DLTS User Meeting 2023 - Q&A
We were delighted to host the third deep level transient spectroscopy (DLTS) User Meeting in 2023. DLTS is a powerful tool to investigate the concentration and activation energy of semiconductor defects by obtaining the relevant time constants involved in the capacitance transients at different temperatures. DLTS has drawn more attention, not only for classical semiconductors but also from other related fields. To continue contributing to DLTS and similar applications, this User Meeting intended to bring together the community who may face common measurement challenges and allow them to share their knowledge and expertise. We thank the speakers and all attendees for making this meeting a success.
In this year's DLTS User Meeting, we focused on:
- Celebrating the success of DLTS researchers at different academic levels;
- Informing the community of the different types of DLTS and related techniques;
- Fostering relations and exchanging ideas via open Q&A sessions and a virtual coffee break.
If you missed the User Meeting, recordings of the talks and tutorials can be viewed here.
This blog post presents a selection of questions raised during the meeting, and were answered live or retrospectively by Dr. Piotr Kruszewski (PK), Dr. Teimuraz Mchedlidze (TM), Ronald Alexander (RA) or me (Sandhya Tammireddy, ST). If you have any follow-up questions, please get in touch with us.
Q&A
Laplace DLTS
When measuring GaN junctions, what voltage ranges do you use for your filling pulses?
Answer from PK: In principle, the voltage pulses to be used in DLTS or Laplace DLTS measurements should be specified in the manner which let the operator to probe the material volume of interest. Depth of the probed volume can be determined from C-V measurements. Typically, I use two different voltage pulses to probe the material close to the surface (VR = - 3 V and VP = 0 V) and “bulk” region (VR = - 10 V and VP = -5 V), where VR and VP are the reverse bias and pulse voltage, respectively.
How do you start when defining your time window?
Answer from PK: I start measurements with 50 or 100 Hz rate window which should enable to observe most trap levels present in the material. If there is an increase or decrease of DLTS signal in measured spectrum but peaks are not present then I slightly modify rate window to “shift” the peaks to lower or higher temperature, respectively.
The capability of Laplace DLTS is impressive for its resolving power. If I want to implement Laplace DLTS, can you recommend a software package?
Answer from PK: Personally, I use commercially available software developed in the Institute of Physics, Polish Academy of Sciences, Poland (http://info.ifpan.edu.pl/Dodatki/WordPress/laplacedlts/) and recommend this package. You can always ask for trial version for tests to check whether software is suitable for your purposes.
However, there are other software packages like commercially available from OriginLab (https://www.originlab.com) or freeware software in Python (https://amorphous.tf.chibau.jp/memo.files/osilap/osilap.html). To be honest I have never tried them.
You say that you average for at least 1000 times, but do you do this on the instrument or in post processing?
Answer from PK: In most cases a number of 100 - 300 transients are collected which gives satisfactory signal to noise ratio (SNR) ~ 500-1000. If the signal consists of few components and have similar emission rates (i.e. e1/e2 ~ 2-3) then SNR as high as 3000 is needed to resolve spectrum into components. This is done by the increase of capacitance transients collected in the experiment which are always captured and averaged on the MFIA first and then processed to Laplace DLTS spectrum.
If I understood correctly to use Laplace, you first need to find the broad peak then go to that temperature and hold that temperature very constant while you acquire 1000 transient measurements for Laplace analysis. How well controlled must temperature be?
Answer from PK: Yes. DLTS peak should be found first and then Laplace DLTS technique around a peak is implemented to see whether capacitance transient is a single or multi emission process. To get a reliable results the temperature drift in the experiment should not be larger than 20 mK.
While obtaining the emission rates as function of electric fields, do you consider uniform trap distribution?
Answer from PK: No. This condition is not required in Double-LDLTS measurements. Moreover, same technique can be used to measure the trap profile as well. (More details can be found here: H. Lefevre, M. Schulz, Appl. Phys. 1977, 12, 45).
Does the choice of the material determine the impact of temperature?
Answer from PK: The material to be studied limits the highest temperature range only. In practice, DLTS-based techniques can detect trap levels located not deeper than 1 eV below and above conduction and valence bands, respectively. In silicon whole bandgap
(Eg = 1.1 V) can be scanned for temperatures below 300 K while for GaN (Eg = 3.4 eV) only small portion of the bandgap can be analysed. Thus it is very important to use DLOS (Deep Level Optical Spectroscopy) technique in wide bandgap materials to detect deeper trap levels.
Can you elaborate on the magnitudes of the steady-state capacitance measured by Boonton and transient capacitance that you measured with MFIA respectively? Are there big differences between them? How do you explain the negative capacitance on page 12?
Answer from PK: In our lab, MFLI (with IA option) is used for DLTS and Laplace DLTS measurements only. Steady-state photo capacitance (SSPC) in DLOS experiments is measured with Boonton series capacitance meters. In DLTS, both meters can easily detects signals in the level of few femtofarads (fF) in our samples. In turn, capacitance signals due to optically stimulated emission as high as 30-50 pF are typically measured in DLOS experiments.
To get the higher sensitivity, differential mode is used to measure DLTS signal by Boonton meters. Thus diode capacitance must be compensated with the tuned air capacitor connected in parallel. The capacitance difference measured by meters can have negative value in this case which is correct and does not affect measured transients. However, this is not the case of MFIA where total diode capacitance is measured by lock-in.
What is the sample contact size?
Answer from PK: The contact size in our diodes is between 0.15 – 0.8 mm2.
Capturing Transients
It’s great to see the difference that a load compensation makes makes to improve the accuracy of your impedance measurements. Is it possible to save and reload these compensation files?
Answer from RA: Yes, it is possible to save and reload files. Once you save the file you can reload it later using the drop-down context menu from the impedance analyzer tab’s calibration section.
MFIA Vs HERA setups
You attributed the second signal in the DLTS spectra from silicon sample to dislocations. Why do you think it cannot be something else?
Answer from TM: We have several arguments here. The first is that the signature of the defect, detected by the “Hera” DLTS system very well coincides with the parameters for the DH6 center (see V. Kveder, Et al., PSS(a), 1982, 72, 701). The second is that reconstructed, in-grown dislocations are usually present in FZ or Cz-grown Si crystals and therefore could be present in our sample as well. The third is the lateral distribution of the detected signal along the sample surface: signal strength changed more than twice from location to location, while the Au-related signal varied for only 3%. The fourth and strongest argument is the dependence of the signal strength on the filling pulse duration: for the Au signal, it was the behavior of a point-like defect, while for the second signal, we observed dependence characteristic for the extended defect. Last but not least, the results obtained on the MFIA DLTS setup and their difference from those obtained on the Hera setup can be nicely explained by the extended character of defects responsible for the second signal. Some figures related to these questions can be found on my page at ResearchGate (https://www.researchgate.net/profile/Teimuraz-Mchedlidze-2), in the latest publications.
Classic DLTS systems use a capacitance bridge to compensate for large sample capacitance and increase sensitivity. How do you manage samples with relatively large capacitance values in the case of MFIA?
Answer from TM: We also expected some problems with samples having large capacitance for the MFIA DLTS system. However, until now we managed to find the proper mode for the measurements for the LCR meter of MFIA (Cp or Cs). Namely for samples with small capacitance (<50 pF) we use Cp mode. For higher capacitance values we choose between the modes depending on the values of Rs and Rp in our system. For example, we used Cs mode for the mesa diodes fabricated from the PV cells with capacitance in the range of hundreds of pF. On the other side, the absence of compensation has its positive sides. Usually, proper compensation is a very difficult problem requiring additional mechanical parts. Our practice showed that the precision of the measurements on the MFIA system was not affected by the absence of compensation.
You mentioned you used an external pulse generator, as opposed to the internal one, was there any reason for this? Do you use it to trigger transient recording/broadcasting as well?
Answer from TM: The only reason for using an additional pulse generator for our measurements is that the internal generator of MFIA allows pulse length durations of 200 µs and larger. Yes, we used the pulse from the external generator to trigger the detection of the transient.
In my software I generally reject 80-100µs of the initial transient due to hardware delay, but you mention that commercial systems are applying 1 or 2 ms of hardware delay, do you think this may have a negative effect on data quality?
Answer from TM: We showed that for the correct detection of transients from extended defects it is important to have the possibility to analyze the data as close to the end of the filling pulse as possible. We expect that for several types of defects (like defects with the strong temperature dependence of the capture cross-section, anomalous dependence on the electric field, all extended defects, and some others) the initial part of the transient may be also very important. Therefore, in general, it is better to cut as less experimental points from the transient as possible. In the best commercial DLTS systems that measure capacitance, the minimal delay time is 1 ms. That will not be satisfactory for the correct detection of the signal in the case of our sample with dislocations, for example. If one adds a delay of 80-100 µs that will be even worse. However, this will concern only the “special” defects listed above. ST: The MFIA does not suffer from a hardware delay when measuring the transient, so can measure close to the end of the filling pulse.
Any requirements for samples? for instance, I work with polycrystalline laser crystals - any way to differentiate the difference between interior properties of the grains and their surface conditions?
Answer from TM: That is a complicated question. The answer depends on the concrete sizes of grains, their interface with the bulk, their density, their surface, and so on. The possibility of using DLTS for the system should be considered in each case separately. In general, DLTS is not the best tool for such systems, it may be used only as an addition to other experimental techniques.
Defect spectroscopy
Are the hysteresis measurements reproducible?
Answer from ST: Yes. The current-voltage sweeps are fully reproducible after 30 s of settling time. This settling time is chosen according to DLTS capacitance transient saturation time. A wide range of sweep rates (200 mV/s-2000 V/s) for 8 devices, as function of temperature was measured and we observed the consistent results.
Why do you choose 80 kHz for your DLTS measurements, and why does standard DLTS use 1 MHz?
Answer from ST: In Perovskites, there is a defect signature associated with dielectric relaxation at around 1 kHz range in the capacitance spectrum. Hence, the device capacitance doesn’t reach it’s geometric value till 80 kHz-100 kHz. By considering the impact of series resistance at elevated temperatures, the frequency for DLTS is chosen to be large enough for the capacitance signal to reach saturation and small enough to avoid the influence of series resistance.