6
Electrical characterization of the Au/InP(100) and Au/InSb/InP(100) structures q B. Akkal a, * , Z. Benamara a , L. Bideux b , B. Gruzza b a Laboratoire de Micro-e ´lectronique Applique ´e, Universite ´ Djillali Liabe `s, 22000, Sidi Bel Abbe `s, Algeria b Laboratoire des Sciences des Mate ´riaux pour l’Electronique et d’Automatique, Universite ´ Blaise Pascal de Clermont II, Les Ce ´zeaux, 63177 Aubie `re, France Accepted 5 January 1999 Abstract In this work, we measure the current–voltage (I V) and the capacitance–voltage (CV) characteristics of the Au/InP(100) and Au/InSb/ InP(100) Schottky type diodes. The InP(n) substrate is restructured by some monolayers of the InSb thin film. We then propose a study of the electrical quality of the elaborated components after the Au/InP interface creation; first without annealing and then after annealing by heating at 5008C temperatures. Analysis of the measured I(V) characteristics for the Au/InP and Au/InSb/InP samples allows the determination of the electrical parameter variations. The saturation current I s , the serial resistance R s , the mean ideality factor n and also the barrier height f Bn , are respectively equal to 2.10 × 10 24 A, 19 V, 1.8 and 0.401 eV for the Au/InP sample and equal to 1.34 × 10 27 A, 175 V, 1.78 and 0.592 eV for the Au/heated InSb/InP. Another good result is that the analysis and simulation of the I(V) and the C(V) characteristics allows us to determine the very important mean interfacial state density N ss(mean) , and is obtained to be equal 4.23 × 10 12 eV 21 cm 22 for the Au/InP sample and equal to 4.42 × 10 12 eV 21 cm 22 for the Au/heated InSb/InP. This work thus permits the evolution study of these electrical parameters related to the restructuration conditions. q 1999 Elsevier Science Ltd. All rights reserved. Keywords: Interfacial state density; Current–voltage; Capacitance–voltage 1. Introduction To resolve the controversy between the role of the inter- face states and the deep level bulk traps, with regard to the electrical characteristics of Au/InSb/InP Schottky diodes, a detailed study of the current–voltage (I V) and capaci- tance–voltage (C V) characteristics in these diodes was undertaken. The forward I V and CV characteristics are analysed to estimate the energy density distribution of the interface states in the band gap. There are several techniques for studying deep donor traps in semiconductors [1,2]. The capacitance energy level spectroscopy of Robert and Crowell [1] is based on the solution of truncated Poisson’s equation to yield capacitance of different sections, within the space charge region of a Schottky barrier diode containing multiple deep donor traps in semiconductors. Only a few studies on metal-semiconductor contacts with the 100 InP orientation have been carried out. This is because of the major difficulty in preparing an iterative process on this kind of surfaces. We propose a study of the electrical quality of elaborated components after Au/ InP(100) interface creation without any annealing and after a heating to a temperature of about 5008C, in order to compare these results with those of an Au/InSb/InP restructured surface. 2. Experimental details The InP(100) substrates used were n-type (Sn doped) wafers at different doping levels (10 15 –10 17 atoms cm 23 ). They were chemically cleaned according to a method based on successive baths of H 2 SO 4 , a solution of 3% bromine in methanol and deionized water [3], and were introduced into an ultra high vacuum (UHV) chamber at a pressure of 10 29 10 210 torr. Before depositing the gold gates in the Schottky diode elaboration process, the InP surface was restructured by an InSb film. This was realised after metallic indium detection on the InP substrate using an Auger electron spectrometer of the considered surface that was obtained after ionic bombardment and cleaning. The restructuration is thus made by depositing antimony reacted with the metallic Microelectronics Journal 30 (1999) 673–678 Microelectronics Journal MEJ 581 0026-2692/99/$ - see front matter q 1999 Elsevier Science Ltd. All rights reserved. PII: S0026-2692(99)00009-9 q Journe ´es Maghreb-Europe: Les Materiaux et Leurs Applications aux Dispositifs Capteurs Physiques, Chimiques et Biologiques (MADICA’98). * Corresponding author. Fax: 213-7-56-30-68.

Electrical characterization of the Au/InP(100) and Au/InSb/InP(100) structures

  • Upload
    b-akkal

  • View
    225

  • Download
    3

Embed Size (px)

Citation preview

Page 1: Electrical characterization of the Au/InP(100) and Au/InSb/InP(100) structures

Electrical characterization of the Au/InP(100) andAu/InSb/InP(100) structuresq

B. Akkala,* , Z. Benamaraa, L. Bideuxb, B. Gruzzab

aLaboratoire de Micro-e´lectronique Applique´e, UniversiteDjillali Liabes, 22000, Sidi Bel Abbe`s, AlgeriabLaboratoire des Sciences des Mate´riaux pour l’Electronique et d’Automatique, Universite´ Blaise Pascal de Clermont II, Les Ce´zeaux, 63177 Aubie`re, France

Accepted 5 January 1999

Abstract

In this work, we measure the current–voltage (I–V) and the capacitance–voltage (C–V) characteristics of the Au/InP(100) and Au/InSb/InP(100) Schottky type diodes. The InP(n) substrate is restructured by some monolayers of the InSb thin film. We then propose a study of theelectrical quality of the elaborated components after the Au/InP interface creation; first without annealing and then after annealing by heatingat 5008C temperatures. Analysis of the measuredI(V) characteristics for the Au/InP and Au/InSb/InP samples allows the determination of theelectrical parameter variations. The saturation currentIs, the serial resistanceRs, the mean ideality factorn and also the barrier heightfBn, arerespectively equal to 2.10× 1024 A, 19V, 1.8 and 0.401 eV for the Au/InP sample and equal to 1.34× 1027 A, 175V, 1.78 and 0.592 eV forthe Au/heated InSb/InP. Another good result is that the analysis and simulation of theI(V) and theC(V) characteristics allows us to determinethe very important mean interfacial state densityNss(mean), and is obtained to be equal 4.23× 1012 eV21 cm22 for the Au/InP sample and equalto 4.42× 1012 eV21 cm22 for the Au/heated InSb/InP. This work thus permits the evolution study of these electrical parameters related to therestructuration conditions.q 1999 Elsevier Science Ltd. All rights reserved.

Keywords:Interfacial state density; Current–voltage; Capacitance–voltage

1. Introduction

To resolve the controversy between the role of the inter-face states and the deep level bulk traps, with regard to theelectrical characteristics of Au/InSb/InP Schottky diodes, adetailed study of the current–voltage (I–V) and capaci-tance–voltage (C–V) characteristics in these diodes wasundertaken. The forwardI–V and C–V characteristics areanalysed to estimate the energy density distribution of theinterface states in the band gap.

There are several techniques for studying deep donortraps in semiconductors [1,2]. The capacitance energylevel spectroscopy of Robert and Crowell [1] is based onthe solution of truncated Poisson’s equation to yieldcapacitance of different sections, within the space chargeregion of a Schottky barrier diode containing multipledeep donor traps in semiconductors.

Only a few studies on metal-semiconductor contacts withthe 100 InP orientation have been carried out. This isbecause of the major difficulty in preparing an iterative

process on this kind of surfaces. We propose a study ofthe electrical quality of elaborated components after Au/InP(100) interface creation without any annealing andafter a heating to a temperature of about 5008C, in orderto compare these results with those of an Au/InSb/InPrestructured surface.

2. Experimental details

The InP(100) substrates used were n-type (Sn doped)wafers at different doping levels (1015–1017 atoms cm23).They were chemically cleaned according to a method basedon successive baths of H2SO4, a solution of 3% bromine inmethanol and deionized water [3], and were introduced intoan ultra high vacuum (UHV) chamber at a pressure of 1029–10210 torr.

Before depositing the gold gates in the Schottky diodeelaboration process, the InP surface was restructured by anInSb film. This was realised after metallic indium detectionon the InP substrate using an Auger electron spectrometer ofthe considered surface that was obtained after ionicbombardment and cleaning. The restructuration is thusmade by depositing antimony reacted with the metallic

Microelectronics Journal 30 (1999) 673–678

MicroelectronicsJournal

MEJ 581

0026-2692/99/$ - see front matterq 1999 Elsevier Science Ltd. All rights reserved.PII: S0026-2692(99)00009-9

q Journees Maghreb-Europe: Les Materiaux et Leurs Applications auxDispositifs Capteurs Physiques, Chimiques et Biologiques (MADICA’98).

* Corresponding author. Fax: 213-7-56-30-68.

Page 2: Electrical characterization of the Au/InP(100) and Au/InSb/InP(100) structures

indium. This operation gives an InSb layer at the devicesurface. The InSb layer evaluated by the Auger electronspectroscopy method and by an EPES (elastic peak electronspectroscopy) method [4]), presented some monolayers andpermitted blockage of the indium surface migration [5].

To realise a metallic gate, we have used a mask in molyb-denum. This mask permits us to obtain electrical measure-ments with gold gates of 1 mm diameter and a thick layer ofabout 1000 A˚ . Zones with and without deposits are created.Consequently the homogeneity study of these deposits andtheir cleanliness is possible.

The I–V measurements are made through a standard set-up (two electrometers model 616), and theC–V ones arerealised by a capacity–voltage set-up (Princeton AppliedResearch) with a constant 1 MHz frequency.

The sample is maintained by a pumping system on acopper support which makes an the electrical contact.

3. Computational model

For a Schottky diode, the thermionic emission current isgiven by [6].

I � Is�1 2 e2qV=kT�eqV=nkT; �1�

wheren, Is are the ideality factor and the saturation currentdensity expressed by

Is � A*T2 exp 2q

kTfBn

� �; �2�

whereT, A*, fBn are the temperature in Kelvin, the Richard-son constant and the barrier height, respectively.

The voltage drop across the rectifying barrierV is definedby

V � VG 2 RsI ; �3�whereVG, Rs are the applied voltage and the series resist-ance, respectively.

For an MIS structure with interface states entirelygoverned by the semiconductor, the expression for thedensity of the interface states as deduced by Card andRhoderick [7] reduces to

Nss� 1q

ei

d�n 2 1�2

es

w

� �; �4�

wheree i, e s andd are the interfacial layer permittivity, thesemiconductor permittivity, the thickness of the interfaciallayer, respectively, andw is the depletion zone width whichis deduced from the experimentalC–V curves.

The energy of the interface states,Ess relative to theconduction band edge,Ec, at the semiconductor surface, isgiven by [8]:

Ec 2 Ess� q�fBn 2 V�: �5�For an intimate metal–semiconductor (MS) contact, the

doping densityNd and the diffusion potentialVd can bedirectly obtained from the slope and the intercept on theVaxis of a high frequencyC22 vs. V plot.

Fonash [9] has shown that when the semiconductorentirely governs the populations of the interface states andthe frequency is sufficiently high, these states cannot followthe ac signal, the slope of theC22 vs. V curve is given by

dC22

dV� 2

2qes

1Nd�1 1 a�

� �; �6�

whereCsc is the depletion zone capacitance, and the par-ametera is given by

a � �qNssd�=ei : �7�According to relation (6), the slope ofC22 vs. V is

constant if the interfacial states densityNss is constant.Otherwise ifNss varies at the interface, the slope does notremain constant and also varies.

If d andNd are sufficiently small, the extrapolated inter-cept,V0 of the high frequencyC22 vs. theV curve is givenby

V0 � �Vd 2 kT=q��1 1 a�: �8�The dependence of the capacitance of a Schottky diode

upon frequency can also arise because of the presence ofdeep donor traps in the depletion region. Goodman [10] hasdeveloped the theory of theC–V characteristics of aSchottky barrier for an n-type semiconductor with multipledeep donor traps in the semiconductor. Their theory predictsthe occurrence of an infection point in the high frequencyC22–V curve, for a single level deep donor traps with

B. Akkal et al. / Microelectronics Journal 30 (1999) 673–678674

Fig. 1. The current–voltage characteristic of the Au/InP structure.

Page 3: Electrical characterization of the Au/InP(100) and Au/InSb/InP(100) structures

uniform density in the depletion region. The slope, dC22/dVat high frequencies is discontinuous at the point where thequasi-Fermi level of the semiconductor goes through thetrap level. Under this condition the energyEt of the deep

trap, relative to the conduction band edge of the semicon-ductor, is related to the diode bias voltage,Vi at the infectionpoint by

Et � q�fBn 2 Vi�; �9�whereVi represents the voltage at which the slope of dC22/dV vs. V changes.

The presence of deep donor traps modifies theC22–Vcharacteristic in which each linear segment corresponds toone deep donor trap.

4. Results and discussion

Fig. 1 shows the measuredI(VG) characteristics of an Au/InP(100) contact. We can see an important reverse currentrelated to the nature of the interface during the interfacialzone formation. A great quantity of defects is observed, andthis seems to be the principal cause of the leakage current.Therefore, the direct characteristics present a bad exponen-tial form.

The experimental values of ln[I/(1 2 e2qVG/kT)] vs. VG isplotted (Fig. 2). It is linear over the voltage range 0–0.15 mV. The best fits of the experimental values ofI–VG

in the linear region between 0 and 0.15 mV (Fig. 2) to thetheoretical Eq. (1) provided the values ofIs and the meanvalues of the ideality factorn which are equal to 2.108×1024 A and 1.8, respectively.

The calculated value ofIs was substituted in Eq. (2) todetermine the value offBn, which is equal to 0.401 eV.

At a forward voltage, considerably higher thanVG .0.15 V, the ln[I/(1 2 e2qVG/kT)] vs. VG curves becomestraight lines and thereby permit determination of the seriesresistanceRs from the slope of the curves [6]. The value ofRs obtained is equal to 19V.

Using Auger spectroscopy, we deduce that during thegold deposition, In and P atoms migrate to the surface.This process can be explained by simple modelling of thephenomena. First, an inter diffusion between Au, In and Patoms occurs and as soon as the quantity of In and Pmigrated atoms becomes constant, Au grows layer-by-layer and carries these In and P atoms to the surface ofthe sample. Fig. 3 represents several steps of this modelling.

Other experiments were performed on InP(100) samplespreviously heated at 5008C. The obtained characteristicseems to be of ohmic contact. This ohmicity is probablybecause of the metallic indium present at the interface. Infact, gold evaporation on a non-heated surface destroys thesubstrate In–P bonds; but, nevertheless, heating can hom-ogenise the InP surface by creating a metallic indium thinfilm.

In continuation of the study of the Au/InP structure theInP(100) surface is restructured by an InSb buffer layer fromindium crystallites, in order to avoid the migration of theindium and phosphorous atoms to the surface. The electricalstudy will hence be done on two samples, one on the Au/

B. Akkal et al. / Microelectronics Journal 30 (1999) 673–678 675

Fig. 2. The ln�I =1 2 e2qV=kT� variations vs. bias of the Au/InP structure.

Fig. 3. Modelling of the Au/InP(100) interaction.

Page 4: Electrical characterization of the Au/InP(100) and Au/InSb/InP(100) structures

non-heated InSb/InP and the other on the Au/heated InSb/InP structure.

Fig. 4 shows theI(VG) characteristic of the Au/non-heatedInSb/InP. This curve is of better quality owing certainly tothe metallic indium presence decreasing at the interface.

Despite several experiments on these restructurations atroom temperature, we have never obtained good and perfectSchottky contacts with a weak reverse current. An explana-tion can be given by the results of the photoemission studyon the antimony atom condensation.

The electrical parameters such as the saturation currentIs,the ideality factorn, the serial resistanceRs, and also thebarrier heightfBn are emphasised related to Au/InP directcontacts. This assertion is demonstrated in Table 1.

To be sure that indium crystallites have all disappearedwithout any antimony presence, and to decrease the anti-mony aggregate sizes and in order to affirm the III–V char-acter of the buffer layer, the antimony excess was eliminatedby heating the substrate at 5008C.

On photoemission, the Au/heated InSb/InP systempresents a great number of Sb–P bonds characterising a

surface phosphorous migration. Moreover, during heatingthe Sb–Sb bond proportions and the aggregate sizes alsodecreased. This InSb thin layer appears to be more stablebecause to observe significant interaction it needs more goldatoms than the non-heated InSb structure.

Therefore, the valence band analysis shows that the5008C heated InSb surface presents an accented and well-marked III–V character. This is confirmed by the electricalmeasures in Fig. 5 as well as by the quality of the non-heatedsurface.

The measured characteristics of the capacitanceC(VG) at1 MHz vs. bias of Au/InP and Au/ heated InSb /InP areshown in Fig. 6.

For Au/InP the capacitance effect appears only for reversebias greater than 2 V, but increases quickly and reaches270 pF near zero bias.

The fluctuation of the ideality factorn related to the biasin the I(VG) measured characteristics for Au/InP and Au/heated InSb/InP can be evaluated by substituting the satura-tion currentsIs and the coordinates ofI(VG) characteristics inEq. (1).

We then evaluatedNssas illustrated in Fig. 7 vs. the (Ec 2Ess) function.

These characteristics are obtained by replacing thenvalues of the correspondingI(VG) curve and the depletionwidth variation determined using the measuredC(VG) curve.Thus, we have takene i � 4.5× e0 [11] andd � 40 A. [12],

B. Akkal et al. / Microelectronics Journal 30 (1999) 673–678676

Fig. 4. The current–voltage characteristic of the Au/InSb/InP structure.

Table 1The experimental parameters obtained for Au/InP and Au/heated InSb/InP

Is (A) n Rs (V) fBn (eV)

Au/InP Non-heated 2.10× 1024 1.8 19 0.401Heated Ohmic Ohmic Ohmic Ohmic

Au/InSb/InP Non-heated 2.21× 1025 1.73 105 0.464Heated 1.34× 1027 1.78 175 0.592

Fig. 5. The current–voltage characteristic of the Au/heated InSb/InPstructure.

Page 5: Electrical characterization of the Au/InP(100) and Au/InSb/InP(100) structures

B. Akkal et al. / Microelectronics Journal 30 (1999) 673–678 677

Fig. 6. The capacitance–voltage characteristic of the Au/InP and Au/heated InSb/InP structure.

Fig. 7. The distribution of state density in the band gap of the Au/InP and Au/heated InSb/InP structure.

Page 6: Electrical characterization of the Au/InP(100) and Au/InSb/InP(100) structures

and have established the correspondence with the interfacialstates density energyEssand bias voltage using relation (5).

The mean interfacial state densities evaluated for Au/InPand Au/heated InSb/InP are found to be equal to 4.23×1012 eV21 cm22 and 4.42× 1012 eV21 cm22, respectively.Thea parameter estimation using Eq. (7) is equal to 0.68and 0.71 for Au/InP and Au/heated InSb/InP, respectively.This parametera expresses the voltage decrease because ofthe effect of the interfacial states density.

Fig. 8 shows theC22(VG) for Au/InP and Au/heated InSb/InP characteristics.

For Au/InP Schottky diode theC22(VG) is not linear,implying the presence of discreet deep donor levels.However, for Au/heated InSb/InP, theC22(VG) curves arealmost linear and do not present deep donor levels. In conse-quence, we can deduce that the doping concentration isuniform.

The slope of the first interpolation segment (1) givesNd

andVd. These two quantities seemed to be equal to 2.95×1015 cm23 and 0.34 V for Au/InP and equal to 7.36×1015 cm23 and 0.45 V for Au/heated InSb/InP.

We then analyse the two segments (2,3) for Au/InP givenby the Goldmman model [10] to determine the position ofthe two donor levels related to the conduction band edgeEc.

The second and the third donor levels for Au/InP arelocated at 0.24 eV and 0.578 eV relative to the conductionband edgeEc. They are ionised and do not respond to the acmodulating signal of 1 MHz frequency.

5. Conclusion

This article presents an electrical study of Au/InP(100)and Au/InSb/InP(100) systems. The electrical properties ofAu/InP are sensitive to interfacial reactions, such as diffu-sion of Au and migration of In and P atoms.

The Au/InP contact is of an ohmic type for a heatedsurface, and of a poor quality Schottky type with an impor-tant reverse current for a non-heated surface.

Our proposed solution would be the creation of an InSbbuffer layer that is well controlled by quantitative Augeranalysis. Heat treatment before Au deposition gave goodresults in electrical measurements.

The studied sample presents an important serial resist-ance caused by the two Au/InSb and InSb/InP contacts.This resistance is directly proportional to the interfaciallayer thickness and inversely proportional to the dopinglevel.

The C(VG) curve for Au/InP is controlled by two donorlevels located at 0.24 and 0.578 eV relative to the conduc-tion band edgeEc.

References

[1] G.I. Roberts, C.R. Crowell, J. Appl. Phys. 41 (1970) 1767.[2] I. Balberg, J. Appl. Phys. 58 (1985) 2603.[3] B. Gruzza, B. Achard, C. Pariset, Surf. Interface Anal. 162 (1985)

202.[4] E. Bauer, J. Vac. Sci. Technol. 7 (1970) 3.[5] L. Bideux, B. Gruzza, A. Porte, H. Robert, Surf. Interface Anal. 20

(1993) 803.[6] A. Singh, Solid State Electronics 26 (1983) 815.[7] H.C. Card, E.H. Rhoderik, J. Appl. Phys. 4 (1971) 1589.[8] C. Barret, A. Vapaille, Solid State Electronics 19 (1976) 73.[9] S. Fonash, J. Appl. Phys. 54 (1983) 1966.

[10] A.M. Goodman, Solid State Electronics 26 (1983) 815.[11] G. Eftekhari, J. Vac. Sci. Technol. B11 (4) (1993) 1317.[12] S.M. Sze, Physics of Semiconductor Devices, 2, Wiley, New York,

1981, p. 245.

B. Akkal et al. / Microelectronics Journal 30 (1999) 673–678678

Fig. 8. The 1/C 2 variations related to bias of the Au/InP and Au/heatedInSb/InP structure.