Roland A. - PVD for microelectronics (779636), страница 19
Текст из файла (страница 19)
Target erosion also brings the target surface physically closerto the source magnets, with the most strongly eroded regions being closestto the magnets below. This tends to increase the magnetic field in groovedregions and serves to accelerate the effect with target life. Finally, due tothe finite geometric size of the source as seen by the wafer and the nondirectional nature of the sputtered flux, a uniformly eroded target will notproduce a uniform thin film. In batch PVD systems, one can scan the waferbeneath the target to address these concerns; however, in single-waferPVD cluster tools this has not been practical. Therefore, the job of themagnetron designer is to tailor the target erosion profile over the target toTHE PLANAR MAGNETRONFIG.
4.2 Development of a narrow sputter erosion groove during target life can change the distribution of sputtered flux.produce an acceptably uniform thin film, while achieving good utilization(e.g., full-face erosion) of the expensive target material.4.3 The Swept-Field MagnetronThe interrelated problems of target grooving, target utilization, and thinfilm uniformity have led to the use of so-called swept-field magnetrons inwhich the permanent magnet array, rather than being stationary, is mechanically rotated behind the wafer in the plane of the target at relativelylow speed ( ~ 40-60 rpm). By tailoring the shape of the magnet array andits center of rotation, designers have been able produce a more uniformlyeroded annular region or uniform erosion over virtually the full face of thetarget.
Since the engineering design rules and manufacturing details of advanced source designs are highly proprietary (or disclosed only in thepatent literature), we only indicate the basic idea.Figure 4.3 shows a circular ring of bar magnets that upon off-axis rotation produces an annular race track with two broad erosion grooves.
To92R. POWELL AND S. M. ROSSNAGELFIG. 4.3 Planar magnetron with a circular target and a rotating ring-shaped magnet array (afterFigs. 3C and 3D in ref. 4.8).erode the central region of the target, more advanced designs have been developed. Figure 4.4 shows a heart-shaped magnet array (Quantum T Msource from Varian Associates), which can be precisely tailored both inshape and magnet configuration to produce extremely uniform, full-faceerosion.
The permanent magnets (MI through Ml4 ) are sandwiched between two parallel, iron keepers (labeled KI and K2), which distribute themagnetic field uniformly along the magnets and serve to define and holdthe contour of the array during rotation. If this heart-shaped magnet is substituted for the ring-type magnet in Fig. 12.3 and rotated about point C, abroad and uniform erosion profile can be obtained, as indicated in Fig. 4.4.Just as there are many different ways of implementing the basic DCmagnetron concept, the swept-field magnetron design has been modifiedand optimized to deal with specific target materials and PVD process conditions. For example, because the angular emission of sputtered flux fromTHE PLANAR MAGNETRON93FIG. 4.4 Planar magnetron with a circular target and a rotating, heart-shaped magnet array (afterFig.
3L in ref. 4.8).an A1 target is different from that of Ti, one would not expect the same target erosion profile (i.e., the same magnetron design) to produce the samethin film uniformity. Hence, it is not unusual to have separate sources forA1, Ti, or TiN, with a source for PVD Ti/TiN further differentiated depending on whether a collimated (e.g., TiN barrier) or noncollimatedprocess (e.g., TiN ARC layer) is being used. Differences in magnetron design also relate to differences in uniformity and step coverage needs at thewafer. For example, a thick, blanket A1 alloy film intended for an interconnect line requires very high target utilization with little need for stepcoverage. This might be accomplished with a magnet design producing abroad, annular erosion groove in the target or a design producing a numberof concentric erosion grooves.
On the other hand, the step coverage forTi/TiN liners is critical to their application, so target utilization could bereduced to achieve the higher priority. In this case, preferential erosionnear the edge might be used to compensate for the fact that outwardly facing via sidewalls at the wafer edge "see" a reduced flux of sputtered material due to the finite size of the target.R. POWELL AND S. M. ROSSNAGEL94In addition the influence of materials and processes on magnetron design, the spacing of the magnetron from the wafer can have a strong effecton uniformity. In fact, as the target erodes and changes both its thicknessand surface contour, the source-to-substrate spacing may need to bechanged for optimum uniformity.
Figure 4.5 shows representative curvesof film uniformity for a 200-mm wafer versus source-to-substrate spacingcalculated for two different designs of swept-field planar magnetrons witha 12-inch-diameter target. The corresponding film thickness profiles arealso shown (a diameter scan of sheet resistance R s would give the inverse1614v12E100"E8--=6J42......................................................................................................................................................0....!Source Substrate Spacing (cm)1"-U~(9t--- ~._.14-0.98--~0.96Or"I---o 0 .
9 4 N~0.92Oz0.9~0.880"',0.5,1,1.5,2'",2.5,3.....,3.5!4Wafer Radius (in)FIG. 4.5 (Upper) Film uniformity over a 2 0 0 - m m wafer versus source-to-substrate separation calculated for two different designs of 12-inch diameter magnetron source. (Lower) Calculated profile ofnormalized film thickness for the two sputter sources.THE PLANAR MAGNETRON95of this profile since R s = p/t is inversely proportional to film thickness t).Such curves can be directly measured or calculated to first order by usingthe physically measured target erosion profile as a source function andthen assuming a simple angular emission from the target (e.g., cosine)without gas scattering.
One can also approximate a symmetric target erosion profile as a finite sum of ring sources and then sum their flux at thewafer location to estimate film uniformity.We see from Fig. 4.5 that either of the two magnetron designs represented can provide excellent uniformity (max-min of ~ _+5%), althoughthe optimum uniformity is obtained at quite different source-to-substratespacing and can be degraded significantly if the spacing changes by as little as 1 cm. Fortunately, by choosing a proper separation (e.g., using awafer table with vertical z-axis motion) and modifying this separation either manually or automatically as the target erodes, good uniformity cangenerally be obtained throughout target life.
In some cases, it is possible todesign a target erosion profile that will keep film uniformity from driftingoutside of a specified upper limit (e.g., 3 o < 5%) for a fixed source-tosubstrate separation throughout target life. Alternatively, one can sometimes modify operating conditions (power, pressure, etc.) enough to compensate for changes in film thickness uniformity associated with targeterosion.Historically, the diameter of planar magnetrons has increased along withthe diameter of the wafers they were intended to coat. For example, targets8 inches in diameter are commonly used for PVD of 150-mm waters, whiletargets 12 inches in diameter are used for 200-mm wafers.
Advancedswept-field magnetron designs are scalable to 300-mm wafers and, basedon the historic trend, we would expect that magnetrons with 18-19 inch diameter will be used for this application. It is possible, however, that a different geometry may be required to obtain acceptable uniformity and costof-ownership for 300-mm wafers.
For example, a rectangular magnetronsource could be used and the wafer moved through the beam m which issimilar to how PVD is used to coat large-area substrates such as flat paneldisplays and architectural glass.4.4 Source ArcingAn arc is the general term used for any low-impedance condition createdduring the PVD process that then appears at the output of the magnetronpower supply. As a result of the abrupt drop in impedance, the arc candump the energy stored in the output of the power supply into the target.96R. POWELL AND S. M. ROSSNAGELArcs are constantly occurring during magnetron sputtering and, if one isnot careful, sufficient energy can be delivered to them to create fine particulates or even melt microscopic amounts of target material that "splat"onto the wafer.
![](https://s.studizba.com/z.php?f=/templates/si/i/noavatar.png&w=100&h=100&t=1"){kind=avatar}
