Modulation spectroscopy characterization of semiconductor microstructures and semiconductor devices

~-

JOURNALOF

=

ELSEVIER

LUMINESCENCE Journal of Luminescence 60&61 (1994) 287-292

Invited paper

Modulation spectroscopy characterization of semiconductor microstructures and semiconductor devices H. Shen1 US Army Research Laboratory, Electronics and Power Source Directorate, AMSRL-EP-EF, Fort Monmouth, NJ 07703, USA

Abstract

Applications of modulation spectroscopy for characterization of semiconductor microstructures and devices are reviewed. Particular emphasis is given to non-destructive, contactiess techniques such as photorefiectance. Semiconductor microstructures and devices have been studied and characterized by a variety of spectroscopic methods such as photoluminescence (PL), cathodoluminescence, photoluminescence excitation spectroscopy (PLE), absorption spectroscopy, photocurrent spectroscopy, spectral ellipsometry, resonant Raman scattering, etc. Each of these techniques provides specific information about the material of interest. For applied work, an optical characterization technique should be as simple, inexpensive, compact, rapid, and informative as possible. Other important issues are the ability to perform measurements non-invasively, at (or above) room temperature, and on whole wafers. Because ofits simplicity and proven ability, modulation spectroscopy is now a major tool for the charactenization of semiconductor microstructure and devices [1—3].Furthermore, the potential for in situ monitoring/control of semiconductor fabrication condition during actual growth and processing has been demonstrated using modulation spectroscopy. ‘Also at: GEO Centers. Inc. Lake 1-lopatcong, NJ 07849, liSA.

Modulation spectroscopy is an analog method for taking the derivative of the optical spectrum (reflectance or transmittance) of a material by penodically perturbing the measurement conditions [1—3].This can be accomplished by varying some property of the sample or the measurement system in a periodic fashion and measuring the corresponding normalized change in the reflectance (transmittance). The observed normalized changes are typically small so that the difference signal is closely related to a derivative of the absolute spectrum with respect to the modifying parameter. The derivative nature of modulation spectroscopy emphasizes structure localized in the photon energy region of interband (intersubband) transitions of semiconductors and suppresses uninteresting background effects. In addition weak features that may not have been detected in the absolute spectra are often enhanced. Owing to the derivative-like nature a large number of sharp spectral features can be observed, even at room temperature. Shown in Fig. I is a schematic drawing of the experimental arrangement for a general modulated

0022-2313/94/507.00 © 1994 — Elsevier Science B.V. All rights reserved SSDI 0022-2313(94)E0496-K

288

H.5/u’,, Journal at Luminescence 60&61

(

1904) 287 292

LAM P I

F,

MONOCHNOMATOR

x -~

0

SAMPLE__________ ~ ....—

PMT DETECTOR

HV

~JMODULATION ~1SOURCE

L~

_4:~I-I

0R+I0~Rcos

_____

Wmt)

SERVOL~ SIGNAL HVf AC SIGNAL REFERENCE (Qm) LOCKIN tiR/R

>

i

Fig. I. Schematic representation of a modulation experimental apparatus.

reflectance experiment. Light from an appropriate light source passes through a monochromator, and is focussed onto the sample. The modulation (electric field, stress, and temperature) is applied to the sample at frequency ~ The reflected light is focussed onto an appropriate detector. The light striking the detector contains two signals: the DC (or average value) is proportional to the DC reflectance R of the material, while the modulated value (at frequency Qm) is proportional to the change in reflectance ~R, produced by the modulation source. The normalized signal AR/R gives a denyative-like spectrum. For characterization purposes the most useful form of modulation is electromodulation (EM), because it provides the shar-

pest structure and is sensitive to surface and interface electric field. The most widely used contactiess mode of EM is termed photoreflectance (PR). The EM can be accomplished either by contact and contactiess modes. In the case of contact modes, the modulation is achieved by applying a bias across the sample. Contactless EM can be performed using PR, contactless electroreflectance (CER) or electron-beam electroreflectance (EBER). In PR [1—3],modulation is caused by photoinjected electron—hole pairs which modulate the built-in electric field of the semiconductor or semiconductor microstructure under investigation. The pump (laser or other light source), with a photon energy above the band gap of the

H. Shen / Journal of Luminescence 60&61 (1994) 287—292

289

semiconductor, is chopped at frequency Qm’ Most PR experiments have utilized a mechanical chopper with maximum Qm = 5 kHz. To achieve modulation frequencies up to 1 MHz, an acoustooptic modulator can be used. PR is not only contactless but requires no special mounting of the sample. CER utilizes a condenser-like system consisting of a thin, transparent, conductive coating (indium—tin oxide or 50—60 A of a metal such as Au or Ni on a transparent substrate (glass, quartz, etc.) which serves as one electrode [4]. A second electrode consisting of a metal strip is separated from the first electrode by insulating spacers. The sample (‘-.. 0.5 mm thick) is placed between these two capacitor plates. There is no pump beam required

with resonances appearing wherever an integral number of de Broglie wavelength fits into the tnangular well formed by the electric field. The period of the FKOs can be used to measure the built-in electric field. There is a considerable amount of experimental work in semiconductor microstructure utilizing modulation spectroscopy. In the space allotted, it is impossible to review all of it. Instead examples of spectra for different line shapes will be provided. Fig. 2 shows the 300 K PR spectra of Glembocki et al. [7] from a thick Al~Gai_~Asexpitaxial layer on GaAs [Fig. 2(a)] and a series of GaAs/ AI~Ga~ _~Asmultiple quantum wells [Fig. 2(b)—(d)] with different well widths (L2). Even at 300 K the

in CER. In EBER the pump beam is a modulated low energy electron beam (‘-.- 200 eV) chopped at about 1 kHz [5]. However, the sample and electron gun must be placed in an ultrahigh vacuum chamber. Other conventional forms of modulation include temperature and stress [1—3]. In temperature modulation, thermoreflectance (TR), the sample may be mounted on a small electric heater while, in stress modulation, peizoreflectance (PzR), it is on a piezoelectric transducer. Although TR and PzR are contactless they require special mounting of the sample. One of the strengths of modulation spectroscopy is the ability to fit the observed data to known line shapes. While it is difficult to calculate a full reflectance spectrum, this is not the case for its derivative because it is localized. Although EM is the most popular modulation form, it has the most complex line shape in modulation spectroscopy, because the electric field can destroy the translational symmetry of the material. Under low electric field, it accelerates unbound electrons and/or holes, producing a sharp, third-derivative line shape [1—3,6]. However, for bound states such as excitons, impurities, and quantum wells the perturbing field does not accelerate electrons and/or holes, but alters the binding energy of the particle (Stark effect) and varies the intensity of the transition and the lifetime of the state, leading to a first derivative [1—3,6]. Under sufficiently large field, the EM spectrum can display an oscillatory behavior above the band gap called Franz—Keldysh oscillations (F’KOs) [1—3,6],

spectra are extremely rich. The notation h~and l~ represents “symmetry” allowed transitions of index n between the quantized conduction and valence subbands of heavy (h)- or light(l)-hole character. In fact all the “symmetry” allowed transitions were observed. From a fit to the first-derivative line shape, it is possible to yield accurate values of transition energies to within a few meV. These transition energies can be used to determine the Al composition x and the quantum well width L~. The work of Glembocki et al. is particularly significant since it demonstrated that intersubband transitions in undoped multiple quantum wells could be observed at room temperature using the contactiess EM method of PR. Since then a large variety of III—V, Il—VI, and SiGe systems have been studied including both lattice-matched and strained layer configuration. Information about well and barrier width, band offset, strain, coupling, etc. have been derived from the sharp derivative spectrum [1—3]. Shown in Fig. 3 are the typical FKOs. This is a 300 K PR spectra [8,9] from an undoped GaAs layer of thickness of 1000 A on a buried doped n’ (~~) buffer. Such configurations have been designated UN~ (UP~) or SIN~ (SIP~). In the n + (p ~)buffer/substrate the Fermi level occurs near the conduction (valence) band edge. At the surface the Fermi level is pinned at some value. Therefore, there exists in the undoped region a large, almost constant electric field F. To deduce the electric field from the FKO, 312 we plotted in the inset the quantity as a function of index n, where (4/3it) (E~ E0) —

H. Ntwtl .lour*lal ol l,tonitlevccm'e 6t)&6l

29(1

i

,

1994, 27¢7 292

i

7-300. --- AIGaAs

O.O

8.01

0

3

6

9

12

cY

"~--hl

~ ~.

Lz =150A

%

A~GaAs

0.0

-8.0 o

500K

..]hJ

i

t.4

-1

GaAs h2

h3 Fig. 3. Photoreflectance spectrum of the SIN* s a m p b of at room temperature. The inset shows a plot of 14.3r~)(E,, E,,I ~ ' as a function of FKOs index n.

L z = 240A

--~

GaAs

[2

~4

A~GaAs r--]

cr21c~

0

/

-!

t

t

-2

3

h5

ha

h2

h2

Lz = 4.60A

Ii

l

|

ff~GaAs I

1 ~- GaAs

°:h2

-1

-2 i

i

1.4

1.8

Energy (eV)

-2

cd

I

1.6

1.5 PHOTON

1.6 ENERGY

1.7 (eV)

Fig. 2. Room temperature photoreflectance spectra for an undoped GaAs,'Ga~ ~A1.As heterojunction [top trace (a)] and three multiple quantum well samples with x z 0.2 [(b) (d)].

E, is the photon energy of the nth extremum, Eo is the energy gap. It can be proven that such a plot gives a straight line with a slope of el~F/(21d ~2, where/t is the reduced interband mass. Therefore.

the electric field F can be obtained directly from the period of FKO. The large number of FKOs from U N ~ : U P ' (SIN+/SIP +) make it possible to use PR and CER as an effective probe of surface/interface states. These structures have been used to study the Fermi leveJ pinning at GaAs and AIGaAs surface [9] and to study the effect of various processing procedures such as dry etching, surface passivation and metallization [9, 10]. Recently, modified SIN ~ structures have been used to study low-temperature grown GaAs [11, I2] and the piezoelectric effect in ( I I [ ) grown lnGaAs/GaAs system [13]. FKOs have also been used to determine the doping concentration in the doping range of 1015 1017cm -3. In this case the FKOs are from the non-uniform field in the space charge region (SCR) of the semiconductor. Under small modulation, the field determined from FKOs is the maximum (surface) field [14] rather than the average lield in the SCR. The carrier concentralion can be obtained from the surface electric field if surface Fermi level is known. Furthermore, the phase of CER can be used to determine the carrier type. Despite its success in the study of many semiconductor microstructures, modulation studies of twodimensional electron gases (2DEG) in GaAs/ GaA1As modulation doped heterojunctions

H. Shen / Journal of Luminescence 60&ó1 (1994) 287—292

(MDHJ)/ high electron mobility transistors (HEMTs) have been the subject of considerable controversy. This is because the observed signals in the spectral vicinity of the direct band gap of GaAs could originate in either the 2DEG or in other GaAs portions of the sample. The problem has been averted by studying the GaAlAs/InGaAs/ GaAs system since in this case the 2DEG signature will be associated with the InGaAs section of the sample, which is spectrally separated from the GaAs/GaAlAs signal. Displayed by solid lines in Fig. 4 are PR spectra of Yin et al. originating in the InGaAs region of a HEMT structure. The line shapes are unusual for modulation spectroscopy, which generally exhibit sharp, derivativelike features (i.e., positive and negative lobes) associated with excitons (see Fig. 2). The traces of Fig. 4 lie on only one side of the baseline. This is due to the screening effect produced by the dense 2DEG, which wipe out these strong excitonic resonances. In such a system the spectrum can be explained on [15]

__________________________

I I

Photoreflectance

/

300 K

/ Sampie

U) ~

/ /

A~ B~ Sample

0

4

1.4

cance of this result is that PRE is a room temperature probe, in contrast to PLE. Other extensions of PR include differential

~

~

0 ~

~

.

~

~amp1e

..~.

~

.

-~

/ j

~.

~

—

A’~ ~ Expt. FD2DJDS taneshape Fit FDL Lineshape Fit I

1....

I

1.3

the basis of the derivative of a broadened step-like two-dimensional density of states and a Fermi level filling factor. From the details of the line shape fit it has been possible to extract the Fermi level position and hence the 2DEG density. By comparing the experimental energies with a theoretical self-consistent Schrodinger—Poisson calculation, they are able to evaluate other important material parameters, such as L~,In composition, and built-in electric field. Other device characterization includes structures such as heterojunction bipolar transistors (HBTs) [16], quantum well lasers, quantum well infrared detectors [17], superlattice optical mirrors [18], resonant tunneling structures [19], solar cells, etc. These studies, the vast majority of which were done at room temperature, have demonstrated the potential of this technique in gaining important device information from the structures. For example, doping profiles as well as built-in electric field have been derived from FKO in HBTs and solar cells. The fundamental transition detected by PR from the QW laser was found within 15—30 A of the lasing wavelength. An interesting extension of PR is the photoreflectance excitation spectroscopy (PRE) [20]. Since the modulation is produced by photoexcited carriers, the intensity of a PR feature, just as in PLE, will depend upon the absorption coefficient of the modulating beam. Strictly speaking, it is proportional to the photovoltage. Plotted in Fig. 5 is the 300 K PRE spectrum from GaAs/AlAs multiple quantum well. In the experiment the probe photon energy was turned to the first heavy hole-to-electron transition, while the photon energy of the pump beam derived from a Ti: Sapphire laser was scanned. Heavy-, light-hole and forbidden transitions have been observed. Also plotted in Fig. 5 by dotted line are 4 K PLE result. The signifi-

~

(~‘\~\

Al

~

D~

~

r./~

291

Energy (eV) Fig. 4. Solid lines are the experimental photoreflectance spectra from the InGaAs region of three GaAIAs/tnGaAs/GaAs MEMT structures. The dashed and solid lines are least-squares fits. The obtained values of the intersubband energies are designated by arrows,

photorefiectance [21] and time resolved photoreflectance [22]. The former uses two pump beams at different wavelengths to perform non-destructive .

depth profiling. The latter uses a digital oscilloscope to detect the transient development of the PR signal and hence the electric field.

ft. ,S'hen Journal o/ Luminescence 60&61 (1994 i .56

1.58

1.60

1.62

1.64

287 292

References

0.4 i

/K

PLE

/ 03 .,~ IlH

0.2

IlL

01

0.0

~%~'/300K PRE

\ / ~

1.46

1.48

i

,

i

1.50

1.52

1.54

1.56

Energy (eV) Fig 5. 300 K PRF spectrum of a 100/~ GaAs/AIAs M Q W compared to 4 K PLE spectrum for the same sample.

This paper has briefly reviewed the use of modulation spectroscopy for the characterization of semiconductor microstructures. Most of the modulation experiments, during the past decade, either used the sharp derivative feature to determine various quantum transitions or used FKOs to determine the built-in electric field. Device structure characterization has held particular interest. Some speculations about future trends are: tl)study of low-dimensional systems, such as quantum wires, boxes, dots: (2) use of FKOs in modified U N + / U P + (SIN+/SIP +) to study surface/interface states and processing induced damage: (3)continued work on actual devices: (4)advances in instrumentation to improve data acquisition time for in situ monitoring.

Acknowledgemem The author is grateful to Fred H. Pollak for his continuing encouragement and support.

[1] V.H. Pollak, in: Encyclopedia of Materials Characteriza tion: Surfaces, Interfaces and Thin Films, eds. C. Evans, R. Brundle and S. Wilson (Butterworth Heinemann, Boston. 1992~ p. 385. [2] O.J. Glembocki and B.V. Shanabrook, in: Semiconductors and Semimetals, Vol. 67, eds. D.G. Seller and C.L. Littler (Academic Press, New York, 1992) p. 222. [3] F.H, Pollak and H. Shen, J. Electron. Mater. 19 ( 19901 399. [4] X Yin and F . H Pollak, Appl. P h y s Lett. 59 (1991)2305. [5] M . H Herman, Proc. Soc. Photo-Optical Instrum. Engineers ISPIE, Bellingham, 1990) 1286 [1990j 39. [6] D.E. Aspnes. in: Handbook on Semiconductors, Vol. 2, ed. M. Balkanski INorth-Holland, New York. 19801 p. 109. [7] O.J. Glembocki, B.Y. Shanabrook, N. Bottka, W.T. Beard and J. Comas, Appl. Phys. Lett. 46 11985) 970. [8] X. Yin, H.-M. Chen, F.It. Pollak, Y. Chan, P.A. Montano, P D . Kirchner, G.D. Pettit and LM. Woodall, Appl. Phys. I,ett. 58 (19911 260. [9] H. Shen, M. Duttm L. k'otiadis, P.G. Newmam R.P. Moerkirk, W.H. Chang and R.N. Sacks, Appl. Phys. Len. 57 (1990) 2118. [10] O.J. Glembocki, J.A. Dagata, E.A. Snow and D.S. Katzer, Appl. Sur[ Sci. 63 11993i 143. I l l ] H. Shen, F.C. Rong, R Lux, J. Pamulapati, M. Dutta, M. Taysing-Lara, E.H. Poindexter, L. Calderon and Y. Lu, Appl. Phys. Lett. 61 11992) 1585. [12] A.C. Warren, J.M. Woodall, P.D. Kirchner, X. Yin, X. Guo. F Pollak and M.R. Melloch, J. Vac Sci Technol. B 10 (19921 1904. [13] H. Shen, M. Durra, W. Chang, R. Moerkik, D.M. Kim, K.W. Chung, P.P. Ruden, M.I. Nathan and M.A. Stroscio, Appl. Phys. Lett. 60 (1992) 2400. [14] H. Shen and F.H. Pollak, Phys. Rev. B 42 (1990) 7097. [15] Y. Yin, H. Qiang, E.H. Pollak, D.C. Streit and M. Wojtowicz, Appl. Phys. Lett. 61 (1992) 1579. [16] X. Yin, F.H. Pollak, L. Pawlowicz, T.J. O'Neill and M. Hafizi, Appl. Phys. Lett. 56 i1990) 1278. [17] P.A. Dafesh, J. Appl. Phys. 71 11992t5154. [18] l.J. Fritz, P.L, Gourley a n d T . J . Drummond, Appl. Phys. Lett. 55 [1989) 1324. [ 19] R.L. Tober, J. Pamulapati, J.E. Oh and P.K. Bhattacharya, J. Electron Mater. 18 (19891 379. [20] H. Shen, X.C. Shen, F . H Pollak and R . N Sacks, Phys. Rev. B 35 (19871 3487. [21] M. Sydor, A. Badakhshan, J.R. Engholm and D.A. l)ale, Appl. Phys. Lett. 58 (1991) 948. [22] H. Shen, M. Dutta, R. Lux, W. Buchwald, L.Eotiadis and R.N. Sacks, Appl. Phys. Lett. 59 (1991} 321.