Characterization of a broad beam Kaufman-type ion source operated with CHF₃ and O₂

Recent developments in the optics industry require high-precision manufacturing processes. One versatile tool for pattern transfer of gratings from a photoresist mask into a hard optical material (e.g., fused silica, silicon or other oxide glasses) is reactive ion beam etching (RIBE) with broad beam ion sources.^1–4 Further applications of broad beam ion sources are ion beam planarization^5,6 and ion beam smoothing.^7,8

There are several different process parameters, e.g., gas composition, ion energy or ion incidence angle, and sample properties, e.g., used materials, line density, or resist pattern shape, which must be considered to achieve the desired grating structure and diffraction efficiency. Often, there is only a small process window to fulfill all the requirements. Usually, the etch selectivity (the ratio of the removal rate of the substrate material to the removal rate of the mask material) is an important value in pattern transfer because it determines the degree of stretching or compressing of the grating during the etching process. Beside this, process times become longer and longer due to the need for processing of large area gratings that can be etched with sophisticated motion algorithms.^1,9 Consequently, it requires highly stable and reproduceable process conditions for RIBE. Therefore, there is a high necessity for the understanding of the involved processes but also for an appropriate characterization of the ion beam source.

Broad beam ion sources can be used either with inert gases (e.g., Ar, Xe, and Kr), reactive gases (e.g., $O 2$⁠, $C F 4$⁠, $CH F 3$⁠, $S F 6$⁠, and $CC l 4$⁠), or a mixture thereof. For pattern transfer into Si or $Si O 2$⁠, a mixture of $O 2$ and $CH F 3$ is a common choice for tuning the etch selectivity to a certain application. Different kinds of plasma generation, such as inductively coupled or capacitively coupled radio frequency plasmas (RF ion sources), microwave heated plasmas [electron cyclotron resonance (ECR) ion source], or hot filament (Kaufman-type ion source) are applied in broad beam ion sources. Every type of ion source has its own characteristic, and the type of plasma excitation strongly influences the ion beam properties. This paper focusses on a Kaufman-type broad beam ion source.

Ion beam characterization is often done with energy selective mass spectrometry (ESMS) to determine the ion beam composition and ion energy distribution (IED). Additionally, measurements with Faraday-probes were done to get the current density distribution. There is only few literature on characterization of ion sources used with reactive gases available, especially of Kaufman-type ion sources. Mayer and Barker have shown that the fragmentation of $C F 4$ strongly depends on the applied external magnetic field.¹⁰ With increasing magnetic field strength, fragmentation is enhanced due to a higher electron density inside the plasma. Similar conclusions were made by Steinbruechel who used chlorine containing gases $( CC l 4 , CC l 2 F 2 , and BC l 3)$ in RIBE of aluminum.¹¹ Zeuner has shown for a Kaufman-type ion source used with Argon that the ion energy distribution (IED) can be very complex due to different origins of the ions. Different locations of ionization (inside the plasma, points with different potentials in the ion optics) and charge transfer collisions in the ion beam lead to low energy and high energy signals relative to the main ion energy peak.¹² 

For this paper, a Kaufman-type broad beam ion source used with a mixture of $CH F 3$ and $O 2$ was characterized with ESMS and Faraday measurements. Additionally, etching experiments with Si, $Si O 2$⁠, and the positive photo resist AZ®1505 were done and discussed in the context of ion beam composition.

The experiments were conducted on a commercial ion beam etching system ISA200 (NTG Neue Technologien Gelnhausen). It is equipped with a three-axis-motion-system and a load-lock. The system is evacuated by a two-stage backing pump and a turbo molecular pump (1700 l s⁻¹) to a base pressure of 2 × 10⁻⁵ Pa. Depending on the feed gas and volumetric mass flow, the working pressure was between 2 × 10⁻³ Pa and 2 × 10⁻⁴ Pa, respectively. A scheme of the ion beam etching system is shown in Fig. 1(a).

The broad-beam Kaufman-type ion source [see Fig. 1(b)], originally developed at IOM, has a graphite plasma vessel and a graphite two-grid extraction system with a clear aperture of 180 mm. Inside the discharge vessel, there is an anode of three graphite rings with 32 mm distance between the rings. The electrically heated cathode is a tungsten filament (filament voltage U_fil). The discharge voltage U_dis between the cathode and the anode is 150 V. That high discharge voltage was for a stable operation of the ion beam source necessary due to the low volumetric mass flow. The filament is inserted into the discharge chamber with a cathode cup via a hole in the ground plate. Depending on the mounting height of the cathode cup, there may or may not be an electrically conductive connection to the discharge chamber. This situation can easily be changed, also unwittingly by users and does not lead to obvious problems in source operation. The discharge and filament power supplies are biased with respect to the beam voltage either via the plasma (without contact) or via the physical connection between the cathode and the discharge vessel (with contact). Therefore, the influence of the contact situation on the ion source performance was investigated. The screen grid is mounted on the discharge vessel so that both objects are at the same potential. Because of the electrical interconnection of the discharge and the screen grid power supply, the average main ion energy is usually estimated to be the sum of discharge voltage and beam voltage. In all experiments, the beam voltage U_beam was set to 700 V and the acceleration voltage U_acc was set to −400 V, respectively.

The discharge chamber is housed in a stainless-steel vessel that contains three rings, each with 60 permanent bar magnets, around the whole anode region corresponding to three anode rings. The magnetic orientation of the bar magnets in the three rings alternates. The resulting magnetic field can be manipulated by shorting bridges that connect the different magnetic rings (Fig. 2). This is usually done to change the shape of the extracted ion beam to get a gaussian-like profile. In our experiments, we used three different options, without any shorting bridges and two quite different shorting bridge configurations (SBC) in following named as configuration black and red. The configurations were found iteratively for stable source operation and a symmetric beam shape, while the facility was modified to use reactive gases (⁠ $CH F 3$ and $O 2$⁠). There is no systematic behind configurations. The usage of shorting bridges reduces the magnetic field strength in the anode region, i.e., it decreases in the order without shorting bridges > configuration black (11 bridges) > configuration red (18 bridges). Detailed schemes of the shorting bridge configurations black and red and the magnetic field across the source axis can be found in Figs. S1 and S2 in the supplementary material.¹⁷ A comprehensive overview of the source setups used (cathode contact situation, shorting bridge configuration) is shown in Table I.

For beam diagnostics, a vertical Faraday stripe consisting of 7 cups with 26 mm distance between each one was used. To get a complete ion current density profile of the beam, the Faraday stripe was moved parallel to grids, and measurements at every 10 mm were done. The distance of the Faraday cups to the accelerator grid was 360 mm.

An energy selective mass spectrometer (ESMS) EQP300 (Hiden Analytical Ltd) was used to measure the IED and ion beam composition. The differentially pumped system has an orifice of 100 μm and was placed ∼1300 mm in front of the ion source. The maximum detectable ion energy is 1000 eV, and the mass range is from 0 amu to 150 amu. The energy analyzer shows a constant transmission for all ion energies, but the transmission quadrupole mass filter exhibits a decreasing sensitivity with an increasing mass to charge ration (m/z). No correction for this mass discrimination was done. Therefore, a real quantitative analysis of beam composition was not made. The acceptance angle of the ESMS ion optic is usually ≤1.5° for ion energies above 100 eV.¹³ It decreases with increasing ion energy, i.e., the energy distribution depicts only ions with nearly vertical incidence. A more detailed description of the used ESMS system can be found in Ref. 12.

The magnetic field was characterized using a custom-built automated magnetic field measurement system. A Teslameter F71 vector Hall probe from Lake Shore Cryotronics was mounted on a three-axis motion system and automated by custom programmed control software. Magnetic fields up to a strength of 35 T can be measured from direct measurements up to a frequency of 50 kHz. Almost the complete discharge vessel with a volume of 7.6 l was measured in a continuous measuring run with a measuring time of 30 h and more than 67 thousand measuring points with a resolution of 5 mm in radial and axial direction.

The experiments were done with different process gases. On the one hand, 4 SCCM $CH F 3$ and on the other hand mixtures of $CH F 3$ and $O 2$ with a fixed ratio of 0.7:1 and a total volumetric mass flow of 4 SCCM to 8 SCCM were used. The beam current I_beam, which is the ion source control parameter, was varied between 30 mA and 110 mA. A beam current of 70 mA and a gas mixture of 1.65 SCCM $CH F 3$ and 2.35 SCCM $O 2$were used as standard conditions for all experiments in which the ion source setup was changed.

Beside the Faraday- and ESMS-measurements, etching experiments were conducted. Therefore, $Si O 2$- and Si-samples with a binary photoresist grating (AZ®1505, 20 μm period, duty cycle 0.5) and a photoresist layer (AZ®1505) on the Si-wafer were used. The samples were fixed in the middle of the sample holder that was placed in the ion beam center at a working distance of 435 mm. After etching the structured $Si O 2$- and Si-samples, the remaining photoresist was stripped using Caro’s acid $( H 2 S O 4 + H 2 O 2)$ at boiling temperature. The etched step heights were measured using a tactile profilometer (Taylor-Hobson Talystep). To determine the etch rate of the photoresist, the film thickness of the pristine sample and the etched sample was measured interferometrically (Carl Zeiss Jena FTP500) and afterward subtracted.

For better understanding of further results, at first some general considerations on ion beam composition and ion energy distribution are made.

Figure 3(a) shows an example of a mass spectrum for a gas mixture of $CH F 3$ and $O 2$ measured at the main ion energy. The ion beam consists of three main kinds of components. At first, there are fragments resulting from ionization and dissociation of feed gases (e.g., $C +, C H +$⁠, $F +$⁠, $H F +$⁠, $C F +$⁠, $CH F +$⁠, $CF 2 +$⁠, $CHF 2 +$⁠, $CF 3 +$⁠, $O +$⁠, and $O 2 +$⁠). The second type of ions result from ionization of background gas and residual gas of former processes (e.g., $H 2 O +$ and $A r +$⁠). The third type of ions results from reactions between the gases inside the plasma (e.g., $CO F +$ and $COF 2 +$⁠) and of plasma components with the graphite discharge vessel or the graphite screen grid (e.g., $C O +$ and $CO 2 +$⁠). It is noteworthy that $C O +$ has the most intense signal in the mass spectrum. Multiple charged ions were not found or are only subordinate species¹⁴ and can be, therefore, neglected. Due to the negative potential of the accelerator grid, negative ions cannot escape the discharge vessel. Possible negative ions in the ion beam are either generated by electron attachment of neutral species or are directly produced on the accelerator grid surface.¹⁵ But this was not in the focus of this work.

Regarding their ion energy distribution (IED), two types of distributions can be distinguished; this is shown in Fig. 3(b). Both IEDs are dominated by primary ions with the main ion energy peak at 810 eV. These ions are produced inside the plasma vessel and accelerated through the sheath region at the screen grid. As already mentioned in Sec. II A, the main ion energy is usually estimated to be to the numeric sum of U_dis + U_beam (150 V + 700 V). The discrepancy to the measured ion energy is discussed later. The secondary ions contribute to the low energy part. On the one hand, there is a broad IED in the low energy part (below main ion energy). This is the case for $C +, C H +$⁠, $O +$⁠, $C F +$⁠, $CH F +, C O +$⁠, $CO F +$⁠, and $CF 2 +$⁠. These low energy ions are produced by impact-induced dissociation outside the ion source due to collisions with the background gas in the vacuum chamber.¹⁶ The energy is then defined as share of the main ion energy corresponding to the mass ratio ion/parent ion [e.g., in Fig. 3(b): $CO 2 + ( 810 eV)→ C O + ( 515 eV ) + O ( 295 eV)$]. On the other hand, some ions $( F + , H F + , A r + , CO 2 + , CHF 2 + , COF 2 + , and CF 3 + ,)$ show no significant intensity in the low energy part of the IED. This is because there are no heavier parent ions that can dissociate or the ions are not the preferred dissociation products. A more detailed discussion on the origin of different ions and peaks in their ion energy distribution and broadening of peaks can be found for inert gases in Ref. 12 and some reactive gases in Ref. 14.

As written before, there are several different external configurations of the ion source that influence the beam composition and IED. At first, we will consider only the option to use the cathode at a floating potential (without electrically conductive contact between the discharge vessel and the cathode) and at a fixed potential (with electrically conductive contact). In Fig. 4, the main ion energy as a function of plasma running time for the use of $CH F 3$ as single feed gas is shown. When the cathode is used at a fixed potential, the main ion energy shifts to a lower value during the first 30 min. This is negligible when the cathode runs at a floating potential. Additionally, the main ion energy is significantly reduced compared to the cathode usage with electrically conductive contact. Simultaneously, the filament heating voltage increases about more than 4 V until it reaches a constant value to provide a constant beam current of 70 mA. It is believed that the fluorination of the cathode filament and of deposits on the anode rings lead to this behavior. A drift of the main ion energy was not observed when the ion source was used with the gas mixture of 1.65 SCCM $CH F 3+2.35 SCCM$ $O 2$⁠.

Figure 5 depicts the influence of shorting bridge configurations, i.e., the magnetic confinement in the anode region, for both cathode contact situations on the main ion energy when a $CH F 3 / O 2$ mixture is used. Depending on the ion source configuration, the main ion energy shifts approximately between 770 eV and 850 eV. It is clear that the main ion energy deviates significantly from the expected value U_dis + U_beam (150 V + 700 V = 850 V). The reason for this behavior is believed to be a varying plasma potential. The different shorting bridge configurations as well as the cathode contact situation strongly influence the connection of the plasma to the discharge vessel. This leads to a plasma potential below the anode potential of 150 V and, thus, to a lower main ion energy. Applying the shorting bridges to the permanent magnets reduces the resulting magnetic field strength, which causes a lower electron density and, hence, a lower plasma density. It also leads to a higher necessary discharge current to extract a constant beam current of 70 mA. As mentioned before, the shorting bridges are usually used to shape the ion beam. As shown in the Faraday measurements (Fig. 6), the influence of the SBC on the ion beam shape is more pronounced when the cathode has an electrically conducting connection to the plasma vessel. The ion beam composition changes as well, but due to the irregular change of the magnetic field depending on the position of shorting bridges (see Fig. S2 in the supplementary material¹⁷), there is no clear trend observable.

The influence of the different operation modes of the ion source on the resulting selectivity of the model systems SiO₂/AZ®1505 and Si/AZ®1505 is shown in Fig. 7. The selectivity is defined as the ratio of the etching rate of the substrate to the etching rate of the mask material. It ranges from approximately 1.4 to 1.8 for SiO₂/AZ®1505 and from 0.9 to 1.2 for Si/AZ®1505, respectively. The reproducibility of etching experiments was determined by four similar experiments over several weeks (SBC black, without contact) with a standard deviation of 4%. From these results, it becomes clear that changes in the SBC effects more than only beam shape and could be a source for process deviations between similar ion beam etching systems.

In addition to the gas mixture, there are several process parameters that influence the etch rate and the selectivity, e.g., ion energy, ion beam current, or volumetric mass flow. The ion beam composition is strongly affected by changing the beam current and volumetric mass flow because this directly influences the plasma density.

At first, the variation in the total volumetric mass flow will be discussed. As it is exemplarily shown for ions $C +$ and $C F +$ in Fig. 8, the main ion energy shifts within 757 eV and 783 eV with varying total volumetric mass flow. There is a significant increase in the relative amount of low energy ions with an increasing total volumetric mass flow. Due to an increasing chamber pressure, the collision-induced dissociation is enhanced. Beside the IED, the ion beam composition is affected by the total volumetric mass flow too. The relative intensity of selected ions is shown in Fig. 9(a). Although the amount of $O 2 +$⁠, $CHF 2 +$⁠, and $C F +$ increases, the intensity of $H F +$ and $C O +$ decreases. It is not possible to distinguish the ions $O 2 +$ and $CH F +$ with the ESMS. But in a pure $CH F 3$ plasma, the ion $CH F +$ is only a minor species and shows in the mass spectrum a relative intensity below 2%. Therefore, it can be assumed that the main contribution to the signal at m/z 32 amu comes from $O 2 +$⁠. With the increasing volumetric mass flow, the discharge current required to extract a constant beam current decreases from 448 mA (4 SCCM) over 398 mA (6 SCCM) to 377 mA (8 SCCM). Consequently, the plasma density inside the discharge vessel decreases as well. Due to this, the reaction of $oxygen$ species with the graphite materials seems to weaken, which leads to a decreased amount of $C O +$ and an increased amount of $O 2 +$⁠. As a result of higher $O 2 +$ amount in the ion beam, the resist is etched faster, which leads to a reduced selectivity for the model systems [see Fig. 9(b)]. Due to an increasing pressure in the ion beam source, the beam profile shape shifts slightly with increasing volumetric mass flow to a lower maximum current density but higher full width at half maximum (FWHM).

The influence of the ion beam current on the beam composition is shown in Fig. 10(a). A similar behavior as it was discussed before can be found. With the increasing beam current, the discharge current and, therefore, the plasma density increase. This leads to an enhanced reaction of oxygen species with the graphite discharge vessel. Consequently, the amount of $O 2 +$ decreases, whereas the amount of $C O +$ increases. This, in turn, leads to an increasing selectivity with increasing beam current [Fig. 10(b)]. Figure 10 also shows the systematic difference in the beam composition and, thus, also the selectivity between the installation conditions of the cathode. In the case of an electrically conducting connection between the cathode holder and the discharge vessel (open symbols), the amount of $O 2 +$ in the ion beam is less than without this connection (filled symbols). Thus, the selectivity with the connection is higher than without. As already mentioned, the discharge current reflects the extent of the reaction of oxygen species with the graphite components. Within all used beam currents, the discharge current is ∼10% higher if the cathode contact is given.

In the context of the presented results, another typical behavior of Kaufman-type ion beam sources used with reactive gases can be understood. This is a decreasing selectivity for SiO₂/photo resist and Si/photo resist with increasing operating period. During operation of the source with $CH F 3 / O 2$⁠, deposits of oxides and fluorides of tungsten (from the filament) and aluminum (from sample holder in front of the ion beam source) cover more and more area and become thicker. Therefore, the graphite surfaces inside the plasma vessel are somewhat passivated and the reactions of oxygen species with the graphite are depressed. This leads to a higher amount of oxygen in the ion beam and, hence, to a higher photo resist etch rate and consequently to a decreasing selectivity. Thus, a periodically abrasive removal of the deposits is necessary.

In this study, we characterized a Kaufman-type broad beam ion source regarding the ion beam composition and ion energy distribution with energy selective mass spectrometry. In this context, the consequences of the ion source configuration and the process parameters on the etching performance, i.e., the etch selectivity of the model systems SiO₂/AZ®1505 and Si/AZ®1505, were evaluated. It was found that the magnetic confinement of the ion source, which can be manipulated by shorting bridges, and the installation conditions of the cathode holder have a significant influence on the selectivity. The process parameters’ total volumetric mass flow of the feed gas and the extracted ion current greatly changed the discharge current. With varying discharge current and, thus, plasma density, the reaction of oxygen species with graphite components of the discharge vessel changes. With an enhanced reaction, the amount of $O 2 +$ in the ion beam decreases, which leads to an increasing selectivity of model systems. It was shown that not only process parameters but also little changes in the ion source configuration is a source of etch performance deviations. This means that identical ion sources from a manufacturer can show different etch results.

Since Kaufman-type ion sources are not the first choice for very long process times due to a need for a regular replacement of the cathode filament and a change in selectivity with processing time, RF broad beam ion sources become more important in material processing. Experiments for a comparable characterization of an RF ion source are still ongoing and will be published later.

The authors want to acknowledge Toni Liebeskind for the preparation of samples for etching experiments. We also thank NTG Gelnhausen for support in rebuilding the ion beam etching system ISA200. This project was supported by the Federal Ministry for Economic Affairs and Climate Action (BMWK) on the basis of a decision by the German Bundestag.

The authors have no conflicts to disclose.

Gregor Dornberg: Formal analysis (lead); Investigation (lead); Writing – original draft (equal). Erik Rohkamm: Investigation (equal); Visualization (equal); Writing – review & editing (equal). Peter Birtel: Investigation (equal); Visualization (equal); Writing – original draft (supporting). Frank Scholze: Resources (equal); Writing – review & editing (equal). Frank Frost: Conceptualization (equal); Funding acquisition (lead); Writing – review & editing (equal).

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