POSITION-SENSITIVE ATOM PROBE ANALYSIS OF MULTI-QUANTUM WELL STRUCTURES

HAL Id: jpa-00229972

https://hal.archives-ouvertes.fr/jpa-00229972

Submitted on 1 Jan 1989

HAL is a multi-disciplinary open access archive for the deposit and dissemination of sci- entific research documents, whether they are pub- lished or not. The documents may come from teaching and research institutions in France or abroad, or from public or private research centers.

L’archive ouverte pluridisciplinaire HAL, est destinée au dépôt et à la diffusion de documents scientifiques de niveau recherche, publiés ou non, émanant des établissements d’enseignement et de recherche français ou étrangers, des laboratoires publics ou privés.

POSITION-SENSITIVE ATOM PROBE ANALYSIS OF MULTI-QUANTUM WELL STRUCTURES

J. Liddle, A. Cerezo, C. Grovenor

To cite this version:

J. Liddle, A. Cerezo, C. Grovenor. POSITION-SENSITIVE ATOM PROBE ANALYSIS OF MULTI-

QUANTUM WELL STRUCTURES. Journal de Physique Colloques, 1989, 50 (C8), pp.C8-437-C8-

442. �10.1051/jphyscol:1989874�. �jpa-00229972�

COLLOQUE DE PHYSIQUE

Colloque C8, Suppl6ment au n0ll, Tome 50, novembre 1989

POSITION-SENSITIVE ATOM PROBE ANALYSIS OF MULTI-QUANTUM WELL STRUCTURES

J.A. LIDDLE, A. CEREZO and C.R.M. GROVENOR

Department of Metallurgy and Science of Materials, University of Oxford. Parks Road, GB-Oxford OX1 3PH, Great-Britain

ABSTRACT The Position-Sensitive Atom Probe (POSAP) has been used to examine a number of InP based quantum well structures in order to elucidate the chemistry and morphology of the interfaces between wells and barrier layers. The interfaces are of critical importance in determining the performance of devices manufactured from such structures. Results have been obtained from three sets of GaInAdInP wells; all the material was grown b y Metal-Organic Chemical Vapour Deposition (MOCVD). The information obtained in these observations will be useful in gaining a fuller understanding of the growth processes of these materials, and, because of good agreement with TEM and HREM, in refioing the models used to interpret data obtained with these techniques and with the standard assessment methods of X-ray diffraction and photoluminescence.

INTRODUCTION

Low dimensional semiconductor structures have a wide range of applications such as lasers, amplifiers and filters in optical communications systems. The confinement of electrons or holes in one or two dimensions can result in improved performance of devices, because of reduced scattering and quantum-mechanical confinement of the charge carriers [ 11. In multilayer structures, charge carriers can be spatially separated from dopant atoms, producing materials with very high mobilities for microwave frequency electronic devices. Electrons and holes can even be separated spatially from each other to manufacture photodiodes with extremely high gain and low noise characteristics 121.

Fundamental to the success of this technology is the accurate control of layer composition and of interface quality.

Interfaces may be described in terms of their morphological roughness and chemical diffuseness. The composition of the wells determines the emitted wavelengths, while the interfacial roughness influences the linewidth of the radiation.

A technique that can successfully distinguish between these two aspects can provide invaluable information on the growth of such multilayer structures, thus enabling processing conditions to be optimised. Previous experimental techniques used to examine these structures, primarily TEM and HREM, suffer from the inability to distinguish between morphological and chemical effects in most cases, because any information obtained is the result of integration through the thickness of a sample.

The atom probe, with its unparalleled spatial and chemical resolution, allows layer compositions to be determined accurately on the required scale (many of the devices mentioned above have active layers only -10nm thick), while the development of the Position-Sensitive Atom Probe (POSAP) [3] means that the morphology and chemistry of the interfaces can be examined separately with very high resolution. Despite the difficulties inherent in producing field-ion specimens from epitaxial layers only 300nm thick which consist of materials with very different polishing characteristics, both POSAP and Pulsed Laser Atom Probe (PLAP) [4] data have been obtained from a number of samples.

This paper reports some of the experimental results obtained from three different sets of quantum well samples (comprising GaInAs wells and InP barrier layers) grown under different conditions.

EXPERIMENTAL

Three sets of GaInAs4nP quantum well structures were examined in the present study, all grown by Metal- Organic Chemical Vapour Deposition (MOCVD). Samples were prepared by selectively etching the substrate material (InP) away from the multilayer stack (using an HCI etchant), and chemically polishing the free standing layer to produce a point (with a dilute H202:H2S04 solution). Details of the sample preparation technique have been described previously [ S ] .

Once suitable field-ion tips had been produced they were analysed using the POSAP. The samples were then examined by conventional PLAP, which, although lacking the spatial resolution of the POSAP, has much better mass resolution, and so allows more accurate determination of compositions,,

POSAP composition maps for Ga and As are displayed for the samples examined. The intensity of each pixel represents the concentration of a particular element, averaged over the area of the pixel and the depth of material analysed. The area of each pixel is approximately (O.lnm)2, and the depth o f material analysed is of the order of 4-6 unit cells (2-3nm). The concentration is calculated in avolume of = 0.02-0.03nrn3.

The position sensitive part of the detector can only record single ion events accurately [3]. The group V ions tend to evaporate as clusters of high mass-to-charge ratio, so they arrive later than the lighter group I11 ions. Group V atoms also tend to evaporate in a rather uneven fashion, so several ions arrive at the detector simultaneously, and again no positional information can be associated with them. In order to achieve the highest accuracy in the determination of the morphology of the wells, only those peaks which have no overlaps with those of another element are used. In the present work the P2+, P+, P4+, P5+,

AS^+

and

AS^^+

have not been used for this reason. A POSAP mass spectrum illustrating this problem is shown in Figure 1. The effects described here all combine to produce non-stoichiometric analyses when selected areas of the image are examined.

Article published online by EDP Sciences and available at http://dx.doi.org/10.1051/jphyscol:1989874

Cognizance should also be taken of the fact that, due to the very different compositions of the wells and bamers, there may be significant local magnification effects, both in field-ion and POSAP images. This could lead to the interface between regions of higher and lower local magnifications being made to appear more diffuse than is actually the case.

RESULTS

The first set of wells (sample 1) examined was grown in a conventional MOCVD reactor, to which a quartz wool baffle had been added in an attempt to promote complete mixing of the gases, and produce homogeneous layer thicknesses across the whole substrate. The GaInAs wells in this sample, in fact, vary in thickness between 3nm and 7nm, while the InP banier layers are

-

lOnm thick (well and barrier layer thicknesses were determined by TEM for all the samples).

Figure 2 shows a field-ion image, taken in argon, of the region analysed from this sample. The GaInAs layers appear as lines of bright spots on this picture. POSAP composition maps for Ga and As are shown in Figure 3. In Figure 3(a) it is clear that the Ga distribution is not at all uniform in this sample, and the overall concentration of Ga in the well is far below what would be expected from a region of pure GaInAs. The As distribution (Figure 3(b)) shows much the same set of features, but appears wider than the Ga distribution. Composition profiles for all four elements are displayed in Figure 3(c). It should be noted that in this and subsequent profiles, the totals of the group I11 and group V concentrations have been normalised to 50% to account for the artificially low group V levels, the reasons for which are described in the previous section. It is apparent from these profiles, that the composition of the wells is far from the nominal GaInAs, but instead corresponds, approximately, to that of a quaternary lattice-matched to InP. This result has been confirmed by conventional PLAP. The PLAP data are summarised in Table I.

The second sample (sample 2) was grown in the same reactor as the first set of wells, but with the quartz wool baffle absent. The GaInAs layers are approximately 9nm thick, while the InP barrier layers are slightly wider at about IOnm. Figure 4 shows the POSAP composition maps for Ga [4(a) and (c)]and As [4(b) and (d)]from the two sides of the same well. Compared with the first sample, the distributions of both Ga and As are much more uniform, and the interfaces between well and barrier are much more clearly defined. The corresponding composition profiles are shown in Figures 4(e) and (0. The level of phosphorus inside the well is much lower in this sample, not rising above 6.0%

(compared with 23.8% in the first sample). This has been confirmed by conventional PLAP and the results of this analysis are given in Table 11. It is possible to see on these profiles that the left hand interface is more abrupt than the right hand one. The As concentration also falls less rapidly than the Ga concentration, giving an As "tail" extending from the well into the barrier.

Figure 5 shows an argon field-ion image taken from the third sample (sample 3) in this study. This material was grown in a reactor of a different design that attempted to minimise the extent of any volumes of recirculating gas and to speed the passage of reactants through the reactor chamber. The GaInAs wells show up a s bright bands on either side of the image, and appear to be over half as wide as the InP region. The wells are approximately lOnm wide, while the barrier layers are 33nm across. POSAP composition maps for Ga [6(a) and (c)] and As [6(b) and (d)] from both interfaces are shown in Figure 6; the corresponding profiles are displayed in Figures 6(e) and 6(f). Figure 7 shows a profile across one well obtained by taking successive PLAP analyses at points spaced across the well. These data show that the left hand interface is relatively abrupt, while the right hand interface is slightly chemically diffuse and rougher morphologically. It is also clear that the concentration of phosphorus in this well is negligible.

DISCUSSION

The three sets of quantum wells examined in this study have one feature in common, the presence of an arsenic tail extending from the well into the bamer layer in one direction. It is also clear that the phosphorus content of the wells decreases markedly from the first set of the wells to the last.

The explanation of these observations lies in the nature of the MOCVD process. Gas streams for the different elements are switched in and out of the reactor to grow the different layers. It takes a finite time for the different reactive gases to pass out of the reactor. Since the substrate remains heated while the gases in the reactor are changed, material of compositions between those of InP and GaInAs is grown. It appears that the group 111 carrying gases (trimethyl- gallium and trimethyl-indium) pass through the reactor vessel rapidly, while the group V ones (arsine and phosphine) do so more slowly.

The distribution of the group V species is a result of two effects. In the growth of materials containing arsenic and phosphorus, the phosphine partial pressure must be an order of magnitude higher than that of the arsine for incorporation of equal amounts of these elements. This means there will be large amounts of phosphine trapped in the recirculating volume, giving a large quantity of phosphorus in the next GaInAs layer. In the first sample the quartz wool baffle clearly acts as a resenroir for phosphine, and in the second the presence of "dead air space" performs the same function, though not to such a high degree. In the case where the reactants pass quickly through the reactor, those species that are more easily incorporated into the growing layer will show up in subsequent layers. In the present work there is evidence for the existence of an arsenic tail on one side of the wells. This is consistent with the fact that arsenic is more easily incorporated into the growing layer than is phosphorus.

" Mass-to-charge ratio 175 Figure I. POSAP mass spectrum illustrating the problem of peak overlap.

WE,LL WELL WELL

+

Figure 2. Ar field-ion image of set of GaInAsJnP quantum wells. The corresponding POSAP data are displayed in Figure 3.

WELL WE.LL

1 b 50nm

Figure 5. Ar field-ion image of G a I n A d n P quantum wells. The corresponding POSAP and PLAP data are presented in Figures 6 and 7 respectively.

0 10

Distance across image (nm)

Figure 3. POSAP composition maps for Ga (a) and As (b) taken from sample 1 (quartz wool baffle present). The corresponding profiles for all four elements present are displayed (c).

0 12.5 0 12.5

Distance across image (nm) Distance across image (nm)

Figure 4. POSAP composition maps for Ga 4(a) and (c) and As 4(b) and (d) taken from both interfaces of a GaInAs well from sample 2 (quartz wool baffle absent). The corresponding composition profiles for all four elements present are displayed underneath each pair of composition maps 4(e) and

(0.

0 14 0 14

Distance across image (nm) Distance across image (nm)

Figure 6. POSAP composition maps for Ga 6(a) and (c) and As 6(b) and (d) taken from both interfaces of a GaInAs well from sample 3 (second reactor design). The corresponding composition profiles for all four elements present are displayed underneath each pair of composition maps 6(e) and

(0.

CONCLUSIONS

It has been demonstrated that the combination of POSAP and PLAP is capable of providing extremely detailed information on both the chemistry and morphology of quantum wells. This knowledge can be used directly to help optimise the growth techniques employed in the production of quantum well structures.

The POSAP provides truly three dimensional data that shows both the chemistry and morphology of interfaces on an extremely fine scale. This information is complementary to that provided by TEM and HREM, which, while they integrate the data through the sample thickness, can examine relatively large areas, and can be considered slightly more routine than the techniques of PLAP and POSAP.

The data acquired by atom probe techniques will also help in the calibration of routine assessment methods, such as X-ray diffraction and photoluminescence, where the existing models used to explain experimental data are in need of refinement.

ACKNOWLEDGEMENTS

The authors would like to thank Professor Sir Peter Hirsch FRS for the provision of laboratory facilities, Mr T.J.

Godfrey for his technical assistance, and STL Technology Ltd. and Plessey for provision of samples. JAL thanks Plessey for the provision of a CASE studentship.

REFERENCES

111 T.P. Pearsall, Surf: S c i . , B (1984) 529.

[2] F. Capasso, Surf Scf.,m(1984) 513.

[3] A. Cerezo, T.J. Godfrey and G.D.W. Smith, Rev. Sci. Instnrm.,

3

(1988) 862.

[4] G. Kellogg and T.T. Tsong, J. Appl. Phys.,

51

(1 980) 1 184.

[5] J.A. Liddle, A.G. Norman, A. Cerezo and C.R.M. Grovenor, I. de Phys.,@ (1988) C6-509.

TABLE I: Compositions from the well and barrier regions of sample 1 (quartz wool baffle present), as determined by PLAP analysis.

PLAP (accurate to f 2%)

%Ga %As %In %P GdAs In/P %Ga+In %As+P

Well 16.0 26.5 33.7 23.8 0.60 1.42 49.7 50.3

Barrier 6.8 10.6 40.4 42.1 0.64 0.96 47.2 52.7

TABLE 11: Compositions from the well and barrier regions of sample 2 (quartz wool baffle absent), as determined by PLAP analysis.

PLAY (accurate to +ZO/oo)

%Ga %As %In %P GdAs In&' %Ga+In %As+P

Well 23.2 41.6 29.2 6.0 0.56 4.87 52.4 47.6

Barrier 4.8 9.5 48.9 36.8 0.5 1 1.33 53.7 46.3

0 Distance .across well (nm) 12.5

Figure 7. Sequential PLAP analyses taken across a GaInAs well from sample 3 (second reactor design).