Two-dimensional Dopant Profiling and Imaging of 4H Silicon Carbide Devices by Secondary Electron Potential Contrast

Microelectronics Reliability 45 (2005) 1499–1504 www.elsevier.com/locate/microrel

Two-dimensional Dopant Profiling and Imaging of 4H Silicon Carbide Devices by Secondary Electron Potential Contrast M. Buzzoa,b, M. Ciappab, M. Stangonib, and W. Fichtnerb a

b

Infineon Technologies, Quality Management and Failure Analysis, Villach, Austria Swiss Federal Institute of Technology (ETH), Integrated Systems Laboratory, Zurich, Switzerland

Abstract Recent publications reported a surprisingly intensive dopant contrast arising in Secondary Electron SEM images of Silicon Carbide devices. This work gives an insight into the physics of the contrast generation and discusses the proper experimental setup to be used for the quantitative two-dimensional delineation of bipolar and homojunctions in Silicon Carbide devices. Ó 2005 Elsevier Ltd. All rights reserved.

1. Introduction Silicon Carbide (SiC) is presently among the most promising semiconductors for power device applications. Dopant profiling and two-dimensional imaging of device cross-sections have a strategic relevance for the development and for the failure analysis of the new SiC-based components. In the framework of a recent study on the assessment of the dopant profiling and imaging performance of high-resolution techniques [1], images of 4H-SiC implanted samples acquired by Secondary Electron (SE) Scanning Electron Microscopy (SEM) have shown an unexpectedly high dopant contrast. Thanks to the observed contrast, it has been possible to delineate quantitatively bipolar junctions with a level of accuracy, which is superior to that provided by dedicated scanning probe techniques such as Scanning Capacitance Microscopy. Due to the technological relevance of the applications of this technique, it has been decided to investigate in more detail the physical mechanisms of the contrast

generation in SiC and the best experimental setup to be used. The contrast between differently doped regions has been reported mainly for silicon samples starting from the late 50s [2, 3]. Even if a comprehensive description of the model on which the contrast mechanism is based is still missing, a crucial role has been attributed to the local variations in the ionization energies [4]. Due to the resulting built-in potential arising in unbiased junctions, the observed contrast can be ascribed to the potential contrast of the SE. In order to validate this assumption three different experiments have been performed. The contrast mechanisms is discussed basing on device simulations, focused in particular on the computation of the stray electric field outside the sample.The efficacy of the SE Potential Contrast is validated by the application on complex SiC structures and its capability to resolve both pn- and homo -junctions as well the potentiality to locate electrical junctions is shown.

0026-2714/$ - see front matter Ó 2005 Elsevier Ltd. All rights reserved. doi:10.1016/j.microrel.2005.07.069

1500

M. Buzzo et al. / Microelectronics Reliability 45 (2005) 1499–1504

Fig. 1 Cross-section of the SiC structure under investigation. 2. Experimental The cross-section of the sample under investigation is represented in Fig. 1. It is a SiC structure realized on a 4H-SiC 8º-off n-type substrate (ND˜10 18cm-3). A buffer layer (ND˜10 17cm-3) and subsequently an epitaxial layer (ND˜10 16cm-3) both doped with nitrogen have been deposited, followed by the implantation of highlydoped aluminium p +-wells (NA˜10 19cm-3). The sample has been cleaved in air. All the experiments have been performed using a cold field emission scanning electron microscope (Hitachi S-4700, sketched in Fig. 2). The first experiment has been carried out at a sufficiently low acceleration voltage of 0,5 kV and by a beam current of 20 pA by using a traditional lateral Everhard-Thornley configuration for the collection and detection of the SE. The obtained SE image is shown in Fig. 3a. The second experiment has been performed with the same sample as before. In this case the SEM has been especially equipped with a through-the-lens E×B detector, capable to detect genuine SE filtered from fast SE. The obtained SE image at an acceleration voltage of 0,5 kV and at a beam current of 20 pA is shown in Fig. 3b. The third experiment has been conducted at the same experimental conditions as for the second experiment excepted for the acceleration voltage, which has been increased up to 20 kV, with a beam current of 500 pA. The related SE image is presented in Fig. 3c.

Fig. 2 Sketch of the SEM showing the location of the lateral Everhard-Thornley (lower) detector and the through-the-lens E×B (upper) detector. Here are also represented the primary electron beam (PE) and the trajectories of secondary (SE) and backscattered electrons (BSE). In order to compare quantitatively the images in Fig. 3, intensity profiles have been obtained from the SE maps along the symmetry axis of an implanted region using the software analiSYS and represented in Fig. 4. The contrast of each image has been normalized according to the formula

C=

I x − I sub ×100 I x + I sub

where Ix represents the unknown intensity level and Isub represents the reference intensity level (substrate). This representation delivers a true contrast value, which is independent on the contrast and brightness settings of the microscope. Simulations have been performed using the device simulator DESSIS-ISE [5]. 3. Results & Discussion Fig. 3a and Fig. 4 (dotted line) show that the SEM renders a very poor contrast between p + and n regions and no contrast between n + and n regions when the

M. Buzzo et al. / Microelectronics Reliability 45 (2005) 1499–1504

1501

(a)

(b)

(c) Fig. 3 (a) SE image acquired with the EverhardThornley detector. (b) SE image acquired with the E×B SE detector. Both images have been acquired at 0,5kV and 20 pA. (c) SE image acquired with the E×B SE detector at an acceleration voltage of 20 kV and at a beam current of about 500 pA. Everhard-Thornley SE detector is used in the conventional configuration. On the contrary, the dopant contrast between bipolar and homojunctions is very intense (Figure 3b and solid line in Figure 4) by using a through-the-lens E×B detector, adjusted to act as a low pass filter for SE with an energy of few eV. In this case, the contrast quality also enables us to resolve the presence of a thin buffer layer between the n + and n epitaxial layers. As shown in Fig. 3c and Fig. 4 (dashed line), the dopant contrast drastically decreases when increasing the acceleration voltage up to 20 kV. Under these conditions, a slight contrast can still be observed even if the overall image contrast is worsened by the signal peak due to the edge effect.

Fig. 4 Contrast profiles taken along the symmetry axis of the implanted region of the sample in Fig. 3. The maximum contrast between differently doped regions is observed using the E×B detector at low beam energy. In order to represent the effect of the built-in potential on the SE, the resulting stray potential and stray electric field outside the specimen have been simulated for an abrupt SiC junction with ND˜10 16cm-3 and NA˜10 19cm-3 (Fig. 5). In the simulation model, the detector has been located at a distance of 400µm from the surface of the sample and its potential has been fixed to ground (boundary condition). In Figure 5, the electric field is plotted according to the conventional representation, where the vectors of the electric field point in the opposite direction as the electrostatic force exerted on the electrons. Thus in this case, electrons are attracted towards the n-type and repelled away from the p-type semiconductor. The built-in potential can reach values up to 3V depending on the doping concentration, according the formula [6]

Vbi =

kT N A N D ln 2 q ni

where k is the Boltzmann constant, T is the absolute temperature in degrees Kelvin, q is the elementary charge, NA and ND are respectively the concentration of acceptors and donors and n i is the intrinsic carrier density. This determines an electric field outside the sample which increases or decreases the kinetic energy of the SE, depending on the location of the emission site. Due to the low strength of the electric field, the

1502

M. Buzzo et al. / Microelectronics Reliability 45 (2005) 1499–1504

(a)

Fig. 5 Stray potential and stray electric field resulting at the surface of an abrupt unbiased SiC junction due to the built-in potential (simulation by DESSIS). The radial thin lines are equipotential lines, while the thick lines represent the field lines. influence is non-negligible only in the case of lowenergy SE. The electrons emitted from the p-region are accelerated out of the surface while the ones emitted from the n-region are retarded. As a consequence, a detector using a low-pass energy filter collects a current of secondary electrons which is lower for the ntype as for the p-type. Thus p-regions are rendered with a brighter contrast as n-regions. The simulation in Fig. 5 shows that the electric field is more intense in the close vicinity of the junction. At the location of the depletion zone, the field lines are almost parallel to the surface. Electrons emitted in this region are either immediately re-absorbed by the sample, or deflected in horizontal direction. This effect is maximized at the electrical junction and turns into a minimum of the SE current collected by the detector. Since the stray electric field (due to the built-in potential) mainly influences the low-energy tail of the SE, the information regarding the doping contrast is carried with a sufficient signal-to-noise ratio only by the genuine SE [7]. Therefore, the dopant contrast is enhanced by maximizing the detection of the lowenergy SE. This is possible by the use of a detector which acts as an energy filter, like the through-the-lens E×B detector and not by the Everhard-Thornley SE detector, which also collects high-energy SE and backscattered electrons. The decrease of contrast by increasing beam

(b) Fig. 6 SE image acquired with the E×B SE detector in the p + region (a). The vertical profile shows a higher intensity for the p-region as the n-region (b). The profile shows a minimum at the location of the electrical junction (EJ). energy observed with the through-the-lens E×B detector in Fig. 3c can be explained by the fact that, when the sample is illuminated by the primary electron beam, a local potential change occurs due to the generation of secondary electron-hole pairs within the semiconductor [7]. This turns into a perturbation of the equilibrium carrier distribution within the sample. The generation of an electron-hole pair in SiC requires about 3 eV. Thus at low accelerating voltages the sample faces the lowest perturbation of its thermodynamic equilibrium, while at high accelerating voltages, the ionization reaches such a level that the carriers are no longer described by the Fermi distribution and the local potential is drastically altered [8]. As consequence the electric field at the surface is distorted and the contrast is noticeably reduced. Furthermore, the intensity peak arising as a consequence of the enhanced edge effect, masks the intensity modulation due to the dopant contrast at the

M. Buzzo et al. / Microelectronics Reliability 45 (2005) 1499–1504

1503

at the same electrostatic potential. This experiment demonstrates that the excellent contrast observed in these images is a pure doping contrast and it is not due to self-biasing effects. The image in Figure 7a has been acquired at an acceleration voltage of 500V. The contrast is excellent even at low magnification, such that both the p +n junction and the nn + homojunction can be revealed very accurately at the same time.

(a)

In homojunctions, the built-in potential depends on the ratio of the doping levels of the adjacent layers according to the formula [9]

kT N (1) Vbi = ln ( 2) q N where N(1) and N(2) represent respectively the higher and the lower doping concentration. This corresponds to 59 mV per decade, which is exactly the built-in potential of the nn + homojunction that has been resolved in Figure 7b.

(b) Fig. 7 SE image acquired with the E×B SE detector for the whole SiC structure (a) and the contrast profile (b). sample surface. The samples imaged in Figure 6a and 7a have been observed immediately after cleavage. This is the reason why they exhibit a better contrast than the sample in Figure 3b, which has been used for several experiments and, as a consequence of the irradiation by the primary electron beam, is coated with a polymeric deposit that reduces the contrast. The image in Figure 6a has been acquired at a higher magnification at an acceleration voltage of 500V. As expected from the previous theoretical considerations, the SE signal originated from the p + implantation is more intense than the signal from the nregion. Furthermore, the related profile (represented in Figure 6b) shows a minimum within the depletion region at the location of the electrical junction. The surface of the sample imaged in Figure 7a and 7b has been sputtered with aluminum before the cleavage, with the scope to keep all the doped regions

From previous considerations, it can be concluded in general that the expected contrast arising in bipolar junctions in SiC is quite high, i.e. in the 30% range. In the case of homojunctions, the contrast has a lower dynamic range. Two adjacent layers with a doping difference by a factor of ten are expected to provide a contrast in the 2% range. Furthermore, the contrast has been shown to be much more intensive for the freshly cleaved samples than for the samp les, which have been imaged several times. The reduction in contrast can be attributed to the cracking of carbon layers at the sample surface by the primary beam. The role of fixed charges, of interface states, as well of other mechanisms leading to the pinning of the Fermi level has to be investigated in more detail. 4. Conclusions It has been shown that the dopant contrast observed in SiC is due to the Secondary Electron Potential Contrast effect resulting from built-in potentials within the semiconductor. The use of a through-the-lens E×B detectorenables us to detect genuine SE filtered from fast SE, which are sensitive to the stray field outside the specimen determined by the built-in potentials. The high sensitivity of the technique renders an excellent contrast both for pn junctions and for n-type

1504

M. Buzzo et al. / Microelectronics Reliability 45 (2005) 1499–1504

homojunctions, where a doping difference of a decade can be resolved. SE potential contrast represents a valuable approach to dopant profiling and imaging which allows precise junction delineation. Further investigations are required in order to improve the quality of the crosssectioned surface and to distinguish the doping type basing on the measured signal. The straightforward approach and the reliability of the results make SE potential contrast the method of choice for dopant profiling on SiC. Acknowledgements The authors are indebted in particular to Dr. Roland Schmidt (Hitachi High-Technologies Europe) for precious discussions. References [1] Buzzo M., Leicht M., Schweinböck T., Ciappa M., Stangoni M., Fichtner W. 2D Dopant Profiling on 4H Silicon Carbide P+ N Junction by Scanning Capacitance and Scanning Electron Microscopy. Microelectronics Reliability 44(2004)1681-1686 [2] Oatley C.W., Everhart T.E. The Examination of p-n Junctions with the Scanning Electron Microscope. Journal of Electronics (1957), pp. 568-570 [3] Turan R., Perovic D.D., Houghton D.C. Mapping electrically active dopant profiles by field-emission scanning electron microscopy. Appl. Phys. Lett. 69(11), 1996. [4] Sealy C.P. , Castell M.R., Wilshaw P.R. Mechanism for secondary electron dopant contrast in the SEM. Journal of Electron Microscopy 49(2), 2000, pp. 311-321. [5] http://www.synopsys.com [6] Sze S.M . Semiconductor Devices Physics and Technology. John Wiley & Sons (1985). [7] Reimer L. Scanning Electron Microscopy. Physics of image formation and microanalysis. Springer-Verlag (1986) [8] Thomas Ch., Joachimsthaler I., Heiderhoff R., Balk L.J. Electron-beam-induced potentials in semiconductors: calculation and measurement with an SEM/SPM hybrid system. J. Phys. Appl. Phys. D37 (2004) 2785-2794. [9] Rhoderick E.H., Williams R.H. Metal-Semiconductor Contacts. Clarendon Press (1988).