Roland A. - PVD for microelectronics (779636), страница 53
Текст из файла (страница 53)
Replacing Wwith AI or Cu will allow one to take advantage of planarized processes(e.g., the two-step process for A1 and dual-damascene process for Cu) sothat the same metal can be used in both the plug and the line. In spite of285286R. POWELL AND S. M. ROSSNAGELFIG. 9.1 Selected material properties of key elements encountered in PVD for microelectronicapplications.FIG. 9.2 Illustration of the application of PVD films in a three-level metallization scheme of a 0.5/.tin very large scale integrated (VLSI) device [9.1 ]. The wiring is characterized by CVD W plugs withPVD Ti/TiN liners and PVD slab A1 lines (i.e., PVD AISiCu or AICu clad with Ti/TiN).
The interlayerdielectrics are based on a combination of spin-on glass (SOG) and thermal or plasma CVD oxides.PVD MATERIALS AND PROCESSESFIG. 9.3287Illustration of the application of PVD films in an ultra-large scale integrated (ULSI) device.encouraging results for hot and/or high-pressure PVD processing (seeChapter 7) and ionized PVD (see Chapter 8), it is an open questionwhether PVD, CVD, or a combined CVD/PVD approach will be used forfilling these high aspect ratio plugs.
Therefore, some comments are included on the integration of PVD and CVD. Finally, this chapter includesa brief discussion of PVD issues of refractory alloys (TiW), refractorymetal silicides (e.g., MoSi 2, WSi 2, TiSi 2 and CoSi2), and the use of PVDin back-end-of-line bonding applications.9.2 MetrologyThe science and technology of thin film metrology for microelectronicshas developed enormously over the past 20 years, driven by the need ofresearch scientists for increasingly sensitive surface analytical tools andby the need of IC technologists to measure key thin film properties withdevice-scale spatial resolution and to map these properties over largediameter wafers [9.2-9.6]. As a result, commercial equipment or analytical services are now available to measure and map critical electrical, mechanical, and optical properties of PVD films, including film thickness,288R.
POWELL AND S. M. ROSSNAGELchemical composition and purity, surface roughness, grain size distribution and orientation, step coverage, electrical resistivity, optical reflectivity, stress, and the size distribution and composition of fine particlesadded by the process or by mechanical handling. In addition, noncontact,nondestructive methods are being developed to measure film propertieson actual product wafers, thereby reducing costs associated with testw a f e r s - a particular issue for 300-mm technology due to the excessivecost per wafer (> $1000). For example, a novel "laser sonar" methodbased on picosecond ultrasonic laser (PULSE) technology has been developed that can simultaneously measure the thickness of a multilayermetal film stack (e.g., TiN/Ti/A1Cu/TiN/Ti/Si02/Si) with high spatial resolution (20 ~m) and sub-Angstrom precision over a wide range of filmthickness of ~ 20 A-5/xm [9.7].There are literally hundreds of analytical methods that can be used tocharacterize PVD films used in microelectronics; however, in the contextof IC production only about a dozen are routinely used for process qualification or failure analysis: (1) full-wafer mapping of electrical sheet resistance (Rs) with a four-point probe or noncontact eddy current method;(2-4) thickness mapping by physical profilometry and, more recently, byX-ray fluorescence (XRF) and thermal-wave methods; (5) microscopiccross-sectional imaging of contacts, vias, and interconnects obtained by acombination of sample cleaving/polishing and secondary electron microscopy ( S E M ) m often with high-resolution field emission (FE) electron sources and elemental information provided by energy dispersive Xray (EDX) analysis; (6-9) surface and in-depth elemental and chemicalanalysis by a complementary combination of the "Big Four" methodsauger electron spectroscopy (AES), secondary ion mass spectrometry(SIMS), X-ray photoemission spectroscopy (XPS or ESCA), and Rutherford backscattering spectroscopy (RBS); (10) particle detection and mapping by laser light scattering; (11) optical properties by reflectometry; (12)crystal structure and orientation by X-ray diffraction (XRD); (13) filmstress by laser reflection; and (14) surface roughness by atomic force microscopy (AFM).
Ultimately, of course, it is device performance and reliability that determines the quality of a PVD film for a microelectronic application. Figure 9.4 shows representative probing areas and detectionsensitivity for several of the principal surface-sensitive analytical tools.In the context of IC production, the most routinely measured PVD metalfilm properties are probably resistivity, thickness, and morphology (e.g.,step coverage of a via).
In this regard, a widely used measurement for PVDequipment qualification is the uniformity of sheet resistance for an unpatterned (i.e., a blanket) PVD film. This uniformity is often specified by thePVD MATERIALS AND PROCESSES289Analytical sensitivity and probing depth of common surface-sensitive tools used in PVDmetrology (courtesy of Charles Evans & Associates, Redwood City, CA).FIG. 9.4supplier in the statistically based unit of standard deviation, sigma or o..For example, the R uniformity of a l-/zm A1 alloy film on a 200-mm waferwith edge exclusion of 3 mm might be given as 3 o = 5% or, equivalently,as 3 o = +_5%.
For a statistically normal, bell-shaped distribution, thiswould imply that about 99.7% of the R data points lie within a range extending from 5% below the mean to 5% above (see Fig. 9.5). Unfortunately, the distribution of thickness R of a PVD film over a wafer is notdominated by random events like the tossing of a coin, but often has obvious patterns (e.g., a W-shaped profile) that reflect the design of the magnetron source and chamber geometry. In addition, rarely are more than100 points collected in routine sheet resistance mapping, so that talkingabout 99.7% of the data points is meaningless unless more than 1000points were collected. As a result, a more realistic way of reporting PVDfilm uniformity is to take the m a x i m u m (M) and m i n i m u m (m) valuesfrom the data set and report the ratio of the data range (M - m) to its sum(M + m).
Using this " m a x - m i n " notation, a film might be stated to have(M - m)/(M + m) - 5% or, equivalently, (M - m)/(M + m) = _+5%,290R. POWELL AND S. M. ROSSNAGEL(Max - Min)....~.X%.._m,..- I(Max : MIn)/2~ I'IflX ---- 100 (Max-Min) /!-15t!Ir-IO. . . . .
. . .-5t<Rs > - x%::::;-::r,9!I!i1~r< Its > + x%FIG. 9.5Comparison of a gaussian distribution with the measured statistical distribution of sheet resistance for a PVD A! film.w h i c h m e a n s that all o f the d a t a p o i n t s lie w i t h i n 5 % b e l o w the m e a n v a l u eo f (M + m ) / 2 to 5 % a b o v e . A s an e m p i r i c a l r u l e o f t h u m b , a n d a s s u m i n gthat o n e c o u l d in fact t r e a t the P V D d i s t r i b u t i o n as n o r m a l , the m a x / m i nr a t i o o f P V D f i l m s t y p i c a l l y has a v a l u e b e t w e e n 2 o a n d 3 o .
E v e n t h o u g hthe d i s t r i b u t i o n o f P V D f i l m t h i c k n e s s R s o v e r the w a f e r is not n o r m a l , rand o m s t a t i s t i c a l p r o c e s s e s m a y b e the d e t e r m i n i n g e f f e c t on the repeatability o f u n i f o r m i t y . In this c a s e , n o r m a l s t a t i s t i c a l n o t a t i o n is a p p r o p r i a t e .The sheet resistance R of a thin film of thickness t is often referred to as"sheet rho " and sometimes improperly called "sheet resistivity." While theGreek symbol rho (p) is used to denote resistivity that has CGS units ofohm-cm, sheet resistance R s = p/t has units of ohm/square, also written asl~/sq, or 1~/I--1.
So "sheet rho" is an oxymoron. By way of illustration, if an8000-/~ AICu film of bulk resistivity p = 3.0/~l~-cm is deposited onto SiO 2,the measured sheet resistance of the film is given by R = p/t = (3.0 • 10 -6D,-cm)/(8 • 10 -5 c m ) = 3.8 • 10 -2 Ddsq.PVD MATERIALS AND PROCESSES291The repeatability of a statistically variable parameter is often expressed in terms of a dimensionless quantity C - - or C, in the most gen9 .pKeral c a s e - which is referred to as the proces~ capablhty index or manufacturability. C is of particular importance to the production use ofsemiconductor ]aardware (including PVD tools) and gives informationabout the relationship between design tolerance and process width.
In particular, C is defined as the ratio of design tolerance (i.e., the spread between upper and lower specified control limits) to the process width (maxmin, 6o-, etc.). A high value of C means that the process is tight and thatstatistical variations are unlikely to produce a defective, out-of-spec product. For example, assume that the thickness of a desired PVD A1 film is targeted to be 1.0/xm, but might be acceptable if its thickness were no greaterthan 1.1/xm and no less than 0.9/xm - - a specified control limit of + 10%.If the wafer-to-wafer repeatability of this deposition on a given PVD toolhas a standard deviation of l o = 3% - 30 nm, then the process capabilityis calculated to be C, = (1.1 /xm - 0.9/xm)/(6 • 30 nm) = 1.1, where aprocess width of 6o- was chosen. For state-of-the-art PVD tools andprocesses, one desires C -> 2, in which case only about 1 film in 106 willPbe outside of the specified control limits.
Part-per-million levels of defective parts in semiconductor fabrication was a concept pioneered byMotorola and is referred to as a "six-sigma" or "zero-defects" quality control methodology.While the uniformity of blanket PVD film thickness or sheet resistanceare often used for process or equipment qualification, it is important tonote that there are a number of other PVD "uniformities" that impact device performance and whose distribution can be quite different from thatof blanket thickness; these include bottom coverage, sidewall coverage,and film composition. For example, early-generation magnetrons (e.g., theCon-Mag TM from Varian) gave extremely high blanket uniformity, but theirsidewall coverage in high aspect ratio vias was not nearly as uniform. Also,differences in the sputtered angular distributions and gas-phase scatteringof an AICu(I%) alloy's component elements may lead to highly nonuniform Cu distribution from center to edge, even though the film thicknessand resistivity may be much more uniform.Another metrology issue relates to the sheet resistance of extremely thinPVD films (<< 500 ~ ) such as Ti and TiN used for contacts, barriers, andadhesion layers.
If the film thickness and/or polycrystalline grain size aresmaller than the electron mean free path, then scattering of conductionelectrons at free surfaces and grain boundaries adds to the intrinsic resistivity. The resistivity p is found to have a dependence on film thickness tof the form p = P0 (1 + aA/t), where P0 is the resistivity inside the crystalR. P O W E L L292A N D S. M . R O S S N A G E Lgrains, A is the mean free path of conduction electrons in the film, and adepends on the grain boundary scattering cross section and the density ofgrain boundaries [9.8]. Therefore, even though the intrinsic bulk resistivity of a thick polycrystalline film of TiN may be P0 ~ 50/zD,-cm, the actual resistivity can be greater for very thin films, leading one to underestimate their thickness from a sheet resistance measurement.
![](https://s.studizba.com/z.php?f=/templates/si/i/noavatar.png&w=100&h=100&t=1"){kind=avatar}
