Ageing and thermal recovery of advanced SiGe heterojunction bipolar transistors under long-term mixed-mode and reverse stress conditions
Introduction
The ongoing evolution toward millimeter-wave circuits with still higher operation frequencies continuously forces SiGe HBT processes to prove their competitiveness against advanced Si CMOS and III–V compound semiconductor technologies. Yet, besides excellent radio-frequency (RF) performance at circuit operation frequencies, meanwhile well beyond 100 GHz [1], customer focus is now also put on operating constraints (“safe operating area”, SOA) and long-term reliability of the devices. Life-times of 10 years or in the case of space applications even 20 years are mandatory. Therefore, the focus of this predominately experimental study is to evaluate the reliability of advanced SiGe HBTs not just under the restricted operational conditions of the SOA but also to study the devices up to their limits. Exceeding these will lead to almost immediate HBT destruction (“electro-thermal run-away”) [2].
If the HBT power load line in a circuit stays within the SOA limits but collector voltage already surpasses open-base breakdown voltage (BVCEO) or the emitter-base (EB) diode is reverse biased a gradual decrease of the HBT current gain β (“ageing”) can be observed. This beta degradation is directly correlated to the increase of the base recombination current caused by creation of recombination centers (“traps”) at dielectric interfaces. These again are being created by hot carrier injection (“HCI”) making the mechanism comparable to the well-known HCI degradation in CMOS devices.
Most studies of SiGe HBT ageing are based on short-term (up to a few hours) application of forward “mixed-mode” (MM) [3] and reverse (REV) stress conditions mainly evaluating the impact of stress-current and stress-voltage on base current degradation. The time evolution of ageing is then assumed to obey a power function law with constant power coefficient. This corresponds to a constant ageing rate, i.e. change of base current degradation per time interval. However, by applying multiple long-term (up to 1000 h) forward MM [3] and reverse stress conditions on two different HBT types we will show that the ageing rate saturates over time which is caused by the complex dynamics of trap generation and annihilation. Additionally, ageing is also studied as function of stress power load and ambient temperature for both MM and REV stress. Concerning the dependence on power load we will show that at later stages base current becomes less dependent on stress-current, an effect which has not been described previously. In the latter case we will separate in situ base current degradation from its resulting effect back at room temperature. This allows to distinguish temperature dependence of HCI and trap generation from thermal shift of the base current and to show for the first time that reverse and mixed-match stress display different thermal behavior.
These results are summarized into an ageing function valid under ambient temperatures Tamb in the range from −40 °C up to +150 °C and during the full HBT life-time period. The main target here is to present a semi-empirical ageing function in a form easily accessible for introduction into HBT compact models (e.g. Gummel–Poon or HICUM) providing circuit designers a practical means to estimate the impact of base current degradation on the performance of their particular circuit.
This is especially important since it offers a valuable alternative to usual definitions of transistor “life-time” based for example on definitions like “10% DC current gain reduction at peak-fT base-emitter voltage”. For example, additional S parameter measurements demonstrate that even more than 60% DC current gain degradation had no measurable impact on capacitances and RF performance above 5 GHz, as well as transit frequencies did not change. Additionally, to evaluate in general whether the results obtained by DC stress tests can be applied to HBT load lines in RF circuit operation simple duty cycles were performed. These demonstrate that the DC MM stress tests represent an upper limit on HBT degradation under RF operation.
Further studies include investigations on high temperature anneal which is an effective means for trap annihilation (“thermal recovery”). It will be shown that forcing high emitter current densities at 150 °C results in a junction temperature around 300 °C but will restore current gain to almost initial values. Most of the recovery occurs within 1 h independent of how much the HBT had been degraded under previous stress. This defines an upper time limit for intentional SiGe HBT recovery.
Section snippets
HBT technology and pre-stress characterization
The HBTs under investigation originate from IHP’s 0.13 μm BiCMOS technology SG13S [4] which offers two HBT types: high-speed (HS) HBTs with peak fT = 250 GHz at a corresponding collector current density JC ≈ 18 mA/μm2 and high-voltage (HV) HBTs with fT = 45 GHz at JC ≈ 0.9 mA/μm2. Emitter areas AE, transit frequencies fT, fmax, and target breakdown voltages of the HBTs under investigation are put together in Table 1.
Before applying forward MM and reverse stress bias the operating constraints of the HBT have
Ageing mechanism and description
The general mechanism leading to the observed base current increase or degradation, respectively, is well-known and can be described as follows [9]: traps created at the EB spacer oxide interface cause an increase in the non-ideal forward-mode base current due to Shockley–Read–Hall (SRH) recombination. These traps are supposed to be mainly so-called PB centers, i.e. recombination centers associated with silicon dangling bonds at the oxide interface. Although the majority of PB centers are
Stress tests
The usual procedure for HBT stress tests is to measure Gummel characteristics (VCB = 0 V) of the HBT in the state before stress, then to start applying the stress conditions {VCB,stress, JE,stress, Tamb} (MM) or {VEB,stress, Tamb} (REV) for certain time intervals Δtstress, and to measure further Gummel characteristics after each particular Δtstress. During stress load the HBT degrades by increase of the base current IB. ΔIB was used to monitor HBT ageing. Since a change of the collector current
Analysis of HBT ageing
In the following section HBT reliability under the influence of a wide range electrical stress load, ambient temperatures from −40 °C up to +125 °C, and in time periods from a few seconds up to 1000 h will be discussed. Yet, before concentrating on the change of the DC parameter IB it had to be clarified whether any RF HBT parameters (capacitances, transit frequencies) are also modified by the applied stress conditions.
Thermal recovery
If thermal load is sufficiently high trap annihilation may dominate generation and base current degradation can be reversed into a decreasing base current. This thermal recovery was already visible for the HV HBT in Fig. 9b where a stress load {14 V, −9 mA/μm2, 27 °C} causes a junction temperature of almost 200 °C. As consequence after 3 h MM stress IB became almost 10% less than in its initial state showing the anneal of pre-existing traps.
Now Fig. 14, Fig. 15 show in situ base current recovery
Duty cycle considerations
All previous MM stress tests are based on DC electrical loads {VCB,stress, JE,stress} which do not resemble the actual transient load on the HBT during RF circuit operation. In typical RF applications the HS HBTs are for example implemented in noise amplifier (LNA) and output buffer of 120 GHz transceivers [1] or in 60 GHz power amplifiers [32]. As sketched in the inset of Fig. 16 during operation their load line shifts between extreme values of high current density and low CB voltage (“ANN”) and
Conclusions
The experimental analysis of base current degradation under electrical stress within and beyond the conditions applied during RF circuit operation showed that ageing is a more complex function of electrical load, local temperature, and particularly of stress-time than expected. The assumption of a constant time power coefficient α = 0.5 holds only for a short time but in the long-term α exponentially decreases toward αlt ≈ 0.2 or even lower. In a similar way the emitter current coefficient is also
Acknowledgements
The authors wish to acknowledge the DOTSEVEN (316755) Project supported by the European Commission through the Seventh Framework Program (FP7) for Research and Technology Development.
References (32)
- et al.
IEEE J Solid-State Electron
(1975) - Sun Y, Marinkovic M, Fischer G, Winkler W, et al. In: Proceedings IEEE solid-state circuit conf (ISSCC), 2013. p....
- Vanhoucke T, Hurkx GAM. In: Proceedings IEEE bipolar circuits and technology meeting (BCTM) 2005. p....
- Zhu C, Liang Q, Al-Huq R, Cressler JD, et al. In: Proceedings IEEE international electron and device meeting (IEDM)...
- et al.
IEEE J Solid-State Circuits
(2010) - Pfost M, Brenner P, Lachner R. In: Proceedings IEEE bipolar circuits and technology meeting (BCTM), 2004. p....
- et al.
Compact hierarchical bipolar transistor modeling with HICUM
(2010) - et al.
IEEE Trans Electron Dev
(1996) Fundamentals of Solid-State Electronics
(1991)- et al.
IEEE Trans Electron Dev
(2012)
Considerations for the reliability estimation of silicon CMOS
J Appl Phys
Microelectron Reliab
Cited by (26)
A unified aging compact model for hot carrier degradation under mixed-mode and reverse E-B stress in complementary SiGe HBTs
2020, Solid-State ElectronicsCitation Excerpt :Investigation of degradation physics in complementary devices has also been studied using TCAD simulation [6]. From the circuit designers’ point of view, the major degradation mechanism that needs to be accurately incorporated in SPICE compact models is hot-carrier degradation (HCD) that predominantly limits the lifetimes of modern SiGe HBTs [3,5–10]. The principal physical degradation mechanism is the generation of hot carriers, through impact ionization under high field or current conditions at the collector base junction, which are accelerated toward the emitter–base (EB) spacer oxide interface where they create traps via Si-H bond-breaking.
Scalable compact modeling of trap generation near the EB spacer oxide interface in SiGe HBTs
2020, Solid-State ElectronicsA physical and versatile aging compact model for hot carrier degradation in SiGe HBTs under dynamic operating conditions
2020, Solid-State ElectronicsCitation Excerpt :The rest of this paper is organized as follows; Section II illustrates the model formulation and implementation; Section III describes the three SiGe HBT technologies under test and their respective stress bias conditions, the aging test results, model validation and a discussion on the evolution of aging model parameters with bias followed by the conclusion. The schematic in Fig. 2 illustrates the spatial dynamics of hydrogen diffusion in the emitter-base spacer oxide due to the hot-carrier-induced degradation in SiGe HBTs and the subsequent Si-H bond breakage at the [2–8,14] oxide interface. While the new implementation ensures model validity under dynamic stress (Fig. 1), in this section, we test the versatility of the model under various aging conditions, close to the SOA of the transistors, through comparison between the aging model simulation and results of aging tests on three different advanced SiGe HBT technologies.
Reliability evaluation of a 0.25 μm SiGe technology for space applications
2019, Microelectronics ReliabilityCitation Excerpt :The objective of this work is to activate and understand the main degradation mechanisms in elementary circuit cells in order to establish a safe domain for the use of the technology in space applications. It is indeed a question of approaching the problems of reliability by analysing the process known data [2] and by testing simple circuits. This reliability preliminary evaluation is in part based on existing space standards which are used on III-V technology (GaAs, GaN).
PdGe contact fabrication on Se-doped Ge
2017, Scripta Materialia