Benefits of using picosecond ultrasonics for RF process monitoring.
The global radio frequency (RF) semiconductor market size is growing rapidly at a compound annual growth rate of 8.5%, with an expected increase from $17.4 billion in 2020 to $26.2 billion in 2025, according to Research and Markets [1]. As many are aware, the rollout of 5G technology and the Internet of Things (IoT), which is enabled by 5G, are the main driving forces for this growth.
Growth or not, 5G devices offer their own set of challenges. These devices require up to 100 filters to isolate each RF band and avoid interferences that can drain battery life, reduce data speeds and cause dropped calls. As such, RF filters are becoming more critical for signal process applications. In the case of 5G devices, this means bulk acoustic wave (BAW) filters, which can work better at higher frequencies.
With increasingly more filters being put into devices, the size of the filters is also shrinking dramatically. In RF BAW filters operating in the 2.5GHz range, multiple channels are packed closely together, and they require frequency tolerances of <0.05% to avoid issues like crosstalk and line drops in mobile communications [2,3].
A BAW filter uses a piezoelectric film, like aluminum nitride (AlN), sandwiched between a top and a bottom electrode. This filter can be of two types: a thin film bulk acoustic resonator (FBAR) or a solidly mounted resonator (SMR). In either case the resonator needs to be acoustically isolated from the substrate. To assure a tight frequency tolerance, a frequency response must be measured. This is determined by the acoustic velocity of the AlN piezoelectric layer and the thickness of the full stack. The final frequency of each device is adjusted by adding a mass load deposited on the top electrode. This causes the downshifting of the resonance frequency. A frequency accuracy of 0.1% or better is required.
Another challenge: the uniformity of each of the layers in the stack is stringent and needs to be tightly controlled. Thin film deposition systems cannot achieve the required layer uniformity. So, tuning steps like ion milling often are introduced to achieve the desired cross-wafer uniformity.
To achieve accurate measurements, metrology techniques should meet sensitivity, accuracy, and stringent repeatability requirements. Several techniques — from sheet resistivity, X-ray fluorescence, ellipsometry and reflectometry — are available for qualifying deposition tools. However, device-level process control is required to meet yield targets. This requires measurements on the actual structures of interest to adjust the thickness via a trimming process.
Picosecond ultrasonics [4] is a non-contact, non-destructive optoacoustic metrology that has found widespread adoption in the characterization of metal film thickness. The small spot size (8µm x 10µm) of the system enables measurements within ~0.5mm edge exclusion at production-worthy throughputs.
In RF process monitoring, picosecond ultrasonics offers significant advantages over other technologies: the ability to simultaneously measure all layers and discriminate repeating layers of the same material in a multilayer stack, the ability to measure pre-trimming to feed data to the ion trimmer for achieving desired thickness uniformity, and the ability to characterize the thickness and acoustic velocity of the piezo layer.
When measuring a multilayer stack using picosecond ultrasonics, the arrival time of the echoes from each layer is used to calculate thickness, without the requirement of calibration standards. Table 1 shows that up to six layers can be measured simultaneously. As demonstrated, picosecond ultrasonics offers excellent repeatability for all the layers (3s < 0.1% for all critical layers).
Table 1: Picosecond ultrasonics measures all the six layers in the stack simultaneously.
Prior to the adoption of picosecond ultrasonics, a metrology strategy required a break in the process to measure the bottom three layers, followed by send-ahead wafers to characterize the process. This was both time consuming and did not provide direct feedback on the thicknesses of individual layers. In addition, during process development, the ability to measure high-resolution wafer maps is desired to understand variations. Measurements typically take a few seconds per site; this allows for the rapid mapping of cross-wafer variation (figure 1).
Fig. 1: 400 measurement site maps for a). bottom electrode Mo, b). AlN and c). top electrode Mo for the tri-layer stack Mo/AlN/Mo.
As mentioned, the thickness adjustment of the piezoelectric layer is very critical to the process as such an adjustment directly defines the resonance frequency. Measurements of transparent and semi-transparent films using picosecond ultrasonics is well documented [5]. The transient optical reflectivity in the signal contains a sinusoidal oscillating component, i.e., Brillouin oscillations (figure 2). Sound velocity can be calculated from the period of the oscillations; the sound velocity and the echo round trip travel time as shown by the arrow in the figure determine the thickness of the AlN layer. If desired, the refractive index of the AlN layer can be obtained from a spectroscopic ellipsometer and be fed forward to the picosecond ultrasonics tool to improve accuracy.
Fig. 2: Typical measurement signal from AlN/Si using picosecond ultrasonics.
Studies [6] have shown the benefits of doping the AlN layer with scandium (Sc) to increase piezoelectric coefficients, soften the material, increase permittivity, and boost the electromechanical coupling of K2. Picosecond ultrasonics detects Sc concentration changes by monitoring changes in acoustic velocity. In fact, studies [7] show an ~15% difference in sound velocity between various Sc concentrations. Figure 3 shows excellent linearity of sound velocity and Sc concentration with R2=0.97.
Fig. 3: Correlation of AlN sound velocity and Sc concentration.
The picosecond ultrasonic system is designed to offer long-term stability and intrinsic tool-to-tool matching (0.5% at site-level), both of which are key requirements for adoption in high-volume manufacturing.
In summary, the RF market is growing rapidly as a result of the proliferation of 5G technology. RF filters are key enablers for signal processing applications, and process control for RF filters demands stringent metrology due to tight process tolerances. Picosecond ultrasonics is uniquely qualified as an in-line metrology technology that can meet these challenges.
References
[1] https://www.reportlinker.com/p05874886/?utm_source=PRN
[2] R. Aigner, 2008 IEEE Ultrasonics Symposium, 2008.
[4] C. Thomsen, H. T. Grahn, H. J. Maris, J. Tauc, Phys. Rev. B, vol. 34, 1986, pp. 4129-4138
[5] J. L. Arlein, S. E. M. Palaich, B. C. Daly, P. Subramonium, and G. A. Antonelli, J. of Appl. Physics, vol 104, 2008, pp 033508 1-6
[6] R. Matloub, A. Artieda, Alvaro, C. Sandu, E. Milyutin, P. Muralt, vol. 99, (2011), Applied Physics Letters. 092903
[7] K. Umeda, H. Kawai, Honda, A. Akiyama, M. Kato, T. Fukura, Proc. of the IEEE Int. Conf. on Micro Electro Mechanical Systems (MEMS), 10.1109, 2013, pp 733-736
 |
 |
 |
 |
 |
 |
 |
 |
 |
 |
Leave a Reply