Chapter 2 Experiments
2.2 Material Analysis and Electrical Measurement
The Current-voltage (I-V) characteristics of devices were measured by a semiconductor analyzer of model Agilent 4156C. The contact resistance was also measured by the same semiconductor analyzer. Sheet resistance was measured by a
24
four point probe system. Furthermore, the samples were analyzed by different kinds of material analysis instrument in order to compare with the results of electrical measurement, for example, X-ray Diffraction (XRD), Transmission Electron Microscopy (TEM), Scanning Electron Microscopy (SEM), and Secondary Ion Mass Spectrometer (SIMS).
XRD was used to analyze the crystallized orientation of NiGe. When X-ray goes through the crystallized structure, the constructive interference occurs. According the Bragg’s Law (nλ=2dsinθ), the constructive interference only occurs for certain θ correlating to a (h k l) plane. Therefore, the crystallized (h k l) plane of NiGe can be inferred. TEM was used to observe the microstructure of NiGe film. TEM depends on electrons beam transmitted through an ultra-thin specimen and the electrons beam interacts with the specimen as it passes through to form the image. The image is magnified and focused on the imaging device with high resolution. Therefore, the quality of NiGe film can be clearly observed. In addition, SEM is used to observe the surface of NiGe film. SEM produces images by means of scanning the samples with electrons beam and the electrons interact with atoms in the samples to detect the signal and show the samples’ surface topography. SIMS was used to analyze the distribution of elements in the samples, such as P, As, and Ni. SIMS analysis relies on sputtering the surface of the samples with a focused ion beam and analyzing the ejected secondary ions from the surface. The kind of elements could be determined by mass/charge ratios of these secondary ions. Moreover, the amount and depth distribution of the elements could also be detected at the same time.
In brief, all the material analysis plays the role of assistant to prove the measured electrical characteristics. As long as the results are consistent, the inference or conclusion would be more convinced.
25
Table 2-1 Main process recipe of conventional junctions.
Implantation Recipe Annealing Recipe Substrate P, 50keV, 5x1012 cm-2
Table 2-2 Main process recipe of IBG junctions.
Implantation Recipe Annealing Recipe NiGe formation Substrate P, 50keV, 5x1012 cm-2
26
Table 2-3 Main process recipe of IAG junctions.
NiGe Implantation Recipe Annealing Recipe Substrate 350oC, 5min As, 10keV, 1x1015 cm-2 500oC, 10s
550oC, 10s 600oC, 10s
Heavily
Table 2-4 Main process recipe of IBG+IAG junctions.
Implantation
Table 2-5 Main process recipe of Contact Resistance of IBG+IAG junctions Implantation
27
1. HF clean
2. 300nm SiO2 deposition 3. Active area defined by
lithography process 4. SiO2 etching by BOE 5. Remove PR by ACE
6. P or As implantation
7. Dopant activation (600oC,60s)
8. Ni(10nm)/TiN(15nm) deposition by Sputter system
9. NiGe formation
10. Unreacted-Ni and TiN were etched by hot H3PO4
28
Fig.2-1 Process flow of IBG+IAG junctions.
13. Al contact defined by lithography process 14. Al etching by HDPRIE 15. Al backside contact 12. Dopant activation 11. As implantation
IAG junctions without first implantation
29
1. HF clean
2. 300nm SiO2 deposition 3. Diffusion region defined by
lithography process 4. SiO2 etching by BOE 5. Remove PR by ACE
6. Implantation:
P, 20keV, 1x1015 cm-2 7. Dopant activation
(600oC,10s)
8. 300nm SiO2 deposition
9. Active region lithography 10. SiO2 etching by BOE 11. Remove PR by ACE
12. Ni(10nm)/TiN(15nm) deposition by Sputter system
13. NiGe formation (350oC, 5min)
14. Unreacted-Ni and TiN were etched by hot H3PO4
30
Fig.2-2 Process flow of Contact Resistance of IBG+IAG junctions.
15. Implantation :
As, 10keV, 1x1015 cm-2
16. Dopant activation
17. Al contact defined by lithography process 18. Al etching by HDPRIE 19. Al backside contact
31
Fig.2-3 Schematic layout of six-terminal Cross Bridge Kelvin Resistor (CBKR) structure.
32
Chapter 3
IBG and IAG junctions
3.1 NiGe formation
In order to find out the lowest thermal budget for NiGe formation, the sheet resistance of the NiGe film formed at different thermal budgets was measured and shown in Fig.3-1 and Fig.3-2. According to Fig.3-1, the sheet resistance of NiGe decreases from 73 Ω/□ after 300 oC annealing to 20 Ω/□ after 325 oC annealing for 5 minutes and still keeps low after 400oC annealing. In addition, in Fig.3-2, the sheet resistance of NiGe decreases from 80 Ω/□ for 3 minutes annealing to 20 Ω/□ for 5 minutes annealing at 325 oC. According to the change of sheet resistance shown above, NiGe starts to growth at 325oC but the growth rate is slow so that at least a 5 minutes annealing is necessary. In the thesis, NiGe formation was performed at 350 oC for 5 minutes to ensure that NiGe could completely form if not specified. According to the formula of the sheet resistance Rsh=V/I x CF, where CF is correction factor determined by the value of d/s, where d is the distance of the probe to the sample edge and the s is the distance between the probes, the sheet resistance of NiGe measured by the four point probe system is under the condition of infinitely large samples and the correction factor is 4.5324. However, the samples’ size used in the thesis is nearly 0.5x0.5 cm2 and the value of d/s is about 3. Therefore, the correction factor is about 2.4575 based on the Table 3-1 and the thickness of NiGe is 20 nm (Fig.3-15); then, the resistivity of NiGe is calculated to be 22.04x10-6 Ω-cm. In addition, XRD diffraction analysis on the NiGe/Ge structure formed by 350C annealing shows polycrystalline NiGe phases as shown in Fig. 3-3. The main NiGe crystal orientations
33
are (2 1 0), (1 1 2), and (1 1 1).
3.2 Conventional N
+/P Junctions
At first, the I-V curves of the conventional junctions with phosphorus ion implantation at 20 keV to a dose of 1x1015 cm-2 and 50 keV to a dose of 5x1013 cm-2 on heavily-doped substrate are shown in Fig.3-4. After dopant activation at 600 oC for 60 seconds, the junctions with phosphorus ion implantation at 20 keV to a dose of 1x1015 cm-2 has a relatively higher forward current and lower reverse current due to the higher doping concentration. The forward and reverse bias current ratio is about 5 orders of magnitude. In addition, the band-to-band tunneling can also be observed when the reverse bias exceeds 0.5 V. Because the high n+ concentration and heavily-doped substrate cause high electric field at the junction edge, the electrons would easily tunnel through the band to the other side as the reverse bias is applied.
Next, the I-V curves of conventional junctions with phosphorus ion implantation at 50 keV to different doses of 1x1015, 5x1013, 1x1013, 5x1012 cm-2 on lightly-doped substrate are shown in Fig. 3-5. After annealing at 600 oC for 60 seconds, the forward current increases with the increase of the ion implantation dose and the junction with the highest dose of 1x1015 cm-2 has the highest forward current. The reason is that more dopants could be activated to form higher concentration of n+ region so that the electron diffusion current increases and the NiGe/Ge contact resistance could be reduced. Moreover, the reverse bias current slightly increases with the increase of dose because higher ion implantation dose may induce more defects to raise leakage current. Therefore, the forward/reverse current ratio also increases with the increase of dose and the current ratio of the junction with a dose of 1x1015 cm-2 is about 103.
34 depletion width is proportional to(𝑁1
𝑎+𝑁1
𝑑)
1
2. In addition, the reverse bias current is directly proportional to the depletion width when the generation current within the depletion region dominates the reverse bias current. Now, Na is the concentration of substrate and Nd is the concentration of the n+ region. If the concentration of n+ region is the same, the concentration of substrate would dominate the reverse bias current.
The lower concentration of substrate would cause the wider depletion width, and the wider depletion width would cause the higher reverse bias current. Comparing the I-V curves of the conventional junctions with phosphorus ion implantation at 50 keV to a dose of 5x1013 cm-2 on lightly-doped and heavily-doped substrate, e.g. Fig.3-4 and Fig.3-5, respectively, it is the reason why the reverse bias current on lightly-doped substrate is larger than that on heavily-doped substrate.
Thirdly, the I-V curves of conventional junctions with arsenic ion implantation at 50 keV to a dose of 1x1015, 5x1013, 1x1013 cm-2 on lightly-doped substrate are shown in Fig. 3-6. After dopants activation at 600 oC for 60 seconds, the junction with arsenic ion implantation to a dose of 1x1015 cm-2 shows the highest forward current and the lowest reverse current; the forward/reverse current ratio is about 300.
However, the junctions with dose of 5x1013 and 1x1013 cm-2 show poor junction characteristics. The reason is that the doping concentration is not high enough so that the arsenic doped layer is depleted.
Comparing the I-V curves of conventional junctions of implantation by phosphorus and arsenic at 50 keV to a dose of 5x1013 cm-2 on lightly-doped substrate, the reverse bias current of the arsenic implanted junction is higher than that of the
35
phosphorus implanted junction because the activation ratio of arsenic is relatively lower than that of phosphorus so that the junction depth of the arsenic implanted junction is inferred to be shallower. Furthermore, the heavier arsenic atom would generate more defects than the lighter phosphorous atom. Therefore, the reverse bias current of the arsenic implanted junction is higher as shown in Fig. 3-5 and Fig. 3-6.
3.3 NiGe on Conventional N
+/P Junctions (IBG junctions)
In this session, we examine the effect of NiGe formation on the preformed N+/P junctions, i.e the IBG junctions. Fig.3-7 shows the I-V curves of the IBG junctions with different NiGe formation temperatures. The preformed junction is implanted by phosphorus at 20 keV to a dose of 1x1015 cm-2 on heavily-doped substrate. All junctions show high reverse bias current and the electrical characteristics are nearly ohmic behavior. The junction with NiGe formation at 300oC for 5 minutes shows lower current and slightly rectifying characteristic due to incomplete NiGe formation.
From Fig.3-1, the sheet resistance is higher than 70 Ω/□ which implies that Ni2Ge formed but not NiGe formed. Other junctions have the same high forward current about 102 A/ cm2 as a result of complete NiGe formation.
Compared with the conventional junctions, before the NiGe formation, the reverse bias current is about 10-4 A/ cm2; however, the reverse bias current rises apparently after NiGe formation and the current increases with the increase of the NiGe formation temperature. It means NiGe formation causes something happened to destroy the preformed junction. According to the temperature dependence, a possible reason may be the fast diffusion of Ni. Fig. 3-8 shows the diffusivity of elements in Ge, Ni has the second high diffusivity among these elements [78]. If Ni atoms enter the depletion region of the junction, they act as generation-recombination center and
36
destroy the junction. Therefore, finding some methods to suppress Ni diffusion is important.
A method of fluorine implantation before silicide formation had been proposed to solve the problem of large leakage current caused by Ni diffusion [79]. Hence, fluorine implantation before NiGe formation was performed in this thesis in order to solve the Ni diffusion problem. Fig.3-9 shows the I-V curves of the fluorine implanted junctions with different NiGe formation temperatures. The preformed conventional junction is formed by phosphorus ion implantation at 20 keV to a dose of 1x1015 cm-2 on heavily-doped substrate. It is observed that fluorine ion implantation reduces both forward current and reverse current. This result confirms that fluorine implantation could suppress the Ni diffusion but affect the NiGe formation at the same time. The reverse bias current of the junction with NiGe formation at 325 oC is reduced by nearly 3 orders of magnitude. But the reverse bias current still increases with the increase of the NiGe formation temperature.
The junctions with deeper junction depth were fabricated to examine whether the fast diffusion of Ni still damages the junction characteristic. The deep junction was formed by phosphorus ion implantation at 50 keV to a dose of 5x1013 cm-2 on heavily-doped substrate. The I-V curves are shown in Fig. 3-10. The forward current is about 102 A/ cm2 and is nearly the same with the forward current of the IBG junctions. Furthermore, the reverse bias current is nearly 10-2 A/ cm2 and is lower than that of the IBG junction with phosphorus ion implantation at 20 keV to a dose of 1x1015 cm-2. It could be inferred that high ion implantation dose may cause more defects than low ion implantation dose does, and more defects may let Ni diffuse faster so that the reverse bias current increases. Fig.3-10 also shows the I-V curves of the IBG junctions with fluorine ion implantation. In particular, the reverse bias current of the fluorine implanted junction is higher than that of the junction without fluorine
37
implantation and also increases with the increase of the NiGe formation temperature.
It could be inferred that fluorine implantation causes additional defects than phosphorous implantation and these defects enhance Ni diffusion.
Next, the IBG junctions are also fabricated on lightly-doped substrate to observe whether the junction is still destroyed by Ni diffusion. Because different implantation doses might induce different amount of defects, the IBG junctions with different implantation doses were designed to observe the effect of Ni diffusion. Fig.3-11 shows the IBG junctions formed with phosphorus implantation at 20 keV to different doses on lightly-doped substrate. The forward current of the four IBG junctions increases with the increase of the ion implantation dose due to the different activated concentrations. However, the reverse bias current of the four IBG junctions are all about 10-1 A/cm2 and do not increase with the ion implantation dose. This observation is not consistent with the IBG junctions on heavily-doped substrate. The cause could be inferred that the junction depth on lightly-doped substrate is deeper than that on heavily-doped substrate with the same ion implantation and activation conditions. The Ni diffusion is deep enough to destroy the IBG junction on heavily-doped substrate, but it is not deep enough to destroy the IBG junction on lightly-doped substrate. In addition, the forward current increases after NiGe formation due to the dopant segregation at the NiGe/Ge interface in comparison with Fig.3-5.
First-principle calculations were introduced to discuss the behaviors of the n-type dopant around the NiGe/Ge interface [81]. First, the realistic polycrystalline phases NiGe/Ge contact is built by including NiGe(112) phase only, so the NiGe/Ge contact is simulated by a supercell connecting 8 NiGe(112) layers, 16 Ge(001) layers, and 12Å -vacuum where the dangling bonds of the Ge surface are saturated by H atoms.
Before exploring whether the n-type dopant such as P and As can be segregated and pile-up at the NiGe/Ge interface or not, the implanted dopant is assumed to be
38
activated and migrate into the substitutional sites. Therefore, one at a time of the Ge atoms around the interface is replaced by a doping atom in the NiGe/Ge interfacial structure and then the doped interfacial structure is relaxed by the DFT-LDA calculations until the forces are less than 0.01eV/Å for atoms within 9 Å from the interface, where the atoms beyond this region approach their corresponding bulk positions. The total energy of the system with one Ge atom being substituted for a P and As are shown in Fig. 3-12(a) and (b), respectively. The P and As dopant concentration here is equivalent to 8.34 x 1020 cm-3.
The result in Fig. 3-12(a) shows that the P doping occurs most stably at the positions labeled as MG3 and SG1a. It implies that the P dopant can be segregated around the NiGe/Ge interface and piled up on the NiGe side. Nevertheless, the As atom prefers to stay on the Ge side near the interface as shown in Fig. 3-12(b) since the Ge-side doping of As is more stable than the NiGe-side. The most stable substitutional site for the As dopant is labeled as SG1c. To sum up, although both P and As dopants can be segregated around the NiGe/Ge interface in our calculations, the former prefer to pile up on the NiGe side but the latter like to stay on the Ge side around the interface.
Finally, IBG junctions formed by arsenic ion implantation at 30keV to a dose of 1x1013, 5x1013, and 1x1015 cm-2 on lightly-doped substrate are fabricated and the I-V curves are shown in Fig. 3-13. The poor junction characteristic is observed as the As dose is 1x1013 cm-2 because the activated dose cannot form good N+/P junction. The IBG junctions formed with arsenic ion implantation to a dose of 5x1013 and 1x1015 cm-2 have better junction characteristic. The forward current also increases after NiGe formation due to dopant segregation. Therefore, the dopant segregation effect could be observed by As implantation as well, and the forward current is slightly lower than the IBG junctions formed by P implantation due to the low activation rate of As.
39
3.4 NiGe Dopant Segregation Junctions
Fig. 3-14 shows the I-V curves of the IAG junctions with different annealing temperatures. Very poor junction characteristics are observed on all IAG junctions.
One reason may be the segregated n+ layer is too thin to maintain good N+/P junction.
The other possible reason is that because Ni diffuses much faster than arsenic, the distribution of Ni would be much deeper than arsenic. Therefore, using IAG process to get a good shallow junction is very difficult. In addition, cross-sectional TEM micrographs of the NiGe/Ge structure with IAG process after annealing at 500C and 550C for 10 sec are shown in Fig.3-15. The highest sustainable temperature of the NiGe/Ni structure is 500 C. This temperature is also the highest allowable temperature for the IAG process.
In order to maintain good junction characteristic, a low energy phosphorus ion implantation was performed before NiGe formation followed by the IAG process. It is called the IBG+IAG junction. The I-V curve of the IBG+IAG junction with 500 C annealing is shown in Fig. 3-16. Comparing the IBG+IAG junction and the Al-contacted conventional junction, the IBG process steps form a good 100-nm-deep shallow junction (Fig. 3-18) while the IAG process improves the turn-on characteristic. Furthermore, the NiGe formation process does not degrade leakage current.
3.5 Contact resistance of IBG+IAG junctions
First-principles calculations were used to calculate the total energy of the NiGe/Ge structure. Replacing Ge atom by P or As atom, Fig. 3-12 shows the total energy of the NiGe/Ge structure as a function of the dopant position. Either P or As
40
atom may segregate at the NiGe/Ge interface, however, the most stable segregation position for P is at the NiGe side and the most stable segregation position for As is at the germanium side. These results imply that NiGe formation would reduce the contact resistance on the P- doped and As-doped junctions (IBG process) while As would be a better choice for the implantation after NiGe process (IAG process).
Therefore, IBG process is preferred for the P-doped junction due to P has higher activation concentration than As and IAG process is preferred for the As-doped junction due to the lowest total energy at the interface. The contact resistance would reduce by virtue of the two times dopants segregation.
Cumulative distribution of contact resistivity extracted by the CBKR structure is shown in Fig.3-17. The contact resistivity of the Al-contacted conventional junction is as high as 10-4 -cm2, because the Fermi-level pinning near valance band at the interface. In addition, the SIMS analysis shows the junction depth is about 125 nm in Fig. 3-18. After the IBG process, the contact resistivity is reduced to 2x10-5 -cm2. According to the SIMS analysis shown in Fig. 3-19, there is a peak concentration at
Cumulative distribution of contact resistivity extracted by the CBKR structure is shown in Fig.3-17. The contact resistivity of the Al-contacted conventional junction is as high as 10-4 -cm2, because the Fermi-level pinning near valance band at the interface. In addition, the SIMS analysis shows the junction depth is about 125 nm in Fig. 3-18. After the IBG process, the contact resistivity is reduced to 2x10-5 -cm2. According to the SIMS analysis shown in Fig. 3-19, there is a peak concentration at