Chapter 1 Introduction
1.2 Dissertation Content
The contents of this dissertation include: literature review, experiment, results, discussion and conclusions. In Chapter 2, the literature of ohmic contact for n-type InGaAs and GaAs-based HBTs are reviewed. In Chapter 3, the study of ohmic contact, the samples preparation for material analysis, and the fully Cu-metallized GaAs HBT device process flow are described. In addition, the results of multilayer interfacial material analysis, the DC and Power characteristics and the reliability tests for the fully Cu HBTs will also be presented. The Pd/Ge/Cu ohmic contact, formation mechanism, the DC and Power characteristics of the HBT devices using the Pd/Ge/Cu ohmic will be discussed in Chapter 4. Finally, the conclusions will be given in Chapter 5.
Table
Table 1 Properties comparisons of possible interlayer metals
Property Cu Ag Au Al
Resistivity(µΩ‧cm) 1.67 1.59 2.35 2.66 Young’s modulus．10-11 dyn/cm2 12.98 8.27 7.85 7.06
Thermal Conductivity (W/cm) 3.98 4.25 3.15 2.38
CTE．106 17 19.1 14.2 23.5
Melting Point (℃) 1085 962 1064 660
Specific heat Capacity (J/Kg•K) 386 234 132 917 Corrosion in air Poor Poor Excellent Good
Deposion
Resistance to Electromigration High Very Low Very High Low
Delay Time(ps/mm) 2.3 2.2 3.2 3.7
Chapter 2
Literature Review
2.1 Cu Metallization
As the dimension of the devices shrink, the intrinsic switching speed, higher package density and higher complexity circuit of the device can be significantly improved. However, the increasing in wiring resistance resulting from the reduction in interconnection lines feature size and the resistance-capacitance (RC) time delays become the major limitations in achieving high circuit speeds. The device performance is severely impacted by interconnect parasitic considerations and to a lesser extent by active device switching speeds. In addition, as the dimensions of the devices shrink, problems associated with electromigration in aluminum-based interconnection lines have serious deleterious effects on device reliability.
Copper is being evaluated for silicon based VLSI metallization because it’s lower bulk electrical resistivity of 1.67 µΩ･cm as compared to 2.66 µΩ･cm of Al and the lower resistivity can greatly improve the RC time delay. Moreover, Cu has superior resistance to electromigration and stress voiding as compared to commonly used Al. Unfortunately, Cu atoms are quite mobile in most metals, as well as in silicon even at modest temperature. Moreover Cu is a deep level dopant in silicon, which resulting in the deterioration of devices such as leakage current and threshold voltage instability.
For III-V-based devices, Au was commonly used as the metallization metal.
If Cu is used as the metallization metal instead of Au, there will be several advantages such as lower resistivity, higher thermal conductivity, and lower cost
as shown in table 1. But as in the Si case, Cu also diffuses very fast into GaAs when Cu is directly contact with the GaAs substrate without any diffusion barrier [8]. Copper is a deep dopant in GaAs, if Cu diffuses into ohmic contact, the passivation layer SiNx and device active region, it will deteriorate the electrical properties of the devices. Figure 1 shows that the Cu-diffusion changes the EL2 centre into a deep donor (T3) with a lower activation energy, 0.7 eV [9].
Therefore, a very effective diffusion barrier is necessary to prevent Cu from diffusing and intermixing into the underlying materials both for Si and III-V systems.
2.2 Ohmic Contact for GaAs-Based Devices
The definition of ohmic contact on a semiconductor is to allow electrical current to flow into or out of the semiconductor freely without barrier [10]. The contact should have a linear I-V characteristic, be stable over time and temperature, and contribute as little contact resistance as possible.
2.2.1 Requirements for A Good Ohmic Contact Material
The requirements for a good ohmic contact include: [11]1. Low contact resistance
The first requirement for ohmic contacts of most devices is low contact resistivity, the resistance must be low enough not to affect the device I-V characteristics. The requirement for the reduction of the contact resistances has been continuing, because as the size of the device shrink to improve the device performance according to the scaling rule, the specific contact resistivity must decrease in order to keep the same contact resistance.
2. Thermally stable:
The second requirement for the ohmic contacts in GaAs devices is thermal stability during device fabrication and device operation.
In addition, a smooth surface, good adhesion, shallow horizontal and verticle diffusion depths, and low metal sheet resistance are required for ohmic contact in GaAs device. The requirements for ohmic contact are illustrated in Figure 2.
2.2.2 Guideline for Low Resistance Ohmic Contact Formation
When a metal deposited on a semiconductor, the Fermi levels in the metal and the semiconductor must be equal. For the Fermi levels to be equal in both sides, an energy barrier eΦB must exist between the metal and semiconductor interface [12]. The carriers can not transport freely because of the energy barrier.
The carrier transport mechanisms through the metal/semiconductor interface are strongly influenced by the doping concentration in the semiconductor and the temperature.
The current density (J) between contact metal and n-type semiconductor is shown below:
J = exp(-qΦB/E00) (2.1) When
E00=(qh/4π) x (ND/εm*) ½ ε:dielectric constant
ND: doping concentration m*:effective electron mass
χ:electron affinity
Equation 2.1 indicates that the current density increases when the doping concentration increases. When the semiconductor is extremely heavily doped (>high-1018cm-3), the electrons can tunnel through the energy barrier between metal and semiconductor to form good ohmic contact. This is called “tunneling mechanism”. The band diagram is shown in Figure 3. These diagrams give us guidelines to design the ideal M/S interfacial structure for low resistance ohmic formation.
Higher doping level is easy achievable in p-type GaAs. Because the dopants used in the p-type GaAs are not amphoteric and DX centers are associated with donors only, the ohmic contact to highly doped p-type GaAs can be easily formed. However, for n-type GaAs, the upper limit of the Si doping concentration achieved by the conventional ion-implantation technique is about 1018 cm-3. This level is limited by the formation of DX centers in n-type GaAs.
Due to this limitation on the of doping concentration, the formation of ohmic contact on n-type GaAs is difficult. The best way to modify the interfacial microstructure to produce low resistance ohmic contact is to form a new intermediate semiconductor layer (ISL) with low energy barrier or high carrier density at the metal/semiconductor interface after heat-treatment as shown in Figure 4. This fabrication process are called “deposition and anneal ohmic contact” [11].
The most common method of forming “deposition and anneal ohmic contact” on n-type GaAs is to apply an appropriate metallization scheme to the heavily doped GaAs followed by annealing process. During the annealing process, one of the constituent metals diffuses into the wafer and dopes the cap
GaAs layer heavily.
This ohmic contact fabrication technique needs a relatively simple fabrication system and with excellent reproducibility. Thus, this technique is suitable for manufacturing devices and used in a wide variety of GaAs devices.
However, the big disadvantage for this technique is that the process parameters can not easily be found, such as the contact metals, thickness of each metal layer, annealing time and temperature, diffusion coefficients, stress, surface energy, etc.
There are many kinds of “deposition and anneal ohmic contact” reported from the literatures. Ge-based ohmic contact, one of the famous ohmic contact systems, will be described in the following section.
2.2.3 Ge-based Ohmic Contact Materials
AuGeNi contact materials were invented by Braslau et al. in 1967. [13], and have been extensively used as n-type ohmic contact materials for advanced GaAs devices over 30 years. Although AuGeNi ohmic contacts provided low contact resistance and excellent reproducibility, this ohmic contact also has several drawbacks such as rough surface morphology, deep reaction depth in GaAs substrate, complex alloying process, and thermal instability after contact formation. These reasons cause the large scale spread of the contact resistivity.
To overcome these problems, many groups develop different ohmic contacts and try to apply them to the future sub-micron GaAs devices. The most popular ohmic contact system is Ge-based ohmic contact materials. Because Ge was found to dope heavily in the GaAs surface after heat-treatment, the contact resistivity of traditional Ge-based ohmic contacts is below 10-6Ωcm2 range.
In order to increase the donor concentration, a small amount of elements
can be added. These additional elements increase the donor concentration in the GaAs surface layer and decrease the energy barrier height at the contact metal/GaAs interfaces.
To increase the donor concentration, “direct” doping elements and
“indirect” doping elements were chosen in the past. The ”direct” doping elements were Sn, Sb, and Te which would increase the donor concentration in the GaAs surface layer by diffusing after heat-treatment. The “indirect” doping elements were Pd, Pt, and Au. Because the mixing enthalpy of the Ga with those elements (M) is smaller than that with As, Ga would form M-Ga phases with these elements in the GaAs surface. And Ge atoms could easily diffuse to the Ga vacancies in the vicinity of the GaAs surface to increase the donor concentration in the GaAs surface layer and reduce the Rc values [14]. In addition, Ag and Cu also belong to the “indirect” doping elements with wide solubility with Ga in a wide temperature range, forming M(Ga) solid solutions [15]. The formation of M(Ga) solid solutions would also incease the Ga vacancy concentration and facilitate heavy doping of Ge atoms in the GaAs surface layer.
Pd/Ge/Cu ohmic contact is also one kind of the Ge-based ohmic contacts, but the formation mechanism is different from other Ge-based ohmic contacts reported in literature. Before introducing the formation mechanisms of Pd/Ge/Cu ohmic contact, PdGe and Cu3Ge ohmic contact will be briefly introduced in next paragraphs.
Marshell et al. [16] developed the PdGe ohmic contact in 1980. PdGe ohmic contact is based on solid phase regrowth, not based on complex alloying process like traditional AuGeNi ohmic contact. The advantages of PdGe ohmic contact are uniform and shallow reaction, good thermal stability, and planer interface. The formation mechanism of PdGe is complex. When annealing at
100℃, Pd layer reacted with GaAs to form PdxGaAs ternary phase. The TEM image is shown in Figure 5 [17]. Because the mixing enthalpy of PdGa is smaller than PdAs, it could creat Ga vacancies in the near-interface region of the GaAs substrate. But at this temperature, ohmic contact is still not formed. After annealing above 300℃, the Pd reacted with amorphous Ge layer to form PdGe compounds and bring in some excess Ge atoms. Also at this temperature PdxGaAs ternary phase decomposed to form Pd atoms with GaAs layers. Then excess Ge atoms can dope the regrown GaAs layer and become highly doping layer and finally excess Ge atoms become epitaxial crystal Ge layer between the regrown GaAs and PdGe layer. The final structure is shown in Figure 6. Table 2 shows the summery of the contact resistivity of PdGe contact from four literatures. The lowest contact resistivity was about 10-6 Ωcm2 on n-type GaAs with Si doping concentration of 1018 cm-3.
M. O. Aboelfotoh et al. [18] have developed the Cu3Ge ohmic contact in 1994. This ohmic contact system is very unique. After annealing, Cu layer reacts with the Ge layer to become Cu3Ge structure. Because the chemical potential of Ga atom in the Cu3Ge is lower than that in the GaAs substrate, Ga atoms diffuse out to the Cu3Ge compound and creat many Ga vacancies. Ge atoms can dope in near-interface region of GaAs to highly doping. It also can be seen by SIMS profile in Figure 7. Besides, this compound is crystal structure and has long range order. The grain boundary of Cu3Ge compound is vertical to the GaAs surface. It can increase conductivity. The TEM image is shown in Figure 8.
Giving a summary of these two ohmic contact systems, the formation of the low resistance ohmic contact has two conditions. First, create Ga vacancies and then Ge atoms doping into near-interface region of GaAs substrate. It can use
some elements reacted with GaAs to form ternary phase. And after annealing, MGaAs phase decomposed to metal and regrown n+- GaAs layer same as the formation mechanism of the PdGe ohmic contact. It uses the chemical potential of Ga atoms in the ohmic compound is lower than in the GaAs so that Ga atoms diffuse out to create Ga vacancies as the formation mechanism of Cu3Ge ohmic contact. Second, it forms the low resistivity metallic compound like Cu3Ge crystal structure. Furthermore, we can combine high doping with low barrier high mechanism to form a good Ohmic contact. The Image shows in Figure 9.
2.3 GaAs Based Heterojunction Bipolar Transistors
The concept of the heterojunction bipolar transistor was first introduced by William Shockley in 1948. A detailed theory related to this device was developed by H. Kroemer in 1957 [23]. Kroemer realized that the use of a wide-band-gap emitter and low-band-gap base would provide band offsets at the heterointerface that would favor the injection of the electrons, in an n-p-n transistor, into the base while retarding hole injection into the emitter. These advantages would be maintained, even when the base is heavily doped, as is required for low base resistance, and the emitter is lightly doped. Thus in an HBT, high emitter injection efficiency would be maintained while parasitic resistances and capacitances would be lower than for a conventional homojunction bipolar transistor.
The cross section of a basic n-p-n AlGaAs/GaAs heterojunction bipolar transistor is shown in Figure 10. The n-type emitter is formed in the wide-band-gap AlGaAs while the p-type base is formed in the lower band gap GaAs. The n-type collector, in this basic device, is also formed on GaAs. To
facilitate the formation of the ohmic contacts, a heavily doped n+-GaAs layer is present between the emitter contact and the AlGaAs layer. The energy band diagram of this device is shown in Figure 11.
Some inherent advantages of HBTs over silicon bipolar transistors are as follows [24]:
(1) Due to the wide-band-gap emitter, a much higher base doping concentration can be used, decreasing base resistance.
(2) Emitter doping can be lowered and minority carrier storage in the emitter can be made negligible, reducing base-emitter capacitance.
(3) High electron mobility, built-in drift fields, and velocity overshoot combine to reduce the electron transit time.
(4) Semi-insulating substrates help reduce pad parasitics and allow convenient integration of devices.
(5) Early voltages are higher and high injection effects are negligible due to high base doping.
Figure 12 shows the band diagram of a homojunction BJT and HBT. The energy band gap difference between the emitter and the base gives the HBT a substantial edge over BJT. When the base-emitter junction of a BJT is forward biased, both the electrons forward-injection into the base and the hole back-injected into the emitter experience the same amount of energy barrier. For the HBT, when the base-emitter junction is forward biased, the holes, which are back inject from the base into emitter, experience a △Eg larger energy barrier than the electrons, which are injected into the base. So, the HBT provides a design freedom meaning that a HBT structure design can have a heavily base dope to reduce the base resistance, while still maintaining a high current gain.
We can quantify the advantage of HBT compared to BJT by calculating the ratio of the collector current to the base current [25].
⎟⎟ ⎠
Equation (2.1), which relates the intrinsic carrier concentration to the energy gap, was used in the derivation. It’s defined as the current gain which is one of the most important parameters in bipolar transistors.
Among several HBT device structures, InGaP/GaAs HBTs are becoming attractive as compared with the AlGaAs/GaAs HBTs in the circuit applications such as high-speed analog-to-digital converters, high-power microwave amplifiers, and high-speed optical communication circuits due to their robust reliability and excellent DC and RF performances. In addition, several advantages have been claimed for this material system, such as large valence-band discontinuity, very low interface recombination velocities with GaAs, significantly less oxidation in comparison with AlGaAs, no DX centers issue, and good selective etch with GaAs [26].
Table
Table 2 Summary of PdGe ohmic contact data from literature
Pd / Ge ( A ) ρc (Ωcm2) Doping (cm-3) Formation Condition Ref.
500 / 1260 ~10-6 1 x 1018 Anneal 325℃ for 30min
[18]
500 / 1260 2.84 x 10-6 2 x 1018 Anneal 330℃ for 30min
[19]
500 / 1250 1.4 x 10-6 3 x 1018 RTA 425 ℃ for 60sec
[20]
200 / 400 5.33 x 10-6 2 x 1018 RTA 350 ℃ for 45sec
[21]
Figure
Fig1. Optical DLTS spectra from n-type GaAs. The rate window was 30s-1.
Fig2. Ideal interfacial structure for the low-resistance ohmic contact
Fig3. Conduction mechanisms through metal/semiconductor interface (a) lightly doped (ND < 1017 cm−3)
(b) intermediate level of 1017–1018 cm−3 (c) heavily doped (ND > high-1018 cm−3)
Fig4. Cross-section of the metal/semiconductor interface with ISL
Fig5. The TEM image of the Pd /Ge contact after annealing at about 100℃
(a)
Fig6. (a) The structure and (b) TEM image of the PdGe contact
n-GaAs
n+-GaAs (regrowth)
epi-Ge (crystal)
PdGe
Fig7. SIMS profiles of an Cu3Ge contact formed at 400℃ for 30min
Fig8. TEM image of an Cu3Ge contact formed at 400℃ for 30min
Fig9. Energy band diagrams of metal/semiconductor interfaces with (a) highly doped ISL (b) low energy barrier ISL.
Fig10. Schematic of the cross section of an HBT structure
Fig11. Energy band diagram of an HBT structure
(a)
(b)
Fig12. The band diagrams of (a) a homojunction bipolar transistor and (b) a heterojunction bipolar transistor.
Chapter 3 Experiment
The contents of our experiments can be divided into three parts. First, the Pd/Ge/Cu ohmic contact formation on n-type InGaAs was studied. Second, the Pd/Ge/Cu ohmic contact was applied on the InGaP/GaAs HBTs as the n-type InGaAs contact metal to form Au-free fully Cu-metallized InGaP/GaAs HBTs.
Third, material analysis and electrical characterizations of the devices were performed.
3.1 Research of Ohmic Contact
The specific contact resistances of the n-InGaAs/ Pd (15nm)/ Ge (150nm)/
Cu (150nm) were carried out by transmission line (TLM) method. The TLM patterns were fabricated by I-line photolithography. The InGaAs mesa was etched by H3PO4/H2O2/H2O solutions. After the conventional organic solvent cleaning process, the substrates were chemically cleaned in a solution of HCl：
H2O (1：1 by volume) to remove the native surface oxide layer, the samples were load into the evaporation chamber. Pd (15nm)/ Ge (150nm)/ Cu (150nm) compositions were then deposited on the substrates using an electron-beam evaporator in a pressure of ~1x10-6 Torr. After the metal evaporation, the samples were immersed into ACE and IPA with the common lift-off process, followed by a high pressure DI water rinse to remove the residues. After lift off, the samples were annealed in a conventional tube furnace at various temperatures from 150oC to 450 oC for 20 minutes.
The thermal stability test of the Pd/Ge/Cu ohmic contact was performed by high-temperature annealing test (250oC) in a N2-ambient tube furnace for 24 hours. The ohmic contact resistances (RC) of the samples were measured using the transmission line model (TLM). The sheet resistance (Rsh) is directly measured using a four-terminal sensing measurement (also known as a four-point probe measurement (Fig18)) for the sample treatment at different temperature.
So as to understand the formation mechanism of Pd/Ge/Cu ohmic contact, the Pd/Ge/Cu multilayer was analyzed by XRD, TEM, AES, and AFM. Phase identification was analyzed by the X-ray diffraction (XRD). The interface microstructure of the n-InGaAs/Pd/Ge/Cu ohmic contact materials was observed by transmission electron microscopy (TEM) and the interfacial elements analysis was studied by using Auger electron spectroscopy (AES). The surface
So as to understand the formation mechanism of Pd/Ge/Cu ohmic contact, the Pd/Ge/Cu multilayer was analyzed by XRD, TEM, AES, and AFM. Phase identification was analyzed by the X-ray diffraction (XRD). The interface microstructure of the n-InGaAs/Pd/Ge/Cu ohmic contact materials was observed by transmission electron microscopy (TEM) and the interfacial elements analysis was studied by using Auger electron spectroscopy (AES). The surface