Journal of Crystal Growth 195 (1998) 617—623 E⁄ect of interface roughness on performance of AlGaAs/InGaAs/GaAs resonant tunneling diodes Jiang Li!,*, A. Mirabedini", L.J. Mawst", D.E. Savage#, R.J. Matyi#, T.F. Kuech! ! Chemical Engineering Department, University of Wisconsin, Madison, WI 53706, USA " Electrical Engineering Department, University of Wisconsin, Madison, WI 53706, USA # Materials Science and Engineering, University of Wisconsin, Madison, WI 53706, USA Abstract The interface roughness of a GaAs/AlGaAs/InGaAs double-barrier-quantum-well structure was controllably altered by changing substrate surface misorientation and growth interruption time at metal organic vapor-phase epitaxy (MOVPE) growth interfaces. Atomic force microscopy (AFM) and X-ray reßectance measurements were used to quantify the interface roughness. The InGaAs quantum wells grown on singular substrates exhibit an island growth mode morphology, while step-bunched growth is observed on the misoriented substrates. A short growth interruption time, of typically &15 s, can decrease the InGaAs/AlGaAs interface roughness. The low-temperature I—» characteristics of the resonant tunneling diodes based on this quantum-well structure are found to be sensitive to the quantum-well interface roughness. The measured interfacial roughness was used as input to a numerical simulation of device perfor- mance. ( 1998 Elsevier Science B.V. All rights reserved. PACS: 81.05.Ea; 81.15.Kk; 85.30.Vw Keywords: Interface roughness; Resonant tunneling diodes; I—» characteristics 1. Introduction The formation of abrupt heterointerfaces is cru- cial for modern epitaxial devices, especially for those which rely on quantum-e⁄ect-based devices. A detailed understanding of the MOVPE process is extremely important in achieving atomically ßat *Corresponding author. Fax: #1 608 265 3782; e-mail:
[email protected]. interfaces. Several key parameters have been identi- Þed that inßuence the formation of the interface structure and morphology. One of the key par- ameters is substrate misorientation. It can greatly inßuence the interfacial structure by promoting dif- ferent types of surface growth structures such as islands, steps, and step bunching during growth [1—3]. Studies on quantum-well lasers have also demonstrated the e⁄ect of growth interruptions between the InGaAs quantum well and GaAs layers on laser performance [4]. 0022-0248/98/$ — see front matter ( 1998 Elsevier Science B.V. All rights reserved. PII: S 0 0 2 2 - 0 2 4 8 ( 9 8 ) 0 0 5 8 1 - 8 Despite all these e⁄orts, the quantitative impact of di⁄erent interface structures formed during MOVPE on structures, such as the resonant tun- neling diode (RTD), has apparently not been experimentally studied in detail. A quantitative experimental determination of the growth morpho- logy is necessary to obtain the detailed information on the interface structures at each interface within the device structure. This information is needed to quantitatively evaluate the impact on the device characteristics. According to our knowledge, there has been very little study accomplished in this topic. The RTD is considered as one of the most impor- tant quantum-e⁄ect devices due to its potential for high-speed device and multi-stage logic application such as analog-to-digital converters, multi-state memory, and mixers [5—7]. Similar to all other quantum-e⁄ect devices, one of the most challenging tasks in fabricating RTDs is the control of the interface roughness in the quantum-well structure. The e⁄ect of the quantum-well interface roughness has been studied both theoretically and experi- mentally [1,2,8—10]. These studies provided a gen- eral framework to understand the impact of interface-roughness scattering on RTD device char- acteristics. However, the lack of the direct experi- mental information on the interfacial structure in such studies makes the comparison of theory to the experimental I—» characteristics diƒcult. There still lacks the incorporation of quantitative inter- facial morphology, obtained by experiment, into the theoretical simulation of the device perfor- mance. In this study, we use a GaAs/Al 0.8 Ga 0.2 As/ In 0.3 Ga 0.7 As double-barrier-quantum-well (DBQW) structure [11] to quantitatively investigate the ef- fect of the interface roughness on RTD perfor- mance. The interface roughness of this DBQW structure is altered by changing the substrate surface misorientation and growth interruption time at the MOVPE growth interfaces. The inter- face morphology in this structure is measured by atomic force microscopy (AFM) and X-ray reßectance. The dependence of interface morpho- logy on crystallographic orientation and growth interruption has been determined and is presented here. This quantitative structural information is then used as input to a numerical model that describes the impact of the interface roughness on the RTD device characteristics. To our knowledge, this is the Þrst time that a quantitative experi- mental interface morphology study was directly integrated into the simulation of the RTD device performance. 2. Material growth and measurement A schematic diagram of the GaAs/ Al 0.8 Ga 0.2 As/In 0.3 Ga 0.7 As DBQW structure is shown in Fig. 1. These structures were grown by low-pressure (50 mbar) MOVPE at a growth tem- perature of 700¡C in an Axitron A-200 reactor. Trimethylgallium (TMGa), trimethylindium (TMIn), and arsine (AsH 3 ) were used as precursors in a H 2 carrier-gas ßow. The V/III ratios for the growth of Al 0.8 Ga 0.2 As and In 0.3 Ga 0.7 As layers were 75 and 110. The growth rates, calibrated from transmission electron microscopy (TEM) micro- graphs, were 36 and 18 nm/min for Al 0.8 Ga 0.2 As and In 0.3 Ga 0.7 As layers, respectively. In order to investigate the e⁄ect of substrate misorientation on the interface morphology, n-type GaAs(0 0 1) substrates with miscuts of 0¡, 0.5¡ and 2¡ towards the (1 1 0) plane were used to create di⁄er- ent initial step densities on the substrate surface. The growth interruption time at the In 0.3 Ga 0.7 As/upper Al 0.8 Ga 0.2 As interface, which allows annealing the In 0.3 Ga 0.7 As layer at the growth temperature, was Fig. 1. Schematic cross section through DBQW structure. 618 Jiang Li et al. / Journal of Crystal Growth 195 (1998) 617–623 varied while all other growth parameters were kept constant to study the e⁄ect of growth interruption on the interfacial structures. A series of samples was initially grown, stopping at di⁄erent layers in the DBQW structure, in order to investigate the evolution of the growth-front morphology. AFM images were obtained for each of these surfaces that typically would represent the corresponding interfaces during growth. Quantita- tive statistical analysis was performed to describe the surface features of the samples. A detailed de- scription of our analysis methods of the AFM im- ages was previously published [12,13]. X-ray reßectance measurements were also taken to com- pare the interface roughness in a multilayer struc- ture with AFM results, because X-ray reßectance measurement requires the use of a multi-quantum- well structure to obtain accurate data on the inter- face roughness. Since the critical thickness of the In 0.3 Ga 0.7 As grown on GaAs is less than 10 nm, we used a multi-quantum-well structure with four periods of In 0.17 Ga 0.83 As/Al 0.8 Ga 0.2 As to study the interfaces between In 0.17 Ga 0.83 As and Al 0.8 Ga 0.2 As layers. Although the In 0.17 Ga 0.83 As/ Al 0.8 Ga 0.2 As interface does not exactly correspond to the In 0.3 Ga 0.7 As/Al 0.8 Ga 0.2 As interface, this multilayer structure is used to reveal any trend in the interfacial roughness and the degree of correla- tion or replication in the structure between di⁄er- ent layers. RTDs grown on 0¡, 0.5¡, and 2¡ miscut substra- tes, with and without growth interruption, were fabricated from this DBQW structure and used to quantitatively assess the impact of inter- face roughness on the device I—» characteristics. The I—» curves were measured at 77 K, to minim- ize the e⁄ects of phonon scattering in measure- ment [14]. The thickness of the quantum well and its bar- riers measured by TEM and X-ray reßectance measurements show excellent reproducibility. The thickness variation from run to run is &0.1 nm. There is also a corresponding good reproduci- bility of RTD I—» characteristics. At least 10 RTDs were measured for each sample to check the uniformity of the wafer. The variation is less than 5% in their I—» characteristics in our measurement area. 3. Results and discussions The AFM images of the surfaces, which corre- spond to the interfaces in the DBQW structure, illustrate the evolution of the interface morphology during the growth. As shown in Fig. 2, the surface of the lower Al 0.8 Ga 0.2 As layer exhibits the typical atomic step-terrace structure on the singular (0¡ miscut) substrate and a step-bunched structure on a miscut substrate. The step-bunched structures consist of multi-atomic steps, two or three mono- layer in height. These structures are typical for the growth conditions used in this study [15,16]. As a result, the average surface roughness for the struc- ture grown on miscut surfaces is rougher than those grown on singular substrates. Unfortunately, the AFM images of the In 0.3 Ga 0.7 As layer do not rep- resent the actual In 0.3 Ga 0.7 As/upper Al 0.8 Ga 0.2 As interface morphology during growth. As discussed later in this paper, InGaAs surface morphology is very sensitive to growth interruption, and it may have been altered during the cool-down process after the growth was terminated leading to a poten- tially smoother surface at room temperature than existed during growth. The AFM images (shown in Fig. 3) of the surface of the upper Al 0.8 Ga 0.2 As layer reveal the change of morphology due to the previous growth of the highly lattice-mismatched In 0.3 Ga 0.7 As layer. A control sample consisting of 6 nm of Al 0.8 Ga 0.2 As grown on GaAs bu⁄er layer presents the same surface morphology as the lower Al 0.8 Ga 0.2 As layer in DBQW structure shown in Fig. 2. The second or upper Al 0.8 Ga 0.2 As layer surface exhibits a dramatic change in the surface morphology due to the underlying In 0.3 Ga 0.7 As layer. A higher density of 2D islands, of one or two monolayers in height, appears on the surface of singular substrate. There is also a corresponding roughening of the island edges. The surfaces of the miscut samples show an enhanced step-bunching structure with a much higher surface roughness relative to the single layer Al 0.8 Ga 0.2 As control sample. The step height on the miscut surface in- creases to more than ten monolayers in height as compared to the two or three monolayer height steps shown in Fig. 2b. The introduction of the highly lattice-mismatched In 0.3 Ga 0.7 As layer has Jiang Li et al. / Journal of Crystal Growth 195 (1998) 617–623 619 Fig. 2. AFM images of lower Al 0.8 Ga 0.2 As barrier grown on a GaAs(1 0 0) substrate with (a) 0¡ miscut, (b) 0.5¡ miscut towards [1 1 0]. The height-scan ranges for the image are (a) 2 nm, (b) 2.5 nm respectively. The 0¡ miscut surface shows typical atomic step-terrace structure. The 0.5¡ miscut surface has a step-bunched structure consisting of closely spaced steps each being 2—3 unit cells in height. The average RMS roughness for these two types of surfaces are 0.2 and 0.3 nm, respectively. caused a change of the upper Al 0.8 Ga 0.2 As surface morphology, in good agreement with previous studies of the strain-induced surface morphology [17,18]. It is probably reasonable to assume that the real In 0.3 Ga 0.7 As upper Al 0.8 Ga 0.2 As interface morphology is quite similar to the surface morpho- logy of the upper Al 0.8 Ga 0.2 As layer. The X-ray reßectance measurement supports this assumption, as discussed later in this paper. The surface morphology of the upper Al 0.8 Ga 0.2 As barrier grown on singular substrate, which is not shown here, reverts to the typical atomic step- terrace structure after a 15 s growth interruption. The surface of the layers grown on miscut substra- tes still exhibit a step-bunched structure with a higher roughness than the lower Al 0.8 Ga 0.2 As layer, although the average surface roughness has been reduced by the growth interruption. As shown in Table 1, the root mean square (RMS) roughness values for all samples decrease due to the growth interruption. The growth interrup- tion has a larger impact on the miscut samples than the structures grown on singular substrates. The analysis also shows that a 15 s growth interruption is suƒcient to signiÞcantly smooth growth surfaces. Therefore, a 15 s growth inter- ruption time was used in the growth of structures for the fabrication of RTDs, and for comparison with the structures grown without a growth inter- ruption. The AFM measurement of the surface morpho- logy at the di⁄erent surfaces indicates that growth interruption (or annealing of InGaAs) can greatly inßuence the formation of the interfacial morpho- logy. The X-ray reßectance measurements were used to quantify and compare the interface rough- ness between di⁄erent layers. For a multi-quantum- well structure grown on a singular substrate with four In 0.17 Ga 0.83 As/Al 0.8 Ga 0.2 As periods without a growth interruption between layers, the total RMS roughness (p 505 ) of each layer and the layer thickness are used as modeling parameter to Þt the experimental data. p 505 was determined to be 0.30$0.05 nm for each of the In 0.17 Ga 0.83 As and Al 0.8 Ga 0.2 As layers. The magnitude of the corre- lated roughness was obtained from rocking curve 620 Jiang Li et al. / Journal of Crystal Growth 195 (1998) 617–623 Fig. 3. AFM images of upper Al 0.8 Ga 0.2 As barrier grown on an underlying In 0.3 Ga 0.7 As layer. These surfaces show a dramatic change of the surface morphology compare to the surfaces in Fig. 2. The average surface roughness for these two surfaces are 0.3 and 0.8 nm, respectively. (a) 0¡ miscut surface exhibits a structure dominated by a high density of two-dimensional islands bounded by steps that are one or two monolayers in height. The height scan range is 3 nm. (b) 0.5¡ miscut surface has an enhanced step-bunching structure. The average step height is about 15 unit cells in height. The height scan range is 6 nm. Table 1 The average surface roughness of the upper Al 0.8 Ga 0.2 As barrier grown on di⁄erent substrates and with di⁄erent growth inter- ruption time. The growth interruption shows a signiÞcant im- pact on the surface roughness. A 15 s growth interruption time is suƒcient to reach the smoothest surface roughness No interruption (nm) 15 s interruption (nm) 30 s interruption (nm) 0¡ 0.3 0.2 0.2 0.5¡ 0.8 0.5 0.5 2¡ 1.1 0.9 0.8 of the 5th-order Bragg peak. We could not detect any correlated roughness in this case. Therefore, the X-ray reßectance measurement indicates that the upper Al 0.8 Ga 0.2 As barrier layer should have the same average roughness as the In 0.3 Ga 0.7 As quantum well, but there is no measured correla- tion between the interfacial roughness of the two interfaces, InGaAs/AlGaAs and AlGaAs/InGaAs, for structures grown on singular substrates. These results are used in deÞning the interface structures for the simulation of the RTD I—» char- acteristics. The presence of interface roughness can modify the performance of an RTD by a⁄ecting the peak- to-valley ratio and the subsequent shape of the I—» curve. Previous models [8—10] have provided a general understanding of the impact of interface roughness on RTD performance, but there is ap- parently no quantitative correlation between the interface roughness and the calculated device I—» characteristics utilizing experimental data of the actual interface structure. Utilizing the quantitative interfacial morphology information obtained by AFM and X-ray reßectance, a simple one-dimen- sional model, based on the work of Ricco et al. [19], is used to simulate the change of the I—» characteristics caused by the speciÞc change of the interface roughness. The lower Al 0.8 Ga 0.2 As bar- rier is considered to be conformal to the GaAs layer underneath, which is a valid assumption according Jiang Li et al. / Journal of Crystal Growth 195 (1998) 617–623 621 to a previous study [20], as well as the morphology measurement presented in this study. The interfacial structure between the upper Al 0.8 Ga 0.2 As barrier and the GaAs layer above is constructed from the data obtained from our AFM images of the Al 0.8 Ga 0.2 As layer. The interfacial structure between the In 0.3 Ga 0.7 As quantum-well layer and the upper Al 0.8 Ga 0.2 As barrier cannot be obtained directly from AFM images due to the potential changes occurred during post growth an- nealing of the sample. The X-ray reßectance measurements have shown that the In 0.3 Ga 0.7 As and the upper Al 0.8 Ga 0.2 As layers have the same total roughness but lack a correlation in the rough- ness for 0¡ miscut sample. The structure of the In 0.3 Ga 0.7 As/upper Al 0.8 Ga 0.2 As interface used in our simulation is then obtained by randomizing the data obtained for the upper Al 0.8 Ga 0.2 As AFM image. For RTDs grown on the miscut substrates, the upper Al 0.8 Ga 0.2 As barrier is considered to be conformal to the underlying In 0.3 Ga 0.7 As layer. Each of these interfaces is divided into 128]128 segments ac- cording to their corresponding AFM images. Each segment is treated as an individual RTD structure. The overall RTD I—» curve is obtained by aver- aging the I—» curves of 100 out of these 128]128 RTD structures. The details of the experimental I—» character- istics of the RTDs are shown in Table 2. RTDs on both singular and 0.5¡ miscut substrates with a growth interruption show a negative resistance region in the measured I—» curve, while those on both the 2¡ and 0.5¡ miscut substrates without a growth interruption have no negative resistance region in I—» curves. The RTDs on singular sub- strates have a much higher peak-to-valley ratio than that grown on 0.5¡ miscut substrate. These device results demonstrate that both substrate mis- orientation and the growth interruption have a sig- niÞcant impact on the device performance. These changes in device performance are consistent with the interface morphology study presented above. Two typical experimental I—» curves for the samples grown with growth interruption are plotted in Fig. 4. The simulation results based on the quantitative interface morphology data are in good agreement with the measured RTD I—» Table 2 The summary of I—» characteristics of the RTDs in this study. Only the 0¡ miscut samples and the 0.5¡ miscut sample with growth interruption exhibit negative resistance region in their I—» curves. The 0¡ miscut samples have much higher peak-to- valley ratio than the 0.5¡ miscut sample No interruption 15 s interruption Substrate miscut 0¡ 0.5¡ 2¡ 0¡ 0.5¡ 2¡ Negative resistance region Y N N Y Y N Peak-to-valley ratio 5.1 NA NA 5.2 1.3 NA Fig. 4. The low-temperature I—» characteristics of RTD struc- tures grown with a 15 s growth interruption, and their corre- sponding simulation results. The 0¡ miscut samples have much high peak-to-valley ratio than that of 0.5¡ miscut samples. The simulation results based on the quantitative interface mor- phological data. The simulation successfully reproduced the change of the I—» characteristics caused by the interface roughness. characteristics, also shown in Fig. 4. The interface roughness between the In 0.3 Ga 0.7 As/Al 0.8 Ga 0.2 As layers is too high for the RTDs grown on the miscut substrates, without a growth interruption, to pos- sess a negative resistance region. The high interface roughness can cause a spatial spread in the quan- tum energy levels laterally throughout the RTD. Previous studies [8,10] have already shown that the increase of the interface roughness leads to a drastic decrease of peak-to-valley ratio in RTD I—» characteristics. Our simulations also show that for the sample grown on 0.5¡ miscut substrate, the 622 Jiang Li et al. / Journal of Crystal Growth 195 (1998) 617–623 15 s growth interruption is suƒcient to decrease the interface roughness and obtain a negative resistance region in the RTD I—» curve. The peak- to-valley ratio obtained from this sample is still much lower than that of the samples grown on singular substrates because of the di⁄erences in the interfacial roughness. 4. Conclusions The interface morphology of a GaAs/AlGaAs/ InGaAs double-barrier-quantum-well structure was quantitatively studied by AFM and X-ray re- ßectance. These measurements reveal that the sub- strate surface misorientation and growth inter- ruption time at the In 0.3 Ga 0.7 As/Al 0.8 Ga 0.2 As inter- face have a signiÞcant impact on the formation of the interface structures. The interface roughness al- ters the low temperature I—» characteristics and hence performance of the resonant tunneling diodes based on this quantum-well structure. Similar trends were observed for room-temperature performance. A simulation was used to quantitatively correlate the interface roughness to the I—» characteristics and subsequent device performance. Device structures grown on singular substrates with a short period of growth interruption time yield smooth interfaces and improved device performance. References [1] M. Suhara, C. Nagao, H. Honji, Y. Miyamoto, K. Furuya, R. Takemura, J. Crystal Growth 179 (1997) 18. [2] M. Shinohara, H. Yokoyama, N. Inoue, J. Vac. Sci. Tech- nol. B 13 (1995) 1773. [3] A. Bhattacharya, L.J. Mawst, S. Nayak, J. Li, T.F. Kuech, Appl. Phys. Lett. 68 (1996) 2240. [4] F. Bugge, G. Beister, G. Erbert, S. Gramlich, I. Rechen- berg, H. Treptow, M. Weyers, J. Crystal Growth 145 (1994) 907. [5] R.C. Potter, A.A. Lakhani, H. Hier, J. Appl. Phys. 64 (1998) 3735. [6] S. Sen, F. Capasso, A.Y. Cho, D.L. Sivco, IEEE Electron Device Lett. EDL-9 (1988) 533. [7] A.C. Seabaugh, Y.C. Kao, H.T. Yuao, IEEE Electron Device Lett. EDL-13 (1992) 479. [8] D.Z.-Y. Ting, S.K. Kirby, T.C. McGill, Appl. Phys. Lett. 64 (1994) 11. [9] R.C. Bowen, G. Klimeck, R.K. Lake, W.R. Frensley, T. Moise, J. Appl. Phys. 81 (1997) 3027. [10] F. Chevoir, B. Vinter, Phys. Rev. B 47 (1993) 7260. [11] A.R. Mirabedini, L.J. Mawst, D. Botez, R.A. Marsland, Appl. Phys. Lett. 70 (1997) 2867. [12] J. Li, T.F. Kuech, J. Crystal Growth 170 (1997) 292. [13] J. Li, T.F. Kuech, J. Crystal Growth 181 (1997) 171. [14] P. Gueret, C. Rossel, W. Schlup, H.P. Meier, J. Appl. Phys. 66 (1989) 4312. [15] M. Kasu, N. Kobayashi, J. Appl. Phys. 78 (1995) 3026. [16] M. Shinohara, M. Tanimoto, H. Yokoyama, N. Inoue, J. Crystal Growth 145 (1994) 113. [17] J. Terso⁄, Appl. Surf. Sci. 102 (1996) 1. [18] V.B. Hill, S. Varma, A. Lorke, B.Z. Nosho, P.M. Petro⁄, W.H. Weinberg, Phys. Rev. Let. 74 (1995) 3209. [19] B. Ricco, M.Ya. Azbel, Phys. Rev. B 29 (1984) 1970. [20] J.M. Redwing, S. Nayak, D.E. Savage, M.G. Lagally, D.F. Dawson-Elli, T.F. Kuech, J. Crystal Growth 145 (1994) 792. Jiang Li et al. / Journal of Crystal Growth 195 (1998) 617–623 623