Chapter 3 Experiment
3.5 DC and RF Measurements
3.5.2 RF Measurements
The RF performance of the HBT devices was characterized by on-wafer-S-parameter measurements using HP8510C network analyzer before and after the thermal annealing test. The S-parameters were measured in frequencies ranging from 1 to 40 GHz. Cutoff frequency (fT) and maximum frequency of oscillation (fmax) are often used to characterize the devices. These two quantities represent the unity gain intercept point of the short circuit current gain (h21) and the unilateral power gain (U). Each parameter can be computed from S-parameter data.
The h21 as a function of frequency were measured under collector-emitter voltage (VCE) of 2V, 2.5V, and 3V. The base current was 0.06, 0.08, 0.1, 0.12, 0.14mA. The fT was extrapolated with -20 dB/decade slope.
Table
Table 3 The typical epitaxial layer structure of the InGaP/GaAs HBT
Layer Material Type Doping Thickness（Å）
Emitter Cap GaAs n+ 5×1018 2000
Emitter InGaP n 3×1017 500
Base GaAs p+ 4×1019 800
Collector GaAs s n- 2×1016 7000
Subcollector GaAs n+ 5×1018 5000
Substrate GaAs
Figure GaAs Cap InGaP Emitter
GaAs Subcollector
GaAs Substrate Figure 11.1 Emitter mesa etch
GaAs Cap InGaP Emitter
GaAs Subcollector
GaAs Substrate Figure 11.2 Base and collector mesa etch
GaAs Cap InGaP Emitter
GaAs Subcollector
GaAs Substrate Figure 11.3 Mesa isolation
GaAs Cap InGaP Emitter
GaAs Subcollector
GaAs Substrate
Figure 11.4 Emitter and collector ohmic contact metal formation
GaAs Cap InGaP Emitter
GaAs Subcollector
GaAs Substrate Figure 11.5 Base ohmic contact metal formation
GaAs Cap InGaP Emitter
GaAs Subcollector
GaAs Substrate Figure 11.6 Silicon Nitride Deposition
GaAs Cap InGaP Emitter
GaAs Subcollector
GaAs Substrate
Figure 11.7 Nitride via etch
GaAs Cap InGaP Emitter
GaAs Subcollector
GaAs Substrate
Figure 11.8 Interconnect metal line
Figure 12 Illustration of transmission line methods (TLM) patterns
Figure 13 Illustration of utilizing TLM identify ohmic contact resistance
Chapter 4
Results and Discussion
In this chapter, the contact resistivity and the thermal stability of Pd (150 Å)/Ge (1500 Å)/Cu (1500 Å) ohmic contact on n-type GaAs are shown. After that, the formation mechanism of the Pd/Ge/Cu ohmic contact was investigated using the results of XRD, SMIS, AFM, TEM and EDX. In the last half of the chapter, the results of DC and RF measurements of Au free fully Cu-metallized InGaP/GaAs HBT using Pd/Ge/Cu ohmic contact are shown.
4.1 Contact Resisitivity of The Pd/Ge/Cu Ohmic Contact
The Pd (150 Å)/Ge (1500 Å)/Cu (1500 Å) multilayer metals are deposited on the GaAs wafer with Si-doped epitaxial layer (2000 Å, 1x1018 cm-3). After annealing in a traditional tube furnace at different temperatures for 20 min, the results of the contact resistivities of the Pd(150 Å)/Ge(1500 Å)/Cu(1500 Å) ohmic contact extracted from the transmission line measurements (TLM) as a function of annealing temperature are shown in Figure 14. Low ohmic contact resisitivity can be obtained when the Pd/Ge/Cu ohmic samples were annealed at 220°C ~ 350°C for 20 min. The lowest specific contact resistivity was 5.73 x 10-7 Ω-cm2 after the sample was annealed at 250 °C for 20 min.
In this study, the effect of the ohmic metal composition on the contact resistance is investigated. Figure 15 shows the ohmic characteristics of the Pd/Ge/Cu samples with different Pd thicknesses, the values of the lowest specific contact resistivity for these samples are listed in Table 4. As these
results indicate, the Pd/Ge/Cu sample with 150 Å Pd layer has the lowest specific contact resistivity and the thickness of Pd layer has a significant effect on the ohmic contact resistance. One may notice that the lowest specific contact resistivity of the Pd/Ge/Cu sample with 50 Å Pd appeared at 200 °C, it may be due to the formation of thinner PdxGaAs compound layer which made the Ge atoms diffuse through the thinner PdxGaAs compound layer into GaAs substrate easily. However, the adhesion between the 50 Å Pd layer and the GaAs surface was very weak, the ohmic metal peeled off easily.
Figure 16 shows the ohmic characteristics of the Pd/Ge/Cu samples with different Ge thicknesses, the values of the lowest specific contact resistivity for these samples are listed in Table 5. As these results indicate, the thickness of Pd layer has also a significant effect on the ohmic contact resistance. The Pd/Ge/Cu sample with 1500 Å and 2000 Å Ge layer have similar low specific contact resistivity, but the Pd/Ge/Cu ohmic contact with 1000 Å Ge layer has no ohmic characteristic. It may be due to that for the thinner Ge layer, there were not enough Ge atoms to react with Cu and diffuse into GaAs surface.
4.2 Formation Mechanism of The Pd/Ge/Cu Ohmic Contact
In this section, the formation mechanism of the Pd/Ge/Cu ohmic contact was investigated by results of XRD, SMIS, AFM, TEM and EDX. Use TEM image and EDX analysis for microstructure observation, X-ray diffraction for phase identification, AFM for surface morphology observation, and SIMS for interfacial elements material analysis. The results are shown below.
4.2.1 TEM and EDX for Microstructure Observation
Figure 17 shows the SEM image of the as-deposited Pd (150 Å)/Ge (1500 Å)/Cu (1500 Å) structure deposited on GaAs substrate. From this figure, the Pd, Ge, and Cu layers can be seen obviously. The thin Pd layer enhances the adhesion of the ohmic metal and the Cr layer was used as the anti-oxidation layer for Cu.
Figure 18 ~ Figure 22 shows the TEM images and EDX profiles of Pd/Ge/Cu ohmic metal structure after annealing at 150°C, 250°C, 400°C for 20min respectively. Figure 18(a) shows the TEM image of the sample after annealing at 150°C for 20 min. From this figure, a small proportion of Ge and Cu started to react with each other as the Pd/Ge/Cu ohmic metal structure annealing was annealed at 150°C for 20 min. However, from the EDX analysis, the Cu3Ge grains were not formed and the Pd layer had no obvious reaction with Ge/Cu and GaAs substrate after 150°C annealing as shown in Figure 18(b).
There was no ohmic behavior in Pd/Ge/Cu ohmic system after 150°C annealing.
The TEM image of the Pd/Ge/Cu ohmic metal structure after annealing at 250°C for 20 min is shown in Figure 19(a). From the figure, the Cu/Ge compound started to form grains with vertical grain boundary and long range order. From the EDX analysis, the grains were Cu3Ge compound as shown in Figure 19(b). Literature shows that the compound has low metallic resistivity and Ga has lower chemical potential in Cu3Ge than in GaAs compound [28]. On the other hand, Figure 20 shows the HRTEM image of the near-interface region between GaAs substrate and ohmic compound after 250°C annealing. The PdxGaAs phases started to appear at the GaAs surface after 250°C annealing.
Due to PdxGaAs compound, it may create more Ga vacancies. The Ge atoms could easily diffuse into the Ga vacancies in the vicinity of the GaAs surface,
resulting in a heavy doping n+-GaAs layer. So the ohmic contact characteristic of the Pd/Ge/Cu appeared after annealing at 250°C for 20 min.
The EDX profiles in Figure 21 show that there is still no Cu atom diffusing into GaAs the substrate near the Pd/GaAs interface after 250 annealing℃ . However, after annealing at 400°C for 20 min, obvious atomic inter-diffusion and interfacial reactions started to occur as can be seen from Figure 22. The ohmic contact characteristics of the Pd/Ge/Cu ohmic system started to degrade after 400°C annealing, the possible reasons for ohmic contact degradation were As atoms diffusing out and Ga atoms diffusing into the Ge layer. (Ga atoms acted as acceptors which reduced the donor concentration.)
4.2.2 X-ray Diffraction for Phase Identification
Figure 23 shows the x-ray diffraction profiles for the Pd (150 Å)/Ge (1500 Å)/Cu (1500 Å) ohmic contact structure as deposited and after annealing at 250
°C and 400 °C for 20 min. It can be seen from the XRD spectra that the diffraction peaks of Pd and Cu remained observable for the as deposited sample.
It indicated that the Pd/Ge/Cu multi-layers did not react with each other for the as deposited sample. However, it is obvious from these data that the diffraction peaks of the Cu3Ge compounds occurred and the diffraction peaks of Cu disappeared as the annealing temperature was higher than 250°C. The ohmic contact behavior was related to the formation of the Cu3Ge compounds as the annealing temperature was above 250°C
4.2.3 SIMS for Interfacial Elements Material Analysis
Figure 24 shows the SIMS depth profiles after the sample was annealed at 250 °C for 20 min. From the SIMS profiles, it can be seen that the Ge atoms
diffused into GaAs and the Ga atoms diffused out into the ohmic layer. As mentioned above, both the formation of Cu3Ge grains and PdxGaAs phases after annealing helped the inter-diffusion of the Ga and Ge atoms.
4.2.4 AFM for Surface Morphology Observation
Figure 25 shows the surface morphologies for the Pd/Ge/Cu ohmic metal system and the traditional Au/Ge/Ni ohmic system after annealing as measured by AFM. The root-mean-square (rms) roughness of the Pd/Ge/Cu sample is 4.541nm. The root-mean-square (rms) roughness of the traditional Au/Ge/Ni sample is 6.566nm. It is obviously that the surface morphology of Pd/Ge/Cu ohmic system is smoother than the traditional Au/Ge/Ni ohmic system.
4.3 Thermal Stability Test for The Pd/Ge/Cu Ohmic Contact
To study the thermal stability of the Pd (150 Å)/Ge (1500 Å)/Cu (1500 Å) ohmic contact, the Pd/Ge/Cu multilayer layer were annealed at 250°C for 24 hours and measured the specific contact resistivity was used TLM patterns. For comparison, the traditional Au/Ge/Ni ohmic system TLM patterns were also tested at the same conditions. Figure 26 shows the long time thermal stability test results. From this figure, it can be seen that there was no obvious degradation on the Pd/Ge/Cu ohmic system after annealing at 250 for 24 ℃ hours. The contact resistivity increased slightly after 250 annealing for 12 ℃ hours, and then kept at 8 x 10-7 Ω-cm2 for annealing time up to 24 hours. And the contact resistivity of Pd/Ge/Cu ohmic contact was still lower than that of traditional Au/Ge/Ni ohmic system after annealing at 250 ℃ for 24 hours. From the results shown above, it is clear that the Pd/Ge/Cu ohmic system to n-type
GaAs has low contact resistance and was quite stable even after long time annealing.
4.4 DC and RF Measurements
4.4.1 DC Measurements
Pd/Ge/Cu ohmic contact was applied to the fully Cu-metallized InGaP/GaAs HBTs as emitter and collector metal. In this fully Cu-metllized HBT, Pt/Ti/Pt/Cu was used as the base metal, SiNx was used for passivation, and Ti/Pt/Cu for interconnect metal with Pt as the diffusion barrier. InGaP/GaAs HBTs with traditional n-type metal (Au/Ge/Ni/Au), and p-type metal (Pt/Ti/Pt/Au) contacts, and interconnect metal (Ti/Au) were also processed on half of the same wafer for performance comparison.
Figure 27 shows the optical microscope images of the fully Cu-metallized InGaP/GaAs HBTs with Pd/Ge/Cu ohmic contact after fabrication. Figure 28 shows the typical common emitter characteristics of HBTs with emitter area of 4 x 20µm2. In the figure, one group of curves belongs to the fully Cu-metallized HBTs and the other one belongs to the traditional Au-metallized HBTs. It can be seen from Figure 28 that these two devices show similar knee voltage and offset voltage. We did not observe an increase in the knee voltage or the decay of the collector current, which indicates that the characteristics of the InGaP/GaAs HBTs with Pd/Ge/Cu ohmic contact are reasonably good. The common emitter current gain is around 130 for both cases. Gummel plots of the HBTs with the traditional Au HBT and Au free fully Cu HBT were also compared. The results are shown in Figure 29. The two HBTs also showed similar behavior.
To test the reliability of the Pd/Ge/Cu as the n-type ohmic metal for the
Cu-metallized HBTs, both copper and gold metallized HBTs with 4x20µm2 emitter area were subjected to current accelerated stress test with high current density of 110 kA/cm2. It is much higher than 25kA/cm2 required for the normal device operation and the purpose is to shorten the stress time so that the stress tests could be performed at wafer level without using any package and the results could be obtained in a few hours [27]. Figure 30 plots the current gain (β) of the fully Cu-metallized HBTs with Pd/Ge/Cu ohmic contact after stressed at the high current density of 110 kA/cm2 with VCE of 2.5V for a period of 24 hours.
The measurements were made at an amibient room temperature of TA = 25°C. It can be seen from the data that the current gain of the device showed no significant change with time. The change in the ratio of the final/initial current gain was less than 4% for the device and was still higher than 115 after 24h of the current-accelerated stress test.
To study the thermal stability of the fully Cu-metallized HBT with Pd/Ge/Cu ohmic contact, the 4 x 20-µm-emitter-area HBT device was annealed at 250°C for 24 hours and tested for electrical performance. Figure 31 shows the common emitter I-V curves of the fully Cu-metallized HBT before and after annealing which shows no change in the offset voltage, knee voltage, and saturation current. Figure 32 shows the current gain (β) as a function of aging time at 250℃ for the 4x20-µm2-emitter-area fully Cu-metallized HBT with Pd/Ge/Cu ohmic contact. It suggested that there was no ohmic degradation, copper oxidation, and copper diffusion in the fully Cu-metallized HBT with Pd/Ge/Cu ohmic contact after the annealing.
4.4.2 RF Measurements
The microwave performance of the fully Cu-metallized HBT with Pd/Ge/Cu ohmic contact was characterized by on-wafer S parameter measurement using a network analyzer. The h21 as a function of frequency were measured under collector-emitter voltage (VCE) of 2V, 2.5V, and 3V. The base current was 0.06, 0.08, 0.1, 0.12, 0.14mA. The fT was extrapolated with -20 dB/decade slope. Figure 33 shows the h21 curve at the VCE = 2.5V and IB = 0.14mA as a function of frequency. The cutoff frequency was about 38GHz. It is clear that the microwave performance of the fully Cu-metallized HBT with Pd/Ge/Cu ohmic contact is very good.
Tables
Table 4 The lowest specific contact resistivities of Pd/GeCu ohmic contact on n-type GaAs with different Pd thicknesses.
Metal composition (Thickness Å )
Pd Ge Cu
The lowest specific contact resistivity
(Ω-cm2) Sample 1 0 1500 1500 ~10-5 Sample 2 50 1500 1500 7.43 x 10-7 Sample 3 150 1500 1500 5.72 x 10-7 Sample 4 300 1500 1500 7.88 x 10-7
Table 5 The lowest specific contact resistivities of Pd/Ge/Cu ohmic contact on n-type GaAs with different Ge thicknesses.
Metal composition (Thickness Å )
Pd Ge Cu
The lowest specific contact resistivity
(Ω-cm2)
Sample 1 150 1000 1500 No ohmic Sample 2 150 1500 1500 5.72 x 10-7
Sample 3 150 2000 1500 5.34 x 10-7
Figures
150 200 250 300 350 400 450
10-6 1x10-5
no ohmic X
Pd(15nm)/Ge(150nm)/Cu(150nm)
Specific Contact resistance (ohm-cm2 )
Annealing Temperature (oC)
Figure 14 The specific contact resistivity of the Pd(15nm) / Ge(150nm) / Cu(150nm) contact on n-type GaAs as a function of annealing temperature.
150 200 250 300 350 400 1E-7
1E-6 1E-5 1E-4 1E-3 0.01
Specific Contact resistance (ohm-cm2 )
Annealing Temperature (oC)
Pd/Ge/Cu= 5/150/150 (nm) Pd/Ge/Cu=15/150/150 (nm) Pd/Ge/Cu=30/150/150 (nm)
Figure 15 The specific contact resistivity as a function of annealing temperature of Pd/Ge/Cu ohmic contact on n-type GaAs with different Pd thicknesses.
200 250 300 350 400 1E-7
1E-6 1E-5 1E-4 1E-3 0.01
Specific Contact resistance (ohm-cm2 )
Annealing Temperature (oC)
Pd/Ge/Cu=15/100/150 (nm) Pd/Ge/Cu=15/150/150 (nm) Pd/Ge/Cu=15/200/150 (nm)
Figure 16 The specific contact resistivities as a function of annealing temperature of Pd/Ge/Cu ohmic contact on n-type GaAs with different Ge thickness.
Figure 17 The SEM image of the Pd(15nm) / Ge(150nm) / Cu(150nm) contact for the as-deposited sample.
(b)
Figure 18 (a) The TEM image of the Pd/Ge/Cu contact cross section, (b) The EDX profiles of the Ge/Cu compound grains after annealing at 150℃ for 20 min.
(b)
Figure 19 (a) The TEM image of the cross section of the Pd/Ge/Cu contact, (b) The EDX profile of the Cu3Ge compound grains after annealing at 250℃ for 20 min.
(b)
(a)
GaAs
Pd/Ga
(b)
Figure 20 (a) The TEM image of the cross section of Pd/Ge/Cu contact, (b) The high resolution TEM image of the interface between the metallic compound in the ohmic metal layer and the GaAs substrate after annealing at 250℃ for 20 min.
(a)
0 5 10
0 500 1000
As Ga As
As Ga
counts
keV
(b)
(b)
Figure 21 (a) The TEM image of the cross section of Pd/Ge/Cu contact, (b) The EDX profile of the near-interface region of the GaAs substrate after annealing at 250℃ for 20 min.
Cu3Ge
Figure 22 The TEM image of the cross section of the Pd/Ge/Cu contact after annealing at 400℃ for 20 min.
10 20 30 40 50 60 70 80 90 GaAs
GaAs
GaAs GaAs
As-dep.
250oC 400oC
Pd Cu Cu Cu3Ge Cu3Ge
Intensity (arb. units)
2 θ (deg)
Figure 23 The X-ray diffraction patterns for the Pd (150 Å)/Ge (1500 Å)/Cu (1500 Å) contact after annealing at 250 °C for 20min, 400 °C for 20min, and the as-deposited sample.
0.0 0.2 0.4 0.6 0.8 1.0
Figure 24 SIMS profiles of the Pd (150 Å)/Ge (1500 Å)/Cu (1500 Å) contact after annealing at 250 °C for 20min.
(a)
(b)
Figure 25 The AFM images of (a) the Pd/Ge/Cu ohmic contact after annealing at 250℃ for 20 min, (b) the Au/Ge/Ni/Au ohmic contact after RTA at 380℃ for 30 sec.
0 2 4 6 8 10 12 14 16 18 20 22 24 1E-7
1E-6 1E-5 1E-4
Pd/Ge/Cu=15/150/150 (nm) Au/Ge/Ni/Au=75/35/20/200(nm) Aging Test at 250oC
Specific Contact Resistivity (ohm-cm2 )
Aging Time (Hours)
Figure 26 The specific contact resistivities of the Pd/ Ge/ Cu ohmic contact and Au/Ge/Ni/Au ohmic contact on n-type GaAs as a function of aging time.
E
C B
Figure 27 The OM images of Au-free fully-Cu InGaP/GaAs HBT with Pd/Ge/Cu as n-type ohmic contact device.
0.0 0.5 1.0 1.5 2.0 2.5 3.0 0
5 10 15 20 25 30 35
40 Fully Cu HBT
Traditional Au HBT IB=50µA/step Emitter area: 4 x 20 µm2
Collector current, I C(mA)
Collector-emitter voltage, VCE(V)
Figure 28 Comparison of the typical IC-VCE characteristics for the emitter area(4 x 20 μm2) HBTs with Cu and with Au metallizations.
0.6 0.8 1.0 1.2 1.4 1.6
Figure 29 Comparison of Gummel plots for the emitter area(4 x 20μm2) HBTs with Cu and with Au metallizations.
0 5 10 15 20 25 0
20 40 60 80 100 120 140 160 180 200
Emitter Area=4x20µm2 VCE=1.5V
JC=110KA/cm2
Current Gain(β)
Stress Time(h)
Figure 30 The current gain (β) as a function of stress time at constant IB for the 4x20-μm2-emitter-area fully Cu-metallized HBT with Pd/Ge/Cu ohmic contact.
0.0 0.5 1.0 1.5 2.0 2.5 3.0 0
5 10 15 20 25 30 35 40
Before annealing
After 250oC 24h annealing
IB=50uA/step Emitter area: 4 x 20 µm2
Collector current, I C(mA)
Collector-emitter voltage, VCE(V)
Figure 31 Common emitter I-V curves measured before and after annealing at 250℃ for 24 h for the 4x20-μm2-emitter-area fully Cu-metallized HBT with Pd/Ge/Cu ohmic contact.
0 2 4 6 8 10 12 14 16 18 20 22 24
Figure 32 The current gain (β) as a function of aging time at 250℃ for the 4x20-μm2-emitter-area fully Cu-metallized HBT with Pd/Ge/Cu ohmic contact.
1 10 0
5 10 15 20 25 30
35 Emitter area = 3 x 20 µm2
V
CE = 2.5 V I
b = 0.14 mA
fT = 38 GHz h 21, gain(dB)
Frequency(GHz)
Figure 33 Current gain H21 curves measured for the 3x20-μm emitter-area fully Cu-metallized HBT.
Chapter 5 Conclusions
In this study, a novel Pd/Ge/Cu ohmic contact to n-type GaAs has been successfully developed. The Au-free fully Cu-metallized InGaP/GaAs HBTs using Pd/Ge/Cu ohmic contact to n-type GaAs also has been successfully fabricated for the first time.
The optimized Pd (150 Å)/Ge (1500 Å)/Cu (1500 Å) metal structure forms a low contact resistivity ohmic contact to n-type GaAs at a low annealing temperature. Low ohmic contact resistivity can be obtained when the Pd/Ge/Cu ohmic samples were annealed at 220°C ~ 350°C for 20min. The lowest specific contact resistivity achieved was 5.73 x 10-7 Ω-cm2 when annealed at 250 °C for 20 min. The thicknesses of the Pd and Ge layers play an important role on the ohmic contact resistivity. The thin Pd layer enhances the adhesion of the ohmic metals and helps the Ge atoms (donor) diffuse into the Ga vacancies. The enough thickness of Ge layer enhances good ohmic contact characteristic.
From SIMS, XRD, TEM, and EDX studies, the low contact resistivity was due to the formation of the Cu3Ge compound and the PdxGaAs compound in conjunction with the outdiffusion of Ga into the ohmic metal and the diffusion of Ge into the Ga vacancies. From AFM results, the surface morphology of Pd/Ge/Cu ohmic contact was smoother than the traditional Au/Ge/Ni ohmic contact. The contact resistivity of Pd/Ge/Cu ohmic contact was also very stable after annealing at 250°C for 24 hours. And the contact resistivity of Pd/Ge/Cu
From SIMS, XRD, TEM, and EDX studies, the low contact resistivity was due to the formation of the Cu3Ge compound and the PdxGaAs compound in conjunction with the outdiffusion of Ga into the ohmic metal and the diffusion of Ge into the Ga vacancies. From AFM results, the surface morphology of Pd/Ge/Cu ohmic contact was smoother than the traditional Au/Ge/Ni ohmic contact. The contact resistivity of Pd/Ge/Cu ohmic contact was also very stable after annealing at 250°C for 24 hours. And the contact resistivity of Pd/Ge/Cu