The feature size of electronic devices is continuously decreasing and plasma-based etching to get high aspect ratio feature in Si (also known as Si trenches) is very important technology in semiconductor industry. In addition to the etching system, process control is also important to obtain deep Si trenches as they are difficult to fabricate. In general, the trench profile at high aspect ratio becomes more positive (top width larger than bottom) because its byproducts at the bottom of trench were more difficult to remove than that at lower aspect ratio trench. This study highlights the use of pulsed power applied to source and bias to control the etch characteristics. Monitoring the ion energy reaching on Si surface using Semion system (manufactured by Impedans Ltd) made easy to develop a relationship between pulsing parameters and etch rates.
A 200 mm diameter inductively coupled plasma (ICP) etching system powered by a 13.56 MHz pulsed generator was used for etching of the silicon trench as shown schematically in Figure 1(a). The pulse signal extracted from the source generator was fed to a signal generator to generate a 12.56 MHz asynchronously pulsed bias power to the substrate. The pulse repetition frequency was fixed at 1 kHz, and the duty ratio of the ICP source was maintained at 25% (250 μs). The bias duty ratio and delay time were varied from 25 to 75% and from 0 to 700 μs, respectively (Figure 1b−g).
Figure 1 (a) Schematic diagram of the ICP etching system and (b−g) pulse conditions for the experiment.
Etch selectivity of Si over the bi-layered hard mask (SiO2/Si3N4) for the pattern width of 185 nm and the ARDE ratio between ER at 185 nm width/ER at 14 nm width were investigated. For the etching of Si, gas mixtures of Cl2, Ar, and additive gases (CF4 or C4F8) were used, and the process pressure was maintained at 5 mTorr. During this measurement, ICP source power and bias power were maintained at 400 and 75 W, respectively. The time-average ion energy distribution (IED) on the substrate was measured using a Semion retarding field energy analyzer from Impedans.
To understand the effect of bias pulsing parameters on etch characteristics, instant bias voltages and ion energy distributions on the substrate were measured under various pulsing conditions. Only Argon gas was used during these measurements to prevent the contamination and corrosion of the RFEA sensor.
Figure 2 (a) Instant DC-self-bias voltage data as a function of time for various duty ratios of bias pulsing from 25 to 75%. (b) Time-averaged IED for various bias pulse duty ratios. (c) Instant DC self-bias voltage data as a function of time for bias pulse delay time from 250 to 700 μs. (d) Time-averaged IED observed for different bias pulse delay times.
Figure 2 (a) shows the instant DC-self-bias voltage data as a function of time for various duty ratios of bias pulsing from 25 to 75%. Figure 2(b) shows the time averaged IED for various bias pulse duty ratios. The IED for the source pulsing only (duty ratio 0%) is also shown as a dotted line. These IED results show the addition of two types of IEDs; a monoenergetic distribution centered at 18 eV resulting from ICP source power, and a broad distribution with higher energies resulting from the bias power. As the bias pulse duty ratio was increased, the distribution with the energy of greater than 100 eV was decreased whereas the IED in the range of 30−65 eV was increased, thus reflecting the decreased instant DC-self-bias voltage shown in Figure 2(a). Figure 2(c) and (d) shows the instant DC self-bias voltage data as a function of time for the bias pulse delay times ranging from 250 to 700 μs and the time averaged IED observed for different bias pulse delay. Both monoenergetic low energy IED from ICP source power and high energy IED from bias power having higher ion energy for a longer bias pulse delay time are seen.
Figure 3 Etch selectivity and ARDE ratio as a function of bias duty ratio
As the bias duty ratio is increased during the asynchronous pulsing, the ion bombardment time is increased, which results in nonselective etching and shows the decreased trend in selectivity . However, the variation in bias duty ratio did not noticeably change the ARDE ratio, which was possibly due to the sufficient energetic ion flux at the bias duty ratio of 25% for the etching of all pattern widths
Figure 4 Etch selectivity and ARDE ratio as a function of bias pulse delay time.
The etch selectivity was the lowest at the synchronous pulsing, whereas it was the highest at the delay time of 250 μs. For asynchronous pulsing conditions, increasing the bias pulse delay time from 250 to 700 μs decreased the selectivity slightly. The ARDE ratio was the highest at the synchronous pulsing while showing an improved ARDE ratio for the asynchronous pulsing. Though, there were differences in ARDE between synchronous pulsing and asynchronous pulsing, but there were no significant differences for asynchronous pulsing with different pulse duty times. From these results, it is believed that the decreased etch selectivity, and improved ARDE at the longer bias pulse delay time are all related to the higher ion energy to the substrate.
In this study, the effects of bias pulsing parameters during asynchronous pulsing on Si trench etch characteristics were investigated. It was found that the bias pulse duty ratio for asynchronous pulsing affected the etch rates and etch selectivity over SiO2/Si3N4 mask without changing the ARDE ratio due to the increased sputtering by the increased ion dose at the higher bias duty ratio. In the case of the bias pulsing delay time, the longer bias pulse delay time decreased the etch selectivity but slightly improved the etch rates and ARDE due to the increase of ion energy by increasing the plasma impedance of the system.
Hee Ju Kim and Geun Young Yeom, ACS Appl. Nano Mater. 2023, 6, 10097−10105 https://doi.org/10.1021/acsanm.3c00807
To read the full paper click here