Effect of Refresh Time on XeF 2 Gas‑assisted FIB Milling of GaAs

Focused ion beam (FIB) machining can be used to fabricate gallium arsenide-based devices, which have a surface finish of several nanometers, and the FIB machining speed and surface finish can be greatly improved using xenon difluoride (XeF 2 ) gas-assisted etching. Although the refresh time is one of the most important parameters in the gas-assisted etching process, its effect on the machining quality of the surface finish has rarely been studied. Therefore, in this work, we investigated the effect of the refresh time on the etching process, including the dissociation process of XeF 2 , the refresh time dependency of the sputter in yield under different incident angles, and the surface finish under different refresh times. The results revealed that a selective etching mechanism occurred at different refresh times. At an incidence angle of 0°, the sputtering yield increased with the refresh time and reached its maximum value at 500 ms; at an incidence angle of 30°, the sputtering yield reached its minimum value at a refresh time of 500 ms. For surface roughness, the incident angle played a more important role than the refresh time. The surface finish was slightly better at an incidence angle of 30° than at 0°. In addition, both F and Xe elements were detected in the processed area: Xe elements were evenly distributed throughout the processing area, while F elements tended to accumulate in the whole processing area. The results suggest that the optimum surface can be obtained when a larger refresh time is employed.


Introduction
There is extensive interest in the use of gallium arsenide (GaAs)-based devices in optoelectronics, microelectronics, and quantum applications.Compared with Si, GaAs has a wider bandgap (1.42 eV), higher electron mobility (approximately six times that of Si), lower power consumption, and many other superior properties with respect to its high temperature and radiation resistance [1][2][3][4].However, high-performance GaAs devices require a surface finish measuring nanometers, and only minimum subsurface damage is tolerated during the fabrication process; this inevitably results in greater demands for process control capabilities during the fabrication processes.
As a versatile micromanufacturing technology, focused ion beam (FIB) machining has been widely used for the fabrication and analysis of GaAs and other semiconductor material devices [5][6][7][8][9].It utilizes a beam of high-energy ions, mostly Ga ions, to directly sputter sample surfaces and has micro and nano-machining capabilities.However, When Ga ions bombard the GaAs substrate, the additional Ga atoms introduced by the incident ions create a Ga-rich environment.Consequently, random Ga droplets are generated on the machined surface during Ga-FIB machining, which severely affects the surface finish and the performance of such devices.
FIB-induced gas-assisted etching (GAE) is an effective method used to suppress or even eliminate the formation of Ga droplets.This method employs the introduction of an auxiliary gas to react with the excessive Ga residues on the processed surface, thereby producing other compounds that can be removed by ion beam sputtering to remove the Ga droplets.Due to the effectiveness of GAE, many studies have been conducted on the related machining parameters and their contributions to machining quality.In this respect, for the GaAs substrate, Dhamodaran and Ramkumar [10] investigated the effect of using XeF 2 as an assist gas on the formation of nanodots and found that nanodots on the GaAs surface were suppressed after processing, resulting in a smooth surface with a mean roughness of 1.2 nm.They also studied a case without the assist gas, and Ga droplets with sizes ranging from 50 to 300 nm were observed on the processed surface.Sugimoto et al. [11] achieved a smooth processed surface by using Cl 2 as an assist gas in the etching of GaAs, and the sputtering yield under the highest flux increased by 20 times compared to that without Cl 2 .In addition, Dolph and Santeufemio [12] found that cooling the sample during FIB milling suppressed the reaction between Ga and the III-V elements in strained layer superlattice materials.They compared the results of low-temperature and room-temperature Ga-FIB etching and found that fewer Ga droplets were formed in the processed area, and the etching rate was increased following the low-temperature treatment.However, when the room temperature was resumed, Ga droplets reappeared in the previously clean area.Xia et al. [13] found that a smooth surface without Ga droplets could be obtained by using Ne-FIB at lower energies, but the milling rate was 5-7 times slower than that using Ga-FIB.
In GAE, the refresh time is the most critical parameter, as it determines the amount of assisted gas that is replenished.Fu et al. [14] studied the effect of the refresh time on the deposition thickness and found that the relationship was nonlinear and the thickness gradually decreased with an increase in the refresh time.Santschi et al. [15] studied a titanium material and found that with the introduction of XeF 2 gas, the etching rate increased seven times when the refresh time was greater than 50 ms, which reduced redeposition and improved the quality of the etching structures.Furthermore, Mahady et al. [16] proposed a simulation method for the GAE of SiO 2 material, and the simulation results showed that a better pore resolution could be achieved with a maximum gas flux than a medium gas flux.
This paper systematically investigated the FIB machining of GaAs and evaluated the contribution of XeF 2 GAE in the material removal process.The FIB milling procedure included outlining a square box and exposing the box area to a focused Ga beam at incident angles of 0° and 30°, respectively.To investigate the contribution of the refresh time during the GAE process, we first studied the material removal process in XeF 2 GAE using different refresh times ranging from 0 to 800 ms.By comparing the sputtering yield and the on-site residual element under different fabrication parameters, we obtained the optimized fabrication conditions for the FIB machining of GaAs.

Experimental and Methodology
A Si-doped single crystal (100) GaAs substrate was used for the FIB experiment.Prior to conducting FIB experiments, the sample was ultrasonically cleaned with ethanol, acetone, and deionized water, respectively.Fabrication was conducted on a dual-beam FIB system (FEI Helios G4 UX) equipped with a liquid Ga ion source and an XeF 2 GAE system.The energy of the ion beam was maintained at 30 keV, the ion beam current was 26 pA, and the corresponding beam diameter (full width at half maximum, FWHM) was 12.5 nm.A serpentine scanning strategy with a constant dwell time of 100 ns was used during the FIB milling process.The surface finish of the micro pockets fabricated by FIB machining was measured using a 3D surface profiler (Zygo 9000) with an RMS resolution of 0.008 nm and a maximum scan depth of 150 μm.To obtain the sputtering yield measurements, the volume of the material removed was measured using an atomic force microscope (JPK Nanowizard 4XP) with a tip radius of 7 nm (model of probe AC20TS).The residual element around the fabrication area was measured using an Electron Probe X-ray Micro Analyzer (JXA-8530F PLUS) at a voltage of 15 kV, which analyzes elements in the range of Be4-U92.During the FIB machining process, the auxiliary etching gas XeF 2 was alternately turned on and off to compare the GAE contribution to the fabrication process.A schematic diagram of the FIB milling process with and without GAE is shown in Fig. 1.

Material Removal Process
During the FIB machining of GaAs, droplet-like hemispheres were formed within the radiation area, as shown in Fig. 1, and these were attributed to the selective etching process.Under Ga ion beam radiation, Ga and As atoms are sputtered out from the substrate separately, and their sputtering yields differ.As shown in Fig. 2, the SRIM Monte Carlo simulations results revealed that the sputtering yields of As atoms were constantly higher than those of the Ga atoms at the different incident angles, and this resulted in the formation of a Ga-rich environment on the surface of the GaAs substrate (Additional file 1: Table S1).The redundant Ga atoms were further aggregated and presented as droplet-like Ga structures that were ultimately eliminated under GAE FIB machining.The application of XeF 2 auxiliary gas to the FIB mining process introduced a series of chemical reactions activated by ion beam radiation, and this effect not only accelerated the material removal process but also balanced the selective etching phenomenon of Ga and As elements.At an ambient temperature, the auxiliary etching gas XeF 2 reacted easily with GaAs, and the chemical reaction procedure was as follows: XeF 2 first dissociated on the surface of GaAs, releasing F atoms and Xe atoms under ion beam radiation.Xe atoms are inert and do not react with the GaAs substrate, but F atoms are highly active and initially attracted by the GaAs surface, and this resulted in the formation of an intermediate monofluoride, GaF, and AsF.As the reaction continued, unstable intermediate difluorides (GaF 2 , AsF 2 ) were formed, which ultimately resulted in the formation of chemically stable trifluorides (GaF 3 , AsF 3 ) [17].The local temperature was high within the ion beam sputtering area, and AsF 3 thus remained in a gaseous state and left the substrate surface through the pumping system.GaF 3 eventually formed into a thin solid layer at the surface and was sputtered off by the primary ion beam.The total amount of GaF 3 was similar to that of AsF 3 because Ga and As atoms were contributed to by the GaAs substrate, where the atomic number ratio of Ga:As was 1:1.We ignored the implantation of Ga atoms from the primary ion beam because their content was much lower than that of the Ga atoms within the GaAs substrate.As the Ga and As atoms were removed separately at a ratio of 1:1, the selective etching phenomenon was effectively suppressed.An illustration of the material removal process is shown in Fig. 3.

Sputtering Yield Variations
During the FIB processing, XeF 2 gas was introduced through the gas injection system (GIS) to the sample surface, and XeF 2 gas molecules were then dissociated to Xe and F atoms by the incident ion beam.With the existence of F atoms, the sputtering process was highly related to the chemical reaction rate.Therefore, it was necessary to re-evaluate the sputtering yield under different chemical reaction conditions.The refresh time is the most important parameter in the GAE process; therefore, we conducted a series of experiments to reveal the relationship between the refresh time and the sputtering yield.In FIB processing, the refresh time generally refers to the time interval between two consecutive etchings at the same pixel point.During this process, the incident ion beam scans the entire surface point by point, and the required time is known as the loop time.The ion beam is then blocked, but the GIS remains open to maintain the supply of the assist gas.We first investigated the effect of the refresh time on the sputtering yield at different incident angles.As shown in Fig. 4, the sputtering yield started from 3.9 at a refresh time of 0 ms under an incidence angle of 0°; it then reached a peak of 5.4 at 500 ms and then decreased to 4.4 at 800 ms.However, with an incident angle of 30°, the change in the sputtering yield differed significantly from that under the 0° incidence angle.In this case, with an increase in the refresh time, the sputtering yield started from 6.7 at a refresh time of 0 ms; it then decreased from 20 to 500 ms, reached a minimum value of 6.3 at 500 ms, and subsequently increased to 6.6.
The results revealed that the refresh time affected the sputtering yield under the different incident angles.With an incident angle of 0°, the enhanced etching effect of XeF 2 on the substrate and the corresponding sputtering yield was increased with an increase in the refresh time.However, there was a maximum limit to this process, and this was probably due to the gas supply, which formed a thin layer of GaF 3 on the surface of the GaAs substrate that hindered the chemical etching of F atoms on the GaAs substrate and reduced the generation of GaF 3 , thereby leading to a decrease in the sputtering yield.
At an incident angle of 30°, the enhancing etching effect of XeF 2 on the substrate gradually increased with the elevated refresh time, resulting in a gradual increase in the sputtering yield.However, within a refresh time range of 20 ms to 500 ms, a higher sputtering yield occurred on the GaAs substrate in relation to the abnormal incidence angle (angle of incidence is not 0°).In addition, ion beam bombardment had a weaker heating effect on the local substrate than at a vertical incidence angle.To compare the energy loss of these two incidence modes, we used an SRIM Monte Carlo simulation, and the results showed that 36.5% more energy was removed at an incidence angle of 30° than at a normal incidence angle (angle of incidence is 0°), which reduced the GaF 3 generation rate and ultimately decreased the sputtering yield (Additional file 1: Fig. S1).A continued increase in the refresh time to over 500 ms resulted in a rise in the XeF 2 gas supplement, which generated a higher amount of GaF 3 that led to an increase in the sputtering yield.
Fig. 3 Illustration of the material removal process: a gas-assisted etching process of XeF 2 ; b chemical process used to etch GaAs substrate by F atoms, where b1, b2, and b3 represent the processes that form fluoride, unstable difluoride, and the final product, trifluoride Fig. 4 Relationship between sputtering yield and refresh time under different incident angles

Surface Quality Analysis of GAE
The essence of FIB sputtering is the removal of substrate materials through physical collisions using high-energy ions.Therefore, even with GAE, ion injection damage occurs on the surface of the FIB processed area, which results in crystal defects (such as vacancies, dislocations, amorphization, and expansion) [18][19][20][21], and this has a significant impact on the performance of quantum devices that have strict surface quality requirements.To investigate the impact of the refresh times on surface quality, we selected two representative refresh time values of 0 ms and 500 ms for use in experimental validation.As shown in Fig. 5, we measured the profile and the roughness of the bottom surface of the FIB machined rectangular cavities at refresh times of 0 ms and 500 ms, respectively, and the surface qualities obtained under the two different refresh times are shown in Table 1.With an incident angle of 0°, the arithmetic mean roughness (Ra, 8.534 nm) of the rectangular cavity at a refresh time of 500 ms was slightly higher than that of 0 ms (8.298 nm), but the root mean square roughness (Rq) and the total height of the contour peak and valley (Rt) were both lower than those of 0 ms.A comparison between the processed surfaces at incidence angles of 0° and 30° showed that the surface roughness at an incidence angle of 30° was systematically better than that at 0° for refresh times of both 0 ms and 500 ms.We observed that no obvious step-like patterns appeared on the bottom surface of the rectangular concave pits at an incidence angle of 30°, and the overall surface quality was slightly superior to that achieved with an angle of 0°.This indicates that when the ion dose remained constant during XeF 2 GAE of Ga arsenide, changes in the incidence angle and variations in the auxiliary gas dose did not cause significant alterations in surface roughness.Therefore, the ion dose remains the primary factor influencing changes in surface quality [22,23].
In addition to surface roughness, we also analyzed the residual elements within the FIB-machined area using EPMA, as shown in Fig. 6.The experimental results showed that there was no significant difference between the Xe element content of the processed and non-processed surface areas.This result indicated that during the GAE process, ion beam radiation had no impact on the adsorption of Xe atoms due to the inert property of Xenon.The measurement result also revealed that within the FIB processed area, the Xe content per unit area was slightly lower at a refresh time of 500 ms (0.92%) than that of 0 ms (0.97%).
Fig. 5 Variations in the cross-sectional profile along the measurement region at the bottom of the cavity, as measured by atomic force microscope (AFM).AFM measurement results for a incidence angle of 30° and a refresh time of 0 ms machined surface; b incidence angle of 30° and a refresh time of 500 ms machined surface; c incidence angle of 0° and a refresh time of 0 ms machined surface; d incidence angle of 0° and a refresh time of 500 ms machined surface This could be attributed to the purging effect from the XeF 2 auxiliary gas, which blew the inert Xe atoms away from the substrate surface during the ion beam refresh period.However, there were significant differences between the distribution of F elements in the processed and non-processed areas.As shown in Fig. 7, the F content per unit area in the FIB processed area was constantly higher than that in the non-processed area.With refresh times of 0 ms and 500 ms, the F content per unit area in the processed area was 0.28% and 0.12%, respectively, while it was 0.18% and 0.05%, respectively, in the non-processed area.The F distribution was therefore more concentrated in the processed region under both refresh times.There may be two reasons for this phenomenon: (i) ion radiation promoted the adsorption of F atoms on the surface of the sample, resulting in more F atoms reacting with the substrate in the processing area and leading to an increase in the concentration of the F element; (ii) the products of the reaction between F element and GaAs were not completely sputtered by the ion beam, leaving a certain amount of fluoride in the processed area.We also found that the areal concentration of the F element in the processing area was lower at a refresh time of 500 ms than at 0 ms.We believe this may have been due to the increased refresh time, which allowed F to react sufficiently with the GaAs substrate and cause GaF 3 to become the main reaction product.GaF 3 was then sputtered by the ion beam, leading to a significant reduction in the F content in the processed area.
The AFM and EPMA results revealed that the surface quality of the GAE processed sample was almost identical at the same incidence angle at refresh times of 0 ms and 500 ms.However, the F element content in the processed area was lower at 500 ms than at 0 ms.Therefore, it is recommended that when conducting GAE machining, an inclined incident ion beam with a longer refresh time should be used to achieve a better surface finish at a higher machining speed.

Summary
In this study, we systematically investigated the effects of two different incident angles and refresh times on the FIB GAE process of GaAs substrate when using XeF 2 as the auxiliary gas.The results revealed the following: (i) the variation in the sputtering yield differed with different refresh times and incident angles.At an incidence angle of 0°, the sputtering yield increased with an increase in the refresh time, but there was an upper limit to the effect of this process.When the auxiliary gas was replenished, a layer of GaF 3 formed on the surface of GaAs, which suppressed the chemical etching of F atoms, reduced the formation of GaF 3 and consequently caused a decrease in the sputtering yield.At an abnormal incidence angle of 30°, a greater amount of the substrate material was removed, which suppressed the heating effect of the ion beam on the substrate, slowed the generation of GaF 3 , and caused a decrease in the sputtering yield within 20 ms to 500 ms.However, with an increase in the refresh time, greater numbers of F atoms participated in the reaction and more GaF 3 was produced, which led to the recovery of the sputtering yield.(ii) At different refresh times, the surface roughness of the bottom of the rectangular cavity fabricated by FIB GAE was similar.However, the surface roughness was improved at an incident angle of 30° compared to 0°. (iii) GAE machining did not affect the adsorption of Xe elements, and Xe was almost uniformly distributed over the entire surface.However, under the condition of a longer refresh time, the Xe content in the FIB machining area was lower (per unit area) due to the removal effect of XeF 2 as the auxiliary gas.Unlike Xe, F elements were concentrated in the FIB machining area.This may have been due to the promotion of ion radiation on the adsorption of F atoms and the residue of reaction products.Furthermore, with an increase in the refresh time, the etching of GaAs by XeF 2 was sufficient; GaF 3 was produced, and it was more easily removed in the GAE process as the main reaction product.This significantly reduced the F content in the machining area.Therefore, machined surface quality can be improved by controlling the refresh time.

Fig. 1 Fig. 2
Fig. 1 Schematic diagram of focused ion beam machining of GaAs with and without GAE.a Gas-assisted etching schematic; b Comparison of surface finish with and without gas-assisted

Fig. 6 Fig. 7
Fig.6 Elemental analysis of the processed surface with refresh times of 0 ms and 500 ms using Electron Probe X-ray Micro-Analyzer.a Surface distribution of F elements with a refresh time of 0 ms.b Surface distribution of Xe elements with a refresh time of 0 ms.c Surface distribution of F elements with a refresh time of 500 ms.d Surface distribution of Xe elements with a refresh time of 500 ms

Table 1
Surface roughness of focused ion beam machined areas under different machining conditions