Roland A. - PVD for microelectronics (779636), страница 47
Текст из файла (страница 47)
By doing so. the net sputtering rate of the coil always exceeds the deposition rate. and the coil isslowly eroded. This approach requires that the coil be constructed of thesame material as the magnetron cathode and that i t be of the same high purity as the magnetron cathode. as atoms from the R F coil will certainly beincluded in the deposited films on the sample. This approach is also not really conducive to water cooling of the RF antenndcoil. High-purity tubingof most desirable materials is not available.
and it would also be difficultto deal with water connections for some materials. Therefore. the approachof a always-eroded RF coil usually allows the coil to operate uncooled,246R. POWELL AND S. M. ROSSNAGELwhich can result in coil temperature of 500~ This can cause substantialradiant heating of the sample and may not be conducive to the ionized deposition of lower-melting-point materials such as A1.Another aspect of the always-eroded, uncooled coil is that it no longerneeds to have a round or hollow cross section. Commercial applications ofthis design use a flat plate to form the coil [8.7].
This leads to an insidefacing side that is heavily eroded and back-facing side that is lightlyeroded. This design intrinsically removes concerns with buildup on theback-facing side and also reduces deposition near the vacuum feedthroughs, resulting in a long operating lifetime before cleaning.The always-eroded coil can potentially lead to problems with directionality near the edge of the sample. Atoms sputtered from the coil differ fromthe atoms sputtered from the magnetron cathode in that (a) they have amuch shorter trajectory from the source to the sample and as such as areless likely to be ionized in the discharge, and (b) since they are coming infrom the perimeter coil electrode, their trajectory is mostly inward-facingas far as the wafer is concerned.
These two issues may result in a lowerstep coverage on the bottom surfaces of features near the wafer edge. Thelocal emission from the coil electrode may also lead to preferential deposition on the inner sidewalls of features near the wafer edge; the oppositeeffect of what is seen near the wafer edge in long-throw sputtering.There has also been some experimentation with spiral coils instead ofperimeter coils. The spiral coil is likely to lead to better plasma uniformity,and spiral coils are routinely used on etch-related inductively coupledplasma tools. The pitch of the spiral coil can be adjusted spatially to resultin excellent uniformity for the plasma as well as allowing plasmas to beformed that are not purely circular but perhaps rectangular in cross sectionfor use with flat panel display substrates.However, spiral coils suffer from a very significant problem with anI-PVD tool: The deposition rate on the side of the coil that faces the magnetron sputtering cathode is much higher than the side that faces the wafer.This means that the top side of the coil is heavily deposited, and the underside (wafer side) of the coil is heavily etched, resulting in significantcontamination of the film by coil material.
This could only be used in thealways-eroded operational mode described above.The exact location of the RF coil has an effect on both the net deposition rate and the uniformity of the directional deposition. One obviousproblem occurs is when the diameter of the RF coil is too small. This cangeometrically block the edge regions of the sample from some fraction ofthe cathode emission surface. The inside diameter of the RF coil should besuch that it does not intersect a line drawn from the edge of the sample toIONIZED MAGNETRON SPUTTER DEPOSITION: I-PVD247the edge of the cathode. Other geometrical problems are primarily foundwith the vertical location of the coil with respect to the sample. At distances less than 3 - 4 cm, emission from the coil can overwhelm depositionfrom the plasma and the edge regions of the sample may have poor orasymmetric directional deposition.
If the coil is located less than 3 - 4 cmfrom the cathode surface, there can be a very strong coupling between thetwo nominally separate plasmas. This results in a much lower dischargevoltage for the magnetron as well as increased tuning difficulties for theRF coil.However, while it is critical that the RF coil not strongly interact witheither the sample or the cathode, it is also important that the cathode-tosample distance not be too large. Because the operating pressure (described below) will turn out to be several tens of mTorr, the overall deposition rate will fall strongly as a function of throw distance due to thehigh levels of gas scattering.
At 40 mTorr, a throw distance of greaterthan 13-15 cm will result in over an order of magnitude reduction in thenet deposition rate compared to a more typical PVD operating pressure of1-2 mTorr. It is desirable that the throw distance be minimized as muchas practical, and it appears that this minimum distance is on the order of10 cm in a practical system.The topics of RF coil design and operating frequency are likely toevolve in each successive generation of I-PVD tools. There are many different approaches for launching RF waves into plasmas, and subsequentdesigns may take advantage of the intrinsic magnetic field or some othergeometry-dependent feature to generate a higher plasma density.
Since themean free path for ionization of the sputtered atoms is directly dependenton the plasma density and coupled to the electron temperature, it may bepossible to increase the relative ionization of the atoms with denser orhigher temperature plasmas.The matchbox for an I-PVD tool differs from that of a conventional RFdiode.
The two primary differences are that the plasma coil antenna functions in place of the inductor in the conventional matchbox, and also thata third capacitor is included to help balance the potential on the coil. Thematchbox is shown in Fig. 8.4. The power supply feeds both a shunt capacitor to ground and a series segment of a capacitor, the plasma electrode,and the third capacitor, which is then grounded.
The capacitors before andafter the plasma coil are typically of the same value (1000-2000 pf, variable air gap) and are tuned such that the voltage on each end of the coil is180 ~ out of phase. This limits the maximum voltage swing on the coil, reduces the amount of sputtering, and also makes it symmetric along the coillength. Depending on the operating conditions, the reflected power to the248R. POWELL AND S. M. ROSSNAGELRF Power Supply(a)(b)1~[~---] #.-71-4 Turn Metal Coil inVacuum SystemFIG.
8.4RF matchbox design for I-PVD tool.power supply may be as high as 20% of the incident power, although it ispossible to reach very low reflected powers. The lowest reflected powermay not be indicative of the best operating conditions, due to the interplaybetween inductive and capacitive coupling. In general, the higher the inductive coupling, the higher the plasma density, which might be seen as anincrease in the ion saturation current to the sample. Higher coil voltagesare consistent with a higher level of capacitive coupling, which would result in a lower plasma density as well as increased coil sputtering.Depending on the desired operating mode (from net deposition on the coilto no deposition to net erosion), the tuning of the circuit will v a r y - aswill the reflected power. In some systems, the best inductive match (maximum plasma density, net deposition on the coil) is consistent with aboutIONIZED MAGNETRON SPUTTER DEPOSITION: I-PVD24920% reflected power.
The best matching conditions (minimum reflectedpower) tend to correlate with increased capacitive coupling and net erosionof the RF coil electrode.Even though the initial manufacturing-scale tools for I-PVD appear tobe based on inductively coupled plasma ionization of conventional magnetron-sputtered metal, several alternative designs have been developedand published. One is based on physical sputtering and ECR (electroncyclotron resonance) plasma ionization [8.9]. This is shown schematicallyin Fig. 8.5.
The ECR plasma in this case is set up by a permanent magnetassembly located at the top of the tool. Microwave power at 2.45 gHz isfed in near the top but from the side, creating a dense plasma in the centralregion of the system. A sputter cathode is located adjacent to the ECR region such that high rates of ion bombardment can occur on the cathode surface, resulting in sputtering of atoms into the ECR discharge region.
Thetool has been primarily used for Cu deposition, but the design would allowthe use of other metals. The ECR tool typically reaches plasma densitiesthat are 2 to 5 times higher than the inductively coupled RF plasmas, whichallows some reduction in the operating pressure of the tool. This is important because of potential arcing and breakdown problems at higher pressures in the microwave-launcher region.FIG.
8.5ECR-based I-PVD tool [8.9].R. P O W E L L AND S. M. R O S S N A G E L250A second, alternative design uses a magnetron sputtering source thathas been configured as a hollow cylinder (Fig. 8.6) [8.10]. In this case,the magnetron discharge is located adjacent to the inner-facing walls ofthe cylinder and the magnetic field (not shown in the figure) is configuredaxially. Because of this geometry, an additional mechanism of electronconfinement m geometrical in nature and known as the hollow cathodeeffect m leads to higher levels of plasma confinement and higher localdensities.
The sputtered metal in this case transits through the denseplasma region and has a much higher probability for ionization than in aconventional, planar discharge. The metal ions are then free to leave thesource region in the form of a metal plasma beam, which diverges out toward a substrate location. This device differs from other I-PVD tools inthat it has no additional active means of ionization for the metal: The ionization level is intrinsic to the hollow cathode design.
![](https://s.studizba.com/z.php?f=/templates/si/i/noavatar.png&w=100&h=100&t=1"){kind=avatar}
