SEMICONDUCTOR APPLICATIONS

Reduction of TiN Nodule Density Through Optimization
of Cathode and Process Variables

Ching-Ping Fang, Arnold Aronson, Corey A. Weiss and Thomas J. Licata
Praxair Materials Research, Semiconductor Equipment Divivsion, Phoenix, AZ

Suresh Annavarapu and Paul Gilman
Praxair Materials Research, Advanced Materials Division, Orangeburg, NY

Yoshiyuki Kawana
Nihon Materials Research Corporation, Tokyo, Japan

ABSTRACT
It is well known that PVD TiN processes can generate particles which reduce device yield. One source of particles is the titanium target surface, which when sputtered reactively in N2, can exhibit TiN nodules up to several millimeters in size. In this paper we report the results of investigations on TiN nodule formation and methods for nodule suppression. The microstructure and areal density of nodules were evaluated under several DC magnetron and target conditions. In particular, DC magnetrons designed to produce various degrees of target erosion uniformity were evaluated, as were Ti targets of various grain sizes, oxygen concentrations and geometries. A standard nitride mode production process of 8 kW and 4 mT was used. Surfaces of the 295-mm planar targets were inspected at 200, 400 and 700 kWh. At 700 kWh, target erosion profiles were measured, and the areal densities of the nodules were characterized. The results show that variations in target composition do not appreciably change the nodule distribution. However, nodule distributions correlate strongly with the uniformity and magnitudes of erosion provided by the magnetron design. A more uniform erosion significantly reduces nodules under our process conditions and is achieved by optimizing the magnet pack and target geometry. Similarly, periodic high power processing that "burnt-off" the nitrided surface layer of the target in an Ar ambient can be used with a less uniformly eroding magnetron to suppress nodule densities to a certain level. In this process, a 1 mm Ti layer which inhibits particle delamination is deposited periodically on non-product wafers.

Introduction
In integrated circuit (IC) device manufacturing, particle contamination on semiconductor wafer surfaces is a major factor that reduces device yield and performance. The advent of state-of-the-art cleanrooms and integrated vacuum processing tools has dramatically reduced in-film particulate contamination. Processing tools, therefore, are the dominant sources of particles in thin film processing. Particles may be deposited on wafer surfaces during any type of process1-5. In PVD film deposition, investigations have shown that particle contamination due to homogeneous nucleation is negligible because of the low pressure typically used (1-20mT). Instead, most particle problems in sputtering processes are attributed to flaking of the deposited material from the chamber wall and from the target during sputtering1,6.

TiN films are important elements in metal interconnects, since they function as diffusion barries, adhesion layers, and antireflective coatings. TiN thin films are produced by reactively sputtering a Ti target in a Ar-N2 ambient. Typical processing pressures are in range of 2 to 8 mTorr with DC magnetron powers between 2 to 10 kW. However, it has been observed that in Ti-based PVD processes, such as W-Ti and TiN, nodules form on target surfaces in weakly eroded areas. The nodule formation mechanism is still under investigation, but the evidence of nodule concentrating in weakly eroded area suggests that physical vapor redeposition is a key factor. To verify the effects of controllable equipment and process variables on the growth and suppression of nodules on target surface during TiN processing, a series of experiments and analyses, described below, have been performed.

Experiments
The experiments were performed in a high vacuum (10-8 Torr base pressure) sputtering chamber. The common process parameters are listed in Table 1.

Table 1. Common Process Parameters
 

Power (kW)

Voltage (V)

Current (A)

Gas flow (SCCM)

Pressure (mT)

Ar

N2

TiN Process

8

625

12.8

41

42

3.5

Target denitriding

8

673

12

41

0

2.7

A typical process consisted of 33 minutes TiN deposition and 100 seconds of Ti processing to denitride the target, simulating 100 wafers of TiN deposition and 5 wafers of Ti processing, respectively. The total processing time for each target before ex-situ analysis was 700 kWh. The nodule densities and distributions on each target were measured and observed at 200 kWh, 400 kWh and 700 kWh. Nodules with diameters greater than 0.2 mm were sampled under a Emerson contact reticle with hole gauges. Sampling fields were blindly chosen from different areas on the target surfaces to prevent biased field selection. A minimum of 30 fields was measured in each area.

Seven variables are highlighted in Table 2. Target IV had lower oxygen concentration. Target II consisting of fine grains (mean grain size = 9 mm) was contrasted with the standard microstructure (mean grain size = 40 mm). Target VI, beveled in a convex fashion (U.S. patent) was included to eliminate line of sight redeposition from the target center. For this target, the ICC magnetron designed for uniform erosion had to be used to obtain stable operation at the standard cathode power. A pulsed power supply was examined with Target V, since previous work indicates this mode of sputtering should reduce the number of non-conductive sites on the target surface that may nucleate nodules. The pulsed supply was operated at a frequency of 100 kHz with from +100 V to -625 V and a 20% of duty cycle. Finally, a high power denitriding process was tested to understand whether nodule formation could be suppressed in a quick and practical fashion.

Table 2 Testing Matrix
  Oxygen Grain Shape Magnetron Power
Target I 450 ppm Standard Standard RMX 12 Standard
Target II 450 ppm Finer Standard RMX 12 Standard
Target III 450 ppm Standard Standard ICC 12 Standard
Target IV 250 ppm Standard Standard RMX 12 Standard
Target V 450 ppm Standard Standard RMX 12 Pulsed
Target VI 450 ppm Standard Beveled ICC 12 Standard
Target VII 450 ppm Standard Standard RMX 12 High power denitriding

click for larger image
Figure 1 Development of nodules on target surface

Results
The development of nodules on a target surface during sputtering in a typical case, Target IV in Table 2, is shown in Figure 1. The formation of nodules started slowly at the beginning, and then developed rapidly even with target denitriding. The nodule density at the edge area of target IV increased from 4.8/cm2 at 200 kWh to 158.5/cm2 at 700 kWh.

click for larger image
Figure 2a
click for larger image
Figure 2b,c,d
click for larger image
Figure 2e Target profiles and nodule densities.

The erosion profiles and nodule densities after 700 kWh processing are expressed in Figure 2. The results from the target V sputtered with pulsed DC power were not plotted in Figure 2 and statistically analyzed, because a tremendous number of nodules on the surface after 700 kWh sputtering made statistic analysis and profile measurement impossible. The result of the high power denitriding (Target VII) was also excluded in Figure 2 due to its different sputtering life (400 kWh) and sputtering chamber. The erosion profiles in Figure 2 can be considered as the erosion intensity of each case, since the same sputtering life was used. The erosion due to DC magnetron processing is typically weak at both the center and the peripheral areas. The nodule densities were observed to be higher at those weaker erosion areas. Target II with finer grain size and target IV with low oxygen (250 ppm) did not result in lower nodule densities. The erosion profile from the ICC magnetron (target III) showed more uniform flat shape and stronger overall erosion than the results from the RMX 12, due to the magnet design. Accordingly, nodules were found in the center and the edge area, but were rare in other areas. Target VI was beveled 15 mm from the target edge and at 10o where nodules have the highest density under the normal operation. The height of the dark space shield was adjusted accordingly. Under these conditions, the nodule density decreased to less than 10/cm2 in a narrow peripheral area of 4 mm from the edge, while the nodule density was greater than 300/ cm2 in the same area of the target III where only ICC magnetron was tested. It is likely that this reduction in nodule density would provide lower particle levels for IC processing, through this has not been attempted in this study.

In the high power denitride testing, Target VII was sputtered in our testing stand with a repetitive process of 33 min TiN followed by 100 seconds of 10 kW denitriding. After 400 kWh, no significant improvement was observed. The target was, then, moved to our Mark IV production tool, which has smaller process chamber than the testing stand, and pasted with Ti for 100 seconds at 12 kW. The comparison of nodule density before and after 12 kW Ti pasting are tabulated in Table 3. The suppression of nodule formation with high power pasting is moderate. The smaller nodules may be eliminated in the pasting, while most existing nodules remain unchanged.

Table 3. Effect of the high power pasting on nodule density
  Nodule density, #/cm2

Area in radius

Before 12 kWh pasting

After 12 kWh pasting

r<1.0 cm

12.1

11.1

1.0 cm<r<5.5 cm

6.6

4.8

5.5 cm<r<11 cm

5.5

5.2

11 cm<r<13.5 cm

0

0

13.5 cm<r

71

60.5

Conclusions
To verify the effect of various factors on nodule formation and suppression during TiN reactively sputtering process, a series of experiments were performed under carefully controlled conditions. The comparison experiments show that variations in target oxygen and grain size do not appreciably change the nodule distribution. However, nodule distributions correlated strongly with the erosion uniformity and magnitudes. A more uniform erosion significantly reduces nodules under our process conditions and has been achieved by optimizing the magnet pack and target geometry. This solution is compatible with large scale manufacturing of ICs. Similarly, improvements in nodule density obtained from standard hardware result from a properly chosen target denitriding process.

References

  1. J.S. Logan and J.J. McGill, J. Vac. Sci. Technol. A. p. 1875 (1992).
  2. J.F. O’Hanlon and H.G. Parks, J. Vac. Sci. Technol. A. p. 1863 (1992).
  3. M. Itano, Particle on Surface, Vol. 3, edited by K. L. Mittal, Plenum Press, p. 35. (1991)
  4. C.A. Weiss, A. Ghanbari and G.S. Selwyn, 12th International VLSI Multilevel Interconnection Conference, June 27, p. 412 (1995).
  5. C.W. Teutsch, B. Miller and C. Fournier, Particle on Surface, Vol. 3, edited by K.L. Mittal, Plenum Press, p 173 (1991).
  6. G.S. Selwyn, C.A. Weiss, F. Sequeda and C. Huang, particle Contamination and Detection in Magnetron Sputtering Processes, J. Vac. Sci. Technol. in print.
   
 

CCH Process
CCH Do Aluminum Alloys
CCH Do and Do* Ti
Co and Ni
Ta and Cu

Thin Films Application Lab
Bonding Technology
Novellus Inova™ Tool
Endura™ & Centura™ Systems
RE-Al PLUS": Sputter Targets
Technical Papers  
  1. Reduction of TiN Nodule Density Through Optimization of Cathode and Process Variables
  2. Effect of Oxygen Content on Particle Generation in TiN Reactive Sputtering
  3. Link From Titanium Target Orientation to Titanium/Titanium Nitride/Aluminum Thin Film Orientations
  4. Technical Papers Available in PDF