Monitoring Cu ECD using laser-based SAW measurements
10/01/2004
An optical technique based on the laser generation of surface acoustic waves allows noncontact monitoring of electroplated copper thickness on product wafers. The technique is extended to monitor variations in copper resistivity and grain size, and its capability to independently monitor within-wafer and wafer-to-wafer variations in thickness, resistivity, and grain size is demonstrated.
The microstructure of electroplated copper varies depending on deposition conditions, and undergoes considerable changes during anneal or self-anneal. It is important to ensure consistency of the microstructure because it affects electrical resistance, stress voiding, and electromigration characteristics of copper wires, as well as the removal rate in chemical mechanical polishing of interconnect structures [1, 2].
Recently developed techniques based on the laser generation of acoustic waves [3] allow noncontact, small-spot measurements on product wafers, significantly improving copper metrology. Although applications of the new techniques have targeted copper thickness measurements, previous research indicated that a technique using laser-induced gratings at the sample surface, known as SurfaceWave or impulsive stimulated thermal scattering (ISTS) [3, 4], can also be used to monitor variations in copper electrical resistivity and microstructure [5, 6]. While traditional four-point probe (4PP) measurements are also sensitive to variations of copper microstructure, their usefulness is limited by the fact that they cannot determine either film thickness or electrical resistivity independently. Surface wave measurements can determine both thickness and resistivity simultaneously, and monitor wafer-to-wafer and within-wafer variations of each parameter.
Surface wave measurement
This technique for surface acoustic wave (SAW) measurement in a small spot configuration allows user-selectable acoustic wavelength. In this method, two sub-nanosecond excitation laser pulses are crossed at the sample surface to form a spatially periodic intensity pattern with a period in the range ~2–11µm, controlled by a measurement recipe. Absorption of optical radiation results in a temperature rise and subsequent impulsive thermal expansion that launches two counterpropagating SAWs with a period equal to the intensity pattern period, as well as a thermal grating associated with the temperature profile. The time-dependent surface displacement caused by the SAWs and the thermal grating is monitored via diffraction of a probe laser beam, and signal waveforms are averaged over ~1000 excitation pulses, which takes ~1 sec. The measurement apparatus is described in greater detail elsewhere [7].
Figure 1 presents several examples of surface-wave signal waveforms. The upper portion of c) shows a signal waveform collected from a 1.5µm-thick electrochemically deposited (ECD) Cu film sample at an excitation period of 10.8µm. Fast oscillations in the signal waveform are caused by SAWs excited at a wavelength equal to the excitation period, while the slow decay of the signal baseline reveals the thermal grating relaxation caused by thermal diffusion. In a standard surface wave measurement [3, 4], the SAW frequency is measured to determine the film thickness; the thicker the film, the lower the frequency.
The decay time of the thermal grating contribution was measured by fitting the waveforms to an exponential function. In the simplest model, assuming that the grating period Λ is much greater than the film thickness and disregarding thermal conductivity of the underlying dielectric, the thermal decay time is given by [8]
 |
null
where k is the thermal conductivity, ρ is the density, cp is the specific heat of Cu, and q is the grating wavenumber equal to 2π/Λ.
In good electrical conductors such as Cu, the ratio of the thermal conductivity k to electrical conductivity σ at a given temperature T is a material-independent constant (Wiedemann-Franz law), k/σ = LT, where L is a constant called the Lorentz number. Thus, the electrical resistivity R is expected to be approximately proportional to the surface-wave thermal decay time:
 |
null
A third independent channel of information is provided by the SAW attenuation at short wavelengths. Figure 1a shows examples of waveforms collected at a SAW wavelength of 1.9µm, with a SAW frequency as high as ~1.1GHz. In this case, the decay of the peak-to-peak oscillation amplitude is due to SAW attenuation caused primarily by the grain structure of Cu (Rayleigh scattering off crystallites), and the acoustic attenuation time τA is determined by fitting a waveform to a harmonic function with exponentially decaying amplitude.
A quantitative determination of grain size from the SAW attenuation is challenging, and the theoretical models of SAW attenuation in polycrystalline materials are complex, involving multiple assumptions. In general, the acoustic damping is expected to increase as the grain size increases and becomes similar to the acoustic wavelength. The significant difference in short-wavelength damping between the two samples shown in Fig. 1 is indeed due to a difference in the microstructure (discussed later). Measuring the SAW frequency and thermal-grating decay time at the grating wavelength of 10.8µm, as well as the SAW damping rate at the wavelength of 1.9µm, provides three independent channels of information for monitoring film thickness, resistivity, and grain size.
Cu variations with plating current
A number of 200mm Si wafers with 500nm thermal oxide were coated with a 31nm Ta barrier and 100nm seed copper, then electroplated to a nominal thickness of 0.1–1.6µm at plating current densities of 3.33, 25.0, and 50.0mA/cm2 (corresponding to 1.0, 7.5, and 15A) in a Novellus Sabre Classic using Enthone Viaform high-acid plating chemistry. The wafers were self-annealed at room temperature. SurfaceWave measurements were performed with a commercially available automated instrument from Philips AMS. In addition to measuring the surface waves, sheet resistance with a 4PP tool was also obtained. Two wafers (nominal thickness 1.6µm, plating currents of 3.33 and 25.0mA/cm2) were selected for grain size measurement by orientation imaging microscopy (OIM), which employs electron backscattering in a scanning electron microscope.
 |
The table shows wafer-average Cu thickness measured with the surface wave instrument on three samples of nominal thickness = 1.6µm. As expected, the thickness of the three samples is about the same. It has been shown previously [6] that surface-wave Cu thickness measurements performed at a wavelength longer than that of the film thickness are relatively insensitive to any variations in Cu microstructure. A diameter scan of the sample deposited at 50mA/cm2 showed large local variations in thickness, which correspond to large local nonuniformities in the central area of the wafer visible with an optical microscope.
The table also shows wafer-average Cu resistivity for the same three samples, with resistivity calculated from the thermal decay time (Eqn. 2). The sample deposited with plating-current density of 3.33mA/cm2 stands out, yielding a significantly higher resistivity — a result in agreement with the 4PP sheet-resistance measurements.
The higher resistivity of the sample deposited at 3.33mA/cm2 indicates that the film did not self-anneal, which is known to be the case for Cu films electroplated at low plating current because of the higher concentration of impurities incorporated from the plating bath [9]. The origin of the higher resistivity must be in a smaller Cu grain size. This conjecture is in agreement with the fact that the attenuation of high-frequency SAWs is much smaller for this sample (Fig. 1). The table also shows the measured SAW attenuation time (i.e., the time in which the acoustic oscillations decay to the 1/e level), which is much longer for the sample deposited at 3.33mA/cm2, indicating a significantly smaller grain size.
Reference measurements of Cu microstructure were performed via OIM scans on two wafers deposited at plating current densities of 3.33 and 25mA/cm2. Figure 1b shows parts of the OIM scans for the two samples. As expected, the crystallite size is much smaller for the sample deposited at the low plating current. Analysis of the scans also shows that this sample has a stronger <111> texture. The microstructure of this sample is similar to that of freshly plated ECD Cu films, thus supporting the assumption that the sample deposited at 3.33mA/cm2 did not self-anneal.
Similar results were obtained on samples deposited to nominal thickness 1.1µm at plating current densities of 3.33 and 50mA/cm2. SurfaceWave measurements yield a higher resistivity for the sample deposited at the low plating current, which is in agreement with the 4PP measurements; attenuation of high-frequency SAWs is also significantly lower for the 3.33mA/cm2 sample.
Within-wafer variations
The samples discussed thus far all exhibited good within-wafer uniformity of the microstructure and resistivity, with both surface-wave and 4PP diameter scans appearing relatively flat. A thinner sample (nominal thickness 0.8µm) deposited at a plating-current density of 25mA/cm2 offered an example of within-wafer variations. Figure 2a shows wafer diameter profiles of the thickness and resistivity of this sample measured with the surface wave instrument. One can see that the center-to-edge variation in resistivity is larger than the thickness variation.
In order to compare the surface-wave data with the 4PP measurements, we divided the surface wave resistivity profile by the surface-wave thickness profile to calculate sheet resistance, and plotted it together with the sheet resistance measured by the 4PP, as shown in Fig. 2b. The two curves correlate well. Fluctuations in the surface-wave resistivity profile are caused by reproducible point-to-point variations rather than a random measurement error. The fact that the 4PP profile is smoother can be explained by the much larger measurement spot of the 4PP instrument.
For the sample shown in Fig. 2, measurements of both thickness and resistivity are required to understand the deposition uniformity, since thickness is very uniform, while resistivity is not. Measurements with 4PP alone would not have been able to identify this.
Conclusion
Surface wave measurements of electroplated Cu films provide three channels of information that can be used to independently monitor variations in the film thickness, resistivity, and grain size. The usefulness of this approach for detecting both wafer-to-wafer and within-wafer process variations has been demonstrated.
The approach can be further developed. The model for the resistivity calculation provided by Eqn. 2 is a rough approximation that can be improved by a more accurate analysis involving a rigorous calculation of the thermal-grating relaxation [8], or by an empirical calibration. Although the measurements were performed on blanket wafers, the small probe-laser spot size (15×30µm) permits measurements on product wafers; the next step in developing this application will be an extension of the proposed approach to measurements on damascene line-array structures.
Acknowledgments
This article is based on a presentation by the authors at the 15th IEEE/Semi Advanced Semiconductor Manufacturing Conference in Boston, May 2004, and published in the conference proceedings as "Monitoring Thickness, Resistivity and Grain Structure of Electroplated Copper Films with Laser-Based Surface Wave Metrology," Copyright IEEE, 2004. SurfaceWave is a trademark of Philips Advanced Metrology Systems, Sabre Classic is a trademark of Novellus, Viaform is a trademark of Enthone, and OIM is a trademark of EDAX-TLS.
References
- D.N. Lee, "Texture and Related Phenomena of Copper Electrodeposits," Proc. Mat. Res. Soc., Vol. 427, pp. 167–178, 1996.
- K. Mirpuri, J. Szpunar, K. Kozaczek, "Microstructure Analysis in As-Deposited and Annealed Damascene Cu Interconnects using OIM," Characterization and Metrology for ULSI Technology: 2003 Intl. Conf., D.G. Seiler, A.C. Diebold, T. J. Shaffner, R. McDonald, Z. Zollner, R.P. Khosla, E.M. Secula, eds., American Institute of Physics, Melville, NY, pp. 499–503.
- M. Gostein, et al., in Handbook of Silicon Semiconductor Metrology, A.C Diebold, ed., Marcel-Dekker, NY, pp. 167–196, 2001.
- J.A. Rogers, A.A. Maznev, K.A. Nelson, "Impulsive Stimulated Thermal Scattering," in Characterization of Materials, E.N. Kaufmann, ed., John Wiley & Sons, NY, pp. 744–759, 2003.
- A.A. Maznev, M. Gostein, J.A. Rogers, G. Kowach, "Independent Measurement of Copper Film Thickness and Resistivity by Opto-Acoustic Transient Grating Technique," Abstracts of 2001 MRS Spring Meeting, p. 233.
- A.A. Maznev, M. Gostein, S.H. Brongersma, "Characterization of Electroplated Copper Films with Laser-Generated Surface Acoustic Waves," Proc. MRS Spring Meeting, Vol. 766, pp. 415–420, 2003.
- A.A. Maznev, A. Mazurenko, L. Zhuoyun, M. Gostein, "Laser-based Surface Acoustic Wave Spectrometer for Industrial Applications," Rev. Sci. Instrum. 74, pp. 667–669, 2003.
- O.W. Kading, H. Skurk, A.A. Maznev, E. Matthias, "Transient Thermal Grating at Surfaces for Thermal Characterization of Bulk Materials and Thin Films," Applied Physics A, 61, pp. 253–261, 1995.
- S.H. Brongersma, E. Kerr, I. Vervoort, A. Saerens, K. Maex, J. Mater. Res. 17, p. 582, 2002.
For more information, contact Michael Gostein, chief technologist at Philips Advanced Metrology Systems, 9600 Great Hills Trail, Suite 150W, Austin, TX 78759; ph 512/231-2295, fax 954/660-8389, e-mail [email protected].