Dopant Profiling of Silicon with the ACSTM

  In order to profile dopant concentrations in silicon using the alternating current STM, we use a difference frequency mixing strategy.  In addition to the conventional DC bias applied to the tunnel junction, two frequencies are introduced from tunable waveform generators.  The frequencies are offset by a small amount (typically 5 kHz) that becomes the detected frequency (f1 - f2 = Df).  The mixing product (Df) is created due to the nonlinear nature of the tunnel junction and sample.  It is selected to be low frequency, so it is conveniently extracted and detected using a lock-in amplifier.  This detection strategy allows us to tune the instrument to fundamental frequencies from 0-20 GHz, while the detected output signal detected remains at a constant frequency.

 To map out the frequency and voltage response as a function of dopant type and concentration, cleaned doped Si substrates were used.  The difference frequency signal was plotted as a function of applied frequency and voltage, for both p- and n-type Si.  The results can be seen in Figure 1, which shows that the difference frequency signal is strongly dependent on the fundamental frequency.  Figure 1a shows data for highly doped p-type Si (0.001 W-cm boron-doped).  A wide peak is centered at -0.7 V sample bias.  Figure 1b shows data for lightly doped n-type Si (1-3 W-cm phosphorous-doped).  The peak for phosphorous occurs close to 0 V bias.  There is a clear difference between the peak location for boron-doped vs. phosphorous-doped Si, demonstrating our ability to distinguish between dopant types.  A large nonlinear effect is seen even at low frequencies providing the difference frequency signal necessary for semiconductor characterization.

Figure 1.   a.  Difference frequency signal as a function of DC bias and modulation frequency for p-type silicon doped with boron with a density of 1.2x1020/cm3.  b.  Difference frequency signal as a function of DC bias and modulation frequency for n-type silicon doped with phosphorous at a density of 4.8x1015-1.5x1015/cm3.

The magnitude of the difference frequency signal has been measured as a function of dopant concentration.  The difference frequency signal dependence on dopant density can be seen in Figures 2c and 2d.  The simultaneously acquired I-V curves are seen in Figures 2a and 2b.  The low resistivity, 1-3 W-cm sample, with high dopant concentration, produced a large peak near 0 V sample bias.  The high resistivity, 20 W-cm sample, with low dopant concentration, produced almost an order of magnitude less signal.  Similar results were observed for varying p-type dopant concentrations.  From these and related results, we ascertain that the magnitude of the difference frequency signal depends strongly and monotonically on the dopant density.  The results also demonstrate the capability of the ACSTM to distinguish between areas of high and low concentration using difference frequency detection.  This capability is important for imaging patterned substrates.
 

Figure 2.  Current (a and b) and difference frequency spectra (c and d) for different dopant densities for n-type Si.  All curves were acquired at fundamental frequencies of 10.000 and 10.005 MHz.  a and c are for silicon with a resistivity of 20 W-cm .  b and d are for silicon with resistivity of 1-3 W-cm.

  After confirming our ability to differentiate between different dopant levels and types, we imaged a patterned surface with ~1 mm x 2 mm p+ pads on n-type Si.  The interface between the pad and surface is shown in Figure 3.  The difference frequency map was recorded with applied modulation signals at 10.000000 MHz and 10.005000 MHz, with a DC sample bias of +1.5 V.  The upper left region of the image is n-type Si, while the lower right area of the image is p-type Si.  This assignment can also be confirmed spectroscopically.

Figure 3.  A 2 mm x 2 mm topographic image (left) and simultaneously acquired difference frequency image of a p+ pad on an n-type silicon substrate.  The image was acquired with a DC bias to the sample of -1.5 V and 0.1 pA tunneling current.  Modulation frequencies of 10.000 and 10.005 MHz were applied.  The upper left region is n-type and the lower region is p-type.

  If you have any questions regarding this ongoing work, please contact me: zjd101@psu.edu.

Zachary Donhauser's Home Page  |  Weiss Group Home Page