1. DLTS Physical Basics1

DLTS Manual

O. Breitenstein, 05.02.2003

Content:p.

1. DLTS physical basics1

2. DLTS hardware3

2.1. Cryostat and T-measurement3

2.2. C-meter4

3. Software6

3.1. Measurement modes6

3.2. DLTS correlation8

3.3. Batch measurements8

3.4. The baseline option9

3.5. The user surface9

3.6. The C-calibration tool14

4. DLTS measurement14

4.1. Electrical connections14

4.2. Performing a DLTS measurement15

4.3. Data display and evaluation16

1. DLTS physical basics

Deep Level Transient Spectroscopy (DLTS) was developed in 1974 by D.V. Lang to investigate energetically "deep" charge trapping levels in semiconductor space charge structures, which may be either pn junctions or Schottky barriers. It utilizes the fact that the rf capacitance of the sample (usually measured at 1 MHz under reverse bias) depends on the charge state of deep levels in the space charge region. In total depletion approximation, the rf capacitance of a sample having a homogeneous doping concentration is:

(1)

Here A is the sample area, ND-NA is the total net charge density in the space charge layer (SCL), Vr is the reverse bias, 0 is the permittivity of the semiconductor material, and e is the electron charge. If the sample is a pn junction, ND-NA refers to the lower doped side of the junction. Vd is the built-in diffusion voltage of the space charge structure, which is the crossing point of the extrapolated 1/C2 plot vs Vr with the Vr-axis. If A is measured in units of mm2, ND-NA in cm-3, C0 in pF, and Vr+Vd in V, eq. (1) leads to the following relation with adapted units:

(2)

If charged trapping levels are existent in the SCL, their space charge has to be added to ND-NA. Assuming a donor-like trap level of concentration Nt in an n-type sample biased under a reverse bias Vr, the capacitance change by recharging these levels is:

(3)

The last identity holds approximatively if Nt < ND-NA holds. Then the trap concentration calculates from the capacitance change C as:

(4)

reverse bias

Vr

timptime

capacitance

high temperature

C0

Clow temperature

t1t2time

Fig. 1:DLTS measurement procedure (see text)

Fig. 1 shows the basic experimental procedure of DLTS. The capacitance of the sample is measured under reverse bias Vr. For performing the DLTS routine, bias pulses are periodically applied to the sample (hence its reverse bias Vr is reduced or even reversed for a certain filling pulse width timp), leading to a periodic recharging of the levels in the SCR. During these pulses the rf capacitance increases, which, however, is not regarded in DLTS. Immediately after each pulse the rf capacitance changes by C, whereby C is negative for majority carrier traps (shown in Fig. 1) and positive for minority carrier traps. If a Schottky diode is used, or if in a pn junction the reverse bias is only reduced by a pulse bias Vimp Vr, only majority carrier traps are recharged. If during the pulse a pn junction is forward-biased into injection, also minority carriers may be recharged. The degree of trap filling depends on the filling pulse width timp and on the capture coefficient of the traps cn;p, which is often formally expressed as the product of the thermal velocity times the capture cross section of the trap for the corresponding carrier type. If the pulse width is large enough (in the order of the pulse repetition period), one speaks from "saturation pulses", hence all deep levels in the SCR should be filled after the pulse. For sufficiently small pulses only some part of all levels are filled, hence the signal gets smaller. The pulse width tcapt, for which the signal height has reduced to 1/e = 0.367 of its maximum value, allows one to measure the capture coefficient cn (for electrons) or cp (for holes):

(5)

Here n and p are the free carrier concentrations of electrons and holes, respectively, during the capture process. Note that (4) and (5) only hold if V is large compared to Vd and close to Vr. For V < Vr the trap concentration calculates for saturation pulses for homogeneous trap incorporation as:

(6)

Hence, for homogeneous trap distribution the DLTS peak height C is expected to be proportional to the filling pulse height V. Note that the equilibrium degree of trap filling in the SCR under reverse bias is zero, hence the levels are ionized. If after a filling pulse traps are filled and the reverse bias is reestablished, the system is in thermal non-equilibrium and relaxes into equilibrium by thermally emitting the trapped charges into the corresponding bands, where they are swept away by the electric field. For isolated point defects this relaxation is exponential in time, but for extended defects like dislocations, precipitates, and interface layers the emission transient (just as the trap filling process) may be non-exponential. Therefore all these relations strictly only hold for point defects. This relaxation is connected with a capacitance transient, which is converted into a measurable transient signal by a C-meter. The time constant e of the thermal emission is governed by the thermal emmission rate en;p, which depends on the trap energy Et and on the temperature T:

(7)

Here Nc;v is the effective density of states in the conduction band (for electron emission) or the valence band (for hole emission, respectively), g is the degeneracy factor of the level, usually assumed to be unity, and kT is the thermal energy. Hence, the thermal emission rate en;p is the inverse of the emission time constant e. The exponential dependence of the emission rate on 1/T (the so-calles Arrhenius plot), with the thermal energy Et determining its slope and the capture coefficient cn;p determining its prefactor, is the signature of each trap, which is used for their identification.

In DLTS filling pulses are applied periodically to the sample, leading to periodical capacitance transients (see Fig. 1). The basic idea of DLTS is to convert the capacitance transients into the DLTS-signal by "correlating" them on-line. The simplest kind of DLTS correlation is the 2-point correlation, where the capacitance is measured at two times t1 and t2 after the end of each filling pulse, and the difference between these two values is displayed:

(8)

Hence, the DLTS signal is scaled in units of capacitance (usually pF). Other kinds of correlation are providing an especially good signal-to-noise ration (exponential correlation) or a better energy resolution (high resolution correlation) than 2-point correlation (see Sect. 3.2). Usually the DLTS measurement starts at low temperature, the temperature is slowly ramped up, and the DLTS signal is recorded. During the T-ramp the relaxation time constants of all levels in the sample are gradually increasing from very large values to very low ones according to (7). As long as T is too low for significant thermal emission until t2, the difference (8) is zero (see "low temperature" trace in Fig. 1). If T is so high that the relaxation is already over at t1, the difference (8) is also zero (see "high temperature" trace in Fig. 1). Only if the emission time constant (or the emission rate, respectively) of one level falls into the so-called "rate window" given by the selection of t1 and t2, a DLTS peak appears. For 2-point correlation we obtain the following condition for the DLTS peak to appear:

(9)

The "rate window", which according to (7) is the inverse of the relaxation time constant e, has the unit of s-1. For 2-point correlation the corresponding time constant is always lying between t1 and t2. The prefactor in (8) is chosen so that (for exponential transients) the DLTS peak height directly allows to read the capacitance change C, which according to (4) and (6) allows one to measure the trap concentration.

2. DLTS hardware

2.1. Cryostat and T-measurement

We are using a simple Cu-block bath cryostate shown in Fig. 2, which allows us to measure from liquid nitrogen temperature up to 400 K. The sample has to be mounted e.g. using silver paste and a tiny wire on so-called TO-4 holders (Fig. 3), which have to be inserted into the Cu-block and fixed with 4 screws. The Cu-block represents one electrical contact, the second contact has to be soldered to the corresponding leg of the sample holder after inserting it. The temperature measurement appears by a small Pt resistance thermometer, which is fixed at a thermally equivalent site at the opposite side of the Cu-block. This thermometer has not to be removed! If the samples are mounted on smaller TO-18 holders, a Cu fitting ring can be used for mounting them.

The cryostate is connected with a Raith temperature controller, which is actually designed for heating and cooling by controlling a nitrogen valve. Hence, this controller may also control flow cryostats. In the small bottom LED display the set (aimed) temperature in °C is shown, whereas in the large upper display the actual temperature is shown in °C. The up and down buttons allow to change the set temperature manually. However, in a DLTS experiment the whole T-control is performed by the computer via RS 232 interface. Note that the regulation steepness of this controller is set to a low value in order to avoid T-oscillations during ramping. Therefore there is usually a difference between set and actual temperature, which is no problem for DLTS since here only the actual temperature is measured. The T-controller displays its activity by a "heating / cooling" LED line. An "on / off" switch allows to switch off the heating activity of the controller. This switch has to be used only in case of emergency, hence in any case of malfunction of the T-control system. As a rule, during all phases of a DLTS investigation this swith has to be "on" and the computer controls the complete T-ramping process (see Sect. 3.5.).

For performing a measurement, the dewar has to be approximately half-filled with liquid nitrogen (LN). For cooling down the sample, the whole Cu-block has to be immersed into the liquid nitrogen of the dewar by using the height adjustment means. If the starting temperature is reached, the Cu-block has to be raised, until only the down-hanging Cu-wire is immersed in the LN for residual cooling. Then the whole T-ramping is controlled by the DLTS software. The ramp parameters can be set in the "settings" menu (see Sect. 3.5.).

There is a special option which allows one to change the internal parameters of the T-controller and to test its function. This is the "comport / test ET818" option in the "options" menu (see Sect. 3.5). Don't use this option without contacting Dr. O. Breitenstein.

2.2 C-meter

The preamplifier of the C-meter has to be attached to the cryostat as shown in Fig. 2 for an n-type sample. The sample sockets are the two lying opposite to the connecting cable. The outermost socket carries the positive bias. The residual two sockets are for externally connecting the sample e.g. to an I-V measurement setup. For investigating a p-type sample, the sample polarity has to be reversed, hence the preamplifier has to be connected in opposite direction, since the C-meter provides only positive bias polarity.

The C-meter allows an electronic compensation of the basic capacitance C as well as of the conductance G of the sample. There is the possibility to compensate the sample capacitance and conductance either manually or automatically. There are three possible compensation ranges (500 pF, 100 pF, and 20 pF), which can be selected manually. The lower ranges provide a better signal-to-noise ratio and a more exact C0-reading, but they cannot be used if the basic capacitance C0 is too large. The sensitivity ranges 50, 5, 5e-1, 5e-2, and 5e-3 pF/V can be set independently on the compensation range. In standard operation mode an rf-signal of 100 mV pk-pk (30 mV rms) is applied for the rf capacitance measurement. To improve the sensitivity, an rf-signal of 1 V pk-pk (300 mV rms) may be applied by pressing the "1V HF Empf./10" button, whereby the sensitivity improves by a factor of 10. The instrument allows one to set the measurement bias and the pulse bias from 0 to 15 V either manually, or to feed them in externally via sockets at the rear panel. The pulse bias is defined here as the filling pulse height V (see Chapter 1), which is applied at the other side of the sample than the bias. Hence, during the pulse the difference between "bias" and "pulse bias" is applied to the sample. For "pulse bias" > "bias" the sample is driven into forward bias during the pulse ("injection pulse"). The bias effectively lying on the sample can be directly monitored at a BNC socket at the front. The instrument allows one to superimpose a certain bias modulation signal to the sample bias, which may be advantageous for certain scaling procedures. This bias modulation signal may be attenuated by a factor up to 1000. There are two general operation modes of this instrument, which are "manual control" and "external control" from the computer. In the manual control mode, a pulse trigger signal (TTL-level) may be applied to the pulse trigger input socket at the front panel, and the delta C signal may be fed out at the front or at the corresponding rear socket. In the "external control" mode, the pulse trigger is fed in, together with other control signals, at the "computer control" connector at the rear. In this mode a number of control means at the front are disabled (see table below).

Fig. 4 shows the front plate of the C-meter. All buttons are lighting, if they are active. The buttons, displays, and sockets at the front plate have the following meaning:

Netzmains switch

Steuer. ext.selecting external (computer) control mode

Vorspannungbias, displayed in V

Impulsspannungpulse bias, displayed in V

Mod. einsuperimposing the modulation bias, inactive in external control mode

Faktorthe attenuation factor of the bias modulation

Modulationinput socket for feeding in the modulation bias

Imp. ausdisabling the bias pulse for performing a baseline measurement, inactive in external control mode

Vorsp. ext.external bias setting; bias and pulse bias are fed in at the rear sockets

Probe ext.sample is connected to "ext" sockets at the preamplifier, e.g. for I-V measurements. Inactive in external control mode.

Vorsp. Ausg.output socket for sample bias, connect with oscilloscope

Impuls Triggerinput socket for TTL pulse trigger in manual control mode, inactive in external control mode

delta C (display)capacitance deviation signal, displayed from -10 V to 10 V

Delta G (display)conductance deviation signal, displayed from -10 V to 10 V

C-Kompensationdisplay and potentiometer for adjusting the capacitance compensation, potentiometer is inactive in "auto compensation" mode. The signal (0 ... 10 V) is accessible at the rear.

G-Kompensationdisplay and potentiometer for adjusting the conductance compensation, potentiometer is inactive in "auto compensation" mode. The signal (0 ... 10 V) is accessible at the rear.

1V HF Empf./10selecting an applied rf-signal of 1 V pk-pk (300 mV eff.) instead of 100 mV pk-pk (30 mV eff.). This measure increases the sensitivity (= reduces the sensitivity factor) by a factor of 10.

auto Komp.selecting "auto compensation" mode, inactive in external control mode

Empfindlichkeitsensitivity selection from 50 to 5e-3 pF/V, referred to delta C output

Komp. bereichcompensation range selection, governs maximum allowed basic capacitance

delta C (socket)output socket for delta C (capacitance change) signal, also present at rear. Connect with oscilloscope.

delta G (socket)output socket for delta G (conductance change) signal, also present at rear

The sockets at the rear have the following meaning:

Computerconnector with computer interface

Vorverstärkerconnector with the preamplifier

Netzmains connector

Delta G Ausg.output delta G (conductance change) -10 V ... 10 V

G-Komp. Ausg.output conductance compensation signal, 0 ... 10 V

Delta C Ausg.output delta C (capacitance change) -10 V ... 10 V, connect with "ADC 0"

C-Komp. Ausg.output capacitance compensation signal, 0 ... 10 V, connect with "ADC 2"

Impulssp. Ausg.output internal pulse bias, 0 ... 15 V

Impulssp. Ext.input external pulse bias, 0 ... 10 V, connect with "DAC 2"

Vorsp. Ausg.output internal bias, 0 ... 15 V

Vorsp. Ext.input external bias, 0 ... 10 V, connect with "DAC 0"

3. Software

3.1. Measurement modes

This system measures and stores complete capacitance transients together with the temperature and the basic capacitance values as the primary data file (see Sect. 4.3). Its computer interface card contains two ADCs (ADC 0 and ADC 2) and two DACs (DAC 0 and DAC 2), a digital signal processor (DSP) and a timer, which organizes and carries out the DLTS routine. After a pulse trigger has been given out, ADC 0 is reading the capacitance change (delta C) values at a rate of 100 kHz (every 10 microseconds) until close to the next bias pulse. The DSP averages these deltaC values in certain time intervals up to the next bias pulse, depending on the choice of the measurement parameters, leading to the C-transient data. The special feature of our system is that no measured data are thrown away, hence all measured transient data are used for the DLTS correlation (except for a blanking time at the beginning of the transient, where a parasitic C-transient may occure, see Sect. 3.4.). There are two timing modes, which are the linear equidistant timing and the logarithmic equidistant timing mode. Figs. 5 and 6 schematically demonstrate the different timing sequences in these linear and logarithmic modes. In both modes after an initial blanking period the data are averaged over certain data periods, whereby the nominal measure times (to which the data points are displayed on the transient) are always lying in the center of each data period. The first (minimum) measure time is called tmin and plays a dominant role in defining the time scale of the measured transient.

bias pulse trigger:

0 20µs020µs

ADC measurements:

148121620242832481216

data periods:123456712

compensation trigger:

030µs

Fig. 5:Timing diagram of the beginning of a measurement in linear mode.

Assumed parameters: timp= 20 µs, tmin = 60 µs, points = 7

Note that the ADC digitizes always at the end of its internal averaging period of 10 µs, hence it measures the analog signal before the digitizing event. According to Fig. 5 the measurement starts 40 µs after the filling pulse, hence the first 4 digitized ADC measurements are skipped. The data for the first C-value are the average over ADC values No. 5 through 8, the second data point is the average over ADC values No. 9 through 12, and so on. So here tmin lies in the middle of data period 1 after the 6th ADC measurement, hence tmin = t1 = 60 µs holds, t2 = 100 µs, t3 = 140 µs and so on. So the time base of the linear measurement (distance between successive measurements) is always 2/3 tmin. After ADC measurement No. 32 one complete transient is measured as 7 data points. Then the so-called compensation trigger is given to the C-meter for a duration of 1/10th of the data acquisition period. Note that in this external (computer) control mode of the C-meter the automatic C-compensation works only during the action of this compensation trigger, but not in the measurement periods. This prevents any distortion of the transient shape, even for slow transients. After the end of the compensation trigger the whole sequence starts again with the next filling pulse, with the measured data being averaged by the DSP with the corresponding data of the first transient and so on. Only after the programmed integration time tint , being typically in the order of seconds, the whole procedure stops, and the seven data values averaged over the whole integration period are submitted to the host computer.