Roland A. - PVD for microelectronics (779636), страница 58
Текст из файла (страница 58)
This relates to the fact that obtaining uniformlythick PVD films generally involves tailoring the target erosion profile tocompensate for the finite geometric size of the PVD source. Unfortunately,FIG. 9.21 S E M micrograph showing the step coverage of a collimated PVD TiN film in a steep contact hole (aspect ratio of collimator = 1.5" 1" aspect ratio of hole = 8:1).PVD MATERIALS AND PROCESSES313high aspect ratio collimator cells tend to image the nonuniform erosionprofile of the target onto the wafer, and sources with extremely uniformerosion are difficult to design.
Therefore, although blanket Ti nonuniformity of 30" = 3-5% over 200 mm is typical of state-of-the-art magnetronsused without collimation, 30" values of ~ 10% are more typical of highlycollimated Ti processing. On the other hand, since Ti films in microelectronics are usually thinner than 300 A, a 30" = 10% value represents a variation of only about 30/~, or about 10 Ti atoms.While collimation improves directionality, it also reduces the specificdeposition rate of Ti at the wafer by removing low-angle material from thesputtered flux; to compensate for this effect, higher magnetron power isused.
For example, while noncollimated Ti might be deposited at ~ 1-2kW, a 1:1 or 1.5:1 collimated deposition might require 5-10 kW for equivalent throughput. Figure 9.22a shows the field thickness obtained whentrying to obtain a 65-A film of Ti at the bottom of a contact with noncollimated, 1:1, and 1.5:1 aspect ratio collimation, and Fig. 9.22b shows thenumber of such films that can be deposited before having to change thecollimator or target (the collimator is changed when buildup of Ti on thecell walls reduces transmission by 50%). Collimation clearly reduces deposition on the field and, in spite of impact on absolute deposition rate,still allows a rather large number of wafers to be processed.Collimation also can impact the microstructure developed in the Ti film.In general, columnar growth arises in PVD films due to limited surface diffusion and competition or shadowing between columns.
The surface diffusion length depends on both substrate temperature and the presence of contamination, while the shadowing is a result of the surface topolology.Given the limited surface mobility of Ti at PVD temperatures and the factthat collimation removes obliquely incident adatoms from the incidentflux, it is not surprising that collimated Ti films have a dense columnar microstructure on the field regions. On the other hand, collimated coatings onvertical sidewalls of high aspect ratio features can have reduced densityand increased porosity due to shadowing of the highly directional Ti fluxby the growing Ti grains [9.20, 9.21].9.5 Titanium Nitride9.5.1 METALLURGICALISSUES FOR PVDWhile the applications for PVD Ti and TiN can be quite different (e.g., Ticontact layers and TiN ARC layers), it is difficult to separate the two materials in a PVD context since TiN is deposited by reactive sputtering of a314R.
POWELL AND S. M. ROSSNAGELFIG. 9.22 (a) Thickness of Ti that must be deposited on the field to obtain 65 ]k of Ti on the bottomof a contact hole for different hole and collimator aspect ratios.(b) Number of Ti films obtained before the end of collimator life (defined as point where transmission of Ti flux through the collimatorhas dropped to 50% of the value when new).Ti target in a nitrogen-containing ambient, typically Ar/N 2. Also, there aremany cases where the complementary cleaning properties of Ti and barrierproperties of TiN favor their use as a Ti/TiN bilayer.
Therefore, much ofthe data used in this section will include both Ti and TiN.PVD MATERIALSAND PROCESSES315The major use of PVD TiN is as a barrier layer, e.g., to prevent diffusionof an A1 or W contact plug metallurgy into the underlying Si substrate.
Forexample, TiN prevents the interdiffusion of Si and A1 at the contact level,which could lead to junction spiking. With regard to CVD W contact andvia plugs, TiN is also widely used as a "glue" layer to promote the adhesion of W to the oxide walls and to help it n u c l e a t e - although in manycases, the TiN is deposited over an intermediate Ti layer that makes the actual bond to the oxide (this glue layer is not required in the case of A1,which adheres well to SiO2). Since the WF 6 precursor commonly used forCVD W reacts strongly with Ti, the TiN also serves as a protective coatingfor the underlying Ti.
However, if this TiN coating has any breaks or delaminations, the volatile reaction of WF 6 + Ti to form TiF 4 combined withthe deposition of CVD W on the peeled-back TiN can give rise to a dramatic defect resembling a miniature volcano (Fig. 9.23). Attack of the underlying Si by the WF 6 is also possible, leading to the subsurface migration of W into the Si and giving rise to a wormhole-shaped structuraldefect. Whether used for barrier, adhesion, or protection purposes, PVDTiN should be pinhole-free and as conformal as possible, particularly atsharp bottom corners where PVD coverage can be reduced and give rise toweak spots as shown in Fig. 9.18.
It should also be noted that different applications require different thickness. For example, while a 50-100-/~-thickFIG. 9.23 "Volcano" defect that is formed by chemical reaction of Ti with the WF 6 chemistry usedin CVD of W. (Reprinted from S. Bothra et al in the February 1997 edition of Solid State Technology,copyright 1997 by PennWell.)316R. POWELL AND S. M.
ROSSNAGELTiN film might suffice as a glue layer for CVD W, a 250-400-/~ film maybe required as the contact diffusion barrier for a high-temperature A1-Cualloy reflow or two-step process.As the atomic concentration of N in Ti is increased, the resulting material evolves from pure Ti, to a solid solution of N in Ti, to the compoundTizN (33% N), and finally to TiN (50% N). At concentrations above 50 at%, the excess N exists in solid solution with stoichiometric TiN.
TiN canaccept large vacancy fractions on both the anion and cation sublattices, andover-stoichiometric TiN x (x > 1) remains single phase in the NaC1 structure with excess nitrogen fractions up to about x - 1.2. But it is the 1"1 stoichiometric TiN phase that is preferred over other compositions due to itssuperior barrier properties and that is readily deposited by reactive sputtering of Ti in Ar/N 2. However, treatments to enhance as-deposited TiNbarrier performance m such as air exposure or in-situ or ex-situ annealingin an oxidizing ambient to "stuff' the grain b o u n d a r i e s - are often donefollowing PVD. While PVD Ti is silver colored, stoichiometric TiN has acharacteristic gold or brownish-gold color under reflected light. This hasled to "gold TIN" sometimes being used as an indicator of 1:1 film composition, although in reality the perceived color depends on both stoichiometry and other film properties in a complex way [9.22, 9.23].It is a popular misconception that TiN is a metal.
However, even though TiNfilms are gold and shiny with electrical conductivity comparable to that of titanium, TiN is not a metal. The high conductivity is associated with a strongoverlap of N 2p and Ti 3d bands, while the gold color arises from interbandtransitions combined with a high reflectance in the red and infrared. [9.22].9.5.2 REACTIVEPVD OF TINTiN is deposited by reactive sputter deposition of a Ti target in the presence of nitrogen, typically by using an Ar/N z admixture. The kinetics ofthe resulting PVD TiN film formation depend on process and hardwareparameters in an interactive way (e.g., N 2 partial pressure, magnetronpower, collimator, and PVD shield design), which has important practicalconsequences.
The basic issue relates to nitridation of the Ti t a r g e t - inparticular, one wants to minimize nitridation of the target surface to increase the sputtered flux of Ti atoms, yet at the same time maximize nitridation of Ti at the wafer surface to produce a stoichiometric TiN film(see also Chapter 3 for a discussion of reactive PVD).PVD MATERIALS AND PROCESSES317The steady-state condition of the Ti target surface during PVD can rangefrom fully metallic to fully nitrided, with the exact ratio of exposed Ti andTiN areas reflecting the detailed consumption and liberation of nitrogen atthe surface - - e.g., gas-phase nitrogen is consumed by the reaction of molecular N 2 with Ti (2Ti + N 2 = 2TIN) while nitrogen bound as TiN is liberated by Ar § or N 2+ ion bombardment, etc.
Regardless of the state of target nitridation and contrary to what one would expect, the primarysputter-ejected particles from a Ti target in an Ar/N 2 discharge are alwaysTi and N [9.24]. That is, even when the target surface is fully nitrided,sputter ejection of molecular TiN is not significant. On the other hand, thesputter yield of Ti from TiN is several ( ~ 3) times less than from Ti, sothe ejected flux of Ti atoms is much less from the nitrided target. Gasphase recombination of the sputtered Ti with nitrogen via a two-body collision does not occur since the heat liberated in the formation of a molecule of TiN cannot be dissipated while simultaneously conserving energyand momentum.
Instead, this occurs at the wafer surface where the sputterdeposited Ti adatoms are nitrided to TiN, primarily by heterogeneous reactions such as the dissociative chemisorption of N 2. In effect, both the Titarget and the wafer act as solid state "pumps" of nitrogen, whose relativepumping speeds reflect their state of nitridation and affect the overall TiN,deposition rate. In addition, the nitridation state of target and wafer are influenced by PVD shields and collimators, which become coated with Tiand themselves behave as dynamic getter pumps for nitrogen.
![](https://s.studizba.com/z.php?f=/templates/si/i/noavatar.png&w=100&h=100&t=1"){kind=avatar}
