Minimizing As/B Autodoping in a CVD EPI Process
03/01/2003
Overview
The autodoping behavior of arsenic and boron in an epitaxial layer is strongly dependent on epitaxial process conditions. It is shown that autodoping could be minimized for both elements simultaneously by using low temperature (900°C) and low pressure (15 torr), at least during the initial phase of epitaxial growth.
Modern, smart-power application ICs on silicon often need arsenic and boron simultaneously as buried layer dopants in adjacent fields on the wafer by implantation. An example is the need for a high concentration of arsenic dopant for the collector connection of a vertical transistor and a nearby boron implant to serve (in conjunction with an n-epitaxial layer) as an electrical barrier providing separation in the lateral direction. To enable small circuit dimensions, it is necessary that the implanted n- and p-type implant fields do not overlap during the subsequent high-temperature steps in the process flow of the circuit.
The next immediate process to follow the implants is epitaxy, which gives rise to the well-known effects of vertical and lateral intrinsic autodoping [1]. Only the first few tens of a micrometer can be influenced by the process conditions during epitaxy. In later stages of the epitaxy (and in furnace processes later in the work flow), the main transport mechanism is solid state diffusion of the dopants, which can hardly be affected by any standard chemical vapor deposition (CVD) process [2].
 Figure 1. Transport mechanisms of autodoping during epitaxial growth: 1) gas phase transport, 2) solid state diffusion, 3) adsorption and desorption, 4) surface diffusion. |
null
In Fig. 1, four main transport mechanisms of the implanted dopants just at the onset of the epitaxial process are sketched: 1) transport in the gas phase just in front of the growing epitaxial layer by convection and diffusion; 2) solid state diffusion from the buried (implanted) dopant peak toward the advancing growth front; 3) adsorption and desorption exchanging dopant atoms or volatile compounds between the gas phase and the solid surface; and 4) surface diffusion of dopant atoms or their volatile compounds (e.g., hydrides such as AsH3).
Experimental data are presented for the parameters of a CVD epitaxial process that gives the desired low lateral autodoping of both n- and p-type implanted buried layer dopants simultaneously with a reasonable epitaxial process flow complexity suited for high productivity. No attempt is made to achieve detailed transport models; many of the still unknown transport parameters were not determined (e.g., diffusivities, kinetic constants). The conditions of the experiment are listed in the table.
 |
Based on earlier work, it is known that epitaxial process conditions to achieve minimum arsenic and boron lateral autodoping at the same time are opposed to one another [3]; i.e., high temperature together with low pressure gives the least amount of lateral autodoping for arsenic, while low temperature together with high pressure is optimal for the least boron autodoping. Based on these conclusions, it seems impossible to use a single epitaxial process to achieve low autodoping for both arsenic and boron simultaneously due to the mutually exclusive process conditions.
The lateral autodoping peaks of arsenic and boron were determined by secondary ion mass spectroscopy (SIMS) at the same well-defined places of a test chip in the vicinity of arsenic and boron implanted windows, respectively. As the As and B implanted windows are on different places of the test chip, the SIMS spots for the elements do not coincide either. Therefore, all arsenic profiles can be compared relative to each other, and similarly, all boron profiles can be compared relative to the other. There is no correlation, however, between arsenic and boron profiles because their individual surroundings are very different. All measurements were done on the chip nearest to the center of the wafer.
null
Figure 2 shows an example of a typical arsenic autodoping peak (boron peaks look very similar) measured by SIMS. A quantitative measure for the amount of lateral autodoping is the integrated dose under the arsenic and boron peaks. Figures 3 and 4 show these doses for boron and arsenic, respectively, for a variety of epitaxial process conditions.
The epitaxial process consists of two stages. First, there is a thin (typically ~0.2µm) cap layer at the beginning using the process conditions indicated. Second, there is a main layer on top of the cap layer about 1.3µm thick, which is nominally undoped. Deposition temperature for the main layer is always 1150°C at 40 torr with a total H2 flow of 25 slm (standard condition liters/min).
 Figure 3. Lateral boron autodoping as a function of temperature and pressure. Cap thickness is 0.2µm with one exception (0.1µm). |
null
Lateral autodoping is reduced by a decrease in temperature and an increase of pressure (Fig. 3), confirming the results of Srinivasan [2]. The cap layers were 0.2µm thick with one exception (100 torr, 0.1µm). From this result it can be seen that a thicker cap layer reduces lateral boron autodoping under otherwise identical process conditions.
null
In the high-temperature regime (>~1050°C), arsenic shows a completely different behavior compared to boron (Fig. 4). An increase in temperature and a decrease in pressure reduces the lateral autodoping of arsenic. Here also, though, a decrease of cap thickness increases lateral autodoping similar to the behavior of boron. There is a remarkable change of this dependence at very low temperatures of CVD (standard) epitaxy at around 900°C, however. The autodoping is reduced at low temperatures and reaches almost the same amount again at 900°C as at 1180°C.
Conclusion
Boron and arsenic lateral autodoping can be reduced simultaneously by using a very low temperature (900°C) at the initial stage (first few 1/10µm, i.e., cap layer) of epitaxial growth. The influence of pressure is still opposite on both elements and has to be fixed to a value given by the relative importance of boron or arsenic autodoping. The results of the model experiment shown were confirmed on runs with a real chip layout.
Acknowledgments
The authors wish to thank Heinrich Geiger and Matthias Stecher of Infineon Technologies AG for helpful discussions, and Ingo Weitzel of Siemens AG for the SIMS measurements.
References
1. G.R. Srinivasan, J. Electrochem. Soc., 1980, 127, 1334.
2. G.R. Srinivasan, J. Electrochem. Soc., 1978, 125, 146.
3. M.W.M. Graef, B.J.H. Leunissen, H.H.C. de Moor, J. Electrochem. Soc., Solid-State Science and Technology, 1985, 132, 1942.
Johannes Baumgartl studied general physics at the U. of München and received his PhD in material sciences from the U. Erlangen-Nürnberg. He worked at IBS GmbH and is now a process engineer for epitaxy at Infineon Technologies Austria AG, Siemensstrasse 2, A-9500 Villach, Austria; ph 43/4242-305-2199, fax 43/4242-305-3657, [email protected].
Klaus Locke received his PhD in chemistry from the University of Stuttgart. Previous company affiliations include STEAG Microtech and Fairchild Technologies. He is currently a process engineer for epitaxy at Infineon Technologies AG, Postfach 100944, D-93009 Regensburg, Germany; ph 49/941-202-7457, fax 49/941-202-3957, [email protected].
Johannes Baumgartl, Infineon Technologies, Villach, Austria
Klaus Locke, Infineon Technologies, Regensburg, Germany