Roland A. - PVD for microelectronics (779636), страница 72
Текст из файла (страница 72)
The combination of atmospheric pressure (thetarget front surface is at mTorr vacuum) plus the water pressure needed forproper flow rate can produce bowing of the target, similar to what happensSPU'ITERING TARGETS383to an edge-clamped wafer with backside gas (see Section 5.5.4). For example, 14.7 psi of atmospheric pressure plus 35 psi of water pressuretranslates into a load force of 5600 pounds over a 12-inch-diameter targetused for coating 200-mm wafers.
Larger diameter magnetrons of the sortbeing developed for 300-mm wafers and flat panel display applicationswill be even more susceptible to pressure-induced target bowing or deformation unless significantly thicker target/backing plate assemblies areused. To deal with this scale-up issue, large-diameter magnetron designsutilize water cooling channels between the target and the backing plate(Fig.
11.3). In addition to cooling large targets without bowing, the waterchannel approach addresses reliability concerns with bathtub-type cooling, such as magnet corrosion and rotating water seals.11.3 Target Burn-InDielectrics on the surface of a metal target can cause electrical arcing andparticle generation in the PVD source, while surface contaminants can poison PVD films. Therefore, whenever a PVD module is vented to atmosphere (e.g., to change target and/or shields or to perform maintenance), theFIG. 11.3 Conduction cooling of a planar magnetron target by use of water-cooling channels between the target and the backing plate.384R.
POWELL AND S. M. ROSSNAGELtarget must be reconditioned to remove adsorbed oxides, nitrides, and contamination before it can be used to deposit high-quality PVD films. Forthis purpose, a "target burn-in and conditioning" process is used wherebythe source is gently ramped up in power to provide in-situ cleaning whileavoiding such things as arc tracks or damage from thermal stress. Improperburn-in can be particularly severe for powder metallurgy targets since theyare often not fully densified and can literally blow apart from gases trappedwithin microvoids. Burn-in is also important for a target that has been sitting idle in the process chamber.
For example, at a base pressure of 1 x10 -8 Torr, residual gas arrival rates are ~ 0.01/k/sec. Assuming a stickingcoefficient of only 0.5, the resulting layer of contamination formed on thetarget in one 8-hour shift can be over 150-A thick.In a typical burn-in procedure, a dummy wafer is placed on the substrateholder and the source is turned on at low power and high pressure, wherethe target voltage is low.
Power is then progressively increased and pressure lowered. The highest burn-in power is typically higher than the process recipe power to ensure that the target sees its highest temperatureprior to film p r o c e s s i n g - similar to the wafer degas strategy discussed inSection 5.3.2. For example, the AI target burn-in for a PVD AI process at9.6 kW might consist of 4.0 kW-hr of deposition with the source rampedup as follows: 15 min each of 1 kW, 2 kW, and 4 kW power deposition at5 mTorr, followed by 10 min of 8 kW deposition at 2 mTorr, and finally5 min of 11 kW deposition at 2 mTorr.11.4 Target CompositionWe begin this section by noting that atomic percent, i.e., the ratio of numbers of atoms, is the common method of specifying chemical compositionand materials purity. For example, MoSi~ is a compound with 33% (1/3)Mo atoms and 67% (2/3) Si atoms. Many surface analytical techniques,such as auger electron spectroscopy (AES), X-ray photoelectron spectroscopy (XPS), and secondary ion mass spectrometry (SIMS) also useatomic percent, which can then be converted into chemical formulas.Unfortunately, sputtering target composition is almost always given inweight percent, which can be quite different from atomic percent.
Weightpercent is simply the normalized ratio of the weights of the constituentcomponents of the target. For example, fabrication of a 100-gm target containing equal weight percents of W and Ti might begin with a mixture of50 gm of W and 50 gm of Ti, which are then sintered into a composite matrix. However, in atomic mass units (AMU), W has an atomic weight ofSPUTTERING TARGETS385183.9 A M U and Ti has an atomic weight of 47.9 AMU. This means that 50gm of W is only 27% of a mole of W (1 mole = 6.02 • 1023 atoms = anAvogadro's number of atoms), whereas 50 gm of Ti is 104% of a mole.Hence, the atomic ratio of the target turns out to be 80% Ti and 20% W(Ti0.gW0.2) which is much different from its 1"1 weight ratio.Weight percent is often written in parentheses after the elemental symbols.
Hence, a W(30)Ti(70) target has 30% W by weight and 70% Ti. In thecase of A1 alloys containing very small amounts of Cu and/or Si, theweight percent is usually omitted from the AI. Hence, an A1 alloy with 1.0weight percent of Si and 0.5 weight percent Cu would be written asA1Si(1.0)Cu(0.5) and not as Al(98.5)Si(1.0)Cu(0.5). The relationship between atomic and weight percent is straightforward to calculate by simplycounting atoms and knowing the atomic weight of each species. Considera binary target AxB~_ x consisting of material A with atomic mass m (inAMU) and material B with atomic mass M.
From the target chemical formula, the atomic percent of element A is 100x. It is then easy to show that100Weight percent of A - wt % A =1 - x))(x)(11.1)100Weight percent of B - wt % t3 =(x)where(wt %A -~-wt %B) =100%As an example of the use of Eq. (11.1), consider a TixW ~_, target withx = 0.3, i.e., Ti0.3W0.7. In this case, A is Ti with m = 47.9 AMU; B is Wwith M = 183.9. Since x = 0.1, the atomic percent of Ti is 100 x 0.3 =30%. Using formula (11.1), we calculate the weight percent of Ti as 100/(1 + (183.9/47.9)(0.7/0.3)) = 100/(1 + 8.96) = 10.04%. Hence, theatomic percent of Ti is 3 times greater than its weight percent, and theTi0.3W0.7 target composition can be written by weight as Ti(10)W(90).
Notsurprisingly, the discrepancy between weight percent and atomic percent isgreatest for species with significant differences in their atomic weights.For example, A1 has an atomic weight of 27 A M U and Cu has an atomicweight of 63.5 AMU. Thus a commonly used A1 alloy target with 0.5weight percent of C u m A 1 C u ( 0 . 5 ) m then turns out to have approximately 0.2 atomic percent of Cu.386R.
POWELL AND S. M. ROSSNAGELRegardless of how one specifies target composition, it may graduallyshift over time. This may be the result of the following effects.Dissimilar Sputter Yields If two elements have different sputteryields, they are also likely to have different yields when present as thecomponents of an alloyed target. The first approximation (often used forsurface analysis measurements involving sputter erosion depth profiling)is simply to assume the bulk, elemental sputter yield for each alloy constituent. This results in a rapid depletion at the cathode surface of thehigher-sputter-yield material and the possibility that the initial films deposited from the target may be enriched in the higher-sputter-yield material.
This process is self-limiting in that the surface composition eventually adjusts to a slightly higher concentration of the lower-sputter-yieldmaterial - - j u s t enough to counteract the higher sputter yield of the othercomponent. The net effect can be formation of an altered surface layer inthe top few tens of angstroms of the target that has a different composition from the bulk. If target temperature is allowed to rise to the pointwhere there can be sufficient volume diffusion ( ~ 500~ for AICu), theentire target may eventually be depleted of the higher-yield material.Fortunately, for most material systems of interest to IC processing (suchas AICu), the altered layer is formed rapidly and will not be an issue following the burn-in procedure used for degassing.Granular Targets Sputter targets that are fabricated with complex stoichiometry (e.g., ternary oxides such as BST or SBT) are usually made bymixing powders of their constituent materials, and then sintering and hotpressing.
Since these grains are randomly oriented and the surface is notcompletely flat, there can be subtle changes in composition over time asthe individual grains are slowly exposed and then sputtered. This, in asense, is a microscopic analogy to the altered-layer problem describedabove in which each grain within the cathode functions as an individual,microscopic target.Oxidation Issues Targets of compound materials, such as Ti-W, thatare formed from hot pressing of powders are sensitive to oxidation of theindividual powder grains.
Depending on the fabrication process environment and control, this problem can be minimized.Redeposition and Transport Most manufacturing-scale magnetronsputtering uses low operating pressure, typically 0.5-4 mTorr. At thesepressures, gas-phase scattering is low and few of the sputtered atoms areSPUTTERING TARGETS387scattered in-flight.
However, these scattered atoms may be deposited backonto the cathode. In the case of alloy cathodes whose constituents havevery dissimilar mass (e.g., TiW and A1Cu), the lighter species may be preferentially redeposited onto the cathode m particularly at higher pressures.This can lead over time to changes in the composition of the target surface.This effect, though, is somewhat self-compensating in a way similar to theyield-related altered layer formation, and is rarely a concern. Conversely,the deposited film composition may be inversely related to this scatteringissue, as the heavier atoms from the target pass through the background gasmore easily and preferentially deposit on the sample [11.5].Nonunity Sticking Coefficient Sputter deposition is generally considered to be characterized by a 1.0 sticking coefficient for the sputteredatoms. That is, the sputtered atoms hit a surface and stick immediatelywithout bouncing off.
There may be some cases, particularly with alloyshaving a large mass mismatch (e.g., TiW and AlCu) in which the stickingprobability could be different for each species. This may depend on the relative masses involved m e.g., light atoms might bounce off of a high-massfilm surface - - or perhaps on the kinetics of the deposition. More massivematerials (e.g., refractories such as Ta or W) tend to have higher kinetic energy and as such may be more likely to reflect from high-angle surfacessuch as the sides of a trench.11.5 Target PurityAll sputter targets m be they elemental (e.g., Ti), binary (e.g., WSi~), oralloy (e.g., A1-Si-Cu, T i - W ) m contain impurities.
![](https://s.studizba.com/z.php?f=/templates/si/i/noavatar.png&w=100&h=100&t=1"){kind=avatar}
