Focused-ion beam patterning of organolead trihalide perovskite for subwavelength grating nanophotonic applications

The coherent amplified spontaneous emission and high photoluminescence quantum efficiency of organolead trihalide perovskite have led to research interest in this material for use in photonic devices. In this paper, the authors present a focused-ion beam patterning strategy for methylammonium lead tribromide (MAPbBr3) perovskite crystal for subwavelength grating nanophotonic applications. The essential parameters for milling, such as the number of scan passes, dwell time, ion dose, ion current, ion incident angle, and gas-assisted etching, were experimentally evaluated to determine the sputtering yield of the perovskite. Based on our patterning conditions, the authors observed that the sputtering yield ranged from 0.0302 to 0.0719 μm3/pC for the MAPbBr3 perovskite crystal. Using XeF2 for the focused-ion beam gas-assisted etching, the authors determined that the etching rate was reduced to between 0.40 and 0.97, depending on the ion dose, compared with milling with ions only. Using the optimized patterning...


I. INTRODUCTION
2][3][4] Recent demonstrations of coherent amplified spontaneous emission 5 and high photoluminescence (PL) quantum efficiency ($70%) that lead to optically pumped lasing 6 in perovskite material systems have initiated research for this material application in light-emitting diodes (LEDs) and laser diodes (LDs).6][7] To prepare perovskite LDs, optical cavity formation using a distributed Bragg reflector (DBR), 6 planar whispering gallery cavities 8 or spherical resonators 9 has been utilized to enhance the light amplification in the perovskite LDs.With the formation of these optical cavities, optically pumped lasing has been achieved for perovskite LDs. 5,6owever, because perovskite is a hybrid organic-inorganic material, it is susceptible to material degradation and decomposition in solution-based and high-temperature device fabrication processing, which hinders the progress in realizing electrical-injection LDs.Defining the device structure (mesa) for index-guiding LDs (mode confinement) or patterning for metallization using photo-or electron beam-lithography techniques causes decomposition of the perovskite during the photoresist baking, development process or thermal evaporation.Alternatively, the focused-ion beam (FIB) technique provides the capability of direct maskless patterning (milling/etching or deposition) on most materials for fast device prototyping.Thus, fabricating surface-emission or in-plane emission perovskite LDs will be straightforward and will not require subsequent etch-back or lift-off device processing, as is required in the lithographic routine.Furthermore, FIB has the additional gas-assisted-etching (GAE) capability 10 to tune the etch rate and improve the morphology of the milled area.
Here, we demonstrate a systematic patterning strategy for the MAPbBr 3 perovskite crystal using the FIB technique, which has not been carried out previously.To optimize the FIB patterning on MAPbBr 3 , we investigated milling parameters such as the dwell time, scan pass, ion dose, ion current, ion incident angle, and GAE using XeF 2 .We determined the sputtering yield and sputtering rate of the MAPbBr 3 perovskite for the range of our experimental FIB parameters.The milling rate was also observed to be controllable with the application of XeF 2 for the GAE FIB.Based on the milling parameters optimization, we demonstrated binary and circular subwavelength grating (SWG) reflectors fabrications with nanoscale precision and excellent uniformity on the MAPbBr 3 perovskite crystal.The SWG reflector typically exhibits high reflectivity (>90%), broadband spectrum (>10% of design wavelength), and polarization selectivity within a very thin layer (submicrons) compared with DBR (few microns). 11Thus, perovskite-based SWG reflector can be utilized to form an optical cavity in a perovskite surface-or in-plane emission LDs, a resonant cavity LEDs or a resonant cavity photodetector.Highly precise and uniform nanopatterning technique utilizing FIB could provide a direct approach to achieving perovskite-based photonic nanostructures or nanodevices, which has not been demonstrated yet.

II. EXPERIMENT A. Sample preparation
The MAPbBr 3 perovskites crystals were fabricated using two steps preparation: (1) precursor synthesis and (2) material crystallization using antisolvent vapor-assisted crystallization.First, the perovskites halide precursors CH 3 NH 3 þ and Br À were synthesized through the reaction of HBr acid with methylamine followed by recrystallization from ethanol.Then, the synthesized CH 3 NH 3 Br and PbBr 2 were dissolved in N,N-dimethylformamide.The MAPbBr 3 perovskites crystals were grown along with the slow diffusion of the vapor of the antisolvent dichloromethane in to the solution.Figure 1 shows the full top-view scanning electron microscopy (SEM) image of MAPbBr 3 perovskite crystal.The fabrication details and material characterizations of x-ray diffraction (XRD) and PL are given in previous published work. 12

B. FIB processing
For The FIB patterning was performed using a FEI Quanta 3DFEG SEM/FIB dual beam system that consisted of both electron-and ion-columns.Rectangular trenches of 7 Â 4 lm were milled to investigate the FIB patterning parameters of the dwell time, scan pass, ion dose, ion current, ion incident angle, and GAE.For the GAE, we used a XeF 2 gas injection system.If not mentioned explicitly, the FIB parameters were set to default values of Ga þ ion source accelerated voltage ¼ 30 kV, probe current ¼ 30 pA (beam diameter-¼ approximately 17 nm), dwell time ¼ 1 ls, beam overlap ¼ 50%, serpentine scan, multiscan pass ¼ approximately 1000, and magnification ¼ 2500Â in quad-view mode.In our terminology, the FIB ion dose (I D ) defined as number of coulombs per micron squared that strike the substrate is calculated as where I ion is the ion beam current, T M is the milling time, and A is the patterning area.We calculated the sputter yield (S Y ) defined as volume of material removed per quantity of incident charge; and sputter rate (S R ) defined as volume of material removed per time; as in Eq. ( 2) and Eq.(3), respectively, where V is the milled volume calculated from the patterning area multiply with the milling depth.These values are determined from SEM measurement and optical surface profiler as described in Sec.II C. The etch factor (E F ) defined as the sputtering yield of GAE per sputtering yield with ions only, is calculated as C. FIB patterning measurement We utilized the SEM capability in the FEI Quanta 3DFEG SEM/FIB dual beam system and Zygo NewView 7300 3D optical surface profiler to measure the cross-section profile and milling volume of the patterned rectangular trenches.

D. SWG reflector design
We utilized rigorous-coupled wave analysis (RCWA) method to compute the reflectivity spectrum and electrical field for the MAPbBr 3 SWG reflector.Figure 2 shows the schematic of the MAPbBr 3 SWG reflector with grating parameters of period, duty cycle and thickness denoted as K, DC, and t correspondingly.The ratio of grating width to period is defined as the duty cycle.We considered the incident light propagate bottom up to the surface-normal plane through the SWG structure (z-direction).The transverse electric (TE)-and transverse magnetic (TM)-polarization are assumed parallel to-(y-direction) and perpendicular to-(x-direction) the grating lines, respectively.The MAPbBr 3 SWG reflector is designed to maximize reflectivity at 570 nm operating wavelength (corresponding to the PL emission peak) 12 with TE-polarization.From our ellipsometry measurement, the refractive index for MAPbBr 3 is around 1.9 at 570 nm wavelength.

III. RESULTS AND DISCUSSION
A. Scan pass and dwell time As observed in Fig. 3, more than 50 multiscan passes (shorter dwell time) produced uniform milling (a symmetrical rectangular profile) and smoother surface, which indicated no redeposition effect.However, for less multiscan (>10 loop) and the single scan, i.e., a longer dwell time, the milling profile was asymmetrical, with the trench being deeper along the direction of the scan (from left to right in Fig. 3).The accumulation of material at the sidewall of the opposite patterning direction (the region milled earlier) is due to the redeposited material from the sputtered perovskite.
Figure 4 plots the sputtering yield of MAPbBr 3 perovskite as a function of the number of loops and dwell time.The sputtering yield is approximately 0.03 lm 3 /pC for loop numbers >1000.When using more than 1000 loops (a shorter dwell time), an adequate multiscan is employed to lessen the redeposition effect as any redeposited material in the perovskite crystal from a previous scan milling pass will be removed in the subsequent milling pass.Moreover, a shorter dwell time can also reduce the amount of redeposition because the ions are not milled excessively in each beam spot.This effect has been reported in Ref. 13, where a higher sputtering yield was obtained for a longer dwell time.A longer dwell time also broadens the beam spot and sputter deeper in the Si case, 14 meaning that more material is sputtered out, which will increase the redeposition possibility.In addition to reducing the material redeposition, a shorter dwell time during patterning can improve the vertical sidewalls 15 that enhance the patterning precision and create an abrupt interface.

B. Ion dose and ion current
To study the effect of the ion dose and ion current on the sputtering yield and sputtering rate, we performed FIB milling using ion currents of 10, 30, and 50 pA.The milling time was increased from 5 to 60 s to vary the dosage of ion bombardment.Table I summarizes the applied ion dose for the respective current level.Figure 5 shows the milled rectangular trenches for each ion dose using the 30 pA ion current as an example.As the ion dose increased, the depth of milling increased.However, the sidewalls became slanted, and the surface became rougher with increasing ion dose because of material redeposition.
Figure 6 plots the milled volume as a function of the ion charge for 10, 30, and 50 pA ion currents.The milled volume is linearly dependent on the ion dose for each ion current.
From the slope of the linear fit curves, the calculated sputtering yields were determined to be 0.0459, 0.0302, and 0.0719 lm 3 /pC for ion currents of 10, 30, and 50 pA, respectively.These calculated sputtering yields are consistent with the constant sputtering yields achieved (above ion doses of 25 pC/lm 2 ) for each ion current, as shown in Fig. 7.
Figure 7 plots the sputtering yield versus the ion dose for 10, 30, and 50 pA ion currents.Initially, the sputtering yields are high but then become constant above an ion dose of 25 pC/lm 2 for all of the applied ion currents.We assumed that increasing the milling time to achieve a higher ion dose would increase the ion beam localization and that more atoms would thus be sputtered.A higher ion dose also contributes to more atoms being sputtered as the ion beam distribution is larger. 16The longer ion beam localization and larger distribution increased the redeposition effect as the ion dose intensified.After a certain milling action, the sputtering yields decreased for all of the ion currents, as described above.Instead of milling deeper, sputtering of the redeposited material was occurring.The reduction of the  sputtering yield and increase of the surface roughness as the ion dose increased are also reported elsewhere. 17,18igure 8 plots the sputtering rate versus ion dose for ion currents of 10, 30, and 50 pA, with sputter rates of approximately 0.50, 0.99, and 3.50 lm 3 /s, respectively.As expected, the sputtering rate is higher as the ion current increased because a larger ion beam bombardment is applied.The ion beam diameter is approximately 13, 17, and 19 nm for the ion currents of 10, 30, and 50 pA, respectively.However, the sputtering rates for all of the ion currents decreased and became constant when the ion dose intensified.This behavior is similar to the effect observed above (Fig. 7), where longer and larger ion beam localization plus distribution resulted in a material redeposition effect.
From Figs. 7 and 8, we observe that both the sputtering yield and sputtering rate become constant for ion doses above approximately 25 pC/lm 2 .To investigate the effect of the ion current only on the sputtering yield of MAPbBr 3 perovskite, various rectangular trenches were milled with ion currents of 1.5, 10, 30, 50, 100, 300, 500, and 1000 pA.The milling time was varied for each ion current to accommodate a constant ion dose of 35 pC/lm 2 .Figure 9 shows the milled rectangular trenches for each ion current described above.
Figure 10 shows that milling with higher yield was achieved for ion currents below 100 pA.However, the sputtering yield was approximately constant for ion currents of 100 pA or more.This result is observed because the redeposition effect is occurring for larger ion beam bombardment (the ion beam diameter range of 24-44 nm for ion currents ranging from 100 to 1000 pA) for the same ion dose.Based on the SEM images in Fig. 9, the surface of the rectangular trenches is overlaid with the redeposition from the sputtered perovskite as the ion current increased from 100 pA onwards.Instead of milling the perovskite deeper, the ions are milling the redeposition from the previous scanning pass.This phenomenon can be observed in Fig. 10, where the sputtering yield remains constant beyond 100 pA, although the sputtering rate is increased with increasing ion current.The same observation (constant sputtering yield) is also reported elsewhere. 18To obtain vertical sidewall patterning, using a low ion current is preferable to obtain finer ion beams, which can reduce material redeposition. 19,20 Incident angle  The sputtering yield is plotted as a function of the incident angle in Fig. 12.The sputtering yield increased exponentially with increasing incident angle.This behavior is due to the reduction of redeposition from the sputtered atoms of MAPbBr 3 perovskite as the incident angle increased, which resulted in deeper milling action.As the incident angle increased from the normal incidence, the possibility of collision between ions and the sputtered atoms of MAPbBr 3 perovskite decreased; hence, the redeposition effect was reduced.More sputtered atoms escaped from the surface that eventually increased the sputtering yield.More than 80 of angle milling is not significant because the sputtering yield will decrease dramatically because of reflection of the ions at the surface.17

D. Gas-assisted etching
To study the effect of GAE on the sputtering yield of MAPbBr 3 perovskite, we performed FIB milling with XeF 2 gas injected on the perovskite surface.The ion current is fixed at 30 pA whereas ion dose was varied from 5.36 to 107.14 pC/lm 2 by increasing the milling time.Figure 13 presents SEM images of the milled XeF 2 GAE sample.
XeF 2 GAE milling resulted in a lower sputtering yield of approximately 0.0118 lm 3 /pC, as calculated from the slope of the linear fit of the line in Fig. 14.In comparison, the calculated sputtering yield for ion milling only is higher at approximately 0.0287 lm 3 /pC.These calculated sputtering yields are within the sputtering yield values plotted versus ion dose in Fig. 15.The introduction of XeF 2 during the milling reduced the perovskite etch rate compared with milling with ions only.
Figure 16 shows that the etch factor is reduced to 0.40-0.97,depending on the ion dose.It is assumed that perovskite reacts chemically with fluorine (from XeF 2 ) to form nonvolatile products that deter sputtering, which results in a reduction of the sputtering yield.This behavior is similar to the H 2 O vapor effect on decreasing the sputtering yield of Si, SiO 2 , Si 3 N 4 , and Al for H 2 O GAE FIB milling. 21The reduction of the sputtering yield of MAPbBr 3 perovskite for XeF 2 GAE application could be useful in controlling the perovskite milling for nanoscale dimension because the obtained sputtering rate of approximately 0.99 lm 3 /s at 30 pA (refer Sec.III B) is considered fast.For comparison, the sputtering rates at 30 pA for Si, GaAs, and PMMA are approximately 0.007, 0.021, and 0.015 lm 3 /s, respectively, based on the sputtering yields of 0.24 lm 3 /nC for Si, 0.69 lm 3 /nC for GaAs, and 0.50 lm 3 /nC for PMMA. 22The reduced etch factor when applying XeF 2 GAE in perovskite   could also provide an etching selectivity possibility for specific application.

E. Nanophotonic applications: Binary and circular SWGs
Utilizing the results from the FIB patterning parameter studies (ion dose, sputtering yield, and sputtering rate), we fabricated binary and circular SWG reflectors on the MAPbBr 3 perovskite crystal based on the computed optimal grating parameters of K ¼ 534.8 nm, DC ¼ 0.49, and t ¼ 204.1 nm.The FIB patterning parameters applied are similar to the default setup described in Sec.II with the exception of the milling time.The milling time is 7 s for the binary SWG reflector and 40 s for the circular SWG reflector.Figures 17(a) and 17(b) demonstrate that the binary MAPbBr 3 SWG reflector is patterned precisely on the perovskite crystal with nanoscale precision (>62.5 nm) compared to the computed design for all grating parameters.The patterned binary SWG also exhibits high degree of uniformity and contrast.
Figures 18(a) and 18(b) present SEM images of the patterned circular MAPbBr 3 SWG reflector.Similar grating parameters of the binary SWG reflector design are utilized with only exception on the grating shape, which is circular.With this circular shape, the reflected light from the circular SWG reflector will have focusing effect. 23As shown in Figs.18(a) and 18(b), a high-uniformity and high-contrast grating are obtained from the FIB patterning on the MAPbBr 3 perovskite crystal.The fabricated grating parameters of period and duty cycle are highly precise (>63 nm) compared with the computed design.However, the fabricated grating thickness is approximately 70 nm thicker (fabricated thickness is 273.7 nm compared to design thickness of 201.4 nm) due to the patterning complexity of realizing high contrast circular profile.Nevertheless, our simulation (result not shown here) using this 70 nm grating thickness fabrication deviation exhibits only 4% reduction in the MAPbBr 3 SWG reflectivity (initially around 97%) whereas the peak wavelength remained the same around 570 nm.
To measure experimentally the reflection spectrum for both binary and circular MAPbBr 3 SWG reflectors, we require large area of at least 30 Â 30 lm patterning to match the beam spot of the light source in our optical measurement setup.However, we faced difficulty in patterning highcontrast and -uniformity of large scale MAPbBr 3 SWG reflector due to fast sputtering rate of the MAPbBr 3 material as discussed in Secs.III B and III D above.As a result, we computed the MAPbBr 3 SWG reflectivity spectrum for the current patterning area (<10 lm 2 ) using the RCWA method.The reflectivity, broadband, and polarization selectivity are the same for the binary and circular MAPbBr 3 SWG reflectors with the only exception of having the focusing effect from the reflected light for the circular SWG. Figure 19 shows the computed reflectivity spectrum for TE-and TM-polarization of MAPbBr 3 SWG reflector with grating parameters of K ¼ 534.8 nm, DC ¼ 0.49, and t ¼ 204.1 nm, respectively.It shows high reflectivity around 97% at 570 nm for the TE-polarization whereas the TM-polarization is being suppressed below 40% reflectivity, indicating the SWG reflector capability for polarization selectivity.The period for the grating is designed at subwavelength to achieve the SWG effect of reflecting the fundamental mode only.The MAPbBr 3 SWG reflector also exhibited around 23 nm broadband centered at 570 nm wavelength for reflectivity >90%.For comparison, we also computed the planar MAPbBr 3 /air interface (without SWG structure, named as control in Fig. 19), which exhibits reflectivity around 10%.Thus, significant reflectivity increment (almost 10Â magnitudes) surrounding the 570 nm operating wavelength is realized by designing SWG reflector for the MAPbBr 3 material.
We employed the RCWA to compute the near-field intensity profile of the electrical-field (jE y j 2 component) for the MAPbBr 3 SWG reflector.Figure 20 demonstrates the crosssection profile of the grating and surrounding media (air) per period for the SWG structure at incident light wavelength of 570 nm with TE-polarization.The leaky mode exhibits fundamental TE-like mode, indicating that only the zero-order diffraction is reflected.High intensity localization between the grating and air interface indicate that the incident light being highly reflected by the grating.There is almost no light propagate through the grating except for evanescent mode from high order diffractions.Most light from the bottom only propagate through the air gap (on both sides of the grating).
We have also performed PL characterization to investigate the material purity and possibility of surface damage at the surface of the FIB patterning area.Although the PL spectrum (result not shown here) indicate intensity degradation (approximately 60%) after the FIB patterning due to surface damage from the Ga ions, the PL peak wavelength is still exactly around 570 nm.This indicates that the area patterned using FIB still consists of all elements of the MAPbBr 3 material system.

IV. CONCLUSIONS
In this paper, a patterning strategy for MAPbBr 3 perovskite crystal using the FIB technique for nanophotonic applications was presented for the first time.Important FIB parameters such as the scan pass, dwell time, ion dose, ion current, and ion incident angle were studied to investigate the sputtering yield and sputtering rate of MAPbBr 3 perovskite.We also reported reduction in the etching rate of the perovskite using XeF 2 GAE during the FIB milling.Based on the FIB parametric study, the optimized FIB conditions were applied for SWG patterning on a MAPbBr 3 perovskite crystal.We patterned both binary and circular SWG reflectors to demonstrate the FIB capability in directly fabricating

Figure 3
Figure 3 presents the SEM images of eight rectangular trenches being milled on the MAPbBr 3 crystal from singleto multiscan pass (number of loops ¼ 1-5000) to investigate the scan pass and dwell time effect on the sputtering yield.The loop numbers of 1, 5, 10, 50, 100, 500, 1000, and 5000

FIG. 6 .
FIG. 6. (Color) Milling volume as a function of ion charge for various ion currents.The dotted lines are linear fits of the respective ion current.

Figure 11
Figure 11 presents an SEM image of rectangular trenches being milled to examine the effect of the ion incident angle on the sputtering yield.Upon increasing the incident angle

FIG. 10 .
FIG. 10. (Color) Sputtering yield and sputtering rate as a function of ion current.The sputtering yield dotted line serves to show the general trend.The sputtering rate dotted line is a linear fitting rate.

FIG. 12 .
FIG. 12. (Color) Sputtering yield as a function of incident angle.The dotted line serves to show the general trend.

FIG. 14 .
FIG. 14. (Color) Milling volume as a function of ion charge for ions only and XeF 2 GAE.The dotted lines represent linear fitting.

FIG. 16 .
FIG. 16. (Color) Etch factor as a function of ion dose for ions only and XeF 2 GAE.The dotted line serves to show the general trend.

FIG. 20 . 8 J
FIG. 20. (Color) Near-field intensity profile of the electrical-field for one period of MAPbBr 3 SWG reflector.

TABLE I .
Ion dose for the applied ion current.