Characterization of an RF-excited broad beam ion source operated with a mixture of CHF₃ and O₂

The ever increasing requirements for high-precision optical elements require a deeper understanding of the manufacturing process itself. When it comes to surfaces with a very low surface roughness in the sub-nm range and a precise transfer of functional structures, ion beam etching is one of the most promising methods. For the layer thickness trimming of bulk acoustic wave or surface acoustic wave filters,¹ final shape correction of telescope mirrors^2–5 or the manufacturing of high-efficient pulse compression gratings,^6,7 ion beam processing is already the method of choice for the highest requirements on precision. The large number of tunable process parameters, such as the process gas, ion energy, or ion incidence angle, results in a fine-tunable process for realizing the desired surface properties. When the inert process gas is substituted by a reactive gas or a mixture thereof, even more possibilities arise for ion beam processing. The usage of reactive gases allows, e.g., the tuning of the selectivity (ratio of the etching rate of the substrate material to the etching rate of the mask material) in a wider range. This is a key parameter that characterizes the type of structure transfer from the mask to the substrate, i.e., whether a stretching or a compression takes place. For an increase in the processed surface area, the long process times require detailed knowledge of the ion beam source performance. Here, the most fundamental and direct way is a measurement of the ion beam composition and the ion energy distributions. These ion beam properties strongly depend on the plasma excitation method employed in the ion beam source (IBS). Kaufman-type, ECR-, and RF-type ion beam sources are the most commonly used plasma excitation methods each having their own advantages and disadvantages. Due to a low demand for maintenance and high extracted ion currents, RF-type ion beam sources are a good choice, when long process times are required. Even though RF-type ion beam sources took a leading position in ion beam figuring industry, there is a lack of publications on energy-selective mass spectrometry measurements for characterization purposes. So far, only some studies can be found in the literature on the characterization of the operation of the RF-type ion beam source with inert gases^8–11 and even fewer literature on the use of reactive gases.¹² There are some publications on the operation of an ion beam source with reactive gases, however, using different excitation methods.^13–15 For the manufacturing of ultra-precise optical devices, a mixture of $CHF 3$ and $O 2$ is used to enable the realization of the desired surface topography.^6,7 Finzel et al.⁷ showed the high dependence of the mixture ratio on the etching result. Even though they used a 3:1 mixture of the process gases, RF-excited ion beam sources show a higher stability for an operation with a higher oxygen content than Kaufman-type ion beam sources. It is, therefore, of special interest to increase the $O 2$-content in the process gas mixture to achieve a higher degree of freedom in the selection of the process parameters. At the same time, the $SiO 2$ made plasma discharge vessel may act as a potential fluorine sink, which is an upper limit for the increase of the $CHF 3$ content in the process gas. Therefore, we performed the experiments with an arbitrary 1:1 mixture ratio of $CHF 3$ and $O 2$ as a starting point for further studies. To achieve detailed understanding of the etching process, the ion energy distributions (IEDs) and ion beam composition have to be investigated for different process parameters. This work lays the foundation by describing the IEDs and composition of the ion beam for one chosen set of process parameters for future investigations.

The experiments were executed in a custom-built high vacuum chamber of size $1.8× 1.8× 1.5 m 3$⁠. The chamber was evacuated using two turbomolecular pumps with pumping speeds of 1500 and 3050  $l / s$ for $N2$ to a base pressure of $p 0= 5 × 10 − 7 mbar$⁠. The pressure during the measurements was $p p r o c e s s≈ 2 × 10 − 4 mbar$⁠.

For the generation of the broad ion beam, an RF-type (⁠ $f= 13.56 MHz$⁠) ion beam source with a diameter of the extraction area of $120 mm$ was used. The experimental setup is almost similar to the one used in a former publication.¹¹ In contrast to the previous setup, the ion beam source is now mounted on a four-axis motion system inside the vacuum chamber and can be moved in $x$-, $y$-, and $z$-direction and rotated horizontally as depicted in Fig. 2. The plasma discharge vessel is made out of $SiO 2$⁠. For ion extraction, a focusing three-grid extraction system made of graphite was used. The ions are extracted by applying voltages $U B$ and $U A$ to the screen and accelerator grid, respectively (Fig. 1). $U A$ is chosen to be negative in order to prevent electrons from outside of the IBS to stream back to the plasma discharge, which would cause instabilities in the plasma generation. The decelerator grid was connected to ground potential. The resulting qualitative potential curve along the ion beam path is shown in Fig. 3. Consequently, ions with charge $Q$ originating from the plasma discharge vessel are expected with an energy $E=Q⋅( U B+ U P)$ with $U P$ being the plasma potential. The process gases were injected into the ion beam source via separately adjustable volumetric mass flow controllers (MFCs). As process gas, a $1:1$ mixture of $CHF 3$ and $O 2$ was used. The resulting process pressure was $2.6 × 10 − 4 mbar$⁠. The ion beam was charge neutralized by using a plasma bridge neutralizer.¹⁷ Here, a negative voltage is applied to the ICP $Ar$ plasma (cathode voltage) and electrons are extracted by applying a positive voltage to the keeper electrode (keeper voltage).

For the characterization of the reactive ion beam, two differentially pumped energy-selective mass spectrometers (ESMS) by Hiden Analytical Ltd were used. They allow the detection of ion masses between 0.4 and 150 $amu$ for the device with $6 mm$ rod diameter and a detection of ion masses up to $1000 amu$ for a rod diameter of $9 mm$⁠. Both allow spectrometry of ion energies from some tenths eV up to $1100 eV$ for single charged ions. If not stated otherwise, the device with a rod diameter of $9 mm$ was used. An ESMS consists of several ion-optical parts. After entering the ESMS through an orifice (⁠ $d O≈50μ m$⁠), the ion beam is collimated and guided by several different apertures and condensator plates. In the $45 °$ energy analyzer, the ions are analyzed concerning their energy-to-charge ratio $E/Z$ with the charge number $Z=Q/e$⁠. Afterward, the ions enter a quadrupole mass spectrometer, where a selection regarding their mass-to-charge ratio $M/Z$ is performed. The energy and mass analyzed ions are then detected using a secondary electron multiplier in the form of a channeltron detector. Due to the low amount of multiple charged ions originating from the RF-excited ion beam source,^9,11 the terms $E/Z$-distribution and $E$-distribution and $M/Z$-distribution and $M$-distribution, respectively, are used synonymously. The measuring distance between the exit plane of the ion beam source and the ESMS orifice was chosen to be $420 mm$⁠. The current density was measured by a custom-built Faraday-cup array consisting of $16×16$ single Faraday-cups. A more detailed description can be found elsewhere.¹¹ 

Comparing the data of the 6 and the $9 mm$ system with the transmission data provided by Hiden and a simple $1/M$-approach (see Fig. 5) shows a significant deviation for low masses, as already shown by Chatain et al. The transmission for the $6 mm$ system is comparable to the $1/M$ approach and the Hiden data for the medium mass range between 20 and $80 amu$⁠. Fundamentally, the data show a significant deviation between transmission data for the 6 and $9 mm$ system.

First, the main ionic species present in the ion beam are discussed. A recorded $M/Z$-distribution for the operation of the IBS with a mixture of reactive gases $CHF 3$ and $O 2$ is given in Fig. 6. The data were recorded using the EQP with $6 mm$ rods. It can be seen that many different ionic species can be detected in the ion beam. Analogously to other publications on the ion beam composition of a reactive ion beam,^13–15 ionization and dissociation products of the process gases can be found at full ion energy. The main species are $CF 3 +$⁠, $CHF 2 +$⁠, $CF 2 +$⁠, $CHF +$⁠, $CF +$⁠, $C +$⁠, $HF +$⁠, $F +$⁠, $O 2 +$⁠, and $O +$⁠. Additionally, reaction with the discharge vessel (made out of $SiO 2$⁠) and the extraction system (made from graphite) leads to $SiF 3 +$⁠, $SiF 2 +$⁠, $SiF +$⁠, $Si +$⁠, $CO +$⁠, $COF +$⁠, $COF 2 +$ being present in the ion beam as well. Furthermore, small amounts of $Ar +$ are found, which originates from $Ar$ gas escaping from the plasma bridge neutralizer and entering the discharge vessel of the IBS.

The recorded energy distributions for the ion masses and corresponding species given in Table I are displayed in Figs. 7–10. All IEDs were normalized to one with respect to their maximum count rate. The IEDs are sorted for oxygen containing species (see Fig. 7), carbon-fluorine species (see Fig. 8), silicon-fluorine species (see Fig. 9), and species of residual gas, process gas of the plasma bridge neutralizer and fluorine (see Fig. 10). The ion beam source was operated with a mixture of $CHF 3$ and $O 2$⁠. The resulting ion current density in the center of the ion beam was $≈1800 μ A / cm 2$⁠. $P R F=246 W$⁠. $U B=700 V$⁠. The following IEDs were recorded with the ESMS with $9 mm$ rods.

In general, IEDs consist of two parts. Ions originating from the plasma discharge vessel and reaching the detector of the ESMS appear at an energy $E=e⋅( U B+ U P)$ with a typical value of the plasma potential $U P$ of some tens of eV. They gathered the full possible energy out of the electrical potential gradient (see Fig. 3). All contributions to the energy distributions at lower energy are a product of processes occurring outside of the plasma discharge of the ion beam source. For the ions at full energy, there is an additional peak at slightly higher energies, which can be related to a modulation of the plasma sheath potential due to an capacitively coupling of the plasma with the RF-field. The sinusoidal time-variation of the extracting plasma sheath potential leads to an energy spread which is, for purely capacitively coupled plasmas, dependent on the ion mass.²⁵ Similar to the processing with inert gases,¹¹ no strict dependence of the energy separation and the magnitude of the two different peaks on the ion mass can be seen, i.e., the energy separation is not strictly decreasing with higher ion mass.

As mentioned before, besides the peaks in the high-energy part of the IED, additional ones can be observed between $0 eV$ and the main peaks at $≈700 eV$⁠. Similar to the measurement of the IEDs for inert gases,^{8,11,14,15,26} a peak at some tens of eV can be seen, which can be related to charge transfer processes in the ion beam outside the plasma discharge vessel. This energy is, approximately $e⋅ U s . c .$⁠, where $U s . c .$ is the space charge potential of the ion beam. The presence of a space charge potential of the ion beam indicates an imperfect charge compensation in the ion beam. This is in accordance with previous studies showing a non-homogeneous beam neutralization.²⁷ Nevertheless, etching experiments with this setting of neutralization showed a sufficiently high neutralization on the sample surface.

Most peaks in the recorded ion energy distributions can be assigned to one or multiple dissociation processes, and the following statements can be made:

• $CF 3 +$ can only be found as a primary particle at the main ion beam energy, i.e., it is not found as a dissociation product from heavier primary particles.

• $CHF 2 +$ only appears at the main beam energy, i.e., it is only extracted directly from the plasma and does appear as a dissociation product due to the fact that no $CHF 3 +$ is extracted from the plasma that would serve as a precursor.

• Although $CHF +$ and $O 2 +$ cannot be distinguished by their mass using the ESMS, the smaller peak in the $M/Z=32 amu$ IED at $≈430 eV$ can be assigned to $CHF +$ originating from a dissociation process of $CHF 2 +$⁠. The smaller peak at the slightly higher energy of $480 eV$ may be related to the bimodal structure in the $CHF 2 +$ IED at main beam energy.

• $C +$ shows a smudged energy distribution in the energy range up to $360 eV$ due to the high number of possible dissociation reactions.

• For $C +$ the peak area associated with fragmentation processes outside, the IBS is significantly higher than the peak area for ions extracted directly from the plasma discharge vessel.

• The bimodal structure of the $O 2 +$ IED (energy peak separation $≈90 eV$⁠) can also be found in the $O +$ energy distribution of the dissociation products with half the energy peak separation (⁠ $≈45 eV$⁠).

• The peak for $Si +/ CO +$ (⁠ $M/Z=28 amu$⁠) at $≈610 eV$ cannot be associated with any fragmentation process, and its origin remains unclear.

• Even though no $SiF 4 +$ could be detected with the ESMS, its fragmentation product $SiF 3 +$ can be found at its dissociated energy.

• A small peak at $≈350 eV$ can be found in the IED of $SiF 3 +$⁠, which can be related to the dissociation of extracted $Si 2 F 6 +$⁠.

• $F +$ dissociation products were only found in the energy range between 200 and $300 eV$⁠, with a relatively low signal. This indicates that the fragmentation processes mainly result in a high amount of neutral energetic $F ∗$-radicals or negatively charged $F −$-ions that are not detectable by the ESMS or Faraday-cup measurements but may have a high etching potential.

For a quantification of the processes outside of the ion beam source, i.e., the charge exchange and fragmentation processes, the IEDs were integrated from $0$ up to $600 eV$ and compared with the full integral from $0$ to $1000 eV$ in order to quantify which fraction of each specie is produced as a result of fragmentation or charge exchange processes. The IEDs were recorded three times in subsequent measurements for each mass, the values for the different ionic species are given in Fig. 11. The error bars represent the standard deviation of the three independent measurements for each ionic specie. It can be seen that for the $SiF n +$- and $CF n +$-species, an increasing fraction of the ions is produced outside of the ion beam source, the lower the ion mass is. This illustrates that those ions that can originate from several dissociation processes of initial process gas or first reaction products of the process gas with parts of the IBS itself are more likely to be present as products of fragmentation processes outside the IBS.

The ionic composition of the ion beam depicted in Fig. 12 was derived by integrating the recorded ion energy distributions taking into account the mass dependent transmission function of the ESMS. For this purpose, the IEDs were recorded in three subsequent measurements for each mass. The error bars in Fig. 12 represent the standard deviation of these measurements.

It should be kept in mind that the ESMS can only detect ions and may, therefore, not allow to establish the full composition of the beam of fast neutral or negatively particles that are relevant for the ion beam processing. In particular, fast neutrals created in charge exchange processes and negative ions are not included in Fig. 12. Furthermore, similar to the change of fragmentation and, therefore, the beam composition in Kaufman-type ion beam sources for different operation parameters,^13,15 this effect can also be expected for RF-type ion beam sources when varying the applied RF-power.

The mass and energy distributions for one exemplary mixture of process gases and one set of operation parameters of the RF-excited broad beam ion source were recorded. Since there is a lack of publications about RF-excited broad beam ion sources operated with a mixture of $CHF 3$ and $O 2$⁠, these findings lay the foundation for further investigations. To quantify the ion beam composition, the mass dependent transmission curves were determined for two ESMS systems. These differ significantly from one another and can have a high discrepancy with the data provided by the manufacturer.¹⁸ Therefore, the determination of the mass dependent transmission curve is mandatory for each individual ESMS and its operation voltages when a quantification of the ion beam composition is determined. It was shown that the injected process gas dissociates into a lot of smaller different ionic species and reaction products with residual gas and parts of the ion beam source. Each may have different etching behaviors, which influences the processing of a sample. Heavy ions with a high $F/ C$-ratio have a higher etching yield compared to the lighter, dissociated $CF n +$-species.^29,30 As the operation parameters of an ion beam source can have a high influence on the fragmentation of the process gas,^13,15 this effect can also be expected for a variation of the applied RF-power for an RF-excited ion beam source or a variation of the measuring distance between the ion beam source and the ESMS. This may be a significant factor for long-time operations of the IBS since the erosion of the extraction system leads to a decrease of the needed RF-power for the extraction of a fixed ion beam current.¹¹ It is important to note that ESMS can only detect positive ions. Hence, further investigation on the beam composition and its correlation with the etching process is required since the negatively charged and neutral compounds of the beam are yet to be studied. However, this study is a first step into a deeper understanding of the reactive ion beam processing techniques, which are expected to gain even more importance in the next years. In addition, the approach presented here for the characterization of reactive ion beams is a possibility to improve the comparability and reproducibility of corresponding technologies and manufacturing systems.

We would like to thank Dr. G. Dornberg and Dr. D. Kalanov for helpful discussions and J. Knipper, M. Mitzschke, R. Woyciechowski, and the workshop of the IOM for technical support. The work is financed by the EU and the BMBF. This project has received funding from the ECSEL (Electronic Components and Systems and European Leadership) Joint Undertaking (JU) under Grant Agreement No. 875999. The JU receives support from the European Union’s Horizon 2020 research and innovation program and the Netherlands, Belgium, Germany, France, Austria, Hungary, United Kingdom, Romania, and Israel.

The authors have no conflicts to disclose.

Erik Rohkamm: Formal analysis (equal); Investigation (equal); Visualization (equal); Writing – original draft (equal). Daniel Spemann: Resources (equal); Writing – review & editing (equal). Frank Scholze: Resources (equal); Writing – review & editing (equal). Frank Frost: Conceptualization (equal); Funding acquisition (equal); Writing – review & editing (equal).

The data that support the findings of this study are available from the corresponding author upon reasonable request.