United States Patent (19. 11 4,128,338 Wong 45 Dec. 5, 1978

54 MODIFIED OPTICAL TRANSMISSION 3,725,135 4/1973 Hager et al...... 148/15 TECHNIQUE FORCHARACTERIZING 3,902,924 9/1975 Maciolek et al...... 148/15 EPTAXAL LAYERS Primary Examiner-John K. Corbin (75) Inventor: Theodore T. S. Wong, Maynard, Assistant Examiner-R. A. Rosenberger Mass. Attorney, Agent, or Firm-Theodore F. Neils; David R. 73) Assignee: Honeywell Inc., Minneapolis, Minn. Fairbairn (21) Appl. No: 807,608 57 ABSTRACT al An improved method of determining the energy band (22 Filed: Jun. 17, 1977 gap of an epitaxial layer on a substrate 51 int. Cl’...... GON 21/22 corrects for an overestimation of energy gap yielded by 52 U.S. C...... 356/432 normal optical transmittance measurements. The over 58 Field of Search ...... 356/201, 202,203 estimation of energy bandgap is caused by a graded (56) References Cited bandgap region which exists between the epitaxial semi U.S. PATENT DOCUMENTSw conductorductor 1layer and the substrate s 3496,024 2/1970 Ruehrwein...... 148/33.5 3 Claims, 10 Drawing Figures

DETERMINE MEASURE

OPTICAL ds ded TRANSMISSION (T )

DETERMINE ol. Eg., x U.S. Patent Dec. 5, 1978 Sheet 1 of 5 4,128,338

X 2 O H () O al O U

THCKNESS U.S. Patent Dec. 5, 1978 Sheet 2 of 5 4,128,338

MEASURE DETERMINE OPTICAL ds de d TRANSMISSION (T )

FIG. 4 DETERMINE c., E.g. x

2

FIG.5

20 40 60 80 IOO I2O 4O 6O 80 20O MICRONS SUBSTRATE GRADED GAP LAYER REGON U.S. Patent

| |

1-01 33NWLLWSNW U.S. Patent Dec. 5, 1978 Sheet 4 of 5 4,128,338

- / O DATA O - - - CALCULATION

O-3

6 7 8 9 O 2 3 WAVELENGTH (Jim) U.S. Patent Dec. 5, 1978 Sheet 5 of 5 4,128,338

FG.O

d = 5 Lim Ol

O-2

d = 20 Jum 2 lo-3 d = 4Olm H de = 6Oum 2 C a o to 4

to 5 3 4. 5 6 7 8 9. O WAVELENGTH (um) 4,128,338 1. 2 ber of advantages over both vapor phase epitaxial MODEFEO OPTICAL TRANSMISSION growth and bulk growth of (HgCd)Te. TECHNIQUE FOR CHARACTERIZING One characteristic of epitaxial film grown by both EPTAXAL LAYERS vapor phase epitaxy and liquid phase epitaxy is a ten dency to exhibit a compositional gradient along the ORIGIN OF THE INVENTION crystal growth direction. This is particularly true when The present invention was made in the course of a CdTe is used as the substrate material. Examples of contract with the Department of Army. compositional profiles through the thickness of epitaxi BACKGROUND OF THE INVENTION ally grown films are shown in FIG.S. 3, 5, 6 and 9 of the O previously mentioned Hager et al. patent (U.S. Pat. No. The present invention is concerned with the charac terization of epitaxial layers. The present invention is 3,725,135) and in FIGS. 4a-4e of the previously men particularly useful in characterizing epitaxial layers of tioned Maciolek et al. patent (U.S. Pat. No. 3,902,924). semiconductor alloys such as mercury cadmium tellu The device formed by epitaxial growth may be consid ride, telluride, indium arsenide antimonide, 15 ered, therefore, to have three regions: the substrate, a gallium arsenide phosphide, and others. graded composition or graded bandgap region, and the For the purposes of simplicity, the present invention epitaxial layer of desired composition. will be described with reference to a particular semi SUMMARY OF THE INVENTION conductor alloy; mercury cadmium telluride. The com mon chemical notation for mercury cadmium telluride, 20 The present invention is directed to an improved (HgCd)Te, or HgCdTe, will be used. method of characterizing an epitaxial layer on a sub Mercury cadmium telluride is an intrinsic photode strate, wherein a graded bandgap region exists between tector material which consists of a mixture of cadmium the epitaxial layer and a substrate. It is based upon the telluride, a widegap semiconductor (E = 1.6ev), with discovery that conventional optical transmission tech mercury telluride, which is a semimetal having a "nega 25 niques are inaccurate because the existence of a graded tive energy gap' of about 0.3ev. The energy gap of the bandgap region causes the transmittance curves to devi alloy varies linearly with x, the mole fraction of cad ate from those expected for homogenous material. The mium telluride in the alloy. By properly selecting ''x'', it deviation, if uncorrected, to an underestimation of is possible to obtain mercury cadmium telluride detec the cutoff wavelength, and thus an overestimation of tor material having a peak response over a wide range 30 the bandgap of the epitaxial layer. of wavelengths. The method of the present invention overcomes this (HgCd)Te is of particular importance as a detector problem by determining the thickness dis of the sub material for the important 8 to 14 micron atmospheric strate, the thickness dig of the graded bandgap region, transmission "window". Extrinsic photoconductor de and the thickness d of the semiconductor layer. The tectors, notably mercury doped germanium, have been 35 energy gap of the epitaxial layer is determined based available with high performance in the 8 to 14 micron upon the values of ds, do, and d and the results of wavelength interval. These extrinsic photoconductors, measurements of the total optical transmittance of the however, require very low operating temperatures layer, graded bandgap region, and substrate as a func (below 30 K). (HgCd)Te intrinsic photodetectors hav tion of wavelength. ing a spectral cutoff of 14 microns, on the other hand, are capable of high performance at 77 K. BRIEF DESCRIPTION OF THE DRAWINGS At the present time, most (HgCd)Te is produced by FIG. 1 illustrates schematically how light transmits bulk growth techniques such as the technique described through a cadmium telluride-mercury cadmium tellu by P. W. Kruse et al. in U.S. Pat. No. 3,723,190. High 45 ride epitaxial structure. quality (HgCd)Te crystals are produced by this bulk FIG. 2 shows absorption, a and cutoff wavelength growth technique. A as a function of thickness Z for a CdTe-(HgCd)Te Epitaxial growth techniques offer a number of poten epitaxial structure. tial advantages over bulk growth techniques. An epitax FIG. 3 shows the linear graded-gap approximation ial layer is a smooth continuous film grown on a sub 50 strate, such that the film crystal structure corresponds used in the calculations. to and is determined by that of the substrate. The de FIG. 4 illustrates the modified optical transmission sired epitaxial layer is single crystal with uniform thick technique of the present invention. ness and electrical property. The substrate has a differ FIG. 5 shows a CdTe-(HgCd)Te epitaxial structure ent composition or electrical properties from that of the 55 which has been angle-lapped at one edge to allow mea epitaxial layer. surement of thicknesses ds, do, and dil. A number of epitaxial growth techniques have been FIG. 6 shows calculated and measured transmittance investigated in an attempt to grow (HgCd)Te layers. as a function of wavelength for bulk (HgCd)Te. Vapor phase epitaxial growth processes which have FIG. 7 shows calculated and measured transmittance been studied are described in a number of patents in of a liquid phase epitaxial layer of (HgCd)Te on a cad cluding R. Ruehrwein (U.S. Pat. No. 3,496,024), G. mium telluride substrated. f Manley et al. (U.S. Pat. No. 3,619,282), D. Carpenter et FIG. 8 shows composition as a function of thickness al. (U.S. Pat. No. 3,619,283), R. Lee et al. (U.S. Pat. No. as measured by electron beam microprobe for the same 3,642,529), and R. Hager et al. (U.S. Pat. No. 3,725,135). sample used for measurements in FIG. 7. Another epitaxial growth technique which has been 65 FIG. 9 shows the effects of epitaxial layer thickness investigated is liquid phase epitaxy ("LPE'). This tech dL on transmittance. V nique is described in R. Maciolek et al. (U.S. Pat. No. FIG. 10 shows the effect of graded bandgap region 3,902,924). Liquid phase epitaxial growth offers a num thickness do on transmittance. 3 4,128,338 4. DETAILED DESCRIPTION OF THE d Eq. 2 PREFERRED EMBODIMENT (I - RL) (I - Rs) exp ? a(Z) dZ The present invention is an improved method of char- T = - - - acterizing epitaxial semiconductor layers by a modified 5 I - RLRs exp(-2 I a(Z) dZ optical transmission technique. Optical transmission o measurements are commonly used to determine the where the subscripts L and S stand for layer and sub energy gap E of semiconductor materials. This is of strate, respectively, and d = d -- do -- d.s. particular importance in alloy such as 10 In order to illustrate the effects of the graded-gap on (HgCd)Te and lead since the energy gap the overall transmittance curve, a linear graded-gap varies with composition of the alloy. The value of the approximation is made (see FIG. 3). This approximation energy gap is particularly important information when allows Eq. 2 to be quantified, since it relates the compo the semiconductor material is to be used as a photode sition x (and therefore E) directly to the thickness Z. tector, since the energy gap determines the wavelengths 15 As a result, the function a(Z) can be explicitly deter to which the material will be sensitive. mined if a(x) is known. The function a(x) has been The present invention is based upon the discovery derived and satisfactory agreement with data was ob tained (M.D. Blue, Phys. Rey, 134, A226 (1964)). The that conventional transmission measurement techniques reflectivity R of CdTe and (HgCd)Te have also been yield inaccurate values of energy gap in those epitaxial 20 measured and are known. structures which have a graded bandgap region be FIG. 4 illustrates the modified optical transmission tween the epitaxial layer and the substrate. For most characterization technique of the present invention. In device applications, it is the energy gap of the material this method, the thicknesses ds, dG, and d are deter at or very close to the surface of the epitaxial layer mined and the optical transmission T is measured. Based which is of interest. The existence of a graded bandgap 25 upon the values of T, ds, do and dt, it is possible to region, however, causes the transmittance curves to solve equation 2 for the absorption coefficient a in the deviate from those expected for homogenous materials. layer numerically. Once a is known, the values of enery This deviation or perturbation, if uncorrected, leads to gap E and composition x for the layer can be readily determined. an underestimation of the cutoff wavelength Aco, and 30 thus an overestimation of the energy gap of the epitaxial FIG. 5 shows a (HgCd)Te-CdTe structure on which layer. the modified optical transmission of the present inven tion is used. The structure includes the CdTe substrate FIGS. 1 and 2 show the effect of the graded bandgap 10, (HgCd)Te epitaxial layer 12 of a desired composi region on optical transmission measurements. In FIG. 1, tion, and a graded bandgap and composition region 14 the bandgap E (Z) is shown as a function of thickness 35 interposed between substrate 10 and epitaxial layer 12. Z. In FIG. 2, the absorption coefficient a(Z) and cutoff In FIG. 5, one edge of the structure has been angle wavelength Mare shown as a function of thickness Z. lapped. This allows electronbeam microprobe measure In FIGS. 1 and 2, ds and d are the thicknesses of the ments to be made from which thicknesses ds, do, and substrate (for example, cadmium telluride) and the epi dL can be determined. The angle-lapping and electron taxial layer (for example, mercury cadmium telluride), beam microprobe measurements are a destructive test respectively. The thickness of the graded bandgap re ing technique and cannot, therefore, be used on the gion is do. entire sample to determine composition. The primary As shown in FIGS. 1 and 2, the energy gap within the advantage of optical transmission measurements, are graded bandgap region is continuously changing from 45 that they are non-destructive, so that detectors can be formed from the material tested. that of the substrate to that of the epitaxial layer with In the present invention, angle-lapping only an edge the desirable composition. When photons are incident of the structure leaves the remainder of the structure on the layer surface, they are absorbed in the layer as available for fabricating detectors. Assuming that the well as in the graded bandgap region, if the layer is thin. 50 thicknesses ds, dG and d are relatively constant over The effects of the graded bandgap on the overall trans the area of the structure, the data obtained from the mittance curve must be determined in order to modify angle-lapping and electron beam microprobe measure the conventional transmittance technique to compen ments at the edge of the structure provides the neces sate for errors caused by the graded bandgap region. sary information to correct the optical transmission Consider a homogenous material having a constant 55 measurements made on other areas of the structure. absorption coefficient a between two optically flat sur From these nondestructive optical transmission tests, it faces. Let P be the radiation power incident on the is possible to determine a, E and X of layer 12 at any material, and P be the power transmitted. The transmit point on layer 12. tance T is (including multiple internal reflection) In order to demonstrate the applicability of the method of the present invention and check the suitabil ity of the values of a and R, calculations were per t--- (I = R exp(- a d) Eq. 1 formed using equation 1 for a x = 0.2 bulk (HgCd)Te P, I - Rexp(-2 a. d) slab having a thickness of 0.500 millimeter. The trans mittance as a function of wavelength in these calcula where R is the reflectivity and d is the thickness of the 65 tions is shown in FIG. 6. In addition, transmittance slab. For a CdTe-(HgCd)Te structure, the absorption measurements from the slab are also shown in FIG. 6. coefficient a (z) is a function of thickness Z, Eq. 1 can Agreement between the calculated values and the mea be written as sured values can be seen to be satisfactory. 4,128,338 5 6 FIG. 7 shows the comparison between calculated and energy gap data on epitaxial films by a non-destruc transmittance based upon equation 2 and measured tive technique. transmittance for a liquid phase epitaxial (LPE) grown Although the present invention has been described with reference to the preferred embodiments, workers (HgCd)Te layer on a CdTe substrate, The value of dL skilled in the art will recognize that changes may be was 20 microns, the value of do was 65 microns, and the made in form and detail without departing from the x value of the layer was 0.19. The integrations required spirit and scope of the invention. For example, the pres in equation 2 were performed by numerical methods. ent invention has been described with specific reference The thickness dependent absorption coefficient a(Z) to (HgCd)Te-CdTe epitaxial structures. The present was determined by the thickness dependent x value 10 invention is equally applicable to other epitaxial struc obtained by electron beam microprobe analysis as tures such as lead tin telluride on sub shown in FIG. 8. It can be seen that the agreement strates, gallium arsenide or gallium arsenide phosphide between the calculated and measured values of trans on gallium phosphide substrates and a wide variety of mittance as shown in FIG. 7 are excellent. This demon other epitaxial structures in which a graded bandgap strates that equation 2 is accurate and can be used to 15 region exists between the epitaxial semiconductor layer determine a, E and x of the layer when ds, dc dL and and the substrate. T are known. The embodiments of the invention in which an exclu FIG. 9 shows the effect of layer thickness don T as sive property or right is claimed are defined as follows: calculated using equation 2. In FIG. 9, curves are plot 1. A method of determining the energy bandgap of an ted in which the thickness of the layer is 1 micron, 5 20 epitaxial semiconductor layer on a substrate, wherein a microns, and 10 microns. In addition, two calculated graded bandgap region exists between the epitaxial curves for bulk material having a total thickness of 40 semiconductor layer and the substrate, the method com microns and 50 microns respectively are shown. In the prising: three LPE curves, the x value of the layer is 0.2, which determining the thickness dis of the substrate, the is the same x value assumed in the bulk calculation. It 25 thickness dog of the graded bandgap region, and the thickness d of the epitaxial semiconductor layer; can be seen that the larger d becomes, the more bulk measuring optical transmittance of the layer, graded like the transmittance curve becomes. bandgap region and substrate as a function of FIG. 10 illustrates the effects of the thickness dog of wavelength; and the graded gap region on the transmittance calculated determining the energy gap of the layer from the with equation 2. For all three curves, the thickness of values of ds, do, and d and the results of the mea the layer d was 5 microns and the x value was 0.2. The suring of optical transmittance. thicknesses of dog of 20, 40, and 60 microns were used. It 2. The method of claim 1 wherein determining thick can be seen from FIG. 10 that as do increases, the nesses ds, do, and d is by a destructive testing tech curves once again become more bulk-like. 35 nique on a portion of the layer, graded bandgap region, In conclusion, the existence of a graded bandgap and substrate. region in an epitaxial semiconductor structure, causes 3. The method of claim 2 wherein the destructive transmittance curves to deviate from those expected for testing technique comprises: homogenous materials. This deviation can lead to an angle-lapping an edge of the layer to expose a portion underestimation of cutoff wavelength and thus an over the layer, the graded bandgap region, and sub estimation of x value and energy gap of the epitaxial strate; and layer. The method of the present invention overcomes performing electron beam microprobe measurements this shortcoming of prior art optical transmission mea on the exposed portion. surement technique and provides accurate composition 45

50

55

65