Roland A. - PVD for microelectronics (779636), страница 73
Текст из файла (страница 73)
It is neither practicalnor cost-effective to require 100% purity in a production sputter target. Onthe other hand, even trace amounts of selected elements can adversely affect thin film properties and device performance. Therefore, suppliers havedevoted much effort to reducing the level of critical impurities in their targets. This drive toward ultrahigh purity material is sometimes referred toas "the nines game," since target purity is generally stated in the languageof "nines." For example, a "four-nines-five" or 4N5 Ti target would have99.9995% purity, with the total level of all impurities being < 5 ppm =0.0005% by weight.
Although 6N purity A1 and AI alloy sputter targets areavailable, the purity of A1 targets used in production are typically 5N to5N5, while Ti target purity tends to be ~ 4N5 to 5N. Since individual elements can be present at much lower levels than the total impurity content, an element-by-element analysis covering most of the elements of the388R.
POWELL AND S. M. ROSSNAGELperiodic table is typically provided for a given target using analyticalmethods with detection limits ~ 0.01-0.001 ppm. Established techniquessuch as spark source mass spectrometry (SSMS), glow discharge massspectrometry (GDMS), and X-ray fluorescence (XRF) mapping are used tocertify the purity of the starting target material or, for a nondestructivemethod such as XRF, of the finished target itself. Figure 11.4 presents arepresentative analysis of a high-purity 4N Ti target.Since one pays a significant premium for an additional nine or evennine-five of materials purity, it is important to focus on the problem elements. Regardless of target chemical composition, alpha-particle emissionfrom heavy elements such as uranium (23su) and thorium (232Th) turns outto be of general concern, since this can result in significant electron-holepair generation in active device regions with subsequent "upsets" or evenpermanent damage to devices.
As a result, the sum total concentration ofthese two elements is typically restricted to < 1 ppb (0.0000001%) in PVDtargets intended for metallization.Representative analysis of a high-purity 4N Ti target (Vacutec TM target from Atramet,Inc., Farmingdale, NY).FIG. 11.4SPUTTERIKG TARGETS389In the case of high-purity A1 targets, the dominant impurities tend to belight elements such as H, C, and 0. with oxygen of particular concern sinceeven trace amounts of oxidants (e.g., H,O partial pressure > lo-' Torr)can retard processes such a s high-tempeiature A1 reflow [11.6].Assumingas a worst case that a11 of the impurity content in a 5 N 5 A1 target is due to0 , does this amount of sputtered 0 pose a problem? The deposition ratefrom an A1 target is typically .= 1 pm/rnin = 160 A/sec. Assuming that therelative sputtered flux of 0 and Al from the target are not too differentfrom their relative bulk concentration leads to an oxygen "deposition rate"at the wafer of about 160 A/sec x 5 ppm = 8 x lo-' A/sec.
This degreeof oxygen bombardment corresponds to an oxygen partial pressure in thechamber of = 8 X 10-'" Torr, which is 100 times lower than the partialpressure needed to affect A1 reftow. Of course, the target will adsorb moisture, nitrogen. hydrocarbons. etc. between the time it is certified and thetime when a PVD film is deposited.
Proper target burn-in is therefore essential to recovering the low contamination level of the as-manufacturedtarget. Also. reducing sources of contamination such as chamber outgassing. permeation of elastomer-seals, etc. may be a better return on investment than going to a higher purity target.With regard to chamber outgassing, i t is worth noting that achieving alow base pressure greatly reduces the impact o f surface oxidation and/orcontamination during the time of the PVD deposition.
For purposes of illustration. assume that a11 of the chamber outgassing at a base pressure of10 Torr is potential contamination and has a sticking coefficient at thesurface of 1.0. At base pressure. about 0.5 monolayer of "crud" will form inabout I sec - this formation time being inversely proportional to pressure.Assuming a deposition rate for A1 of 1 pm/min ( = 67 monolayers/sec) andthat the Ar working gas introduces no impurities of its own, then the residual gas contamination present during deposition will lead to an impuritylevel in the Al film of = 0.5167 = 0.75%. By reducing the base pressure inthe PVD module to, say. 10-Vorr, through better vacuum practices and improved vacuum hardware.
the contamination s o introduced into the film= 100 - to 0.0075% orwill be reduced by a factor of I W hTorr /I 0"orr75 ppm."11.6 Target UtilizationA target is a consumable item, and how effectively it is utilized duringPVD has a strong impact on cost-of-ownership. In this discussion, we define target utili~atiotzas the percent of starting target material that is usedR. POWELL AND S. M.
ROSSNAGEL390up before the target must be replaced at its "end-of-life." There are otherways of defining or discussing target utilization, such as the percent of material eroded from the target that is actually deposited on the wafer; however, as a practical matter the weight of target material remaining at theend of its useful life is easy to quantify. Target utilization of 100% is neverachieved in practice; nevertheless, high-purity targets can be rather costly,and changing targets too frequently affects tool productivity. Therefore,one would like both high target utilization and long target life ~ whichturn out to be interrelated issues, as shown in Fig.
11.5.When one considers the angular spread of the sputtered material and thefinite solid angle subtended by the target at the wafer, it becomes clear thatuniform target erosion cannot produce uniform thin film deposition. Forexample, the center of the wafer "sees" a greater amount of target materialthan does the edge of the wafer, so that a uniformly eroded target wouldproduce a deposition profile thicker at the center. Therefore, the magnetarray of the DC magnetron source is designed to produce a nonuniformerosion profile across the target that compensates for the target-to-wafergeometry and takes into account such process-related variables as gasphase scattering.
For example, in some cathodes, a radially symmetric"W"-shaped erosion profile is utilized having relatively greater erosionnear the edge (Fig. 11.6), whereas others utilize multiple, concentric etchtracks.As a planar target is sputtered, the nonuniformity of the erosion profileincreases because the parallel component of the magnetic field at the target surface (which determines the local plasma density and therefore theFIG. 11.5Nonuniform target erosion can be used to produce more uniform films, but this has a neg-ative effect on target utilization.SPUTTERING TARGETS391FIG. 11.6 Obtaining an optimum balance between film uniformity and target utilization has led tothe use of tailored target erosion profiles in both circular and rectangular planar magnetrons (courtesyof Sierra Applied Sciences, Boulder, CO).local Ar + flux) increases as the eroded target surface gets closer to themagnets behind the backing plate.
Target life is then determined by thelesser of two times: (1) the time it takes the fastest eroding spot on the target to hit the backing plate, or (2) the time it takes the target profile tohave evolved to the point that a key film property such as nonuniformityof thickness or step coverage can no longer be maintained. Advanced planar magnetron sources (c. 1995) designed for 200-mm wafer coating andfilm uniformity of 3o- ~ 5% are capable of depositing > 6000/xm of AIwith target utilization > 50%. Since AI films are typically ~ 0.8-1-/zmthick, the number of processed wafers through AI target life would thenbe ~ 6000-9000.
In general, target life and film uniformity have an inverse correlation, with higher film uniformity requiring a less uniformerosion profile, leading to lower target utilization. For example, target lifefor 3o- = 3% film deposition could be 2 times less than for 3 o = 5%.Therefore, whenever comparing target life of different PVD sources, thesame film uniformity should be used. Similarly, since film thickness canchange strongly in the vicinity of the wafer edge, the number of wafersduring target life that meet a given film uniformity spec will depend onthe specific edge exclusion used when measuring that uniformity.392R.
POWELLAND S. M. ROSSNAGELOnce the target has reached end-of-life and is removed. recycling of thespent target material and/or reuse of the target assembly with a new targetis sometimes used to further reduce costs.11.7 Microstructural EngineeringTarget microstructure can have a direct influence on both the uniformity andquality of sputtered films. Thus target suppliers continue to engineer thepreferred crystallographic orientation (texture), grain size, and grain distribution of the target to produce optimum sputtered film properties, improverepeatability, and reduce particle generation [ l 1.7-1 1.101.The general trendin advanced microelectronic applications of PVD is toward homogeneous,fine-grained sputter targets with random crystallographic orientation.Cry.~tullogruphicOrientation Sputtered atoms tend to leave the surface of a single-crystal target (or a single microcrystal in a polycrystallinetarget) preferentially atong close-packed or nearest-neighbor directionsthat correspond to high-atomic density (see Fig.
![](https://s.studizba.com/z.php?f=/templates/si/i/noavatar.png&w=100&h=100&t=1"){kind=avatar}
