IN
DEGREE PROJECT INFORMATION AND COMMUNICATION TECHNOLOGY,
SECOND CYCLE, 30 CREDITS ,
STOCKHOLM SWEDEN 2019
Study of ohmic contact formation
on AlGaN/GaN heterostructures
KAI-HSIN WEN
KTH ROYAL INSTITUTE OF TECHNOLOGY
D F
KTH Royal Institute of
Technology
Study of ohmic contact formation on
Al-GaN/GaN heterostructures
Master’s thesis in Nanotechnology
Kai-hsin Wen
Master’s thesis 2019
Study of ohmic contact formation on
AlGaN/GaN heterostructures
Kai-hsin Wen
Information and Communication Technology KTH Royal Institute of Technology
Study of ohmic contact formation on AlGaN/GaN heterostructures Kai-hsin Wen
© Kai-hsin Wen, 2019.
Supervisors: Niklas Rorsman, Chalmers University of Technology Ding-yuan Chen, Chalmers University of Technology Examiner: Mattias Hammar, KTH Royal Institute of Technology
Master’s Thesis 2019
Information and Communication Technology KTH Royal Institute of Technology
SE-100 44 Stockholm Telephone +46 8 790 60 00
Cover: The contour plot of the obtained Rc from the laser focus/dose matrix.
Typeset in LATEX, template by David Frisk
Printed by KTH
Study of ohmic contacts formation on AlGaN/GaN heterostructures Kai-hsin Wen
Information and Communication Technology KTH Royal Institute of Technology
SE-100 44 Stockholm
Abstract
It is challenging to achieve low-resistive ohmic contacts to III-nitride semiconductors due to their wide bandgap. A common way to reduce the contact resistance is to recess the ohmic area prior to metallization. In the minimization of the contact resistance, parameters like the recess depth, anneal temperature and design of the metal stack are commonly optimized. In this work, three other approaches have been evaluated. All experiments were performed on AlGaN/GaN heterostructures. The fabricated ohmic contacts were recess etched, metallized with a Ta/Al/Ta stack, and annealed at 550-575◦C.
Firstly, it is shown that the laser writer intensity, transmittance and focus offset during optical lithography affect the contact resistance. The reason is believed to be the variation in the resist profile, which has an impact on the metal coverage. At the optimum intensity/transmittance/focus condition, which generates a relatively medium undercut, a contact resistance of 0.23 Ωmm was obtained.
In the second approach, the metal layer of annealed contacts was removed by wet etching, followed by the re-deposition of a metal stack and annealing. The purpose was to increase the amount of N vacancies in the AlGaN, which are responsible for the contact formation. A minimum contact resistance of 0.41 Ωmm was achieved with this method, compared to 0.28 Ωmm with the regular method (without re-metallization).
In the last approach, the bottom Ta layer was sputtered, whereas evaporation was used in all other cases. The minimum contact resistance was found to be 0.6 Ωmm, which was higher than for the evaporated contacts. The reason was assumed that the thickness of sputtered Ta should be thinner than the evaporated Ta due to its higher density. Moreover, the obtained lower sheet resistance is assumed to caused by the atomic scale damage due to the high energy ions during sputtering.
Sammanfattning
En utmaning med III-nitrid-halvledare är att uppnå låg-resistivitetskontakter, på grund av deras breda bandgap. Ett konventionellt tillvägagångsätt för att reducera kontaktresistansen är att fördjupa ohmska ytan före metallisering. I strävandet av att minska den ohmska resistansen sker vanligtvis en optimering av följande parame-trar, recessddjup, anlöpningstemperatur och metallagersdesign. I detta arbete så har samtliga tre parametrar evaluerats. Alla experiment utfördes på AlGaN/GaN-heterostrukturer. De tillverkade ohmska kontakterna var recesssetsade, metalliser-ade med ett Ta/Al/Ta lager och anlöpt vid 550-575◦C.
Den primära undersökningen, visar att laserritarintensitet, transmission och -fokusförskjutning under optisk litografi inverkar på kontaktresistansen. Anledningen antas vara variation i resistprofilen, vilket påverkar metallbeläggningen. Vid opti-mal intensitet/transmission/fokus-förhållanden, (som genererar en underskärning), blev den resulterande kontaktresistansen 0.23 Ωmm uppmätt.
I en sekundär undersökning, avlägsnas ohmska kontaktens metallager genom våtet-sning, följt av en återdeponering av ett nytt metallager, samt anlöpning. Syftet var att öka mängden N-vakanser i AlGaN-lagret, som formar ohmska kontakten. Min-sta kontaktresiMin-stansen uppmätt var 0.41Wmm, att jämföras med 0.28 Ωmm, som uppnåddes genom den konventionella metoden (utan återmetallisering).
Acknowledgements
This thesis work is pursued at the Department of Microtechnology and Nanoscience - MC2, Microwave Electronics Laboratory. It is a great experience for me to work here and during this period I have gained so much knowledge.
I would like to sincerely thank my supervisor Niklas Rorsman for giving me the chance to come to Chalmers and work with him. Despite his tight schedule, his door is always open for discussion whenever I have any question. Furthermore I would like to express my deepest gratitude to my daily supervisor Ding-yuan Chen for his patience and guidance. Thank you for teaching me the processing techniques in the cleanroom and all the measurements used in this thesis. I would also like to thank Hans Hjelmgren for helping me build a simulation model and teaching me to use TCAD simulation. Also, I would like to appreciate all the people in Microwave Electronics Laboratory for creating such a fantastic working environments.
Finally, I am deeply grateful to my family and all my friends for always being supportive encouraging me to push on.
Contents
List of Figures xiii
List of Tables xv
1 Introduction 1
1.1 Thesis objectives and summarized results . . . 2
1.2 Thesis outline . . . 3
2 Ohmic contact technology 5 2.1 Metal-semiconductor contact . . . 5
2.1.1 Current transport mechanisms . . . 6
2.2 Ohmic contact mechanism . . . 8
2.3 Ohmic contact types . . . 9
2.3.1 Planar contacts . . . 9
2.3.2 Recessed contacts . . . 10
2.3.3 n+-GaN regrowth contacts . . . 12
2.4 Physical modelling . . . 13 3 Fabrication Process 17 3.1 Standard Cleaning . . . 17 3.2 LPCVD . . . 18 3.3 Mesa . . . 18 3.3.1 Photolithography . . . 19 3.3.2 Plasma Ashing . . . 20 3.3.3 Plasma Etching . . . 21 3.4 Ohmic Contact . . . 21 3.4.1 Metal Deposition . . . 21 3.4.2 Re-metalization . . . 22 3.4.3 Sputtered Ta . . . 23 4 Characterization 25 4.1 Scanning electron microscopy . . . 25
4.2 Transmission line method . . . 26
4.2.1 TLM structure . . . 26
4.2.2 Epi-layer sheet resistance Rsh . . . 27
Contents
5 Results 31
5.1 Laser writer focus/intensity test . . . 31 5.2 Ohmic contact re-metallization . . . 33 5.3 Sputtered Ta . . . 36
6 Conclusion and future work 39
List of Figures
1.1 Schematic of GaN HEMT structure. . . 1 2.1 Band diagram of metal-semiconductor in equilibrium [18] . . . 5 2.2 Schematic of three different carrier transport mechanism for different
Nd. [24] . . . 7
2.3 E00 plotted as a function of doping concentration for GaN at T= 300K. 8
2.4 Illustration of three different recess depth cases. (a)the barrier is still present (b)the barrier is still present but it is too thin to retain 2DEG under it (c)the barrier is completely removed in etching process. . . . 11 2.5 The simulation model structure(a)model from software (b)schematic
of the model . . . 13 2.6 Schematic of the simulation model structure and the obtained results
of various doping depth, concentration and the temperature. . . 15 3.1 Schematic of the ohmic structure in this work. . . 17 3.2 The schematic of the position of photoresist and the laser beam. . . . 19 3.3 The etchant versus etching target material (X: The target material
can not be etched by the etchant, : The target material can be etched by the etchant, –: Not found from the literature). . . 23 4.1 (a)Schematic TLM strcture (b)Microscope image of the TLM structure 26 4.2 (a)A schematics shows the different components of Rtot. (b) Total
resistance Rtot plotted as a function of isolation distance dx . . . 27
4.3 Current flow through contact for high and low ρc . . . 29
5.1 Illustration of the laser focus/transmittance/intensity matrix with the exact energy received on the substrate. . . 31 5.2 The contour plot of the obtained Rc from the laser focus/dose matrix. 32
5.3 The schematic of (a) larger sidewall angle (b) smaller sidewall angle and (c) the GaN depletion region due to too small sidewall angle. . . 32 5.4 (a)Rc comparison of original and re-deposited ohmic metal stack.
(b)The plot of Rc versus annealing time and the comparison of
orig-inal and re-deposited ohmic metal stack. . . 33 5.5 The plot of Rc versus annealing time while first Ta layer thickness
differs. . . 34 5.6 The schematic shows the enlargement of the photoresist and the cross
List of Figures
5.7 Cross sectional SEM image of the sample with the measured Ta thick-ness. . . 35 5.8 The obtained (a)Rc and (b)Rshversus annealing time with different
List of Tables
2.1 Comparison of Rc values and different metal schemes on different
heterostructure . . . 10
2.2 Literature values of Rc for recess contacts on different heterostructure 12 2.3 Rc of regrowth n-GaN contacts on different heterostructure . . . 13
2.4 The obtained ρc of different doping concentration . . . 15
2.5 The obtained ρc of different doping depth . . . 15
1
Introduction
The gallium nitride (GaN) based high electron mobility transistor (HEMT) has attracted considerable attention during the past two decades and has become an attractive candidate for high frequency and high power applications [1, 2]. GaN is a semiconductor material which possesses a wide bandgap, a high breakdown field and a high saturation electron drift velocity. Furthermore, the GaN HEMT is based on a heterojunction, commonly AlGaN/GaN, which generates a two-dimensional electron gas at the interface with enhanced electron mobility compared to doped GaN. Grown on silicon carbide (SiC) substrate, high thermal conductivity is also possible. Fig. 1.1 shows a cross-section of a SiNx-passivated AlGaN/GaN HEMT with the three
terminals gate, source, and drain marked out. As shown in the figure, the source-and drain terminals are ohmic contacts, while the gate is a Schottky contact.
Figure 1.1: Schematic of GaN HEMT structure.
Since the ohmic contacts constitute parasitic elements in the GaN HEMT, it is essential to minimize the contact resistance (Rc) in order to promote a good
per-formance. However, the wide band gap of AlGaN between 3.4 (GaN) and 6.2 eV (AlN) makes this a challenge. The standard ohmic metal stack for GaN HEMTs is Ti/Al/Ni/Au [3, 4, 5, 6]. However, with the Ti-based metal scheme, a high anneal-ing temperature of 800 to 900◦ C is required. Such high anneal temperature may lead unwanted effects on the heterostructure, such as increasing of sheet resistance
Rsh as well as rough surface morphology and poor edge acuity due to the formation
1. Introduction
of Mo/Al/Mo/Au and a pre-treatment with SiCl4 plasma, a very low Rc of 0.15
Ωmm was obtained at an annealing temperature of 650◦ C.
Ta- based metal schemes have also been investigated in several studies. The metal stacks of Ta/Ti/Al [8, 9, 10] require annealing temperature of 700 to 950◦ C. An-other Ta- based metal scheme,Ta/Al/Ta, was developed demonstrating a lower an-nealing temperature. The lowest Rc of 0.06 and 0.28 were achieved with different
Al thickness and annealing temperature of 550 and 575◦, respectively [11]. Lin et al demonstrated recessed Ta/Al/Ta ohmic contacts, where the ohmic contact is formed on the recess sidewall. The lowest Rcobtained is 0.24 Ωmm with a tilt angle of 10◦
during evaporation of the first Ta layer and an annealing temperature of 575◦C [12]. To improve the ohmic contacts, several studies have focused on recessed ohmic contact. Zhang et al performed recess etching by inductive coupled plasma etching (ICP) before metal deposition and obtained a Rc of 0.3 Ωmm [13]. Wang et al
reported that with 2DEG totally removed, a very low Rcof 0.26 Ωmm was obtained
due to direct contact between ohmic electrode and 2DEG [14]. Although the recess etching yields a low Rc, the ohmic recess process needs to have excellent depth
control and be low-damage since ion damage on the epitaxial layer may lead to worse Rc. Regrown ohmic contact can produce excellent ohmic contact with an Rc
of 0.16 Ωmm [15, 16]. However, the processing is complicated and the associated costs much higher compared to recess etching, which may make it less suitable for large scale processes. Consequently, recess ohmic contacts are preferable and an ohmic recess below 2DEG is performed in order to avoid the requirement of precise control of recess depth.
1.1
Thesis objectives and summarized results
The main purpose of this thesis is to investigate and evaluate three different meth-ods intended for decreasing Rc. In the first method, the parameters of the optical
lithography are optimized in order to investigate the impact of the resist profile on recessed contacts. The second method is the removal of the annealed metal, followed by re-metallization to increase the formation of TaN. The third method is to sputter the first Ta layer to improve the metal coverage of the sidewall and the in-situ Ar cleaning is also performed to improve the Rc.
Electrical characterization was performed by the transfer length method (TLM). Scanning electron microscopy is also used to check the cross section of ohmic con-tact. Additionally, technology computer-aided design (TCAD) simulations were per-formed to study the impact of doping concentration, doping depth and temperature on the formation of ohmic contact.
Characterization results show that the lowest Rcof 0.23 Ωmm can be achieved with
1. Introduction
that the idea of enhancing N atoms extraction mechanism is not achieved in this work. The third approach, sputtered first Ta layer, shows the highest Rc of 0.6
Ωmm in this work. Assumptions are made according to the observed Rc, Rsh and
the SEM characterization, and these assumptions may be the factors that lead to the failure of lowering the Rc. Further study on the adequate first Ta thickness as
well as the in-situ Ar cleaning are required to decrease the Rc.
1.2
Thesis outline
This thesis is organized in six chapters:
• The topic and a brief background are introduced in Chapter 1 followed by the aim and outline of the thesis.
• Theory concerning the mechanism of ohmic contact formation, current trans-port mechanism and different types of ohmic contacts are presented in Chapter 2.
• The fabrication process is described in Chapter 3.
• The main characterization methods are introduced in Chapter 4. • The results are presented and discussed in Chapter 5.
2
Ohmic contact technology
2.1
Metal-semiconductor contact
Contacts between metal and semiconductor is an essential part of all the electronic devices. A low contact resistance Rc is an important factor in ohmic contacts.
However, ohmic contact formation in III-nitride semiconductors, eg. (Al)GaN, has been an issue due to its wide bandgap. In this project, we investigate various aspects, which potentially could improve the formation of ohmic contact, including electron beam evaporation and sputtering and re-deposition of ohmic metal stack.
Metal-semiconductor contacts can be divided into two types. One is Schottky con-tact acting as a diode with the rectifying property. The other is a low resistive ohmic contact linear I-V characteristic. Generally, the Schottky barrier is formed by a metal contacting with an undoped or low-doped semiconductor. In contrast, when metal contacts with a highly-doped semiconductor and an additional thermal treatment can form ohmic contacts.
2. Ohmic contact technology
The band diagram of metal contacting an n-type semiconductor is shown in Fig. 2.1. The Fermi level, EF, should be flat when there is no bias applied across the
junction [19]. This results in band bending and causes a depletion region with the width W. The Schottky barrier height, φB, is a barrier to electrons. φB depends on
the metal work function, φM, and electron affinity, χ, of the semiconductor. The
metal work function φM is the the work that is required to remove the electron
from the metal surface to the vacuum level, Evacuum. The semiconductor electron
affinity, χ, is the energy difference between Evacuum and the conduction band of the
semiconductor.The Schottky barrier height is given in Equation.2.1.
φB = φM − χ (2.1)
In order to reach the full potential of semiconductor device technologies, it is nec-essary to minimize parasitic losses. Therefore, ohmic contacts with low resistance and linear I-V characteristic are needed.
In III-nitride devices it is not enough to only deposit metals on the semiconductor to form an ohmic contact. Thermal annealing is required to form a metal nitride at the interface [22, 23].
Specific contact resistivity ρc is commonly used to characterize ohmic contacts. ρc
is defined in Equation 2.2, where J is the current density and the voltage is at zero bias. In this project, contact resistance per width, Rc, is calculated with the unit of
Ωmm, and is used to evaluate the performance of the ohmic contacts.
ρc = ∂J ∂V !−1 v=0 (2.2)
2.1.1
Current transport mechanisms
There are three different mechanisms for carrier transport which is dependent on the n-type doping concentration, Nd. The depletion width is proportional to Nd
(Equation 2.3).
W ∝ √1
Nd
(2.3) For low-doped semiconductor, Nd < 1017 cm−3, thermal energy for the carriers is
required to overcome the Schottky barrier, and therefore this mechanism is named thermionic emission (TE, Fig. 2.2a). If the semiconductor is highly-doped, the Schottky barrier height is constant, but with a much narrower depletion width (Equation 2.3). Due to the small W, it is possible for carriers to tunnel through the barrier. This tunnel mechanism is called field emission (FE, Fig. 2.2b). Field emission occurs when the doping concentration Ndis higher than 1019 cm−3. For the
moderately-doped semiconductors, Nd between 1017 cm−3 and 1019 cm−3, the
2. Ohmic contact technology
field emission (TFE, Fig. 2.2c). The depletion width is too large for carriers to tun-nel through, but with the extra thermal energy, carriers can be thermally excited to an energy above EF where W is thin enough to tunnel through the barrier.
(a)Thermionic emission (b)Field emission (c) Thermionic field
emission
Figure 2.2: Schematic of three different carrier transport mechanism for different
Nd. [24]
The dominant mechanism can be predicted by the calculated value of kT
qE00 (Equation 2.4). This calculation gives ratio between TE and the other two mechanisms.
E00= h 4π( Nd mε) 1 2 (2.4)
where h, m and ε is Plank’s constant, effective mass and the dielectric constant respectively. The calculated value of qEkT
00 gives the ratio between TE and the other two mechanisms. When kT /qE00 1 , field emission is the dominating mechanism
for carrier transport. Thermionic emission dominates when kT /qE00 1 while
kT /qE00 = 1 indicates that thermionic field emission is dominant. This behavior
described in Fig. 2.3, in which E00 is plotted as a function of doping concentration
2. Ohmic contact technology
Figure 2.3: E00plotted as a function of doping concentration for GaN at T= 300K.
2.2
Ohmic contact mechanism
The mechanism of forming ohmic contact at AlGaN/GaN heterostructures is com-plicated and has not been fully understood. Many studies were performed to inves-tigate the mechanisms of obtaining the low Rc in ohmic contacts with Ti/Al based
contacts. The most frequent and acceptable explanation is that nitrogen atoms are extracted from AlGaN layer and thus leaving N-vacancies as n-dopants [25, 26, 27]. Another explanation for the mechanism of ohmic contact formation is that with the low work function compound/alloy formed at the metal-semiconductor interface, a low barrier height is formed which enhance the thermionic or thermionic field emis-sion [28]. Luther et al proposed that the diffuemis-sion of native oxide on GaN and Al through Ti will be reduced, and therefore the ohmic contact with low work function Al-Ti intermetallic phase is formed [28].
The formation of TiN is the most common mechanism adopted to explain the low Rc.
Nitrogen atoms are extracted from AlGaN to form TiN, which results N-vacancies in the barrier layer acting as n-dopants. Therefore, the barrier layer becomes heavily doped, which increases the tunneling probability and lowers the Rc [27]. Moreover,
it was proposed by Luther et al that during the process, not only TiN but AlN was also formed [28]. Formation of AlN also creates N-vacancies, leaving heavily doped interface that narrows the depletion region. The barrier height might also be de-creased because of the potential across AlN [29]. Chaturvedi et al developed a model to study the mechanism of Ti/Al/Mo/Au metal stacks forming ohmic contacts on AlGaN/GaN heterostructure and the model disclosed that Ga diffused throughout the metal and reacted with Mo [30]. The Ga-vacancies lead to a charge imbalance and therefore, N atoms nearby replace the vacant Ga site in the lattice causing N-vacancies acting as n type dopants as described above.
2. Ohmic contact technology
removing TaN by wet chemical etching and the re-depositing the metal stack were anticipated to increase the extraction of N from AlGaN.
2.3
Ohmic contact types
2.3.1
Planar contacts
Approaches to fabricate ohmic contact on AlGaN/GaN heterostructure includes planar, MBE regrowth and recess etched contacts. Planar contact is the simplest and the standard method to fabricate ohmic contacts on GaN HEMTs. A planar ohmic contact is formed by metallizing the contact area followed by annealing. Values of
Rc from the literature are shown in Table. 2.1.
It is clearly found that typical annealing temperature are over 800◦C with an excep-tion of an Mo/Al/Mo/Au metal stack that the required annealing temperature of 650◦C. In addition, for Ti/Al based planar ohmic contacts, higher Al concentration in AlGaN barrier leads to higher Rc, which might due to the larger energy barrier.
2. Ohmic contact technology
Metal Stack Barrier layer Annealing T Rc Ref
(◦C) (Ωmm)
Ti/Al/Ni/Au GaN/Al0.28Ga0.72N/AlN 820 0.45 [4]
Ti/Al/Ni/Au Al0.24Ga0.76N 830 0.2 [5]
Mo/Al/Mo/Au In0.17Al0.83N/AlN 650 0.15 [7]
Ti/Al/Ni/Au In0.18Al0.82N/AlN 900 0.15 [32]
Ta/Si/Ti/Al/Ni/Au In0.18Al0.82N/AlN 825 0.36 [33]
Ti/TiN GaN/Al0.2Ga0.8N/AlN 850 0.13 [34]
Ti/TiN GaN/Al0.35Ga0.65N/AlN 850 0.6 [34]
Table 2.1: Comparison of Rc values and different metal schemes on different
het-erostructure
2.3.2
Recessed contacts
For planar contacts, the distance between ohmic metal stack and 2DEG is large and makes it difficult to obtain low Rc. In order to reduce the distance between
2DEG and ohmic metal, and avoid the required high anneal temperature of planar contacts and the complex process of regrown n+-doped contacts, recessed contacts
has been extensively studied. Prior to the ohmic metallization, an recess etching step is performed, which is able to reduce the annealing temperature.
Čičo et al reported the Rc of 0.39 Ωmm obtained by recess etching before metal
de-position and annealing with temperature of 700◦C while the conventional annealing temperature was above 800◦C [39].
2. Ohmic contact technology
aggravates the formation of ohmic contact. Fig. 2.4(b) and (c) show that the barrier is still presented but it is so thin that there is no 2DEG formed under it and the barrier is completely removed during the etching process, respectively. In this work, to avoid the difficulty of controlling the etching depth precisely, the recess is etched below 2DEG. Therefore, the deposited ohmic metal can have direct contact to the AlGaN and GaN layer at the sidewall.
Figure 2.4: Illustration of three different recess depth cases. (a)the barrier is still present (b)the barrier is still present but it is too thin to retain 2DEG under it (c)the barrier is completely removed in etching process.
Bergsten demonstrated a very low Rcof 0.14 Ωmm with the almost removed InAlN
barrier [40]. Lin et al reported a low Rc of 0.24 Ωmm with the completely removed
AlGaN barrier layer[12]. Buttari et al obtained the Rcof 0.27 Ωmm with the barrier
only slightly etched [41]. Generally, there is no clear conclusions of the recess depth since etching depth is not uniform. Table. 2.2 shows some literature values of Rc
2. Ohmic contact technology
Metal Stack Barrier layer Annealing T (◦C) Rc (Ωmm) Ref
Ti/Al/Ni/Au In0.18Al0.82N 700 0.39 [39]
Ta/Al/Ta In0.17Al0.83/AlN 550 0.14 [40]
Ta/Al/Ta Al0.25Ga0.75N 575 0.24 [12]
Ti/AlMo/Au GaN/Al0.3Ga0.7N/AlN 850 0.26 [42]
Table 2.2: Literature values of Rc for recess contacts on different heterostructure
2.3.3
n
+-GaN regrowth contacts
Regrowth contacts process starts from etching past the barrier layer, and a lattice matched, highly doped n-GaN is then grown in the recess. Saunier et al utilized the regrowth n-doped GaN contacts to achieve extremely low Rc of 0.1 Ωmm [35],
which is similar in Tang’s study [36]. Regrowth n-GaN in the recess improves the
Rc due to the direct contact to 2DEG. This fabrication method can offer extremely
low Rc value, but the process is complicated and the cost is high, which possibly
makes this method unsuitable for mass production. Moreover, to obtain low Rc, a
large density of 2DEG is required while the standard electron sheet concentration
ns of AlGaN/GaN heterostructure is about 1013 cm−2 which makes it more difficult
to achieve. Literature values of the contact resistance on different heterostructure are shown in Table. 2.3. From those papers, higher ns in the barrier decreases the
contact resistance while the obtained value of Rc from barriers with moderate ns
2. Ohmic contact technology Berrier Rc ns Ref (Ωmm) (cm−2) In0.17Al0.83N/AlN 0.16 1.92 × 1013 [15] (2.5/1.5 nm) In0.18Al0.82N/AlN 0.10 — [35] (8/1 nm) In0.17Al0.83N/AlN 0.16 2 × 1013 [37] (5.6/1 nm) GaN/In0.17Al0.83N/AlN 0.22 1.6 × 1013 [38] (2/3.5/1 nm)
Table 2.3: Rc of regrowth n-GaN contacts on different heterostructure
2.4
Physical modelling
TCAD simulation was made to promote further understanding of the ohmic con-tact formation. The simulation model was based on the thermionic field emission mechanism. In the simulation the contact on the top of the structure is modelled as a Schottky contact while the bottom contact is a perfect ohmic contact (Fig. 2.5b). The GaN layer and the lower part of the AlGaN layer are assumed to have a background doping of 1015cm−3.
The n-doping, representing the N-vacancies, is distributed in the upper part of AlGaN layer where the dopants are. The simulations study the impact of n-doping (N-vacancies), doping depth and working temperature.
(a) (b)
2. Ohmic contact technology
As expected, the contact resistance is inversely proportional to the doping concen-tration (Fig. 2.6a and b). The specific contact resistivity is obtained from the slope in Fig. 2.6 and is listed in Table. 2.4. For doping concentration higher than
> 1019cm−3, field emission (FE) is dominant. The depletion width becomes thinner while the Schottky barrier keeps the same height. Therefore, it becomes easier for carrier to tunnel through the barrier and hence, the ρc is decreased. For thermionic
field emission (TFE), the doping concentration is simulated from 1018 to 1019cm−3.
The carriers with enough thermal energy can tunnel through the midsection of the barrier, which implies a direct temperature dependence. The relation between Rc
and doping concentration of FE and TFE is shown below [43].
Rc∝ exp φ B √ ND (F E) (2.5) Rc∝ exp " φB √ NDcoshEkT00 # (T F E) (2.6)
From the equations, it is clearly shown that for FE and TFE, Rc is proportional to
1 √
ND. The simulation results is consistent to the experimental results reported by
Yu et al [43]. The investigation of re-metallization is motivated by the assumption that the creation of N-vacancies is limited by the availability of Ta near the metal and semiconductor interface. By removing the metal nitride, more N-vacancies may be created in the barrier layer, and Rc is anticipated to be decreased.
The depth of the region where N-vacancies are created is unknown. From the sim-ulations, it is clear that a larger doping depth promotes a low contact resistivity (Fig. 2.6c and Table 2.5). The doping depth starts at the surface and with larger doping depth, the dopants spread deeper in the AlGaN layer, which increases the effective area. Due to the current crowding effect, which is introduced in Chapter 4, the resistance has an inversely proportional relation to the effective area [46]. Consequently, with a larger doping depth, lower resistance is obtained.
The temperature dependence of the contact resistivity is simulated with a constant doping concentration of 1019cm−3 while the temperature is varied from 300K to
450K. From the Fermi-Dirac distribution function, more carrier are excited above the Fermi level at higher temperature. Therefore, current density increases and leads to smaller Rc. Chang et al reported that for moderate doping concentration,
Rc has strong dependence on temperature [44]. The reported results are consistent
2. Ohmic contact technology
(a) (b)
(c) (d)
Figure 2.6: Schematic of the simulation model structure and the obtained results of various doping depth, concentration and the temperature.
Doping concentration (cm−3) 1 × 1018 5 × 1018 9 × 1018 1 × 1019 1 × 1020 K 1 × 1021 ρc(Ωcm2) 6.98 1.16 × 10−4 1.18 × 10−6 1.04 × 10−6 8.11 × 10−9 4.47 × 10−9
Table 2.4: The obtained ρc of different doping concentration
Doping depth (µm) 0.01 0.025
ρc (Ωcm2) 1.79 × 10−6 5.57 × 10−7
Table 2.5: The obtained ρc of different doping depth
Temperature (K) 300 350 400 450
ρc (Ωcm2) 1.79 × 10−6 1.07 × 10−6 6.18 × 10−7 3.53 × 10−7
3
Fabrication Process
The fabrication process for ohmic contacts on AlGaN/GaN heterostructures is in-troduced in this chapter. The process starts from standard cleaning. Then a silicon nitride passivation layer with a thickness of 55 nm is deposited by low-pressure chemical vapor deposition (LPCVD). Device isolation is achieved with mesa etching by inductive coupled reactive ion etching (ICP-RIE) plasma after photolithography definition of the mesa structure. Then the ohmic structure is patterned with laser writer where an image reversal photoresist AZ5214 is utilized for lift off. After the ohmic recess etching with ICP-RIE, the ohmic metal stack Ta/Al/Ta is deposited by electron beam evaporator with the same resist, making the ohmic metal self-aligned to the recess. In the final step, lift-off is performed to obtain the TLM structure (Fig. 3.1).
Figure 3.1: Schematic of the ohmic structure in this work.
3.1
Standard Cleaning
The 15 × 15mm2 samples used in our project is diced from the same epitaxial wafer,
which the AlGaN/GaN epitaxial heterostructures are grown on SiC by metal-organic chemical vapor deposition (MOCVD). Before the high temperature LPCVD SiNx
deposition, samples must be cleaned by standard RCA cleaning process to remove all particles and metal contamination on the sample surface to prevent pollution of the LPCVD chamber. Samples are first immersed in remover 400, isopropanol and deionized water for five minutes respectively to remove the photoresist used for protection during dicing. Remover 400 is usually used for photoresist removal and is composed by 1-methyl-2-pyrrolidone. This is followed by cleaning in SC1 solution
N H4OH(25%):H2O2:H2O=1 : 1 : 5 at 80◦C for 10 minutes to remove organics; then
immersed in SC2 solution HCl : H2O2 : H2O=1 : 1 : 6 at 75◦C for 2 minutes to
strip metallic particles. Subsequently immersed in N H4OH(25%):H2O=1 : 20 for
3. Fabrication Process
3.2
LPCVD
Low-pressure chemical vapor deposition is one of the most common manufacturing process in thin film deposition. LPCVD reactors can be divided into hot wall and cold wall system [47]. Temperature distribution in hot wall system is more uniform than cold wall systems and the convection effects can be reduced in hot wall systems [48]. With the advantages of better uniformity, reduced convection effect and good conformal step coverage [49], hot wall LPCVD is used in this study for depositing
SiNx passivation layer.
Silicon nitride is a dielectric material that has high electrical resistivity, high chem-ical resistance to acids, bases, salts and molten metals, as well as the ability to withstand elevated temperature exposure. Therefore, silicon nitride thin films have been widely used in device processing such as passivation layer, mechanical protec-tive mask, diffusion barrier and gate dielectrics [50, 51, 52]. For instance, they could be used as passivation layer because they are good barriers to water and sodium, and they can serve as masks during selective oxidation process since they oxidize very slowly.
After cleaning, the wafers were placed on quartz boat in the furnace. Dummy silicon wafers were also placed on the quartz boat in order to stimulate a fully loaded boat and present a consistent thermal mass [48, 53]. A 55nm SiNx passivation layer
was grown under temperature of 820◦C, a pressure of 250 mTorr with dichlorosilane (DSC) and ammonia (N H3) flows of 224 sccm and 23 sccm, respectively to supply
the silicon and nitrogen. The reaction is described by the formula: 3SiCl2H2+ 4N H3 → Si3N4+ 6HCl + 6H2
3.3
Mesa
Device isolation is commonly achieved by ion implantation or mesa etching. Mesa etching is often performed with Cl2 based plasma dry etching to define the active
3. Fabrication Process
3.3.1
Photolithography
Photolithography is a technique that transfer a desired pattern to the photoresist on the wafer surface and is extensively used in IC fabrication. In this work, photolithog-raphy was performed using a non-contact laser writer (Heidelberg Instruments DWL 2000). The lithography process includes spin coating, exposure and development. A laser writer is used instead of stepper or mask aligner since maskless lithography process could increase the precision and flexibility, and direct writing system could obtain better control of the photoresist profile through the control of the exposure dose and focus settings compared to the mask lithography, where mask contact introduces uncertainty in dimension control and resist profiles. The irradiance of laser beam can be described by Gaussian distribution, and the focus offset is the deviation of best exposure focus from default lens position. Focus off set is optimized to ensure that the photoresist is positioned at the best focal plane (Fig. 3.2).
Figure 3.2: The schematic of the position of photoresist and the laser beam.
3. Fabrication Process
Mesa isolation and recess ohmic contact are both achieved by ICP-RIE. However, the photoresists used in both process are different. Positive photoresist is used in mesa etching, where the chemical structure of exposed area on the photoresist becomes more soluble and can be removed easily in developer. An image reversal photoresist is applied on the wafer during ohmic contact lithography process. Since ohmic contact is formed at the sidewall between the 2DEG channel and ohmic metals by e-beam evaporation followed by lift-off process [12], the applied photoresist becomes an important factor.
A positive photoresist (S1813) was applied for the definition of the mesa, while a negative, image reversal resist (AZ5214) is used for the ohmic contact definition. The AZ5214 is exposed by laser writer and the exposed area becomes soluble as normal positive photoresist. However, a special crosslinking agent in AZ5214 be-comes active after reversal baking at 125◦C for 1 minutes. The originally exposed area becomes almost insoluble and no longer light sensitive while the unexposed area still remains as normal positive resist. After flood exposure, the originally unexposed area becomes soluble and is developed in spin developer.
In this work, the resist is deposited with a rotation speed of 4000 rpm for 30 s. After spin coating, samples are softbaked at 110◦C for 2 minutes and 1 minute for the S1813 and AZ5214 resist, respectively. Softbake is performed to drive off most of the solvent in the resist since the dissolution rate in the developer can be affected by the solvent concentration in the photoresist. Then the resist coated sample is exposed by laser writer developed in developer MFCD26 for 90 seconds and 351B for 25 s for the S1813 and AZ5214 resists, respectively. Finally, the developed areas are subjected to plasma ashing.
3.3.2
Plasma Ashing
3. Fabrication Process
3.3.3
Plasma Etching
The mesa etching is performed by an inductive coupled plasma reactive ion etching (ICP-RIE) system, which is a common type of high density plasma etching system. ICP-RIE is an anisotropic etching process that can be used to etch a great variety of materials; for instance, semiconductors, metals, dielectrics and polymers. Two separate energy sources independently control the ion density and energy in the system. One is applied to generate bias power to the lower electrode to control the ion energy, while the other applies power to ICP coil and hence can control the ion flux. An oscillating magnetic field is generated by the RF current through the coil. By controlling the induced magnetic field, the plasma density can increase significantly [47]. Different reactive gases are used when etching different materials; for instance, N F3 is commonly used for SiN and Si etching, argon plasma is used as
physical etching gas, and Cl2 and SiCl4 gases are often used in III-V etching.
In our work, N F3 gas was used to etch silicon nitride layer with 50 sccm flow rate.
The heterostructure was etched with Cl2 gas with flow rate of 20 sccm and Ar gas
with 10 sccm flow rate. After mesa etching, an oxygen plasma striping of 100W for 2 minutes was carried out in order to remove etching residues. Then, the samples were cleaned in remover 400, isopropanol and deionized water for five minutes respectively in order to fully remove the photoresist S1813.
3.4
Ohmic Contact
Similar to mesa etching, ohmic recess is also etched by ICP-RIE. Silicon nitride is etched by N F3 with the flow rate of 50 sccm while the heterostructure is etched
by Cl2, SiCl4, and Ar gases with the flow rate of 39 sccm, 1 sccm, and 10 sccm,
respectively. An oxygen plasma ashing is also carried out with the RF power of 40W for 30 seconds to remove etching residues. Subsequently, ohmic metal deposition and lift-off are performed.
3.4.1
Metal Deposition
E-beam evaporation is a form of physical vapor deposition in which the electron beam from charged tungsten filament bombards the target material which is then evaporated to the gaseous state and deposited on the wafer surface. When heating metals in the evaporator, the electron beam commonly sweeps over the target in order to heat the target material with a even heat distribution.
Prior to metal deposition, the wafer is immersed in HF : H2O=1:10 solution for
4 minutes and HCl : H2O=1:10 solution for 60 seconds to strip oxides and other
3. Fabrication Process
sweep mode since the melting point of Al is 660◦ C which is lower than Ta and easier
to evaporate. The second 20 nm Ta layer is deposited to prevent the oxidation of Al. After the metalization completed, lift-off was performed in remover400 at 65◦ C and ultrasonic bath for 5 minutes.
3.4.2
Re-metalization
The re-metalization step starts from the removal of the original ohmic metal stack and the TaN formed during annealing. Tantalum’s chemically inert and is resistant to most chemicals. The most common method used to etch Ta is dry etching with
CF4/O2 gas [68]. Although dry etching is also high selective, coating photoresist
as a mask needs to ensure the alignment during the exposure which makes it more complicated compared to wet etching. Therefore, wet chemical etching is chosen in this work. Wet chemical etching is a process that immerse the sample in a bath of liquid chemicals in order to remove a certain type of material. Wet etching is a simple and low cost etching technique. However, etching with liquid chemicals is usually isotropic and less controllable and hence the reproducibility is low compared to dry etching. Since the aim is to etch Ta/TaN, while other layers such as the herterostructure AlGaN/AlN, the buffer layer GaN and the substrate SiC should be remain unaffected.
As listed in table (Fig. 3.3), the heated solution of KOH/NaOH with water can etch most of the material including the AlGaN/GaN heterostructure, making it an unsuitable etchant in this investigation. In this work, we choose HF (5%) : HN O3 :
H2O(20%) =1 : 2 : 1 to selectively etch the ohmic contact at room temperature.
This solution is suggested in [59] without stating the etching rate. We repeated the Ta etching for several times and obtained two methods. The first way is immersing the sample in Ta etchant for over 50 minutes. However, it is too time-consuming to perform. Since the ohmic metal stack formation was Ta/Al/Ta, another way is using standard Al etchant composed of H3P O4 : CH3COOH : HN O3= 60% : 3.5% :
3. Fabrication Process
Figure 3.3: The etchant versus etching target material (X: The target material can not be etched by the etchant, : The target material can be etched by the
etchant, –: Not found from the literature).
3.4.3
Sputtered Ta
Sputtering is a type of physical vapor deposition techniques which is based on ion beam bombardment. Sputtering is a commonly used method of thin film deposition, which can be divided into several types such as DC sputtering, magnetron sputtering, and RF sputtering etc. In the DC sputtering system, the target material is placed on the cathode while the substrate is placed on anode. Usually, the sputtering gas filled in the chamber is inert gas (typically Ar). The applied DC voltage between electrodes maintains the glow charge in which the Ar+ is generated. The Ar+ ions
are accelerated at the cathode and thus sputter the target material on the cathode resulting the thin film deposition of the target material on the substrate.
4
Characterization
In this work, two main characterization techniques are used. The first one is trans-mission line method (TLM), which is extensively used to examine the ohmic contact behavior by characterizing the contact resistance and sheet resistance. The other is the scanning electron microscopy (SEM), which is used to check the cross section of ohmic contact. Both methods are introduced in this chapter.
4.1
Scanning electron microscopy
Scanning electron microscope (SEM) is a widely used instrument to examine surface characteristics. Instead of light as the illumination source, a focused electron beam is emitted from the electron gun and focused on the sample surface. A SEM image is formed by collecting the scattered electrons. The electro-optical path of the SEM contains several electromagnetic lenses and deflection coils to control the diameter and astigmatism of the electron beam. Due to the interaction between the elec-tron beam and the sample surface, secondary elecelec-trons (SE) can be generated from the inelastic scattering, while backscattered electrons (BSE) can be produced from elastic scattering which often retain 60 to 80% of the incident electrons’ energy. SEs usually have the energy of several electron volts, and SEs can only escape from the volume near the sample surface of the interaction zone. However, BSEs have higher energy and can escape from deeper level of the interaction zone compared to SEs. Consequently, images generated by SEs will have higher resolution of surface topography while images produced by BSEs demonstrate clear compositional con-trast if the sample contains more than one chemical elements. The compositional contrast depends on the backscatter coefficient, η, which increases with the atomic numbers of the chemical elements in the sample.
η = nBSE ni
(4.1)
η represents the ratio of the number of the BSEs escaping from the sample nBSE
to the number of incident electrons ni [69]. Thus, the sample that contains higher
4. Characterization
the cross-sections of the samples in backscattered electrons mode to identify the epitaxial layers and metal stacks.
4.2
Transmission line method
Numerous of techniques can be used to obtain contact resistance, the transmission line method (TLM) is the most common and popular method. It was originally pro-posed by Shockley [70] and was further studied and extended by Berger [71], which provided a convenient and simple method to determine resistive performance of the ohmic contacts, including the contact resistance (Rc), specific contact resistance,
(ρc), sheet resistance (Rsh) and transfer length.
4.2.1
TLM structure
The ladder structure consisting of more than three rectangular metal contacts, which are identical in the area, are used as the test structure for TLM (Fig. 4.1). These metal pads are separated with adjacent contact pads with incrementing spacing dis-tances (ranging from 5 to 30 µm (d1 to d5) in this work). The resistance between
every contact pads are measured with four probes. The outer two probes give the current and the inner two probes measure the voltage difference, and the resistance values are given. Four-probe measurement is utilized instead of two-probe mea-surements since the current and voltage are separated which eliminate the lead and contact resistance from the measurement [72].
(a) (b)
4. Characterization
4.2.2
Epi-layer sheet resistance R
shThe resistance of a homogeneous semiconductor can be expressed by the sheet re-sistance if the semiconductor is homogeneous rectangular. The rere-sistance and sheet resistance is expressed as:
R = ρ L
tW (4.2)
Rsh =
ρ
t (4.3)
where ρ, L, t, and W represents the resistivity, length, thickness, and width of the semiconductor, respectively. Introducing the Rsh enables to consider resistors
as numbers of squares since the resistance can be simply obtained by multiplying by Rsh. However, it is hard to consider semiconductor as a homogeneous material
since the semiconductor is usually a diffusion or epitaxially grown layer. Hence, an average resistivity ¯ρ can be calculated as a first approximation by considering the
doping profile of the layers.
4.2.3
Contact resistance R
cand specific contact resistivity
ρ
cThe plot of Rtot as a function of dx is shown in Fig. 4.2b. Rtot is the sum of
several components, in which Rm is the resistance of contact metal, Rsemi is
resis-tance due to the semiconductor, and Rcis the contact resistance between metal and
semiconductor interface.
(a) (b)
Figure 4.2: (a)A schematics shows the different components of Rtot. (b) Total
resistance Rtot plotted as a function of isolation distance dx
From the Fig. 4.2b, Rc and Rtot can be obtained. An extrapolation at the point
of zero distance d=0 represents 2Rc. In this work, Rc is scaled with the width of
the contact, and hence has the unit of Ωmm. Moreover, Rtot can be expressed by
4. Characterization
Rtot = 2Rm+ 2Rc+ Rsemi (4.4)
In general, the contact metal resistance Rm is usually much smaller than contact
resistance Rc, and hence can be ignored. The slope of the fitting line (Fig. 4.2b)
represents the sheet resistance Rsh. Consequently, Rtotcan be expressed by Equation
4.7. Rc>> Rm (4.5) Rsemi = Rsh W d (4.6) Rtot = 2Rc+ Rsh W d (4.7)
The transfer length is defined as where the voltage, caused by current transferring between semiconductor and metal contacts, drops to 1/e of its maximum value, which is usually at the contact edge. The transfer length is dependent on specific contact resistance and sheet resistance, and the relation can be expressed as:
LT =
s
ρc
Rsh
(4.8) If L ≥ 1.5LT, the effective contact area, Aef f, can be treated as LTW . Due to the
current crowding effect, LT can be much smaller than L, and consequently, Aef f can
also be much smaller than the contact metal area. Hence, this means that in the close proximity of the contacts, the current density would be higher than expected and therefore needs to be considered.
Specific contact resistivity,ρc, is usually measured in Ωmm2 which independent of
contact area or geometry. It can be defined as in Equation 4.9:
ρc = ∂J ∂V !−1 v=0 (4.9) The specific contact resistance contains the contact resistivity of both the interface and the regions above and below it. ρccan also be determined directly by the contact
resistance Rc, transfer length LT and the width of contact pads. It can be expressed
as:
ρc= RcLTW = RcAef f (4.10)
Aef f in the equation is the effective contact area. The contact area is always different
from the Aef f due to the current crowding effect which is presented by Kennedy et al
[73]. The current going through the semiconductor is uniform; however, the current flow is not uniform through the metal contacts. Since the current flows through the metal contacts with low resistance, and therefore, the current density at the edge of the contacts are higher and will drops to zero at the far edge. For small ρc, only
4. Characterization
expanded for high ρcbecause of the high transition resistance which is shown in Fig.
4.3.
(a)High ρc
]
(b) Low ρc
5
Results
The main results of this project will be presented in this chapter and can be divided into three sections, including laser writer intensity and focus offset optimization, re-metallization ohmic contact, and initial results of ohmic contacts fabricated by sputter deposition.
5.1
Laser writer focus/intensity test
The optimal laser writer focus offset and dose enable to obtain better control of the photoresist, laser writer focus offset was from 5% to 35%, dose intensity from 30% to 60% and the transmittance keeps constant at 25%. The focus/transmittance/in-tensity matrix settings and the equivalent energy are shown in Fig. 5.1.
Figure 5.1: Illustration of the laser focus/transmittance/intensity matrix with the exact energy received on the substrate.
5. Results
Figure 5.2: The contour plot of the obtained Rcfrom the laser focus/dose matrix.
Lin et al reported that the sidewall angle can be controlled by reversal baking temperature and laser writer exposure intensity [12]. For steeper sidewall angle, Rc
increases due to the smaller contact area. However, Rc also increases with smaller
sidewall angle since the increase of depleted GaN leads to decreasing AlGaN and hence, hinder the formation of 2DEG [40] (Fig. 5.3). Moreover, photoresist profile is an important factor to determine the sidewall angle. Consequently, the optimal exposure intensity is studied in this work to control the photoresist profile through varying the laser writer focus offset and dose while the transmittance is constant. It can be seen from the plot that the lowest Rc obtained is around 0.23 Ωmm and
from Fig. 5.1, the corresponding laser energy received on the substrate is 9 mJ/cm2.
This optimal laser writer focus/transmittance/intensity is applied in the subsequent parameter testing.
5. Results
5.2
Ohmic contact re-metallization
The metallization process includes etching away the ohmic metal stack and re-depositing the metal again. The re-deposition metal stack annealed at 600◦C was
characterized by TLM. Rc and Rsh before and after re-metallization process are
plotted in Fig. 5.4.
(a)Rc versus annealing time (b) Rsh versus annealing time
Figure 5.4: (a)Rc comparison of original and re-deposited ohmic metal stack.
(b)The plot of Rc versus annealing time and the comparison of original and
re-deposited ohmic metal stack.
The lowest Rc of the re-deposited metal contacts obtained in Fig. 5.4a is 0.5 Ωmm
while the lowest Rc of original metal contacts is 0.28 Ωmm. It is clearly shown
that the Rc increases after re-metallization process. The Rsh versus annealing time
(Fig. 5.4b) shows that the sheet resistance keeps similar values, which is around 267 Ω/, after the etching followed by metal re-deposition process. The similar Rsh
indicates that the epi-layer is not damaged during the wet etching step and hence, the increase in Rc is not caused by epi-layer damage.
The sample was originally deposited with the metal stack Ta/Al/Ta with thickness of 15/280/20 nm respectively, and annealed at 600◦C. Due to the increase in Rc
and unaffected Rsh, we assumed that the cause of the increasing Rc might be the
5. Results
(a)5 nm (b) 10 nm
Figure 5.5: The plot of Rc versus annealing time while first Ta layer thickness
differs.
Fig. 5.5 is the plot of Rc and Rsh versus annealing time for the Ta thickness of 5
and 10 nm respectively. Compared to the results of 15 nm Ta in Fig. 5.4a, value of the lowest Rc for 10 nm Ta decreases from 0.5 Ωmm to 0.41 Ωmm while the
lowest Rcfor 5 nm Ta increases from 0.5 Ωmm to 0.67 Ωmm. This might be related
to the larger mask we applied during the re-metallization process. During the first metallization step, the applied mask was smaller, and hence the samples had to be evaporated with a tilt angle. However, the mask was enlarged in the re-metallization step to ensure the correct alignment. Fig. 5.6a illustrates the original mask in the first metallization process and the larger mask in re-metallization process and the cross section after re-metallization is shown in Fig. 5.6b.
(a)The schematic of the larger pho-toresist .
(b) The cross section of first and re-metallization.
5. Results
Because of the smaller mask and the tilt angle during the first metal evaporation, the sidewall Ta thickness is thinner than the bottom Ta thickness which is expected to be 15 nm (Fig. 5.7). The exact sidewall thickness measured with SEM is shown to be 13.95 nm. However, with the larger mask used in the re-metallization step, the re-metallization sidewall Ta thickness is thicker than the sidewall thickness in the first metallization step. Consequently, the expected re-deposited sidewall Ta thickness for 10 nm is close to the original deposited sidewall Ta thickness while 5 nm Ta is too thin for the contacts. This is the possible explanation for the decreased
Rc for 10 nm and the increased Rc for 5 nm.
5. Results
5.3
Sputtered Ta
To improve the sidewall metal coverage, the first Ta layer were deposited by sput-tering for three different thickness (10, 15, 20 nm) with the following Al/Ta layer (280/20 nm) deposited by e-beam evaporation. The contact resistance was also been characterized and is shown in Fig. 5.8a.
(a)Rc (b) Rsh
Figure 5.8: The obtained (a)Rc and (b)Rshversus annealing time with different
metal thickness.
Compared to the lowest Rcof evaporated metal stack (0.23 Ωmm), the lowest Rcof
sputtered Ta was 0.6 Ωmm with the Ta thickness of 10 nm and annealing at 575◦C for 64 minutes. However, the sample with thickness of 15 nm did not perform as ohmic contacts even after annealing for 32 minutes while the lowest Rc obtained
at the annealing time of 16 minutes was 6.65 Ωmm. Similarly, the lowest Rc of
the sample with Ta thickness of 20 nm was obtained at the annealing time of 16 minutes, which is 1.72 Ωmm. The Rc started to rise after 16 minutes annealing for
the samples with 15 and 20 nm Ta. The sheet resistance of 10 nm Ta sample is similar to the evaporated sample which is around 280 Ω/. However, Rsh of 15 nm
sample is around 215 Ω/ which slightly lower than the evaporated sample while
Rsh of 15 nm is much lower than other samples, seen in Fig. 5.8b.
There are two assumptions that we made for the observed high Rc of all sputtered
samples and the slightly lower Rsh of 20 nm Ta. The first assumption is that since
the sputtered thin film is denser than evaporated thin film due to the incident momentum [74], the sputtered thin film should be thinner than the evaporated thin film. Consequently, the sample with 10 nm sputtered Ta obtained the lowest
Rc obtained the lowest Rc while the optimized thickness of evaporation is 15 nm.
Another assumption was made due to the lower Rsh of 20 nm Ta. The high energy
5. Results
6
Conclusion and future work
The aim of this thesis is to optimize different aspects of ohmic contact formation. There are three types of ohmic contacts, planar contacts, regrowth n+-regrowth
contacts and recess etched contacts. In this thesis, recess contacts are used and in order to avoid the depth control problem, the recess was etched beyond the 2DEG. The ohmic contact optimization in this work includes laser writer focus/-transmittance/intensity test, re-metallization as well as electron beam evaporation and sputtering were also investigated.
The optimal laser writer focus/transmittance/intensity test was tested on a sample, which composed of a focus/transmittance/intensity matrix. The lowest Rcobtained
in this work is 0.23 Ωmm and the corresponding optimal laser focus/intensity is 25%/35% while transmittance keeps constant at 25% i.e. the laser power received on the substrate is 9 mJ/cm2.
One of the widely accepted explanations of ohmic contact mechanism is that the formation of TiN extracts the nitrogen atoms from AlGaN layer, and hence the N-vacancies are created and act as n-dopants, which makes the barrier layer heavily doped. In this work, we assumed that the formation of TaN between Ta and GaN would block the extraction mechanism of nitrogen atoms. Therefore, wet chemical etching was performed to remove the original ohmic metal stack and the formed TaN. Then the re-deposition of ohmic metal stack was carried out in order to extract more nitrogen atoms from GaN i.e. creating more N-vacancies in barrier layer.
The Rsh obtained from the characterization of re-metallized samples indicates that
the epi-layer was not damaged during the wet etching process. However, com-paring the lowest achieved Rc between original metallization (0.28 Ωmm) and
re-metallization (0.41 Ωmm), it can be concluded that a lowering of the contact resis-tance was not achieved by re-metallization described in this report.
Different sputtered Ta thickness (10/15/20 nm) was investigated in this project. The lowest contact resistance of the sputtered Ta was 0.6 Ωmm when the Ta thickness was 10 nm, annealed for 64 minutes at 575◦C. Although the samples with 20 nm Ta showed ohmic behavior, the Rsh was about 65 Ω/ lower than the evaporated
Bibliography
[1] Bright, A., Thomas, P., Weyland, M., Tricker, D., Humphreys, C.,Davies, R.: Correlation of contact resistance with microstructure for Au/Ni/al/Ti/AlGaN/-GaN ohmic contacts using transmission electron microscopy. Journal of Applied Physics89(6) (2001) 3143–3150
[2] Ambacher, O., Foutz, B., Smart, J., Shealy, J., Weimann, N., Chu, K.,Murphy, M., Sierakowski, A., Schaff, W., Eastman, L., et al.: Two dimensional electron gases induced by spontaneous and piezoelectric polarization in undoped and doped AlGaN/GaN heterostructures. Journal of applied physics87(1) (2000) 334–344
[3] Jacobs, B., Kramer, M., Geluk, E., Karouta, F.: Optimisation of the Ti/Al/Ni/Au ohmic contact on AlGaN/GaN FET structures. Journal of Crys-tal Growth241(1-2) (2002) 15–18
[4] Mahajan, S.S., Dhaul, A., Laishram, R., Kapoor, S., Vinayak, S., Seh-gal, B.: Micro-structural evaluation of Ti/Al/Ni/Au ohmic contacts with different Ti/Al thicknesses in AlGaN/GaN HEMTs. Materials Science and Engineering: B183(2014) 47–53
[5] Xin, H., Poust, S., Sutton, W., Li, D., Lam, D., Smorchkova, I., Sandhu,R., Heying, B., Uyeda, J., Barsky, M., et al.: Optimization of AlGaN/GaN HEMT ohmic contacts for improved surface morphology with low contact resistance. In: CS MANTECH Conf. (2010) 149–152
[6] Liu, Q., Yu, L., Deng, F., Lau, S., Chen, Q., Yang, J., Khan, M.: Study of con-tact formation in AlGaN/GaN heterostructures. Applied physics letters71(12) (1997) 1658–1660
[7] Lee, J., Yan, M., Ofuonye, B., Jang, J., Gao, X., Guo, S., Adesida, I.:Low resis-tance Mo/Al/Mo/Au ohmic contact scheme to InAlN/AlN/GaN heterostruc-ture. physica status solidi (a)208(7) (2011) 1538–1540
[8] Qiao, D., Jia, L., Yu, L., Asbeck, P., Lau, S., Lim, S.H., Liliental-Weber,Z., Haynes, T., Barner, J.: Ta-based interface ohmic contacts to AlGaN/GaN heterostructures. Journal of Applied Physics89(10) (2001) 5543–5546
[9] Kim, K.H., Jeon, C.M., Oh, S.H., Lee, J. L., Park, C.G., Lee, J.H., Lee,K.S., Koo, Y.M.: Investigation of Ta/Ti/Al/Ni/Au ohmic contact to AlGan/GaN heterostructure field-effect transistor. Journal of Vacuum Science & Technology B: Microelectronics and Nanometer Structures Processing, Measurement, and Phenomena23(1) (2005) 322–326
Bibliography
[11] Malmros, A., Blanck, H., Rorsman, N.: Electrical properties, microstructure, and thermal stability of ta-based ohmic contacts annealed at low temperature for GaN HEMTs. Semiconductor Science and Technology26(7)(2011) 075006 [12] Lin, Y.K., Bergsten, J., Leong, H., Malmros, A., Chen, J.T., Chen,
D.Y.,Kordina, O., Zirath, H., Chang, E.Y., Rorsman, N.: A versatile low-resistance ohmic contact process with ohmic recess and low-temperature an-nealing for GaN HEMTs. Semiconductor Science and Technology33(9)(2018) 095019
[13] Zhang, J., Wang, L., Wang, Q., Jiang, Y., Li, L., Zhu, H., Ao, J.P.: Plasma-assisted ohmic contact for AlGaN/GaN heterostructure field-effect transistors. Semiconductor Science and Technology31(3) (2016) 035015
[14] Wang, L., Kim, D.H., Adesida, I.: Direct contact mechanism of ohmic metal-lization to AlGaN/GaN heterostructures via ohmic area recess etching.Applied Physics Letters95(17) (2009) 172107
[15] Yue, Y., Hu, Z., Guo, J., Sensale-Rodriguez, B., Li, G., Wang, R., Faria,F., Fang, T., Song, B., Gao, X., et al.: InAlN/AlN/GaN HEMTs with regrown ohmic contacts and fTof 370 GHz. IEEE Electron Device Letters33(7)(2012) 988–990
[16] Guo, J., Cao, Y., Lian, C., Zimmermann, T., Li, G., Verma, J., Gao, X.,Guo, S., Saunier, P., Wistey, M., et al.: Metal-face InAlN/AlN/GaN high electron mobility transistors with regrown ohmic contacts by molecular beam epitaxy. physica status solidi (a)208(7) (2011) 1617–1619
[17] Joglekar, S., Azize, M., Beeler, M., Monroy, E., Palacios, T.: Impactof recess etching and surface treatments on ohmic contacts regrown bymolecular-beam epitaxy for algan/gan high electron mobility transistors.Applied Physics Let-ters109(4) (2016) 041602
[18] Grover, Sidhant Effect of transmission line measurement (TLM) geometry on specific contact resistivity determination. Master’s thesis, Rochester Institute of Technology (2016)
[19] Tung, R.T.: Recent advances in schottky barrier concepts. Materials Science and Engineering: R: Reports35(1-3) (2001) 1–138
[20] C.A.J. Ammerlaan, A. Chantre, P.W.: Science and Technology of Defectsin Silicon. Volume 9. Elsevier Science Publisher B.V. (2014)
[21] Lyubomirsky, I., Lyahovitskaya, V., Cahen, D.: Dopant accumulation dur-ing substitutional–interstitial diffusion in semiconductors. Applied physics let-ters70(5) (1997) 613–615
[22] Nikitina, I., Vassilevski, K., Wright, N., Horsfall, A., ONeill, A., Johnson,C.: Formation and role of graphite and nickel silicide in nickel based ohmic contacts to n-type silicon carbide. Journal of Applied Physics97(8) (2005)083709
[23] Kuchuk, A., Kladko, V., Guziewicz, M., Piotrowska, A., Minikayev, R.,Stonert, A., Ratajczak, R.: Fabrication and characterization of nickel silicide ohmic contacts to n-type 4H silicon carbide. In: Journal of Physics:Conference Series. Volume 100., IOP Publishing (2008) 042003