For semiconductor PVD, 4N (99.99%), 5N (99.999%), and 6N (99.9999%) designations correspond to a roughly tenfold drop in metallic impurity content at each step, and that ratio, not the label, is what governs whether a film passes a 300 mm wafer's Rs uniformity test [S4][S5].
The grade you specify is dictated by the film function: barrier layers and gate electrodes tolerate 4N5 to 5N, while Cu seed layers and Al-0.5%Cu lines at sub-28 nm nodes effectively require 6N to keep mobile-ion and deep-level defect counts below the device yield cliff [S1][S3].
How Purity Grades Map to Common Semiconductor Target Materials
Suppliers publish per-material grade recommendations that converge across vendors: tantalum at greater than 99.9998% (5N8) for Cu barrier layers, copper and copper alloys at greater than 99.9999% (6N) for seed and UBM, aluminum and aluminum alloys at 99.999% (5N) for metallization, tungsten and tungsten alloys at greater than 99.995% (4N5 to 5N) for plug and electrode films, titanium at greater than 99.999% (5N) for barrier and hard-mask stacks, and nickel and nickel alloys at greater than 99.99% (4N) for less contamination-sensitive layers [S1][S3].
Silicon targets follow the same ladder, with 5N (99.999%) as the entry tier for general semiconductor work and 6N (99.9999%) specified for leading-edge processes where any incorporated metallic contaminant migrates as a recombination center in the active device region [S2]. The "4N5" in-between label, 99.995%, is a real commercial grade, not marketing: titanium at 4N5 is widely sold for interconnect barrier deposition, with 5N and 5N5 offered as upgrades where Rs uniformity budgets are tighter [S3].
Decision Matrix: 4N vs 5N vs 6N for Specific Films
Match grade to film function rather than chasing the highest number on the data sheet. For 99.9% (3N) the typical use is general industrial coatings, with standard film performance and relaxed contamination limits [S5]. 99.99% (4N) covers most electronics and many barrier films; cost per kilogram is materially lower, and the impurity load is acceptable when the deposited layer is not an active device region [S4][S5]. 99.999% (5N) is the workhorse for interconnect barrier, hard mask, gate electrode, and UBM stacks in mainstream logic nodes, and most Al and Ti production targets land here [S1][S3]. 99.9999% (6N) is reserved for Cu interconnect seed layers, advanced Al-0.5%Cu lines, and any film whose mobile-ion or deep-level defect budget is below 1 ppm equivalent contamination in the deposited stack [S1][S3].
Cost roughly scales with each N: in the open market, moving from 4N to 5N copper roughly doubles target price per kilogram, and 5N to 6N adds another comparable multiplier, with the spread driven by refining yield rather than machining [S1]. For 99.999%+ (5N+) semiconductor fabs, maximum Rs precision and lowest particle counts are the return on that spend, not the label itself [S5].
What "Purity" Actually Covers: Trace Elements That Decide Yield
Bulk assay is only the start; the elements that kill yield in PVD films are O, S, P, C, N, and a short list of metallic getter species, and these are typically reported as ppmw or ppbw on the target certificate rather than rolled into the N number [S3][S4]. JX's Cu-target datasheet, for example, highlights "O, S, and P" suppression as the primary particle-control mechanism for 6N Cu, which is why Cu alloy targets (CuMn at 1 at% Mn or below) are specified in addition to pure Cu for advanced seed layers [S3].
Even parts-per-million levels of certain impurities can introduce deep-level defects that compromise device performance, with the effect being non-linear: a Cu film at 99.99% bulk assay can still fail a 300 mm Rs uniformity spec if the O pickup during sputtering pushes the film resistivity above the integration budget [S4]. For this reason, leading fabs pair the purity grade with a maximum allowable O and S specification on the certificate, and they audit that number on every lot, not on a sample basis [S1][S3]. Sputter yield consistency and the formation of particulates or nodules on the target surface are tied to grain size, density, and microstructural uniformity, so two 5N targets from different suppliers can produce visibly different particle counts even with identical bulk assay [S4].
Microstructure and Density: The Hidden Half of Target Spec
Relative density close to theoretical maximum minimizes porosity and ensures a predictable erosion profile over the lifetime of the target, and is the reason high-grade targets are hot-pressed or vacuum-melted rather than cold-formed [S4]. Grain size, grain-size distribution, and texture orientation drive the sputter yield drift as the target erodes, which is why suppliers such as JX publish "uniform grain diameter" as a controlled property for 6N Cu, not just the bulk assay [S3].
Aluminum and aluminum alloys at 5N are typically specified with controlled chemistry to keep Fe, Si, and Cu within tens of ppm, because these species segregate to grain boundaries and can flake as particles during long campaigns [S1]. Tungsten and tungsten alloys at 4N5 to 5N are paired with homogeneous microstructure control in the sintering step, since W has a high melting point and refining cost is dominated by the sintering route, not bulk assay [S1][S7].
Use Cases and Limits: When 6N Is Overkill, and When 4N Is a False Economy
Specifying 6N Cu for a non-semiconductor decorative coating is wasted spend, and conversely, using 4N Ti in a sub-10 nm gate stack is a known yield risk, so the decision needs to be tied to the device node and the film function rather than the supplier's default grade [S2][S3]. For barrier films that are subsequently capped with a diffusion layer, 4N5 titanium and tantalum are routinely accepted by leading fabs, because the barrier does not carry current and its Rs uniformity tolerance is wider than that of a seed layer [S1][S3].
For Cu interconnect seed layers at 28 nm and below, 6N Cu or 6N CuMn is the practical floor; 5N Cu can be qualified for some mature 0.18-0.13 µm analog and power-device lines, where the device pitch is wide enough that the defect density per unit area is tolerable [S1][S3]. The transition from planar to rotatable targets improves material utilization by 20-30% in many production environments, and the per-kilogram purity choice stays the same, since utilization is a geometry decision, not a purity decision [S5]. Sinter-hardened powder-metal components in related tooling routes are covered separately in this encyclopedia's powder-metallurgy section, but the purity logic is the same: grade is set by the most contamination-sensitive film in the stack, not by the cheapest acceptable layer [S1].
Selection Checklist for Spec Writers
Match grade to film function first: barrier, hard mask, and gate electrode at 4N5 to 5N; interconnect seed, UBM, and advanced metallization at 5N to 6N; the active device region, where any mobile ion would be a yield killer, at 6N [S1][S2][S3]. Require a per-lot certificate with O, S, P, C, N, and key metallic getter species reported in ppmw or ppbw, not just an assay headline [S3][S4]. Specify relative density at 99% or higher of theoretical, average grain size, and grain-size distribution, since these govern particle generation and Rs drift over target life [S4]. Confirm backing-plate compatibility and bonding metallurgy before locking the grade, because a 6N target bonded with a non-conformal interface will still produce particles that mask the purity benefit [S1][S5]. Plan for grade audit at incoming inspection, and keep a qualified-bill-of-materials for at least two suppliers, since refining capacity for 6N Cu and 5N8 Ta is concentrated in a small number of producers worldwide [S1].
Trackable next signals for procurement and process engineers: monitor the published grade mix per material from the major target suppliers (Linde, JX, Honeywell, Materion, Sumitomo, and the Chinese mid-tier) for any shift in the 4N5 vs 5N offering, and watch for changes in O/S/P specification language on lot certificates, which is the leading indicator that a refining process change has occurred. The 6N Cu premium is the most volatile line item, so refresh it quarterly against spot refining quotes rather than relying on annual contracts.
Detailed specification references: sputtering target, pressure transmitter, and flow meter.
See also our earlier report, Mag-Drive vs Canned Motor Pump: Zero-Emission Spec Cutoff.