Plasma processes are of major importance in modern manufacturing. In a plasma, gas atoms are ionized and excited to higher energy states by external sources such as electric fields, or flames to name a few. These ionized gas clouds find use in various industries including the semiconductor, medical or food industry in which either thin films are deposited/etched, or surfaces are altered for sterilization by tailoring the surface chemistry of packaging material from hydrophilic to hydrophobic through a plasma treatment. As is common to all processes, accurate control over plasma-based processes is of paramount importance.

For decades, optical emission spectroscopy has been a key technology providing insights and thus process feedback over the atomic, chemical and physical processes within a plasma. Conventionally, optical emission spectroscopy relies on high resolution spectrometers to detect the sharp atomic emission lines of ionized atoms. Although they have known advantages, high resolution emission spectrometers are expensive with a price tag in excess of US$ 20k and difficult to maintain. As cost for a plasma tool may range from US$ 10k – 1m, the cost for process spectrometers is cost-prohibitive and many applications can’t afford a precise control of plasma processes. Here, we introduce a cost effective, compact type of plasma process spectrometer overcoming the cost disadvantage while providing high spectral resolution for process monitoring.

Contrary to conventional UV-VIS spectrometers which rely on the dispersion of a collimated beam of electromagnetic radiation into its spectral elements by a high precision, finely ruled and costly grating, we utilize a static array of unique interferometer channels producing a distinctive wavelength dependent interference pattern on a detector array (Fig.1). Each interference pattern is unique to a given wavelength and can be mathematically translated into spectral information. Figure 2(a) demonstrates a typical interference pattern recorded by our spectrometer capturing the oxygen plasma emission.

Unlike the dispersion of light e.g. by a grating (requiring a wave to travel comparably long distances until it separates into spectral components), the interference allows us to construct extremely compact spectral solutions thereby saving cost and space. Furthermore, our interferometer array allows large field of views while offering the multiplex advantage known for conventional Fourier transform interferometers.

The interferometer chip is placed it in front of a CCD or CMOS array detector into a collimated beam of light coming from the plasma source. In its essence, our interferometer chip converts a standard monochrome camera into a spectrometer engine. The spectral working band of the camera thereby determines the spectral bandwidth of the device. Its spectral resolution is tailored by the design of the chip and mainly governed by the maximum optical path difference generated in the interference structure of the chip. In the current device the peak full-width-half-maximum is 2.4 nm, its spectral bandwidth is determined by the sensitivity curve of the CMOS detector and ranges from 380 nm to 1020 nm whereby our chip enables a 0.4 nm wavelength interval throughout. Neither order sorting filters nor long collimators are required further reducing the complexity of the optical system to a bare minimum. The dimensions of the device are 30 x 30 x 60 mm3. The compact spectrometer, therefore, can be easily adapted to any viewport using the free-space optics in the plasma chambers available from different vendors. The direct mounting on the view port of the plasma chamber using free space optics circumvents all problems associated with fiber optics such as fiber coupling and transmission losses.

To validate our devices for plasma process monitoring, we attached our VIS-NIR spectrometer (refer Fig.3) to the optical viewing port of a plasma cleaner (NTI RIE-2321 Reactive Ion Etching System). Our spectrometer recorded the emissions from the plasma for two different experiments showcasing our ability to monitor the plasma processes.  In the first experiment, we spin-coated a 4-inch diameter silicon wafer with Polymethyl Methacrylate (PMMA) photoresist (MicroChem’s A5) and employed the plasma cleaner with oxygen plasma at 100 W RF power to remove the photoresist while monitoring the oxygen emission line at 776.4 nm over time. Oxygen gas flow was maintained at 20 SCCM for a total plasma duration of 2 minutes while we captured a spectrum every 4 seconds.

In a second experiment, we left the plasma chamber empty and mixed argon gas (1, 2, 5, 10 and 20 SCCM) gas into the gas flow of oxygen (maintained at 20 SCCM) at an RF power of 100 W.

Figure 4 shows the data collected by our spectrometer following the etching of polymer layer on a 4-inch wafer. During the etching process, the emission spectrum of the plasma shows clear differences with a higher emission in the visible part of the electromagnetic spectrum most likely due to the carbon emissions of polymer removal. Furthermore, it appears to shift emission lines in the near infrared part of the electromagnetic spectrum at around 880 and 930 nm, which may be attributed to carbon or carbon-oxide emissions (Fig.4a).

To follow the etching process in real-time, we plot the emission of an oxygen line e.g. 776.4 nm over the duration of the etching process in Figure 4b. The figure plots normalized peak emission for 600 nm(red) and 700 nm thick PMMA layer as well as pure oxygen plasma over time. The plasma was ignited after 30 seconds and maintained for a total duration of 2 minutes.  The emission of oxygen at plasma ignition spikes up and decays exponentially during the etch of the polymer layer. Once the polymer layer is removed completely, the oxygen emission increases again as the overall amount of pure oxygen in the plasma increases. Clear differences for a 100 nm thicker photoresist layer (green curve) can be seen whereby the plasma process takes approximately 10 seconds longer to remove the thicker layer.

Figure 5 demonstrates the capability of capturing fine emission changes to an oxygen plasma (20 SCCM O2, 100W) when argon gas is introduced into the plasma reactor. Pure argon and oxygen plasma are illustrated in Figure 5a whereas Figure 5b shows the emission changes when both gases are mixed in the spectral range from 760 nm to 850 nm. Clear changes in the plasma emission are visible while the oxygen lines are maintained at 776 nm and 842 nm (Fig.5b). As the argon gas flow is increased from 0 to 20 SCCM, Argon lines at 796.4 nm, 809.6 nm, 830.4 nm, 840 and 842 nm appear and increase with increasing flow rate. Figure 5(c) follows the evolution of argon line at 809 nm with respect to oxygen line at 776 nm as argon is introduced. The area under the curve for the emission line peaks is plotted for different flow rates of argon gas. The oxygen peak trend remains relatively flat as the argon peak trend increases with increasing argon gas flow rate.