- Email: [email protected]
Low frequency noise studies of AlAs — GaAs — AlAs quantum well diodes at 77 K
Low frequency noise studies of AlAs - GaAs - AlAs quantum well diodes at 77 K X . M . Li, M.J. Deen, S.P. Stapleton, R.H.S. Hardy and O. Berolo* School of Engineering Science, Simon Fraser University, Burnaby, BC, Canada V5A 1S6 *Communications Research Center, Ottawa, Canada K2H 8S2 The low frequency noise of AlAs - GaAs - AlAs quantum well diodes (QWD) was measured at both 77 and 300 K. The experimental results show that, in the frequency range from 2 Hz to a few kHz, the noise is ~- 1/f at both temperatures. At room temperature, the current noise spectrum S~(f) in A 2 Hz -1 is proportional to the square of the current through the QWD. However, the noise spectral density at 77 K was not proportional to the square of the current. At 300 K, S~(f)/I 2 versus bias voltage is fairly flat up to the negative differential resistance (NDR) region, but beyond the NDR region, a small jump in Si (f)/I 2 of a few dB was observed. This increase in Si (f)/I 2 beyond the NDR region was found to be 20 dB continuous increase at 77 K. The results also show that below the NDR region, the noise levels are about the same at both 300 and 77 K. However, the noise increases more at 77 K than at 300 K as the bias voltage increases beyond the NDR region.
Keywords: low temperature electronics; quantum well diodes; resonant tunnelling devices
In recent years, resonant tunnelling devices made with I I I - V compounds have attracted considerable attention 1-8 because of their interesting physics and their potential for both optical and electronic applications. This interest is largely because of the fast charge transport through the heterostructures and the potential for both high speed analogue and digital applications. The possibility for high speed applications is based principally on the negative differential resistance (NDR) region of these devices. As an example, the NDR provides the basis for high frequency oscillators, mixers, negative resistance amplifiers or detectors 1-6. In addition to the twoterminal resonant tunnelling structures, several threeterminal devices using resonant tunnelling are currently being studied 6-8 for various digital applications. The structure of a single quantum well resonant tunnelling diode (RTD)o generally consists of a layer of undoped GaAs (< 100 A*) sandwiched between two barrier layers of a larger band gap semiconductor material such as AlxGa~_~As. In this double-barrier structure, carriers can tunnel through the larger band gap material into the quantum well and out through the second barrier. When the energy (corresponding to the d.c. bias) of the tunnelling carriers matches closely a bound state energy level within the well, there is a peak in the transmission probability that is measured by a peak in the current of the I - V curve. This peak in the current occurs when the d.c. bias voltage is approximately twice the equilibrium bound state energy within the quantum well. Further "1 ~ = 1 0
I nm
0011 -2275/90/121140- 06 © Butterworth-Heinemann Ltd
1140
Cryogenics 1990 Vol 30 December
increase in the bias voltage results in a drop in the tunnelling current, giving rise to the negative differential resistance region, until thermionic currents over the barrier dominate the tunnelling currents through the barriers, and the current begins to increase again with applied voltage. Most of the early researches on resonant tunnelling diodes were with AlxGal_xAs barriers. However, more recently, the attention of researchers has been focused on studying new materials with increased barrier heights such as AlAs for the GaAs-based quantum wells and also because AlAs has the highest tunnelling barrier height for F-valley electrons. With the increased barrier heights, excess currents are reduced and the peak-to-valley current ratios are enhanced. New materials are also being studied for improving the peak current densities of RTDs. Most previous electrical studies on quantum well diodes were on their d.c. I - V characteristics, or their high frequency properties. There have been very few studies 9-tl on their low frequency noise characteristics despite the fact that low frequency measurements provide a powerful tool for studying non-idealities such as defects, trapping levels and differences in heterojunction qualities at the various interfaces. Two previous studies 9'~° on AIGaAs/ GaAs/A1GaAs diodes concentrated on analysing defect assisted tunnelling and on determining the activation energies and capture cross sections of the trapping centres at a fixed bias voltage. The other studyHon Si/Ge RTDs just reported on experimental results. In this paper, a different approach is taken. Noise measurements in AIAs/GaAs/A1As quantum well devices were measured. These devices promise better peak-to-
AlAs-GaAs-AIAs
and the Si-doped GaAs layers have ohmic contacts for
valley current ratios and higher current densities than AIxGa~__,As/GaAs/Al.~Ga~_xAs diodes. The differences in their noise current spectral densities at 300 and 77 K are studied as a function of the biasing voltages. The analysis considers all noise contributions from the circuit components external to the RTD, so that the noise of the quantum well diode can be determined. The range of values of o~ in the 1/f~ noise dependence is also of interest. The measurements at 77 K also allowed a study of the effects of reduced excess currents and increased tunnelling currents on the noise characteristics.
electrical measurements.
Device and experimental details The devices used in this study were grown by molecular beam epitaxy (MBE) and their structure is shown schematically in Figure 1. These quantum wells were mesa isolated structures of area 10 x 10 #m 2. The two barrier layers are 20 ,X, AlAs (7 monolayers) and the quantum well layer is 45 ,~ GaAs (16 monolayers). Above the first barrier is a 15 ,& spacer layer of undoped GaAs and then a layer of 0.5 /~m n ~ GaAs. Below the second barrier layer is a thick 140 ,~ spacer layer of undoped GaAs and then a 1.5/~m layer of n ~ GaAs. The n + GaAs substrate
500A GaAs (Si-doped) 0 . 5 u m n+ G a A s 15 A G a A s ( U n d o p e d ) 20 A AlAs 45 A GaAs 20 A AlAs 140 A GaAs (Undoped) 1.5um n+ G a A s
Figure 1 Cross-sectional v i e w of a single q u a n t u m well tunnelling diode structure
0.35 . i III 0.30
& g~
/ 0.25
.
::'.
II
/
1 30
........V ( ....
0.20
rr
o.lo
./
•
/W
/
,//
/--/"
/"
.7
/-'/ O.00 ~-~" , 0.00 0.10
- -10
-3o
0.05
-50 ,
0.20
,
0.30
r i 0.40
,
,
0.50
0.60
T Ov LU © Z <
'
JR2
-
Ac coo.,.d
Signal
=
W rr E < Z >-
1
I
cO
A
P"
I/ ,lr~: I / 'il~i~ l/ /'~J
Typical c u r r e n t - voltage curves of one of the quantum well diodes (QWD) at 300 and 77 K are shown in Figure 2. These measurements were made with an automated HP4145A Semiconductor Parameter Analyzer (SPA). Also shown in Figure 2 are the differential resistances and bias points at which the noise measurements were made. The peak current density at 300 K was 2.14 x 104A cm : and at 77 K was 2.24 × 104 A c m .2, which are similar to those reported in Reference 11 for similar RTDs. The peak and valley voltages at 300 K were 0.324 and 0.422 V and at 77 K were 0.356 and 0.44 V, respectively. The current peak-to-valley ratios were 1.45 at 300 K and 2.25 at 77 K. The test devices show significant negative resistance at both 77 and 300 K and this negative resistance makes it difficult to bias the QWD by a simple current source. Because of the NDR region, the bias voltage is not a unique function of the bias current. However, the bias current is a unique function of the bias voltage, and so a voltage source is required to bias the QWD. For the biasing circuit, a simple voltage divider network is used to provide a constant voltage to the QWD. To avoid introducing any complex noise spectrum from the bias circuit, two metal film resistors are used for the simple voltage divider. The noise from the bias circuit was subtracted from the measured noise level to determine the true QWD noise. To accomplish this, two measurements were taken at each bias point, one with the device and another without the device. Using the two sets of data, the noise of the bias circuit can then be subtracted to give the QWD noise. Figure 3 shows the experimental test setup for the noise measurements. The gain of the ultra-low noise preamplifier (PAR Model 133) was set at 100. Its noise level is below - 160 dB, which is the thermal noise ground level of the system. The dynamic signal analyser (HP3561A) is a.c. coupled to the ultra-low noise amplifier during the measurements. The signal analyser has a dynamic range of 80 dB and it can measure signals down to - 130 dB (V 2 Hz ~). Combining the low noise pre-amplifier with the signal analyser, the system can measure noise levels down to - 1 6 0 dB (V 2 Hz-~). All measurements were made from the noise spectrum at each biasing voltage. Figures 4a and b show the equivalent noise circuits for the test set-up with and without the QWD, respectively. The quantum well diode is considered as a current noise
/ / f 50
I%I
:
•
0
1 70
I
quantum well diodes: X.M. Li et al.
i :-~±~ •
QWD
i
[
Analyzer i L
(3
-70 0.70
IEEE488
Metal Box
GPIB
i v
BIAS VOLTAGE (VOLTS) Figure 2 Measured d.c. current - voltage characteristic of a device w i t h area 100/.tin 2 at 3 0 0 and 77 K. Also s h o w n is the differential conductance at both temperatures and the biasing voltages used in the noise measurements• - , 3 0 0 K; - - - , 77 K; 3 0 0 K; • • . , 77 K
Plotter
-
Personal Computer
Figure 3 Experimental test set-up used to measure the l o w frequency noise spectrum
Cryogenics 1 9 9 0 Vol 30 December
1141
AIAs-GaAs-AIAs
quantum well diodes: X.M. Li et al.
Experimental results and discussions
Figures 5a and b show the measured Si(f) and Si(f)]l 2 versus frequency for seven bias voltages at 300 K. As
(a)
(b)
Figure 4 Noise equivalent circuit of measurement system (a) with and (b) without the quantum well diode
source Si(f) (in A z Hz -x) in parallel with its differential resistance Ro. The other resistors are treated as voltage noise spectrum sources (white noise sources). For a resistor of value R, the voltage noise spectrum is (4kTR) ~/2. The battery was modelled as a voltage noise source Sv(f)b in series with its internal resistance Rb. Since Rb "~ R2, its noise contribution can justifiably be ignored. Using these assumptions, the system voltage noise spectrum Sv(f)s and the thermal noise from resistors can be subtracted from the measured noise spectrum Sv(f), which is measured when the QWD is in the test system. The noise spectra measurements without the device are denoted as Sv(f)o. A detailed derivation of the noise subtraction formula is given in Reference 12. Using Figure 4a Sv(f) can be determined from
Svq)
-
Sv~0s
=
(RIR2RD)2[ 4kT(11
12) + Sl(f)l + (RDRl)2Sv(f)b
(RIR2 +
R2RD +
RdRI) 2
(1) Similarly, the noise spectrum without the QWD can be determined from Figure 4b as
Sv(f)o - Sv(f)s = 4kT
R l R2
R1 + R2
+ Sv(f)b
R~ (R1 + R2)2
(2)
Since the system noise spectrum Sv(f)s ( - 1 6 0 dB) is well below the noise level of the bias circuit and QWD, it can be neglected in Equations (1) and (2). Subtracting the noise of the battery in the above equations, the current noise spectrum of the QWD is then obtained as
expected, an increase in the bias voltages results in an increase in Sl(f). Figure 5b shows that the Si(f)/I z versus frequency curves seem to be grouped into two sets. The first set consists of measurements at a bias voltage of up to 0.35 V, i.e. up to the end of the NDR region, and the second set at biases of 0.45 V and higher, above the NDR region. The differences in Si(f)/l 2 between the two sets was = 4 dB for frequencies between 2 and 700 Hz. Also, the Sl(f)/l 2 values for the two lowest bias voltage measurements at 0.15 and 0.2 V indicate that there might be generation-recombination activity due to trapping levels, possibly located at the heterojunction interface of the first barrier layer because of the 'Lorentzian' nature ~2-x4 of their noise spectra. However, the other bias voltages do not clearly indicate the presence of traps, even out to 100 kHz. From Figure 5b, the noise level is related to the magnitude of the current level, and the noise due to the higher currents might be dominating noise contributions from trapping levels that are observed at the lower bias currents. This is supported by the fact that the noise levels are higher than those reported in References 9 and 10, in which trapping levels were measured for a bias voltage just beyond the NDR region. Using the test set-up, the resistors are carefully chosen in the bias voltage divider to eliminate any spontaneous oscillation in the NDR so that the diode can be biased in the negative differential region. The experiment shows that the smaller the R~, the better the stability of the bias voltage. However, it could be a heavy load on the battery. The current noise spectrum at the bias voltage VB in the NDR (0.35 V) was 1/f 1°~ over the entire frequency range. This frequency dependence of S~(f) and others at different biases are summarized in Table 1. They are in agreement with the noise spectra results reported in References 9 and 10. Also included in Table 1 are the biasing voltages and currents, and the d.c. and a.c. resistances. The value of ot in 1/F was 1.15 for biases above the valley current and = 1.08 for biases below the peak current. To examine the 300 K results from a different perspective, in Figures 5c and d, S1(f) and Si(f)/l 2 a r e plotted versus bias voltages at several frequencies. The shapes of these two curves are very similar and they indicate that for a bias voltage in the NDR region there is a dip in the noise level. The noise at the bias voltages just after the NDR region was 2 - 4 dB higher over the entire frequency range. Also, the differences between S~(f) (and also Table 1
Biasing conditions of the QWD for noise measurements
at 3 0 0 K Bias Bias D.c. voltage current resistance (V) (mA) (~2) (Vd/Id)
(RIR2 + R2Ro +
RDR1)2Sv(f) -
(R1 + R2)2R2Sv(f)o
(R1 R2 RD)2
(3) In the experiments, the dynamic resistance R D of the diode is calculated from the static I - Vcharacteristic curve.
1142
Cryogenics 1990 Vol 30 December
0.15 0.20 0.30 0.35 0.45 0.60 0.65
7.97 11.71 20. 19 18.66 15. 30 23.34 27.84
18.91 17. 10 15.02 18.66 29.42 25.73 23.47
RD(dVd/dld) (~2) 14.21 12.48 14.0 -30.76 35.42 13.06 10.61
c~ in Sl(f)1/f ~ (f: 2-- 200 Hz) 1.08 1.10 1.05 1.01 1.15 1.15 1.16
AlAs-GaAs-AlAs -110
-80
-120
-90
-130 < 03
quantum well diodes: X.M. Li et al.
-100
-14o -110 -150 -120
*+~9{¢^'*,;'lli~.
-160
"'**:'";:III
-170
-130
-180
-140 10
a
100 Frequency
1000
10000
2
10
-110
-80
-120
-9O
-130
1000
10000
Frequency (Hz)
b
(Hz)
1O0
~4
~.~+/
-100
rn
-14o -110'
CO
d-
-16o:
C
-120
h k _~ •
~
~"
:
-
__A--A ~ - A
-L~-~.
~ o
-130
--170' -180 O. 15
--O ~ ._.-.-.-4-
-15o;
i 0.25
i 0.35
t 0.45
L 0.55
I 0.65
-140 0.15
0.75
BIAS VOLTAGE (V)
d
i
i
i
i
i
0.25
0.35
0.45
0.55
0.65
0.75
BIAS VOLTAGE (V)
Figure 5 (a) & ( f ) v e r s u s f r e q u e n c y at different bias voltages below, in and above the N D R region and at 3 0 0 K. Note the increase in Sdf) with bias voltage. + , 0 . 1 5 V ; zx,0.20V; O,0.30V; +,0.35V; 0,0.45 V; • , 0 . 6 0 V ; v,0.65V.(b) S d f ) / / 2 v e r s u s frequency. +,0.15V; zx,0.20V; O,0.30V; + , 0 . 3 5 V; v , 0 . 4 5 V ; 0,0.60V; [ ] , 0 . 6 5 V. (c) & ( f ) versus bias voltage at several frequencies b e t w e e n 2 H z a n d 10 kHz and at 3 0 0 K. + , 2 . 0 Hz; i x , 1 0 H z ; O , 2 0 H z ; + , 1 0 0 Hz; • , 2 0 0 H z ; •, 1.0kHz; +,2.0kHz; ix, 10kHz. (d) Sl(f)/I 2 v e r s u s bias voltage at several frequencies b e t w e e n 2 Hz and 10 kHz and at 3 0 0 K. Symbols as in (c)
Sl(f)/I 2)
at 2 Hz and at 10 kHz increase with bias voltages, and this increase in Si(f) also supports our earlier conjecture that the noise level is related to the current through the diode. Figures 6a and b shows the measured Si(f) and S[(f)/I 2 versus frequency for seven bias voltages at 77 K. The deviation from 1/f noise behaviour suggests there are generation-recombination ( g - r) centres in the device. In the frequency region measured, there is only one turning point (one Lorentzian component) in the spectrum for each bias voltage. This means that only one g - r centre was measured at each bias voltage for our QWD 12'13. Plotting S1(f)/l ~ versus frequency was also tried to determine the best value for current dependent coefficient/3. However, the narrowest range of Si(f)/I ~ was found with/3 = 2. Because of the almost 20 dB range in S [ ( f ) / l 2 versus frequency curves at the different bias voltages, other factors such as the biasing voltages or temperature should also be considered to get a better SI(j~/I 2 versus frequency variation with biasing voltages. A comparison of the SI(f) and S i ( f ) / I 2 versus frequency curves at 77 K with those at 300 K gave:
The noise spectra of S~(f) and Si(f)/12 versus frequency both indicate that there is generationrecombination activity, possibly due to trapping levels
2
3
4
at the first barrier heterointerface. It therefore seems that at 300 K the noise from the higher excess currents and the resonant tunnelling currents were dominating the noise from the trapping levels. This conclusion is also supported by the d.c. I - V curves at 300 and 77 K shown in Figure 2. The noise at 77 K does not show the same dependence on current as it does at 300 K. However, S~(f)/I 2 versus frequency curves seemed still to be grouped into two sets, as at 300 K. The first set has a spread of = 5 dB over the frequency range 2 - 2 0 0 Hz, while the second set has a larger spread of = 10 dB. S ] ( f ) / l 2 versus frequency for the 0.37 and 0.4 V biases in the NDR region have different curvatures from the other curves at biases outside the NDR region. The noise spectra of S j ( f ) / l 2 versus frequency at the first two biases of 0.15 and 0.2 V are approximately the same at both temperatures.
To complete the comparisons with the 300 K results, Figures 6c and d show the 77 K results of Si(f) and S~(f)/l e versus bias voltages. The shapes of these two curves are also very different from those at 300 K. While the noise at the low bias voltages is similar, S](f) at the other biases increases with bias voltage by almost 15 to 20 dB. Comparing the Sl(f)/I 2 versus bias voltages at
Cryogenics 1 9 9 0 Vol 3 0 December
1143
AIAs-GaAs-AIAs
q u a n t u m well diodes: X . M . Li
et al.
-110
-80
-120
-90
tiio
-130
-100
< m
&&IVv~''O + ÷ i i &k VV v~ee e Q I 0 ÷+ Ai-Vv V • I + 0 0 ÷÷ ÷ ÷ ~ ~& &&--V V V VI 0 e e ~&+V^
-140
"
±
-150 .
=&A +
~°o ~A-00
+++
u? -160
++++ ili
~999
+++Ai.
9~.
÷& A~A^O 0 +÷i&_ ~i~ + ~A O0 ÷+~& +++ AA 0 0 ÷+A + A A O0 +A A ++++ ~A~:O000 ÷÷:i +++ &AAAO000 ÷ ~
e~
-110
CO
-120
++++++_ ~ER~ *+++++++¥~
-170
i i
~÷--
~a^t
V;
i.
&_
I÷
-VvOi
-130
-180
-140 10
2
100
a
1000
10000
2
10
b
F r e q u e n c y (Hz) -110 [
1000
10000
F r e q u e n c y (Hz) -80
- 120
~+/
-13o
+
~ + ~ + ~A-----------LX
-90
/+/-
/
/
-110 ~
o
v co-
i 0.25
C
~
~
o
A
i 0.35 BIAS
,jf
+
-t20 -130
~o.__._o~
-180 0.15
~ - - ~ - o ~
^ ~
-100"
~- -15o ~ ~ + ~ _÷// Z ' / ~ .o " v ~ ~- -o~- - - - - - - -16o ~ . . / / ~ - 17 0
100
i 0.45
I 0.55
VOLTAGE
i 0.65
-140 0.15
0.75
d
(V)
I
I
i
i
i
0.25
0.35
0.45
0.55
0.65
BIAS
VOLTAGE
0.75
(V)
Figure 6 (a) S l ( f ) v e r s u s frequency at different bias voltages below, in and above the NDR region and at 77 K. Note the increase in S=(f) with bias voltage. + , 0 . 1 5 V ; A , 0 . 2 5 V ; O, 0 . 3 0 V ; + , 0 . 3 7 V; A , 0 . 4 0 V ; o , 0 . 6 0 V ; v , 0 . 7 0 V . (b) Sl(f)/I 2 v e r s u s frequency at 77 K. Symbols as in (a). (c) S t ( f ) v e r s u s bias voltage at several frequencies between 2 Hz and 10 kHz and at 77 K. +, 2 Hz; zx, 10 Hz, 0 , 20 Hz, + , 100 Hz; • , 2 0 0 Hz; • , 1 kHz; v , 2 kHz; 0 , 10 kHz. (d) Se(f)/I 2 v e r s u s bias voltages at several frequencies between 2 Hz and 1 0 k H z at 7 7 K . +, 2 H z ; zx, 1 0 H z ; o , 2 0 H z ; + , 1 0 0 H z ; • , 2 0 0 H z ; • , 1 kHz; +, 2 kHz; A , 1 0 k H z
300 K with those at 77 K showed even more substantial differences. For example, there is a larger dip in Si(f)/l 2 before the peak current followed by a large increase in Si(f)]I 2 after the peak current. Table 2 summarizes the biasing voltages and currents, the d.c. and a.c. resistances, and a in the 1/F dependence of S~(f) at 77 K. Within the frequency range 2 - 2 0 0 Hz, c~ increases with biasing voltage before the peak voltage from 1.2 to 1.4, but is approximately 0.9 for biases after the peak voltage. These variations in o~ indicate that, when tunnelling currents are dominant, Table 2 at 77 K
Biasing conditions of the QWD for noise measurements
Bias Bias D.c. voltage current resistance (V) (mA) (fD (Vd/Id)
> 1, but that, when the excess currents (mostly thermionic) are dominant, c~ < 1. These values for ~x are similar to those in References 9 and 10, but are different from that in Reference 11, where c¢ was found to be = 2. To add further credence to the observations stated above, noted below are the main differences between the 300 K QWD characteristics and its 77 K characteristics: 1 reduced excess currents due to the lower temperature; 2 slightly reduced parasitic resistances in series with the QWD; 3 an improved peak-to-valley current ratio, but a slightly reduced valley-to-peak voltage; and 4 significantly reduced thermionic currents for bias voltages beyond the NDR.
~ in RD(dVd/dld)
Sdf)
(~)
(f: 2 - - 2 0 0 Hz)
~ 1/f ~
12.83 11.03 11.69 -10.40 -22.99 31.04 24.04
1.17 1.18 1.40 0.89 0.91 0.89 0.85
Conclusions 0.15 0.25 0.30 0.37 0.40 0.60 0.70
1144
4.93 13.65 18.17 18.45 16.68 13.58 17.21
31.03 18.24 16.53 20.05 23.85 44.22 40.66
Cryogenics 1990 Vol 30 December
Detailed experimental results have been presented on low frequency noise measurements on A 1 A s - G a A s - A I A s quantum well diodes. The noise was shown to be basically of the form l / f and to increase with biasing voltage. At 300 K, S l ( f ) / l 2 v e r s u s frequency was found to be almost constant, indicating that the noise current spectral density
A l A s - G a A s - A l A s quantum well diodes: X.M. Li et al.
was p r o p o r t i o n a l to 12 t h r o u g h the d i o d e . H o w e v e r , at 77 K, the b e h a v i o u r o f Sl(f) and S l ( f ) / I 2 versus c u r r e n t o r bias v o l t a g e was m o r e c o m p l e x , e v e n t h o u g h the n o i s e
at biases below the peak current was almost the same at both temperatures. At 77 K, because of the significantly reduced excess currents, the noise spectra indicated that there was g e n e r a t i o n - recombination activity due to traps located at the first barrier heterointerface. The measurement showed that when thermionic currents are dominant there is a weak dependence of S~(f)/l 2 on RD, with the lower RDs having slightly higher noise levels. At both 300 and 77 K, the measured noise spectrum was = 1/f or flicker noise.
3 4 5
6 7
8
Acknowledgements T h e a u t h o r s a r e p l e a s e d to e x p r e s s t h e i r a p p r e c i a t i o n to Bell N o r t h e r n R e s e a r c h ( B N R ) , O t t a w a f o r g r o w i n g the w a f e r , and to the A d v a n c e d D e v i c e s G r o u p at the C o m m u n i c a t i o n s R e s e a r c h C e n t e r ( C R C ) , O t t a w a for w a f e r p r o c e s s i n g . T h e y w o u l d e s p e c i a l l y like to t h a n k W a y n e C o y n e , Elizabeth Bala and D o n D a v i d s o n o f C R C for their g e n e r o u s a s s i s t a n c e in f a b r i c a t i o n o f the q u a n t u m w e l l devices.
9 10 11 12
References 1 Sollner, T.C.L.G., Tannenwald, P.E., Peck, D.D. and Goodhue, W.D. Quantum well oscillators Appl Phys Lett (1984) 45 1319- 1321 2 Sollner, T.C.L.G., Brown, E.R., Goodhue, W.D. and Le, H.Q. Observation of millimeter-wave oscillations from resonant tunneling
13 14
diodes and some theoretical considerations of ultimate frequency limits Appl Phvs Lett (1987) 50 332-334 Whittaker, J.F., Morou, G.A., Sollner, T.C.L.G. and Goodhue, W.D. Picosecond switching time measurement of a resonant tunneling diode Appl Phys Lett (1988) 53 385-387 Stapleton, S.P., Deen, M.J., Berolo, E. and Hardy, R.H.S. Experimental study of the microwave reflection gain of AIAs/GaAs/AIAs quantum well structures, Electron Lett (1990) 26 84-85 Brown, E.R., Goodhuen, W.D. and Sollner, T.C.L.G. Fundamental oscillations up to 200 GHz in resonant tunneling diodes and new estimates of their maximum oscillation frequency from stationary state tunneling theory J Appl Phys (1988) 64 1519-1529 Luryi, S. Quantum capacitance devices Appl Phys Len (1988) 52 501 - 503 Capasso, F., Mohammed, K. and Cho, A.Y. Resonant tunneling through double barriers: perpendicular quantum transport phenomena in superlattices, and their device applications IEEE J Quant Electron (1986) QE-22 1853 Capasso, F., Sen, S., Beltram, F., Lunardi, L.M., Vengurlekar, A.S., Smith, P.R., Shah, N.J., Malik, R.J. and Cho, A.Y. Quantum functional devices: resonant-tunneling transistors, circuits with reduced complexity, and multiple valued logic IEEE Trans Electron Dev (1989) ED-36 2065-2082 Weichold, M.H., Villareal, S.S. and Lux, R.A. Low frequency noise measurements on AIGaAs/GaAs resonant tunnel diodes Appl Phy Lett (1989) 55 1969-1971 Weichold, M.H., Viilareal, S.S. and Lux, R.A. Analysis of defectassisted tunneling based on low frequency noise measurements of resonant tunnel diodes Sol St Electron (1989) 32 155l-1555 Okada, Y., Xu, J., Liu, H.C., Landheer, D., Buchanan, M. and Houghton, D.C. Noise characteristics of a Si/Ge resonant tunnel diode Sol St Electron (1989) 32 797-800 Bosman, G. and Zijlstra, R.J.J. Generation-recombination noise in p-type silicon, Sol St Electron (1982) 25 273-280 Van Rheenen, A.D., Bosman, G. and Van Vliet, C.M. Decomposition generation-recombination noise spectra in separate Lorentzians Sol St Electron (1985) 28 457-463 Van Rheenen, A.D., Bosman, G. and Zijlstra, R.J.J. Low frequency noise measurements as a tool to analyze deep-level impurities in semiconductor devices Solid St Electron (1987) 30 259-265
Cryogenics 1990 Vol 30 December
1145
Keywords: low temperature electronics; quantum well diodes; resonant tunnelling devices
In recent years, resonant tunnelling devices made with I I I - V compounds have attracted considerable attention 1-8 because of their interesting physics and their potential for both optical and electronic applications. This interest is largely because of the fast charge transport through the heterostructures and the potential for both high speed analogue and digital applications. The possibility for high speed applications is based principally on the negative differential resistance (NDR) region of these devices. As an example, the NDR provides the basis for high frequency oscillators, mixers, negative resistance amplifiers or detectors 1-6. In addition to the twoterminal resonant tunnelling structures, several threeterminal devices using resonant tunnelling are currently being studied 6-8 for various digital applications. The structure of a single quantum well resonant tunnelling diode (RTD)o generally consists of a layer of undoped GaAs (< 100 A*) sandwiched between two barrier layers of a larger band gap semiconductor material such as AlxGa~_~As. In this double-barrier structure, carriers can tunnel through the larger band gap material into the quantum well and out through the second barrier. When the energy (corresponding to the d.c. bias) of the tunnelling carriers matches closely a bound state energy level within the well, there is a peak in the transmission probability that is measured by a peak in the current of the I - V curve. This peak in the current occurs when the d.c. bias voltage is approximately twice the equilibrium bound state energy within the quantum well. Further "1 ~ = 1 0
I nm
0011 -2275/90/121140- 06 © Butterworth-Heinemann Ltd
1140
Cryogenics 1990 Vol 30 December
increase in the bias voltage results in a drop in the tunnelling current, giving rise to the negative differential resistance region, until thermionic currents over the barrier dominate the tunnelling currents through the barriers, and the current begins to increase again with applied voltage. Most of the early researches on resonant tunnelling diodes were with AlxGal_xAs barriers. However, more recently, the attention of researchers has been focused on studying new materials with increased barrier heights such as AlAs for the GaAs-based quantum wells and also because AlAs has the highest tunnelling barrier height for F-valley electrons. With the increased barrier heights, excess currents are reduced and the peak-to-valley current ratios are enhanced. New materials are also being studied for improving the peak current densities of RTDs. Most previous electrical studies on quantum well diodes were on their d.c. I - V characteristics, or their high frequency properties. There have been very few studies 9-tl on their low frequency noise characteristics despite the fact that low frequency measurements provide a powerful tool for studying non-idealities such as defects, trapping levels and differences in heterojunction qualities at the various interfaces. Two previous studies 9'~° on AIGaAs/ GaAs/A1GaAs diodes concentrated on analysing defect assisted tunnelling and on determining the activation energies and capture cross sections of the trapping centres at a fixed bias voltage. The other studyHon Si/Ge RTDs just reported on experimental results. In this paper, a different approach is taken. Noise measurements in AIAs/GaAs/A1As quantum well devices were measured. These devices promise better peak-to-
AlAs-GaAs-AIAs
and the Si-doped GaAs layers have ohmic contacts for
valley current ratios and higher current densities than AIxGa~__,As/GaAs/Al.~Ga~_xAs diodes. The differences in their noise current spectral densities at 300 and 77 K are studied as a function of the biasing voltages. The analysis considers all noise contributions from the circuit components external to the RTD, so that the noise of the quantum well diode can be determined. The range of values of o~ in the 1/f~ noise dependence is also of interest. The measurements at 77 K also allowed a study of the effects of reduced excess currents and increased tunnelling currents on the noise characteristics.
electrical measurements.
Device and experimental details The devices used in this study were grown by molecular beam epitaxy (MBE) and their structure is shown schematically in Figure 1. These quantum wells were mesa isolated structures of area 10 x 10 #m 2. The two barrier layers are 20 ,X, AlAs (7 monolayers) and the quantum well layer is 45 ,~ GaAs (16 monolayers). Above the first barrier is a 15 ,& spacer layer of undoped GaAs and then a layer of 0.5 /~m n ~ GaAs. Below the second barrier layer is a thick 140 ,~ spacer layer of undoped GaAs and then a 1.5/~m layer of n ~ GaAs. The n + GaAs substrate
500A GaAs (Si-doped) 0 . 5 u m n+ G a A s 15 A G a A s ( U n d o p e d ) 20 A AlAs 45 A GaAs 20 A AlAs 140 A GaAs (Undoped) 1.5um n+ G a A s
Figure 1 Cross-sectional v i e w of a single q u a n t u m well tunnelling diode structure
0.35 . i III 0.30
& g~
/ 0.25
.
::'.
II
/
1 30
........V ( ....
0.20
rr
o.lo
./
•
/W
/
,//
/--/"
/"
.7
/-'/ O.00 ~-~" , 0.00 0.10
- -10
-3o
0.05
-50 ,
0.20
,
0.30
r i 0.40
,
,
0.50
0.60
T Ov LU © Z <
'
JR2
-
Ac coo.,.d
Signal
=
W rr E < Z >-
1
I
cO
A
P"
I/ ,lr~: I / 'il~i~ l/ /'~J
Typical c u r r e n t - voltage curves of one of the quantum well diodes (QWD) at 300 and 77 K are shown in Figure 2. These measurements were made with an automated HP4145A Semiconductor Parameter Analyzer (SPA). Also shown in Figure 2 are the differential resistances and bias points at which the noise measurements were made. The peak current density at 300 K was 2.14 x 104A cm : and at 77 K was 2.24 × 104 A c m .2, which are similar to those reported in Reference 11 for similar RTDs. The peak and valley voltages at 300 K were 0.324 and 0.422 V and at 77 K were 0.356 and 0.44 V, respectively. The current peak-to-valley ratios were 1.45 at 300 K and 2.25 at 77 K. The test devices show significant negative resistance at both 77 and 300 K and this negative resistance makes it difficult to bias the QWD by a simple current source. Because of the NDR region, the bias voltage is not a unique function of the bias current. However, the bias current is a unique function of the bias voltage, and so a voltage source is required to bias the QWD. For the biasing circuit, a simple voltage divider network is used to provide a constant voltage to the QWD. To avoid introducing any complex noise spectrum from the bias circuit, two metal film resistors are used for the simple voltage divider. The noise from the bias circuit was subtracted from the measured noise level to determine the true QWD noise. To accomplish this, two measurements were taken at each bias point, one with the device and another without the device. Using the two sets of data, the noise of the bias circuit can then be subtracted to give the QWD noise. Figure 3 shows the experimental test setup for the noise measurements. The gain of the ultra-low noise preamplifier (PAR Model 133) was set at 100. Its noise level is below - 160 dB, which is the thermal noise ground level of the system. The dynamic signal analyser (HP3561A) is a.c. coupled to the ultra-low noise amplifier during the measurements. The signal analyser has a dynamic range of 80 dB and it can measure signals down to - 130 dB (V 2 Hz ~). Combining the low noise pre-amplifier with the signal analyser, the system can measure noise levels down to - 1 6 0 dB (V 2 Hz-~). All measurements were made from the noise spectrum at each biasing voltage. Figures 4a and b show the equivalent noise circuits for the test set-up with and without the QWD, respectively. The quantum well diode is considered as a current noise
/ / f 50
I%I
:
•
0
1 70
I
quantum well diodes: X.M. Li et al.
i :-~±~ •
QWD
i
[
Analyzer i L
(3
-70 0.70
IEEE488
Metal Box
GPIB
i v
BIAS VOLTAGE (VOLTS) Figure 2 Measured d.c. current - voltage characteristic of a device w i t h area 100/.tin 2 at 3 0 0 and 77 K. Also s h o w n is the differential conductance at both temperatures and the biasing voltages used in the noise measurements• - , 3 0 0 K; - - - , 77 K; 3 0 0 K; • • . , 77 K
Plotter
-
Personal Computer
Figure 3 Experimental test set-up used to measure the l o w frequency noise spectrum
Cryogenics 1 9 9 0 Vol 30 December
1141
AIAs-GaAs-AIAs
quantum well diodes: X.M. Li et al.
Experimental results and discussions
Figures 5a and b show the measured Si(f) and Si(f)]l 2 versus frequency for seven bias voltages at 300 K. As
(a)
(b)
Figure 4 Noise equivalent circuit of measurement system (a) with and (b) without the quantum well diode
source Si(f) (in A z Hz -x) in parallel with its differential resistance Ro. The other resistors are treated as voltage noise spectrum sources (white noise sources). For a resistor of value R, the voltage noise spectrum is (4kTR) ~/2. The battery was modelled as a voltage noise source Sv(f)b in series with its internal resistance Rb. Since Rb "~ R2, its noise contribution can justifiably be ignored. Using these assumptions, the system voltage noise spectrum Sv(f)s and the thermal noise from resistors can be subtracted from the measured noise spectrum Sv(f), which is measured when the QWD is in the test system. The noise spectra measurements without the device are denoted as Sv(f)o. A detailed derivation of the noise subtraction formula is given in Reference 12. Using Figure 4a Sv(f) can be determined from
Svq)
-
Sv~0s
=
(RIR2RD)2[ 4kT(11
12) + Sl(f)l + (RDRl)2Sv(f)b
(RIR2 +
R2RD +
RdRI) 2
(1) Similarly, the noise spectrum without the QWD can be determined from Figure 4b as
Sv(f)o - Sv(f)s = 4kT
R l R2
R1 + R2
+ Sv(f)b
R~ (R1 + R2)2
(2)
Since the system noise spectrum Sv(f)s ( - 1 6 0 dB) is well below the noise level of the bias circuit and QWD, it can be neglected in Equations (1) and (2). Subtracting the noise of the battery in the above equations, the current noise spectrum of the QWD is then obtained as
expected, an increase in the bias voltages results in an increase in Sl(f). Figure 5b shows that the Si(f)/I z versus frequency curves seem to be grouped into two sets. The first set consists of measurements at a bias voltage of up to 0.35 V, i.e. up to the end of the NDR region, and the second set at biases of 0.45 V and higher, above the NDR region. The differences in Si(f)/l 2 between the two sets was = 4 dB for frequencies between 2 and 700 Hz. Also, the Sl(f)/l 2 values for the two lowest bias voltage measurements at 0.15 and 0.2 V indicate that there might be generation-recombination activity due to trapping levels, possibly located at the heterojunction interface of the first barrier layer because of the 'Lorentzian' nature ~2-x4 of their noise spectra. However, the other bias voltages do not clearly indicate the presence of traps, even out to 100 kHz. From Figure 5b, the noise level is related to the magnitude of the current level, and the noise due to the higher currents might be dominating noise contributions from trapping levels that are observed at the lower bias currents. This is supported by the fact that the noise levels are higher than those reported in References 9 and 10, in which trapping levels were measured for a bias voltage just beyond the NDR region. Using the test set-up, the resistors are carefully chosen in the bias voltage divider to eliminate any spontaneous oscillation in the NDR so that the diode can be biased in the negative differential region. The experiment shows that the smaller the R~, the better the stability of the bias voltage. However, it could be a heavy load on the battery. The current noise spectrum at the bias voltage VB in the NDR (0.35 V) was 1/f 1°~ over the entire frequency range. This frequency dependence of S~(f) and others at different biases are summarized in Table 1. They are in agreement with the noise spectra results reported in References 9 and 10. Also included in Table 1 are the biasing voltages and currents, and the d.c. and a.c. resistances. The value of ot in 1/F was 1.15 for biases above the valley current and = 1.08 for biases below the peak current. To examine the 300 K results from a different perspective, in Figures 5c and d, S1(f) and Si(f)/l 2 a r e plotted versus bias voltages at several frequencies. The shapes of these two curves are very similar and they indicate that for a bias voltage in the NDR region there is a dip in the noise level. The noise at the bias voltages just after the NDR region was 2 - 4 dB higher over the entire frequency range. Also, the differences between S~(f) (and also Table 1
Biasing conditions of the QWD for noise measurements
at 3 0 0 K Bias Bias D.c. voltage current resistance (V) (mA) (~2) (Vd/Id)
(RIR2 + R2Ro +
RDR1)2Sv(f) -
(R1 + R2)2R2Sv(f)o
(R1 R2 RD)2
(3) In the experiments, the dynamic resistance R D of the diode is calculated from the static I - Vcharacteristic curve.
1142
Cryogenics 1990 Vol 30 December
0.15 0.20 0.30 0.35 0.45 0.60 0.65
7.97 11.71 20. 19 18.66 15. 30 23.34 27.84
18.91 17. 10 15.02 18.66 29.42 25.73 23.47
RD(dVd/dld) (~2) 14.21 12.48 14.0 -30.76 35.42 13.06 10.61
c~ in Sl(f)1/f ~ (f: 2-- 200 Hz) 1.08 1.10 1.05 1.01 1.15 1.15 1.16
AlAs-GaAs-AlAs -110
-80
-120
-90
-130 < 03
quantum well diodes: X.M. Li et al.
-100
-14o -110 -150 -120
*+~9{¢^'*,;'lli~.
-160
"'**:'";:III
-170
-130
-180
-140 10
a
100 Frequency
1000
10000
2
10
-110
-80
-120
-9O
-130
1000
10000
Frequency (Hz)
b
(Hz)
1O0
~4
~.~+/
-100
rn
-14o -110'
CO
d-
-16o:
C
-120
h k _~ •
~
~"
:
-
__A--A ~ - A
-L~-~.
~ o
-130
--170' -180 O. 15
--O ~ ._.-.-.-4-
-15o;
i 0.25
i 0.35
t 0.45
L 0.55
I 0.65
-140 0.15
0.75
BIAS VOLTAGE (V)
d
i
i
i
i
i
0.25
0.35
0.45
0.55
0.65
0.75
BIAS VOLTAGE (V)
Figure 5 (a) & ( f ) v e r s u s f r e q u e n c y at different bias voltages below, in and above the N D R region and at 3 0 0 K. Note the increase in Sdf) with bias voltage. + , 0 . 1 5 V ; zx,0.20V; O,0.30V; +,0.35V; 0,0.45 V; • , 0 . 6 0 V ; v,0.65V.(b) S d f ) / / 2 v e r s u s frequency. +,0.15V; zx,0.20V; O,0.30V; + , 0 . 3 5 V; v , 0 . 4 5 V ; 0,0.60V; [ ] , 0 . 6 5 V. (c) & ( f ) versus bias voltage at several frequencies b e t w e e n 2 H z a n d 10 kHz and at 3 0 0 K. + , 2 . 0 Hz; i x , 1 0 H z ; O , 2 0 H z ; + , 1 0 0 Hz; • , 2 0 0 H z ; •, 1.0kHz; +,2.0kHz; ix, 10kHz. (d) Sl(f)/I 2 v e r s u s bias voltage at several frequencies b e t w e e n 2 Hz and 10 kHz and at 3 0 0 K. Symbols as in (c)
Sl(f)/I 2)
at 2 Hz and at 10 kHz increase with bias voltages, and this increase in Si(f) also supports our earlier conjecture that the noise level is related to the current through the diode. Figures 6a and b shows the measured Si(f) and S[(f)/I 2 versus frequency for seven bias voltages at 77 K. The deviation from 1/f noise behaviour suggests there are generation-recombination ( g - r) centres in the device. In the frequency region measured, there is only one turning point (one Lorentzian component) in the spectrum for each bias voltage. This means that only one g - r centre was measured at each bias voltage for our QWD 12'13. Plotting S1(f)/l ~ versus frequency was also tried to determine the best value for current dependent coefficient/3. However, the narrowest range of Si(f)/I ~ was found with/3 = 2. Because of the almost 20 dB range in S [ ( f ) / l 2 versus frequency curves at the different bias voltages, other factors such as the biasing voltages or temperature should also be considered to get a better SI(j~/I 2 versus frequency variation with biasing voltages. A comparison of the SI(f) and S i ( f ) / I 2 versus frequency curves at 77 K with those at 300 K gave:
The noise spectra of S~(f) and Si(f)/12 versus frequency both indicate that there is generationrecombination activity, possibly due to trapping levels
2
3
4
at the first barrier heterointerface. It therefore seems that at 300 K the noise from the higher excess currents and the resonant tunnelling currents were dominating the noise from the trapping levels. This conclusion is also supported by the d.c. I - V curves at 300 and 77 K shown in Figure 2. The noise at 77 K does not show the same dependence on current as it does at 300 K. However, S~(f)/I 2 versus frequency curves seemed still to be grouped into two sets, as at 300 K. The first set has a spread of = 5 dB over the frequency range 2 - 2 0 0 Hz, while the second set has a larger spread of = 10 dB. S ] ( f ) / l 2 versus frequency for the 0.37 and 0.4 V biases in the NDR region have different curvatures from the other curves at biases outside the NDR region. The noise spectra of S j ( f ) / l 2 versus frequency at the first two biases of 0.15 and 0.2 V are approximately the same at both temperatures.
To complete the comparisons with the 300 K results, Figures 6c and d show the 77 K results of Si(f) and S~(f)/l e versus bias voltages. The shapes of these two curves are also very different from those at 300 K. While the noise at the low bias voltages is similar, S](f) at the other biases increases with bias voltage by almost 15 to 20 dB. Comparing the Sl(f)/I 2 versus bias voltages at
Cryogenics 1 9 9 0 Vol 3 0 December
1143
AIAs-GaAs-AIAs
q u a n t u m well diodes: X . M . Li
et al.
-110
-80
-120
-90
tiio
-130
-100
< m
&&IVv~''O + ÷ i i &k VV v~ee e Q I 0 ÷+ Ai-Vv V • I + 0 0 ÷÷ ÷ ÷ ~ ~& &&--V V V VI 0 e e ~&+V^
-140
"
±
-150 .
=&A +
~°o ~A-00
+++
u? -160
++++ ili
~999
+++Ai.
9~.
÷& A~A^O 0 +÷i&_ ~i~ + ~A O0 ÷+~& +++ AA 0 0 ÷+A + A A O0 +A A ++++ ~A~:O000 ÷÷:i +++ &AAAO000 ÷ ~
e~
-110
CO
-120
++++++_ ~ER~ *+++++++¥~
-170
i i
~÷--
~a^t
V;
i.
&_
I÷
-VvOi
-130
-180
-140 10
2
100
a
1000
10000
2
10
b
F r e q u e n c y (Hz) -110 [
1000
10000
F r e q u e n c y (Hz) -80
- 120
~+/
-13o
+
~ + ~ + ~A-----------LX
-90
/+/-
/
/
-110 ~
o
v co-
i 0.25
C
~
~
o
A
i 0.35 BIAS
,jf
+
-t20 -130
~o.__._o~
-180 0.15
~ - - ~ - o ~
^ ~
-100"
~- -15o ~ ~ + ~ _÷// Z ' / ~ .o " v ~ ~- -o~- - - - - - - -16o ~ . . / / ~ - 17 0
100
i 0.45
I 0.55
VOLTAGE
i 0.65
-140 0.15
0.75
d
(V)
I
I
i
i
i
0.25
0.35
0.45
0.55
0.65
BIAS
VOLTAGE
0.75
(V)
Figure 6 (a) S l ( f ) v e r s u s frequency at different bias voltages below, in and above the NDR region and at 77 K. Note the increase in S=(f) with bias voltage. + , 0 . 1 5 V ; A , 0 . 2 5 V ; O, 0 . 3 0 V ; + , 0 . 3 7 V; A , 0 . 4 0 V ; o , 0 . 6 0 V ; v , 0 . 7 0 V . (b) Sl(f)/I 2 v e r s u s frequency at 77 K. Symbols as in (a). (c) S t ( f ) v e r s u s bias voltage at several frequencies between 2 Hz and 10 kHz and at 77 K. +, 2 Hz; zx, 10 Hz, 0 , 20 Hz, + , 100 Hz; • , 2 0 0 Hz; • , 1 kHz; v , 2 kHz; 0 , 10 kHz. (d) Se(f)/I 2 v e r s u s bias voltages at several frequencies between 2 Hz and 1 0 k H z at 7 7 K . +, 2 H z ; zx, 1 0 H z ; o , 2 0 H z ; + , 1 0 0 H z ; • , 2 0 0 H z ; • , 1 kHz; +, 2 kHz; A , 1 0 k H z
300 K with those at 77 K showed even more substantial differences. For example, there is a larger dip in Si(f)/l 2 before the peak current followed by a large increase in Si(f)]I 2 after the peak current. Table 2 summarizes the biasing voltages and currents, the d.c. and a.c. resistances, and a in the 1/F dependence of S~(f) at 77 K. Within the frequency range 2 - 2 0 0 Hz, c~ increases with biasing voltage before the peak voltage from 1.2 to 1.4, but is approximately 0.9 for biases after the peak voltage. These variations in o~ indicate that, when tunnelling currents are dominant, Table 2 at 77 K
Biasing conditions of the QWD for noise measurements
Bias Bias D.c. voltage current resistance (V) (mA) (fD (Vd/Id)
> 1, but that, when the excess currents (mostly thermionic) are dominant, c~ < 1. These values for ~x are similar to those in References 9 and 10, but are different from that in Reference 11, where c¢ was found to be = 2. To add further credence to the observations stated above, noted below are the main differences between the 300 K QWD characteristics and its 77 K characteristics: 1 reduced excess currents due to the lower temperature; 2 slightly reduced parasitic resistances in series with the QWD; 3 an improved peak-to-valley current ratio, but a slightly reduced valley-to-peak voltage; and 4 significantly reduced thermionic currents for bias voltages beyond the NDR.
~ in RD(dVd/dld)
Sdf)
(~)
(f: 2 - - 2 0 0 Hz)
~ 1/f ~
12.83 11.03 11.69 -10.40 -22.99 31.04 24.04
1.17 1.18 1.40 0.89 0.91 0.89 0.85
Conclusions 0.15 0.25 0.30 0.37 0.40 0.60 0.70
1144
4.93 13.65 18.17 18.45 16.68 13.58 17.21
31.03 18.24 16.53 20.05 23.85 44.22 40.66
Cryogenics 1990 Vol 30 December
Detailed experimental results have been presented on low frequency noise measurements on A 1 A s - G a A s - A I A s quantum well diodes. The noise was shown to be basically of the form l / f and to increase with biasing voltage. At 300 K, S l ( f ) / l 2 v e r s u s frequency was found to be almost constant, indicating that the noise current spectral density
A l A s - G a A s - A l A s quantum well diodes: X.M. Li et al.
was p r o p o r t i o n a l to 12 t h r o u g h the d i o d e . H o w e v e r , at 77 K, the b e h a v i o u r o f Sl(f) and S l ( f ) / I 2 versus c u r r e n t o r bias v o l t a g e was m o r e c o m p l e x , e v e n t h o u g h the n o i s e
at biases below the peak current was almost the same at both temperatures. At 77 K, because of the significantly reduced excess currents, the noise spectra indicated that there was g e n e r a t i o n - recombination activity due to traps located at the first barrier heterointerface. The measurement showed that when thermionic currents are dominant there is a weak dependence of S~(f)/l 2 on RD, with the lower RDs having slightly higher noise levels. At both 300 and 77 K, the measured noise spectrum was = 1/f or flicker noise.
3 4 5
6 7
8
Acknowledgements T h e a u t h o r s a r e p l e a s e d to e x p r e s s t h e i r a p p r e c i a t i o n to Bell N o r t h e r n R e s e a r c h ( B N R ) , O t t a w a f o r g r o w i n g the w a f e r , and to the A d v a n c e d D e v i c e s G r o u p at the C o m m u n i c a t i o n s R e s e a r c h C e n t e r ( C R C ) , O t t a w a for w a f e r p r o c e s s i n g . T h e y w o u l d e s p e c i a l l y like to t h a n k W a y n e C o y n e , Elizabeth Bala and D o n D a v i d s o n o f C R C for their g e n e r o u s a s s i s t a n c e in f a b r i c a t i o n o f the q u a n t u m w e l l devices.
9 10 11 12
References 1 Sollner, T.C.L.G., Tannenwald, P.E., Peck, D.D. and Goodhue, W.D. Quantum well oscillators Appl Phys Lett (1984) 45 1319- 1321 2 Sollner, T.C.L.G., Brown, E.R., Goodhue, W.D. and Le, H.Q. Observation of millimeter-wave oscillations from resonant tunneling
13 14
diodes and some theoretical considerations of ultimate frequency limits Appl Phvs Lett (1987) 50 332-334 Whittaker, J.F., Morou, G.A., Sollner, T.C.L.G. and Goodhue, W.D. Picosecond switching time measurement of a resonant tunneling diode Appl Phys Lett (1988) 53 385-387 Stapleton, S.P., Deen, M.J., Berolo, E. and Hardy, R.H.S. Experimental study of the microwave reflection gain of AIAs/GaAs/AIAs quantum well structures, Electron Lett (1990) 26 84-85 Brown, E.R., Goodhuen, W.D. and Sollner, T.C.L.G. Fundamental oscillations up to 200 GHz in resonant tunneling diodes and new estimates of their maximum oscillation frequency from stationary state tunneling theory J Appl Phys (1988) 64 1519-1529 Luryi, S. Quantum capacitance devices Appl Phys Len (1988) 52 501 - 503 Capasso, F., Mohammed, K. and Cho, A.Y. Resonant tunneling through double barriers: perpendicular quantum transport phenomena in superlattices, and their device applications IEEE J Quant Electron (1986) QE-22 1853 Capasso, F., Sen, S., Beltram, F., Lunardi, L.M., Vengurlekar, A.S., Smith, P.R., Shah, N.J., Malik, R.J. and Cho, A.Y. Quantum functional devices: resonant-tunneling transistors, circuits with reduced complexity, and multiple valued logic IEEE Trans Electron Dev (1989) ED-36 2065-2082 Weichold, M.H., Villareal, S.S. and Lux, R.A. Low frequency noise measurements on AIGaAs/GaAs resonant tunnel diodes Appl Phy Lett (1989) 55 1969-1971 Weichold, M.H., Viilareal, S.S. and Lux, R.A. Analysis of defectassisted tunneling based on low frequency noise measurements of resonant tunnel diodes Sol St Electron (1989) 32 155l-1555 Okada, Y., Xu, J., Liu, H.C., Landheer, D., Buchanan, M. and Houghton, D.C. Noise characteristics of a Si/Ge resonant tunnel diode Sol St Electron (1989) 32 797-800 Bosman, G. and Zijlstra, R.J.J. Generation-recombination noise in p-type silicon, Sol St Electron (1982) 25 273-280 Van Rheenen, A.D., Bosman, G. and Van Vliet, C.M. Decomposition generation-recombination noise spectra in separate Lorentzians Sol St Electron (1985) 28 457-463 Van Rheenen, A.D., Bosman, G. and Zijlstra, R.J.J. Low frequency noise measurements as a tool to analyze deep-level impurities in semiconductor devices Solid St Electron (1987) 30 259-265
Cryogenics 1990 Vol 30 December
1145