Analysis of Tunable BV CEO in Horizontal Current Bipolar Transistor with Floating Field Plates

Transcription

1 Analysis of Tunable BO in Horizontal Current Bipolar Transistor with Floating Field Plates Marko Koričić *, Josip Žilak *, Željko Osrečki * and Tomislav Suligoj * * University of Zagreb, Faculty of Electrical Engineering and Computing, Department of Electronics, Microelectronics, Computing and Intelligent Systems, Micro and Nano Electronics Laboratory, Zagreb, Croatia marko.koricic@fer.hr Abstract - Open base breakdown voltage (BO ) tuning in Double-Emitter Horizontal Current Bipolar Transistor with floating field plates (FFPs) is analyzed by the device simulations. Tuning is obtained by stepping of the FFPs potential which is controlled via capacitive coupling between neighboring FFPs. When coupled, they serve as the field plates for shaping of the electric potential distribution, introducing the additional drift regions in base-collector depletion region. Breakdown mechanisms in case of partial and complete shielding of the collector E-field by the FFPs are identified. In case of partial shielding, the inner electric field peaks are responsible for breakdown, whereas in the case of complete shielding, breakdown occurs at the outermost drift region. In completely shielded, device very high BO can be obtained by stacking a large number of FFPs. Devices with BO up to 7 V are obtained. I. INTRODUCTION In bipolar transistors, the open-base breakdown voltage (BO ) is usually to times lower than the open-emitter breakdown voltage (BV CBO ) [] and in general represents a tighter constraint on the safe operation limits. High voltage transistors can be used as components that can extend the application of the technology, which is especially attractive if they can be added to standard processes without significant costs. Most attractive applications would be in circuits where the higher voltage swing is desirable, such as in input-output (IO) circuits, power amplifiers and linear analog circuits. Horizontal current bipolar transistor (HCBT) is integrated with 8 nm CMOS [] in a simple process resulting in a low-cost BiCMOS technology suitable for the fabrication of RF communication circuits in the frequency range up to GHz []. The technology is further improved by the addition of high-voltage devices, which are integrated at zero-cost. The improvement of the breakdown voltages is obtained by the geometry manipulation, which is accomplished by the lithography mask design and the application of existing process steps. The principle of the breakdown voltage improvement is based on reduced-surface-field (RESURF) effect [], which is used to shape the potential and electric field distribution in the base-collector depletion region. Highvoltage (HV) HCBT structures with BO of. V [],. V [] and V [7] are demonstrated. This work was supported by the Croatian Science Foundation under contract no. 9. The latest reported HV HCBT structure is doubleemitter (DE) HCBT with floating-field-plate (FFP) with BO = V [8]. Floating field regions are used for a junction termination in power devices [9]. In [8] they were used to improve the breakdown voltage of highperformance bipolar transistor structure suitable for RF applications. The tuning of the BO with multiple FFPs was demonstrated by simulations. In this paper, the breakdown mechanisms and the tuning principle are thoroughly analyzed. Practical limits of tuning and its impact on electrical performance is investigated. II. SIMULATION STRUCTURE AND PRINCIPLE OF OPERATION D model of the quarter of the DE HCBT with FFP is shown in Fig. [8]. FFP is implanted in the form of the p+ region, together with the extrinsic base and therefore those two regions are self-aligned. The full structure is obtained by mirroring the structure along the CB and BE planes in Fig.. Since the breakdown physics is determined by the shaping of the potential distribution in CB-plane, a D simulation structure similar to the one reported in [8] is used. The breakdown of standard DE HCBT [] is determined by the peak electric field in the drift region which is formed below the extrinsic base marked as DR in Fig.. The FFP is floating on the potential which is set by the collector potential and the Fig.. D model of DE HCBT with FFP [8]. 7 MIPRO 8/MEET

2 Concentration, (cm - ) Depth, (µm) Acceptor Concentration Donor Concentration Fig.. Doping profile below FFP and the extrinsic base. junction built-in potential since the current across the junction is zero. As the collector voltage increases, the depletion region spreads from the extrinsic base laterally in the direction of the FFP and the extrinsic collector. At the certain voltage, the part of the collector along the distance between the extrinsic base and FFP (d fp ) is fully depleted (i.e., it is in punchthrough). At that point the FFP is coupled to the extrinsic base via depletion capacitance and its potential is set by the extrinsic base potential and the voltage drop across the region marked by d fp. Once the FFP potential is pinned, the junction between the FFP and the n-collector is reverse polarized. Hence, the part of the collector below the FFP can be fully depleted from the FFP side forming the additional drift region (DR in Fig. ). The FFP then acts as a field plate shielding the lateral electric field. In this way, the maximum electric field in the DR is limited and the breakdown voltage is improved. By adding multiple FFPs this principle can be extended and BO tuned in discrete steps as shown in [8]. III. DEVICE SIMULATIONS Since the potential of the FFP is determined by the full depletion of the collector between the extrinsic base and the FFP, the distance between them d fp is the important geometrical parameter. d fp controls the total amount of donor charge in that region, which in turn sets the voltage drop across that region. Vertical doping profile below the extrinsic base (and FFP) is shown in Fig. and it is similar to the one reported in [8] with slightly higher collector concentration. Acceptor concentration in the extrinsic base and the FFP is cm - and donor concentration in the top part of the collector is. 7 cm -. Due to the lower doping, the depletion region of the extrinsic base-collector junction dominantly spreads on the collector side. The depletion region width can be approximated by step-graded junction with uniform doping concentrations: ε d = + V V q N A ND ( ) B bi R, () Voltage V R, (V) Junction to junction - d fpj Mask to mask - d fp Distance, (nm) Fig.. Calculated punchthrough voltage dependency on the distance between extrinsic base and FFP. V R is voltage between the anode and the chatode of the p + n- junction. where ε is dielectric constant of silicon, q is electron charge, N A and N D are doping concentrations in the extrinsic base and in the collector, respectively, V bi is built-in potential and V R is voltage applied at the junction. With N A >> N D, () reduces to: ε VR db ( Vbi VR ) = db, () qn V D where d B is the depletion region width at the equilibrium. Since the FFP is floating, the current across the junction is zero, it is in equilibrium condition and the depletion region width is d B. The full depletion of the top portion of the collector occurs when two depletion regions merge: d fpj = db + db, () where d fpj is the distance between two junctions, d B is the width of the FFP depletion region and d B is the width of the extrinsic base depletion region. By combining () and (), we can calculate the punchthrough voltage V R needed for the full depletion of the collector between the extrinsic base and the FFP: V R d = Vbi d fpj B. () For the given doping concentrations, the built-in potential at room temperature is V bi =.97 V, and the depletion region width in equilibrium is d B nm. The dependency of V R on d fpj calculated from () is shown in Fig.. It should be noted that the actual distance between two junctions (d fpj ) is smaller than the distance between two lithography mask segments (d fp ) used in fabrication due to the lateral spread of dopands. If lateral spread of dopands is taken into account the curve in Fig. is shifted by twice the lateral dopand spread which is approximately nm. When the d fp is increased from nm to nm, the value of V R increases from.9 V to 7.7 V, meaning that the punchthrough voltage V R is rather sensitive to the d fp value. This is the consequence of the high collector doping concentration in the top part of the transistor. bi MIPRO 8/MEET 7

3 Fig. Forced-V BE simulation of DE HCBT with FFPs showing the electric potential distribution at sufficient for fully depleted collector. Potential of FFPs is stepped and almost linear distribution in the drift regions is achieved. The length of FFPs is l fp= nm and the distance between the extrinsic base and the FFPs is d fp= nm. This simple calculation was taken to roughly estimate the distances d fp between the extrinsic base and the FFP as well as the distances between the following FFPs when multiple FFPs are used. Collector between two regions should be fully depleted for V R smaller than the junction breakdown voltage, which is around V for given collector concentration. The mask distance d fp = nm is taken for initial simulations since it would give the punchtrough voltage V R =7.7 V, which is the value smaller than the junction breakdown voltage. The length of the FFPs is set to l fp = nm, the value sufficient for electric field shielding and small enough to obtain good high frequency characteristics [8]. A. Shaping of the Potential Distribution Forced-V BE simulation of the electric potential distribution of the structure with FFPs at V BE =.7 V and = V, which is smaller than but near the BO value, is shown in Fig.. It can be observed that the collector below FFPs is fully depleted and that the depletion region extends all the way to the extrinsic collector. Since the electric field is perpendicular to the potential contours, it can be concluded that vertical component of the electric field exists between the collector below and the FFPs. Therefore, FFPs serve as field plates and are responsible for the formation of additional drift regions marked in Fig.. It can also be seen that each consecutive FFP is on the higher potential with the potential difference between two being proportional to number of potential contours between them, which equals the punchthrough voltage V R. The dependencies of the electric potential on the applied at the top of the structure and in the drift regions marked by dashed cut-lines in Fig. are shown in Fig.. Potential distribution at the top of the structure across the FFPs is shown in Fig. a. For = V, it can be observed that the outer FFPs are floating on the potential which is set by the collector potential and the junction built-in potential. It can also be seen that the depletion region has reached the first FFP coupling its potential to the extrinsic base. When FFP is completely enclosed by the depletion region, its potential is constant, which can be observed for the innermost FFP at higher. For = V (condition from Fig. ) marked by the dash-dot line, the depletion region has spread all the way to the extrinsic collector. The height of potential stairs is equal to the punchthrough voltage V R, which is well Electrostatic Potential, (V) Electrostatic Potential, (V) NFFP= TOP NFFP= DRIFT V BE =.7 V Distance, (µm) a) V BE =.7 V Distance, (µm) b) = V = V = V = V = V = V = V = V = V = V = V = V = V = V = V = V = V = V Fig.. Dependency of the electrostatic potential on the applied : a) at the top of the structure (TOP-cutline in Fig. ), b) in the drift regions (DRIFT-cutline in Fig. ). Red dash-dot line corresponds to the bias condition from Fig.. Dashed lines are for >BO. predicted by the results from Fig.. In properly designed device, after the full depletion, all the FFPs are on the constant potential and E-field peaks associated with voltage drops between them are constant and kept below the critical value. The only potential drop that increases with is the one between the last FFP and the extrinsic collector (the last potential stair in Fig. a), and the critical E-field is achieved there causing the breakdown eventually. If is increased beyond BO, all the potential stairs increase (see dashed line curves in Fig. a) due to mobile avalanche carriers generated at E-field peaks placed at the pn-junctions corresponding to parts with steep potential gradient. Potential distribution across the drift regions is shown in Fig. b showing almost linear potential change and uniform electric field. B. Tuning of the BO Results of the forced-v BE simulations of BO for structures with different number of FFPs and the distance between FFPs d fp = nm are shown in Fig.. In forced- V BE simulation, V BE is constant, the emitter is grounded and collector voltage is swept. BO is determined as the where the base current (I B ) reverses its sign (i.e., 7 MIPRO 8/MEET

4 VBE=.7 V Base Current, (na/µm) Collector Voltage, (V) Fig.. Forced-VBE BVCEO simulations of structures with lfp= nm and dfp= nm with number of FFPs as a parameter. Collector Current, (µa/µm) 7 IB= na/µm Fig. 8 Forced-IB simulation of structure with FFPs at the onset of the breakdown. a) Impact ionization rate showing the top-peaks and drift region-peaks. Due to partial shielding the higest rate is observed at the inner peaks. b) Hole current density showing the avalanche hole path from the E-field peaks through the FFPs to the extrinsic base. Structure with lfp= nm and dfp= nm. Collector-Emitter Voltage, (V) Fig. 7. Simulated forced-ib (common-emitter) output characteristics of structures with lfp= nm and dfp= nm with number of FFPs as a parameter. becomes zero). For each additional FFP, BVCEO increases in almost discrete steps equal to VR. For the given dfp, tuning is accomplished up to FFPs with the maximum BVCEO around 8 V. The IB in forced-vbe characteristics reduces due to avalanche current generated in the base collector depletion region, which is subtracted from the current determined by the recombination and injection component set by the constant VBE. In Fig. the IB reduces considerably for VCE larger than V and practically BVCEO cannot be improved by adding more than FFPs. This means that all the E-field peaks in the transistor are not properly shielded and that at least one of them reaches the critical value for breakdown. Forced-IB output characteristics are shown in Fig. 7. Discrete tuning of BVCEO can be observed up to 8 V and FFPs. Due to the incomplete shielding, avalanche current is generated and added to the base terminal current causing the increased injection of holes to the emitter region and consequently, due to the emitter efficiency, increased collector current (IC) for larger VCE. Therefore, for the larger VCE softbreakdown behavior can be observed in the output MIPRO 8/MEET characteristics. Also, the hard-breakdown is seen for the IC larger than µa/µm, which occurs at the voltages determined by the discrete VR steps. The observation about incomplete shielding is confirmed by the simulation of the impact ionization rate at the onset of BVCEO shown in Fig. 8a. Simulations are done with forced-ib boundary condition in order to include the positive feedback of the BVCEO mechanism. Breakdown is not taking place at the outermost FFP but is associated with inner FFPs. Peak impact ionization rate increases moving towards inner FFPs, having maximum value at the extrinsic basecollector junction (see Fig. 8a). Avalanche holes generated at E-field peaks flow towards the lower potential in the direction of the electric field (i.e. perpendicular to the potential contours) as shown in Fig. 8b. Therefore holes generated at drift region peaks and at top peaks flow towards the FFP and then down the potential stairs (see Fig. a) to the extrinsic base. When they flow through the depleted top part of the collector, they increase the total positive charge and modulate the peak E-field. This effect is the most pronounced at the extrinsic base-collector junction because practically all avalanche holes generated at all peaks have to flow there (see Fig.8b). Since the forced-ib simulation is shown, the holes reaching the extrinsic base are injected to the emitter, causing the large electron back-injection and closing the positive feedback loop of classical BVCEO mechanism. Impact of the avalanche generated carriers can be compensated either by the larger length of the FFPs (lfp) or by reducing the dfp. Larger lfp results in the better shielding property of the FFPs and the smaller dfp limits the amount of donor charge placed between FFPs, which in turn reduces the voltage drop needed for its full depletion (VR) as well as the associated E-field peaks. By increasing the 7

5 NFFP= VBE=.7 V Base Current, (na/µm) dfp= nm dfp= nm - dfp= nm - dfp=7 nm dfp= nm - Collector Voltage, (V) Fig. 9. Forced-VBE BVCEO simulations of structures with FFPs, lfp= nm and dfp as a parameter. For dfp 7 nm good shielding is observed. VBE=.7 V Base Current, (na/µm) Collector Voltage, (V) Fig.. Forced-VBE BVCEO simulations of structures with lfp= nm and dfp= nm with number of FFPs as a parameter. Collector Current, (µa/µm) IB= na/µm 8 7 Collector-Emitter Voltage, (V) Fig.. Simulated forced-ib (common-emitter) output characteristics of structures with lfp= nm and dfp= nm with number of FFPs as a parameter. lfp, total depletion region width is increased resulting in the longer base-collector depletion region transit time, which degrades the high frequency performance. It also results in 7 Fig. Forced-IB simulation of structure with FFPs at the onset of the breakdown. a) Impact ionization rate. Due to good shielding the higest rate is observed at the outermost drift region peak. b) Hole current density showing the avalanche hole path from the outermost drift region E-field peak through the FFPs to the extrinsic base. Structure with lfp= nm and dfp= nm. the less linear potential distribution and non-uniform Efield in the drift regions On the other hand, according to the results from Fig., a small change in the dfp can significantly reduce the punchtrough voltage VR and reduce the electric field peaks but at the penalty that BVCEO tuning by adding multiple FFPs is accomplished in smaller discrete VR steps. This means that in order to achieve the same value of BVCEO, more FFPs are needed, which also increases the depletion region transit time and degrades high-frequency characteristics. However, simulations have shown that better tradeoff is obtained if the shielding is controlled via dfp. Forced-VBE simulations of BVCEO for structures with FFPs and different dfp are shown in Fig. 9. BVCEO practically remains the same if dfp is reduced from nm to 7 nm. At the same time, avalanche is suppressed, which is seen as a constant IB for larger VCE. If the dfp is further decreased, the BVCEO drops due to the reduction of the value of the punchthrough voltage VR. As an example of the proper shielding the forced-vbe simulations of the structures with dfp= nm with different number of FFPs are shown in Fig.. Discrete tuning of BVCEO is observed, with each additional FFP improving the BVCEO by approximately. V, which is in a good agreement with the VR results from Fig.. The shielding of the E-field peaks by the FFPs is efficient, which is observed as the constant IB at larger VCE. Simulated output characteristics are shown in Fig.. Due to the good shielding, no soft-breakdown behavior at larger VCE is observed. Efficient shielding is confirmed by the forced-ib simulation of the impact ionization rate at the onset of BVCEO shown in Fig. a. The inner peaks in the drift regions are well shielded having smaller impact ionization rates. The BVCEO is triggered by the peak Efield in the outermost drift region (DR in Fig. ). Avalanche holes generated at this peak flow to the FFP MIPRO 8/MEET

6 and then down the potential stairs through all the FFPs toward the extrinsic base (see Fig. b). Due to constant I B, holes are injected to the emitter and back-injected electrons end up at DR E-field peak closing the positive feedback loop. Since that peak is placed along the electron current path, impact ionization is sharply increased due to the transistor current gain β. Modulation of the E-field by the avalanche holes at the top-peaks can also be observed with the highest peak appearing at the extrinsic basecollector junction. However, the difference between peak values at the top is less pronounced compared to the Fig. 8a because the E-field peaks in the inner drift regions are now contributing a smaller amount of avalanche holes (see Fig. b). It should also be noted that similar values of the BO are obtained for structures in Fig. 8 and Fig. but at the expense of one extra FFP in case of the device with efficient shielding, which in turn increases the depletion region transit time and degrades the highfrequency performance. Nevertheless, with proper shielding, the device is scalable to much higher BO 's offering the big technological advantage. C. Discussion on the BO Tuning Limits The voltage drop in the base-collector depletion region can be easily controlled in the lateral direction by stepping of the FFPs potential via V R and breakdown occurs at the outermost FFP where the extrinsic collector charge is not limited. We expect that even higher BO can be obtained by adding more than FFPs, but so far we couldn't overcome the convergence problems in simulations when stacking a large number of floating regions. To the best of our knowledge, the problem is of numerical rather than physics nature. Our theoretical reasoning about BO limits is that the breakdown is limited by the coupling of the FFPs to the substrate potential, which depends on the amount of collector charge between the FFP and the substrate. When that part of the collector is fully depleted vertically from the substrate, the FFP potential is set by the associated punchthrough voltage. Since the FFP potential is then constant, further increase in the collector voltage above vertical punchthrough voltage, causes the reverse polarization of the FFP-collector junction. That voltage is dropped laterally in the top part of the collector. Breakdown then occurs when that lateral voltage drop reaches the value of FFP-collector junction breakdown voltage. For the given collector doping profile, simulations show that the vertical punchtrough voltage and the breakdown voltage due to the coupling of FFP to the substrate potential are approximately 9 V and V, respectively. They are fabricated together with the extrinsic base allowing the best possible control of the distances between them as well as zero-cost integration. Due to unique structure of HCBT, where the additional p + -regions can be placed within the n-collector, multiple FFPs can be fabricated on the top surface from where they can shape the potential distribution along the current path in the collector-base depletion region. Potential stepping of FFPs is obtained by the coupling via depletion capacitance and is controlled by the distance between them. Such technique is impossible to be implemented in the standard vertical-current bipolar transistors in a low-cost manner. In order to achieve the highest possible BO values, all E-field peaks in the base-collector depletion region should be properly shielded. In that case breakdown occurs due to the E-field peak in the outermost drift region. BO as high as 7 V is demonstrated by the device simulations. The theoretical BO limit for the given collector doping profile is estimated to be around V. REFERENCES [] Y. Taur, T. H. Ning, Fundamentals of Modern VLSI Devices, Cambridge University Press, 998, pp [] T. Suligoj, M. Koricic, H. Mochizuki, S. Morita, K. Shinomura and H. Imai, "Horizontal Current Bipolar Transistor With a Single Polysilicon Region for Improved High-Frequency Performance of BiCMOS ICs," in IEEE Electron Device Lett., vol., no., pp. -, June. [] J. Žilak, M. Koričić, T. Suligoj, H. Mochizuki, S. Morita, "Impact of Emitter Interface Treatment on the Horizontal Current Bipolar Transistor (HCBT) Characteristics and RF Circuit Performance," in Proc. Bipolar/BiCMOS Circuits Technol. Meeting, Boston, MA,, pp. -. [] J.A. Appels, and H.M.J Vaes, High Voltage Thin Layer Devices (RESURF Devices), in IEDM Tech. Dig., 979, pp.8-. [] M. Koričić, J. Žilak, T. Suligoj, "A High-Voltage Single-Emitter Reduced-Surface-Field Horizontal Current Bipolar Transistor for BiCMOS Integration," IEEE Transactions on Electron Devices, vol., No. 7, pp 9-, July 7. [] M. Koričić, T. Suligoj, H. Mochizuki, S. Morita, K. Shinomura, and H. Imai, Double-Emitter HCBT Structure A High-Voltage Bipolar Transistor for BiCMOS Integration, IEEE Trans. Electron Devices, vol. 9, no. pp. 7, Dec.. [7] M. Koričić, J. Žilak and T. Suligoj, "Double-Emitter Reduced- Surface-Field Horizontal Current Bipolar Transistor With V Breakdown Integrated in BiCMOS at Zero Cost," in IEEE Electron Device Lett., vol., no., pp. 9-9, Feb.. [8] M. Koričić, J. Žilak and T. Suligoj, "Improving the Horizontal Current Bipolar Transistor Breakdown Voltage by Floating Field Plates," in Proc. Bipolar/BiCMOS Circuits Technol. Meeting, Miami, FL, 7, pp. -. [9] W. Sung, B.J. Baliga, A Comparative Study -V Edge Termination Techniques for SiC Devices, IEEE Trans. Electron Devices, vol., no. pp. 7, April 7. IV. CONCLUSIONS Tuning of the BO in Double-Emitter HCBT is accomplished by addition of floating field plates (FFPs). MIPRO 8/MEET 77