Ion Implantation Surface Modification Simulation

In real ion implantation process, the ion-material interaction is swift and in atomic scale, which can hardly been observed. Nanometric processing involves changes of the material in only a few nanometers from the surface. It is difficult to observe and illustrate accurately the structural changes during the process. Molecular dynamics (MD) simulation method becomes an important tool to analyze the mechanism of ion implantation surface modification and the follow-up processing.
MD simulations are carried out to study the mechanism of monocrystalline silicon surfaces modified by ion implantation. Lattice transformation is observed during the implantation process to examine the relationship between the damage mechanism and the implantation parameters. Nanoindentation simulations are also carried out, and it is proved that the ion implantation modification indeed influences the mechanical property of monocrystalline silicon and the mechanism of the mechanical processing of the material. Experiments of ion implantation, nanoindentation, and nano-scratching are employed to verify the simulation results.

MD Simulation Model

The simulations are carried out using large-scale atomic/molecular massively parallel simulator (LAMMPS) (Plimpton 1995). The use of molecular dynamics simulation to study atom irradiation effects of solids (Pelaz et al. 2004) and nanometric processing (Fang et al. 2005, 2007) has become an effective approach. Three-dimensional model is employed in the simulation. Figure 3 is the 3-dimensional MD model of the system. In the model, the workpiece is silicon consisting of 57,600 atoms and has been divided into three different layers: boundary layer, thermo layer, and Newton layer. Atoms in the boundary are fixed in position to avoid unexpected movement of the workpiece during simulation and reduce the boundary effects. Thermo layer is used to conduct away the outward heat from the interaction region by maintaining a constant temperature of 300 K. The motion of the atoms in the Newton layer is determined by the forces produced by Newton’s equation of motion. Figure 3a is the ion implantation simulation model. The implantation position for ions is randomly chosen. Perpendicular incident is adopted for computationally simpler handling. To modify the surface to a few micrometers, heavy ions and MeV implantation energy are adopted in the experiment. Considering the scale of the simulation, 30 keV is chosen. The simulation time step is prolonged to 1 f. to adapt to the energy. Figure 3b is the nanoindentation model. The nanoindenter in the simulation is regarded as rigid body so that the atoms are fixed relatively to each other and with no interaction force between the atoms. Details of the workpiece and simulation conditions are summarized in Table 1.

MD Simulation Potentials

There are two different atomic interactions in the MD simulations processes: Eq. 1 the interaction in the workpiece and Eq. 2 the interactions between the workpiece and the implant ions and between the workpiece and the diamond indenter.
For the interaction between the silicon atoms in the workpiece, Tersoff potential (Tersoff 1988) is used. The interactions between silicon atoms and implant ions and between silicon atoms and carbon atoms of the diamond indenter were described by Morse potential.
In ion implantation simulation, Si atoms interact with each other normally via the Stillinger-Weber (SW) potential (Stephenson et al. 1996). However, the standard potential is Tersoff for nanoindentation simulation to observe the lattice evolution (Cheong and Zhang 2000), and the mechanism changes of nanometric processing is the main interest. SW potential is sacrificed in order for the continuous and quantitative result in simulation.

MD Analysis Algorithm

Radial distribution function (RDF) and a geometric criterion for damage detection (Caturla et al. 1996) are used in the simulation to analyze the simulation process. RDF, also called g(r), is a useful index to judge phase transformation by evaluating the number of neighbor atoms and obtaining atomic distribution. RDF quantitatively analyzes the structural changes of the workpiece during ion implantation in histogram form by evaluating the statistical information of bond length variation.
The nearest-neighbor distance in c-Si, rNN = 2.35 Ǻ, is used for damage detection. A sphere around each atom with radius rNN/2 is defined, and all the atoms at the initial undisturbed lattice of occupied atoms sites are regarded as ordered, which follows the concept of the Lindemann sphere (Hensel and Urbassek 1998). A Si atom that is outside all the spheres is categorized as interstitial, while within the spheres, when no atom is found, a spot categorized is defined as a vacancy. Here the concept of lattice site is not adopted for disordered zones, for the atom displacements are large in both high-energy bombardment and mechanical processing.

Ion Implantation Process

High-energy ions bombard the material surface and interact with silicon atoms, as shown in Fig. 4. Figure 4a illustrates the ion-substrate interaction in ion implantation process, while Fig. 4b is the simulated implantation process. One part of ions collides with the surface atoms and is backscattered; the other part penetrates into the material, and cascade collides with atoms inside the material, resulting in a collision cascade. The penetrated ions gradually lose their energy, finally rest in a certain depth of the material, and form ion implantation. Substrate atoms are displaced or excited by atom collision. A number of atoms affected increase exponentially, where some are even bounced into the inner layer or sputtered out of the material. Secondary electron and X-ray are also emitted during the process.
Ion implantation surface modification is created by implanted ions and bounced atoms penetrating into the surface. The lattice is disturbed under the impact of the incident ions, resulting in dislocations and forming ion implantation damages. The structural changes in the implantation layer are the result of the critical balance between the damage generation and its annihilation. The depth of the modified layer is related to the incident energy, incident angle, incident ion species, and surface conditions. Ions are energy free when reaching the ion resting layer and the lattice structure in this region is influenced by the resting ions. The structural changes are smaller than the modification layer. The substrate is not influenced by the implantation.

Figure 5 shows the RDF results of the lattice transformation during different ion implantation periods. Different peak distributions of RDF curves illustrate the changes in atomic arrangement. The width of the peak indicates the density of a specific crystal form; the narrower the peak is, the more lattices are in perfect order and vice versa.
Before implantation, silicon exhibits a narrow and clear single peak, indicating that the substrate is perfect diamond cubic crystal. Due to the cascade collision, the crystal structure of the material is significantly been influenced during ion implantation. The peak is shortened and broadened, a series of second peaks appear, and the peak shape is no longer clear, which shows the existence of non-ideal crystal structure. Although how far the disordered zones contribute to the well-known phenomenon of ion beam-induced amorphization has not been resolved up to now (Pelaz et al. 2004), considering the absence of a significant peak, it can be predicted that the material has the tendency of becoming amorphous. After implantation, due to the self-annealing phenomenon of silicon, the material recrystallizes and monocrystal peak reappears. However, the peak is wider, and also with a second peak, indicating that even self-annealing happens, the structure changes still exist.

Ion Implantation Damage

Cascade collision introduced by ion implantation induces lattice structure changes. According to the result of RDF analysis, the damage would not completely heal even after implantation. The factors affecting implantation include implantation energy, ion fluence, and beam current. Implantation energy has an approximately linear relationship with the depth of modification layer, which could be adjusted to fulfill the application requirements. Beam current relates closely to ion fluence, but it would not be considered in the simulation, since the thermo layer is adopted to constrain temperature, and beam current is strictly restricted in the real experiment for reducing self-annealing. Figure 6 shows the vacancy and interstitial detected by geometric criterion under ion implantation with ion fluence of 1 x 1013, 5 x 1013, and 1 x 1014 ion/cm2. Perfect lattice structure is processed to be semitransparent and the simulation border is defined.
Ion implantation would also introduce defects in the form of clusters due to the cascade collision. With an increase in the implantation fluence, the number of defects rises correspondingly, and the defect clusters are more obvious and concentrated, while the percentage of monocrystalline structure diminish accordingly. When ion fluence reaches to a certain extent, it can be assumed that the material has the tendency of becoming an amorphous state. Since the ion implantation location is randomly chosen, the defects can be regarded to be evenly distributed as a whole, and the mechanical property changes of the substrate are uniform and controllable.

Mechanical Property Testing Simulation for Nanoindentation of Ion-Implanted Surfaces

Molecular dynamics are carried out on both normal and implanted silicons for three-dimensional nanoindentation simulation to analyze the changes of mechanical property and the mechanism of the changes. For nanoindentation simulation, round diamond indenter with a diameter of 15a (a = 3.57 A°) is used to carry out constant load tests on material surfaces. The loading speed is 4 m/s. Due to the scale of the simulation, indentation depth is set to be 0.6 nm.
The indenter is set to be rigid to ensure that they would not interact with the testing material, which makes the simulation results closer to reality. The loading force is calculated using Morse potential, which is the composite force workpiece atoms applied on the indenter.
In nanoindentation, diamond indenter produces hydrostatic pressure on the workpiece. Single-crystal silicon has perfect diamond lattice at ambient temperature. The hydrostatic pressure generates lattice slip and lattice disorder, which could be seen in Fig. 7b. Defect-free atoms in the workpiece are minimized for visualization, and lattice defects can be observed. Layer by layer, the lattice slip extends to the scale of the indentation depth. As lattice slip extends to the threshold, the indentation process steps into an equilibrant state; lattice deformation only exists around the indenter, as shown in Fig. 7c.

After ion implantation, the lattice structure of the surface layer of the monocrystalline silicon has been transformed due to the ion bombardment. The silicon atoms are no longer arranged in an ordered structure. The internal damage created by ion implantation acts as the core to absorb the indenter-induced energy and expands during the indentation process which prevents lattice slip and dislocation from carrying out.
As the indenter gradually presses deep into the material, the damage inside the workpiece disappears due to self-relaxation. The lattice deformation only occurs around the diamond edge. The process is shown in Fig. 7. Figure 7a–c illustrate the indentation process of the monocrystalline silicon, while Fig. 7d and e show the structural changes with modified silicon. Low ion fluence of 1 x 1013 ion/cm2 was chosen to illustrate the phenomenon. The ion-implanted silicon reduces the resistance against load. The brittleness is reduced and the elasticity is enhanced.
The load–displacement curves are shown in Fig. 8. Compared with monocrystalline material, the silicon surface modified by ion implantation requires smaller loading force when reaching identical loading depth. Therefore, the mechanical property of implanted silicon is improved and the hardness is reduced. With an increase in the implantation ion fluence, the loading force decreases correspondingly, indicating that more lattices are transformed to amorphous structures and the mechanical property is enhanced more significantly. The displacement discontinuity (pop-out) of single-crystal brittle material disappears by observing the unload-displacement curves, illustrating that the brittleness of the material is reduced and the plasticity enhanced after ion implantation, which is beneficial for generating damage-free surface finish. Nanometric cutting of the single-crystal brittle material after ion modification shares a similar mechanism.

Ion Implantation Experiment

Ion implantations are carried out on monocrystalline silicon in order to prove the validation of the method, choosing implantation depth correspond to the manufacturing scale and low ion fluence to minimize defects quantity. The crystalto-amorphous transition is examined by transmission electron microscopy (TEM), and mechanical property changes are examined using large penetration depth nanoindentation.

Ion Implantation

Ion implantation are performed on (100)-oriented monocrystalline silicon samples at room temperature using a van de Graaff accelerator, with an implantation angle of 70 respective to the surface direction to minimize channeling effects. Specimens are bombarded under vacuum at an energy of 10.0 MeV and a fluence of 1 x 1013, 5 x 1013 and (c) 1 x 1014 ion/cm2. Implantation angle of 70 to the surface direction is adopted to minimize channeling effects.

Focused Ion Beam Preparation of TEM Samples

Cross-sectional transmission electron microscopy (TEM) sample lamellae is prepared to study the structural changes of the modification layer using a focused ion beam (FIB) system equipped with an in situ Kleindiek rotational nanomotor. The angle between the ion and electron beam is 520. Electron beam is used for real-time accurate observation during the operation which introduces least damage to the sample.
The energy and current of both beams (ion beam and electron beam) are optimized to reduce obvious specimen damage during TEM specimen preparation. The accelerating voltage and current of the incident e-beam is maintained at 5 kV and 98 pA, throughout the entire process. An incident ion beam accelerating voltage of 30 kV, with an ion current ranging from 0.5 nA to 10 pA, is used during the FIB milling process.
The specimen preparation follows the in situ H-bar method, which is more efficient and flexible than the conventional technique (Mayer et al. 2007) for improving the yield and reducing the damage during FIB TEM specimen preparation. The implanted surface layer is mechanically trimmed to about 10 μm and glued on the copper grid. The TEM copper grid is then clamped by nanotweezer and attached to the rotational nanomotor, as shown in Fig. 9.

The nanomotor can be rotated by 3600 continuously with 10-7 rad resolution, which is used for tilting the lamellae surface vertical to the ion beam inside the FIB chamber.
Before thinning, a Pt line should be deposited on the surface to protect the top portion of the specimen from Ga ion implantation. For e-beam deposition introduces least damage to the specimen (Decoster and Vantomme 2009), electron beam-induced Pt deposition is initially used to deposit a thin film (about 35–50 nm). Then, focused ion beam-induced deposition (30 kv, 30 pA) is used to further deposit a thick film (about 500 nm) for sufficient protection from the subsequent FIB cross-section process. After cutting two deep trenches on either side of the lamella, the specimen is less than 100 nm thick, which is ready for the TEM observation, as shown in Figs. 10 and 11a. The TEM specimen is then characterized using a TEM at acceleration voltage of 200 kV.

Figure 11b shows TEM and SAD patterns of the modification layer, showing the implantation effect on lattice structure. Both halo rings and diffraction spots can be seen, but the spots are weak, indicating the structure is almost amorphous. The critical influence for amorphous is determined by ion energies when ion mass is settled (Decoster and Vantomme 2009). When the ion energy reaches 10 MeV, 1 x 1014 ion/cm2 reaches the critical influence. In the ion resting layer, ions are energy free when reaching the resting layer and the lattice structure in this region is influenced by the resting ions. The structural changes are smaller than the modification layer. The substrate is not influenced by the implantation.
After implantation, the surface layer of the silicon is modified to a depth up to 4 μm, as shown in Fig. 12. Cross-sectional TEM results are shown in Fig. 13.

After modification, the sample exhibits different crystal structures according to the implantation fluence. Different crystallographic orientations can be seen on different parts of the TEM image in Fig. 13a, indicating that the perfect crystal structure has been divided into several parts. In Fig. 13b, when the ion fluence increases to 5 x 1013 ion/cm2, crystal structure could only be observed in a few places. As the ion fluence rises to 1 x 1014 ion/cm2, the sample becomes significantly amorphous, and the crystal structure totally disappears, as shown in Fig. 13c.

Nanoindentation and Nano-scratching Analyses

Nanoindentation measurements are carried out with a nanoindenter using highprecision Berkovich diamond tip to measure the mechanical property changes of the implanted materials. Surface oxide layer is removed by hydrofluoric acid to exclude the influence. Tests are conducted under indentation penetration depth of 3 μm to minimize the influence of the bulk material and measure the mechanical property at nano-manufacturing scale. The results are shown in Fig. 14, which show that the actual surface modification effect by ion implantation is more obvious comparing to the simulation results. With the same displacements, monocrystalline silicon requires a larger load (up to 130 mN) to reach the identical penetration depth. Displacement discontinuity (pop-out) is observed in unloading curves of the silicon, which is related to the density change caused by the high-pressure phase transformation (Fang et al. 2007). While implanted silicon exhibits a smooth curve, indicating no phase transformation occurs during unloading, as the phase transformation already takes place during the implantation process. With an increase in the implantation fluence, the hardness decreases accordingly. It should be noted that the loading force reduction rate is not in direct proportion to the ion fluence, but getting smaller when the fluence is relatively large, indicating the degree of the amorphous becomes saturated, close to the reduction limit.
Figure 14 shows the load–displacement curves of the indentation tests. With the same displacements, normal silicon requires a larger load (up to 130 mN) to reach the identical penetration depth, but the implanted silicon only needs about 35 mN. Displacement discontinuity (pop-out) is observed in unloading curves of normal silicon, which is related to the density change caused by the high-pressure phase transformation, while implanted silicon exhibits a smooth curve, indicating no phase transformation occurs during unloading, for phase transformation have already been carried out during implantation process.

After ion implantation, the structure of the monocrystalline silicon’s surface layer turns amorphous due to the ion bombardment. The bonding force between atoms greatly reduced, while the Si-Si bonding has even been broken. Therefore, the implanted silicon performs a different mechanical behavior. The resistance against load was lowered, the brittleness reduced, and the elasticity enhanced. Figure 15 shows the results of hardness and elastic modulus measurements of 3 μm indentations on normal and implanted silicon. Depth-sensing indentation hardness is evaluated from the indentation load-displacement curve using the Oliver and Pharr method. Young’s modulus is deduced from the unload-displacement curve using a Poisson’s ration υ = 0.25, with a Young’s modulus of 1,141 GPa and υ = 0.07 for the diamond indenter.
Surface topography after nanoindentation can be observed by SEM, and the results are shown in Fig. 16. After ion implantation, the pressure mark exhibits smoother surface and the indenting area shows smaller brittle cracking after indentation. From the results, it is predicted that ion injection offers a possibility of reducing surface roughness and tool wear in micromachining process. With an increase in the implantation fluence, the hardness decreases accordingly. It should be noted that the loading force reduction rate is not in direct proportion to the ion fluence, but getting smaller when the fluence is relatively large, which indicates the degree of the amorphous becomes saturated, close to the reduction limit, as shown in Fig. 17.

In situ nanomechanical test system is used to carry out the nano-scratching tests with Berkovich diamond tip. Surface topography after nano-scratching is observed in situ AFM scanning, and the results are shown in Fig. 18. After ion implantation, the scratching mark exhibits smoother surface and smaller brittle cracks. The lateral force measured during scratching shows similar results as nanoindentation. According to the results, it is predicted that ion implantation offers a possibility for the high-quality nanomachining process through enhancing the plasticity of the brittle monocrystalline silicon.
Ion implantation surface modification provides a novel approach for manufacturing brittle monocrystalline materials. The mechanical properties of the surface layer of the material are modified in a depth up to a few micrometers by decrystallization using high-energy ion bombardments; the brittleness is reduced and the plasticity is enhanced. The productivity is greatly improved by avoiding the fractures during the surface processing.

Potential Developments

Since the ion implantation surface modification method for ultra-precision machining has been proved to be valid, applications of the method for actual production need to be considered. Low-cost and large-implanted area should be taken priority. Figure 19 shows various radiation methods and their damage to the substrate. After neutron irradiation, it should be waited for a half-life of 3–6 months due to the activation phenomenon, which is a great obstacle to the productivity. Electron irradiation and gamma irradiation have been applied to industrial production. The effectiveness of the gamma-ray irradiation is taken into consideration first.

Gamma Irradiation

It is found that gamma irradiation can only generate effective modification in n-type silicon. Ultra-precision turning is used to conduct taper cutting in order to test the brittle-ductile transition of gamma-irradiated silicon. The test results are examined by atomic force microscope. Nanoindentation results show that only the low-doped n-type (100) silicon wafers have a significant reduction of the modulus (Table 2).
Test results of atomic force microscope on the brittle-ductile transition point are shown in Fig. 20, which is the low-doped n-type (100) silicon with 600 Gy gamma irradiation. It can be seen from the figure that the brittle-ductile transition depth is 136.4 nm.
However, since the changes only occur in some types of silicon, this result may not be valid to prove the effectiveness of gamma irradiation. The reason for this phenomenon is considered to be the doping on silicon wafers; some doped composition can be activated by gamma irradiation, resulting in internal lattice structure changes, and the mechanical properties change; the silicon atom itself has not received much impact. For the widely used high resistance silicon which does not exist doped, gamma irradiation effect is not obvious.
Therefore, it is concluded that gamma irradiation surface modification for certain types of silicon can be effective.

Low-Energy Swift Ion Implantation

Low-energy ion implantation experiments are conducted on high-resistivity silicon to verify the validity of this method.
Industrial ion implantation machine can be used for low-energy swift ion implantation for a large area to improve the implantation efficiency. The implantation beam current is 40 μA/s, and the implantation can be completed in a short time. The results of H+ and He ions into silicon of nanoindentation and mechanical properties testing are shown in Table 3.
Low-energy ion implantations have no obvious influence on the surface mechanical properties of materials regardless of hardness or modulus. The phenomenon from the principle of low-energy ion implantation of the material could be predicted, where the damage is mainly due to nuclear energy loss, while the high-energy ion implantation damage is dominated by the electronic energy loss. In order to achieve the effect F-ion 10 MeV 1 x  1014 ion/cm2 introduced, low-energy light ion implantation requires a higher dose, even multi-implantation.

Multi-implantation

With single low-energy light ion implantation, the damage layer stays from the surface with a depth of 2–3 μm. Figure 21 shows the SRIM simulation results of N ion with 550 keV nuclear energy loss.
It is believed that through a combination of different energies, different fluences, and different particles, multi-implantation can be flexible and controllable to achieve the uniform damage layer into the material surface. The following is an example of multi-implantation simulated by the nuclear energy loss, with the comparison with high-energy F-ion implantation (see Fig. 22).

In the example, the F-ion 10 MeV 5 x 1014 ion/cm2 nuclear damage is set as the target for multi-implantation simulation. He and B ions are used in the simulation to form a superimposed damage. The results show that the nuclear damage by four times implantation could meet the requirements for a uniform implant layer.