 11.2. In this case, cathode components such as permanent magnets need to be encased (potted) in a-382R.
POWELLAND S. M. ROSSNAGELFIG. 11.2 Conduction cooling of a planar magnetron target by use of a bathtub-type arrangement located behind the backing plate.water-resistant material to prevent corrosion, and deionized water shouldbe used to prevent electrolytic corrosion between the electrically biasedbacking plate and the grounded water supply. The electrical conductivityof any cooling fluid in contact with the cathode should also be low enoughto minimize current leakage to ground when maximum voltage is applied.Assuming a conservative design in which only a small increase in coolingwater temperature is allowed (output temperature < 35~we showedearlier that a relatively high water flow rate of > 1.0 gal/min per 4 kW ofapplied power is required.
![](https://s.studizba.com/z.php?f=/templates/si/i/noavatar.png&w=100&h=100&t=1"){kind=avatar}
