Setting Up Drive Level Capacitance Profiling Measurements

Introduction

Drive level capacitance profiling (DLCP) is a powerful technique to accurately determine the charge carrier density in semiconductors when deep level states are present. It is often considered an improvement over the widely applied characterization technique known as capacitance-voltage (C-V) profiling, which assumes that all charge carriers at the depletion edge respond quickly to an applied AC test signal (V_ac). As this assumption is not always correct, it can lead to an overestimation of the carrier density. This blog post illustrates how to set up a DLCP measurement on an LED device using the MFIA Impedance Analyzer with the MF-PID upgrade option.

DLCP Setup

DLCP is based on a more rigorous (albeit still not perfect) assumption than that of standard C-V, in that it separates the carriers into two categories: interfacial states that are not fast enough to respond to V_ac and gap states that do [1]. All interfacial states are located within a certain distance from the semiconductor junction position. To keep this distance constant, it is necessary to output a voltage with a constant maximum level. This means that, in DLCP, when the DC bias (V_dc) is swept, the AC test signal (V_ac) needs to be increased or decreased correspondingly. One can achieve this goal manually or with an API. A more elegant solution is to make good use of the MF-PID option on the MFIA.

Setting up the DLCP requires only a proportional gain P (with I and D both at 0); additionally, we need to pay attention to the conversion factor between root-mean-square (rms) and peak units, √2 = 1.414 (see this blog post). Figure 1 shows the settings in the PID module in LabOne, where the input 'Demod X' in channel 3 represents V_ac (if the MF-MD option is not installed, it will be channel 2). We choose 'Demod X' rather than 'Demod R' as the latter cannot be negative. 'Signal Offset' in the output represents V_dc. The P-only feedback in Figure 1 is described by this relation:

\(V_{dc} = (V_{ac, set} - V_{ac, rms}) * 1.414\)

Bearing in mind that the output voltage is the sum of both V_dc and V_ac in peak values (not rms), we also have that

\(V_{out, max} = V_{dc} + V_{ac, rms} * 1.414 = V_{ac, set} * 1.414\)

Hence the setpoint of V_ac solely determines the maximal output level. In our case, to keep it at -50 mV, we write -50 mV/1.414 = -35.4 mV into 'Demod X' - this is the only parameter that needs adjusting in DLCP measurements. The swept range of V_ac can be more easily set with the LabOne Sweeper module. 

DLCP Results

Figure 2 shows the results of DLCP. When V_ac is swept, we see that the maximal output level is fixed at -50 mV, whereas the amplitude of the sine wave keeps increasing. Interestingly, we also find that the capacitance changes nonlinearly with respect to V_ac. Linear fitting (given by the dashed line) displays a mismatch: as explained in [2], this is because the capacitance is supposed to have a higher-order dependence on V_ac. To calculate the exact carrier density from the DLCP result, interested readers should take a closer look at [3] for a detailed analytical model.

Comparison with Standard C-V Measurement

In standard C-V_dc measurements, V_ac is kept constant. In reverse bias where the semiconducting LED is highly resistive, a small V_ac may only produce a little amount of current that leads to a noisy measurement of capacitance, as shown in Figure 2.

It is worth mentioning that C-V_dc measurements can be smoothened by proper range and filter settings on the MFIA. However, we intentionally skipped this optimization step to take all measurements in the same conditions for a fair comparison.  

Comparison with C-V_ac Measurement

At this point, one may wonder what happens if we sweep Vac without the PID control, so that V_dc is kept constant. In the small perturbation assumption, the capacitance of a semiconducting device should remain constant at fixed reverse bias. This is exactly what we see in Figure 4, where the measured capacitance only slightly increases (by 0.1%) even for a value of V_ac equal to 300 mV. This result is often less useful compared to DLCP.

Conclusion

We described the procedure to run a DLCP measurement on a semiconducting LED. Compared with standard C-V, DLCP proves extremely helpful when deep level states are present and the capacitance changes with respect to the AC test signal (V_ac), possibly even nonlinearly. Compared with another frequently used semiconductor characterization technique, deep level transient spectroscopy (DLTS), DLCP is coarser with respect to temporal resolution. However, DLCP does not rely on the relative change of capacitance and so offers a great advantage in terms of accuracy by probing absolute values [3].

The MFIA can support measurements with all three approaches mentioned in this post. If you are interested to know more, please get in touch with us to set up a demo.

References

[1] Heath, J.T., Cohen, J. D. & Shafarman, W.N. Bulk and metastable defects in CuIn 1− x Ga x Se 2 thin films using drive-level capacitance profiling. Journal of Applied Physics 95, 1000-1010 (2004).

[2] Warren, C.W. et al. An improved method for determining carrier densities via drive level capacitance profiling. Applied Physics Letters 110, 203901 (2017).

[3] Johnson, D.C. Novel Capacitance Measurements in Copper Indium Gallium Diselenide Alloys: Final Subcontract Report, 1 July 1999-31 August 2003, No. NREL/SR-520-35614. National Renewable Energy Lab., Golden, CO (US) (2004).

Acknowledgments

I would like to thank my colleague Michitoshi Noguchi for providing the initial idea and for fruitful discussions.