Method for Imaging Electrical Barrier Layers Such as Pn-Junctions in Semiconductors by Means of Processing Particle-Beam-Induced Signals in a Scanning Corpuscular Microscope

The invention relates to a method for imaging electrical barrier layers such as pn junctions in semiconductors by means of processing particle-beam-induced signals in a scanning corpuscular microscope.

Precise knowledge of the potential distribution or dopant distribution in the inside of the semiconductor crystal is of significance for the development of VLSI components. Depth-dependent dopant profiles can be calculated by means of computer simulation or, for example, can be measured by means of secondary ion mass spectrometry (SIMS). The depth of a pn-junction can be identified by means of electrical measurements.

Due to increasing miniaturization, however, lateral effects produced by scattering or diffusion of the dopant, for example under-diffusion at mask edges, are also of great significance. The measurement of these lateral effects with sufficiently high topical resolution (approximately 0.1 .mu.m) is a hitherto unresolved problem.

It is known to image pn-junctions with the assistance of the induced current in a scanning electron microscope (EBIC: electron beam induced current). The definition of a pn-junction, however, cannot be represented with the required precision given such a known method.

An object of the present invention is to specify a method of the type initially cited which also enables the imaging of electrical barrier layers, for example pn-junctions, in semiconductors with very high precision even when these electric barrier layers are aligned perpendicularly or obliquely relative to the surface of the specimen. For example, the course of pn-junctions should be identified in cross-sections through semiconductor components with an uncertainty of at most 0.1 .mu.m.

This object is achieved by defining specific image-points having defined properties with respect to their surrounding using these properties for the localization of the barriers.

FIG. 1 shows a prior art arrangement for imaging a pn-junction with the assistance of an induced test current (EBIC);

FIG. 2 schematically shows the result of a line scan perpendicular to the pn-junction with the prior art arrangement according to FIG. 1;

FIGS. 3 to 5 illustrate the principle underlying the invention enabling a locating of a pn-junction with required precision;

FIGS. 6 to 8 explain exemplary embodiments of the invention for the locating of electrical barrier regions;

FIG. 9 illustrates a block diagram of an arrangement for performing the method according to the invention; and

FIG. 10 illustrates a flow chart of the computer program attached herewith as Appendix A.

FIG. 1 shows an arrangement for imaging a pn-junction with the assistance of the induced specimen or test current (EBIC) in a scanning electron microscope as well as the wiring of the specimen. Primary electrons PE having an energy of, for example, 20 keV penetrate into the specimen. The primary electrons PE are thus scattered and lose their energy in a volume G assumed to be spherical in FIG. 1. The primary electrons PE create electron hole pairs in this "scatter volume" G due to ionization. When the scatter volume is situated at a sufficiently great distance from the pn-junction, the electrons and holes recombine on the average after they have traversed a path corresponding to their respective diffusion length.

When the primary electron beam moves across the specimen surface in the x-direction and approaches the pn-junction, then finally diffused charge carriers proceed into the range of influence of the electrical field of the space charge region SCR. The respective minority carriers are accelerated by this electrical field to the respective other side of the pn-junction (electrons to the n-side, holes to the p-side). Given a closed external circuit, a "charge separation current" I.sub.i (EBIC) is present, this being measured, amplified and, for example, used for brightness modulation of the picture tube of the scanning electron microscope.

FIG. 2 schematically shows the result of a line scan perpendicular to the pn-junction. The charge separation current I.sub.i is illustrated in a logarithmic scale in FIG. 2. At a sufficiently great distance from the pn-junction, the curve progression of the charge separation current I.sub.i is described by

(I.sub.i denotes induced specimen current of EBIC; I.sub.im denotes the greatest possible value of I.sub.i ; L denotes the diffusion length of the respective minority carriers).

For small absolute values of x, the scatter volume G and the space charge region SCR overlap and the above equation for the charge separation current I.sub.i is in fact invalid. In particular, I.sub.i (x=0) is usually smaller than I.sub.im. The actual size of the induced current at the pn-junction depends on the diameter of the scatter volume G, on the width of the space charge region SCR, and on the influence of the surface recombination.

Knowledge of the size of the charge separation current I.sub.i is not required in order to localize a pn-junction. One can first proceed on the basis that the particle beam induced signal I.sub.i is greatest at the pn-junction, i.e. at the location of the greatest field strength. This certainly applies to a given symmetrical field distribution. It is then only necessary to find that location at which the charge separation current exhibits its maximum value. The exponential paths or variations of the charge separation current I.sub.i at both sides of the pn-junction can therefore be electronically suppressed in an inventive manner. A dark level setting that suppresses the exponential paths of the charge carrier current I.sub.i for values below a specific value DL is employed therefor. As a result, the pn junctions appear in imagings as narrow, bright lines. Fluctuations of the maximum charge separation current I.sub.i during the scanning of a pn-junction limit the applicability of this method. More far-reaching measuring methods according to this invention are proposed with reference to the following Figures in order to suppress the influence of these irregularities and in order to enable an even more precise localization.

The charge separation current I.sub.i can be represented as an EBIC profile I.sub.i (x), as shown in FIG. 2, or as a two-dimensional EBIC imaging I.sub.i (x,y). In order to obtain a two-dimensional EBIC imaging I.sub.i (x,y), the brightness of the picture tube of the scanning electron microscope is controlled with the assistance of the charge separation current I.sub.i (x,y). A bright line appears along the pn-junction, its width depending on the charge separation current profile and the corresponding level DL for the suppression of the exponential portions of the charge separation current profile.

FIGS. 3, 4, and 5 schematically show a specimen produced in silicon gate technology. The silicon substrate covered with a Si.sub.3 N.sub.4 layer is coated with an aluminum layer Al. FIGS. 3, 4 and 5 show an emitter (n.sup.+ -E)/base (p-B) junction. The EBIC signal I.sub.i disappears in the proximity of the Si.sub.3 N.sub.4 passivation layer because the size of the maximum of the EBIC signal I.sub.i in the proximity of said passivation layer is lower than the level DL (see FIG. 2) due to the increased recombination at the boundary layer and, therefore, the EBIC signal I.sub.i is completely suppressed in the proximity of the passivation layer (FIG. 4). A reduction in the level DL (see FIG. 2) in FIG. 4 would lead to a further spread of the EBIC signal R in those regions of the pn-junction that are not adjacent to the passivation layer. As a result thereof, however, the locating of the pn-junction in these regions not adjacent to the passivation layer would become even more imprecise. In order to avoid these problems with effects of boundary layers and surfaces, and in order to obtain higher topical resolution, various methods are proposed below wherein the EBIC signal I.sub.i is processed with the assistance of a computer. For example, 324.times.243 points can be stored and processed with a 12-bit resolution given the assistance of a desktop calculator HP9826. After the entire two-dimensional EBIC image I.sub.i (x,y) has been written into the calculator, the maximums of the EBIC signal I.sub.i (x,y) can be identified. When the angle between the direction of the line scan and the pn-junction is not too small, the two-dimensional EBIC signal I.sub.i (x,y) can, for example, be identified line-by-line in the simplest case. Too small an angle between the scanning direction of the scanning probe and the pn-junction can usually be avoided by turning the specimen or the scanning direction of the scanning probe. The disruptive influence of noise can be suppressed with the assistance of a corresponding algorithm. Results identified with the assistance of the computer can be output on a plotter or on the picture screen of the scanning electron microscope. In the latter instance--as given known EBIC images--, a sub-region of the imaging can be generated with the assistance of secondary electrons, as shown in FIG. 5. In contrast to the illustration in FIG. 4, the pn-junction S in FIG. 5 is shown with constant contrast and with constant width over its entire length. The line width derives from the method used to generate the imaging. Since the maximum of the EBIC signal I.sub.i is not sharp, the digitization of the EBIC signal results in a plateau of about two or three points that correspond to a line width of about 50 nm in FIG. 5 for the maximum of said EBIC signal. An additional spread of the line width is caused by the digitization of the image field.

It is presumed with the employment of a computer that the scanning electron microscope has a digital scanning generator which, for example, allows 1000 lines per image and 1000 points per line to be employed. With a computer, the EBIC signal levels of all 1000 image points of the corresponding line would have to be deposited given every scan line. After the line scan, the computer must identify the point or--given a plurality of pn-junctions--the points of the highest EBIC signal and would mark it on the picture screen by means of a bright spot only. This marking can ensue by means of a rebound on the same line or at the next line scan. In the latter instance, the localization error given 5000.times. magnification would only amount to 20 nm and could be neglected.

FIGS. 6, 7 and 8 show farther-reaching methods for the precise locating of electrical barrier regions. Given the method according to FIG. 6, for example, so-called contour lines, i.e. lines on the x-y plane that connect points at which the values of the EBIC signals approximately coincide to one another, can be identified by means of line-by-line scanning of the specimen surface and by means of corresponding roll-in of the resulting EBIC signals I.sub.i into a computer. The maximums in x-direction and y-direction can in turn be specified on the basis of these contour lines. These maximums can be employed as a 0.sup.th approximation for the localization of a pn-junction S. In case the pn-junction exhibits curvatures, this pn-junction S can be divided into different regions, each of said regions exhibiting a more or less definable curvature with a more or less definable center of curvature. A respective point P(x,y) is defined in the proximity of such centers of curvature, a respective line being scanned proceeding from said point P(x,y) on, for instance, the shortest path over various points H that lie on the sub-region of the pn-junction S belonging to this center of curvature. An EBIC signal I.sub.i is recorded for each such line scan. The maximum is identified for each of said EBIC signals I.sub.i. Averaging is therefore undertaken over a respective series of successively disposed points along each and every line scan in order to eliminate disruptive effects. The plurality of successively disposed points over which the respective averaging is carried out can thus be varied until the addition or the omission of a single point in the averaging produces practically no change of the result (when averaging) over a specific plurality of successively disposed points.

A further point P(x+.DELTA.x, y+.DELTA.y) in the proximity of the point P(x,y) is then selected and one proceeds with said further point in accordance with the method described with respect to the point P(x,y). If a localization of the sub-region of the pn-junction belonging to this curvature region thereby results which coincides with that localization that had been found given the first method step proceeding from point P(x,y), then this localization of the sub-region of the pn-junction S is already reliable. If the two successively identified localizations of the sub-region of the pn-junction do not coincide with one another, then further points P in the proximity of a point utilized for the immediately preceding method step must be sought and employed for a further method step until at least two localizations of the sub-region of the pn-junction that were executed in immediate succession practically coincide.

Given the method according to FIG. 7, the absolute maximum of the values of the EBIC signal I.sub.i of the entire image is first identified. Fields are formed in the environs of this absolute maximum M(x,y) by means of combining respective neighboring points in order to eliminate disruptive effects. The mean value of the EBIC signal I.sub.i is identified for each of these fields. Since the absolute maximum M(x,y) lies on the pn-junction S, that field around the point N(x,y) that exhibits the second-highest value of the EBIC signal I.sub.i after the absolute maximum also lies on the pn-junction S. Proceeding from the field around the point N(x,y), a further field is then sought whose value of the EBIC signal I.sub.i exhibits the highest value of all EBIC signals I.sub.i identified in the immediate proximity of the point N(x,y), whereby the values of points that have already been identified as belonging to the pn-junction S may no longer be taken into consideration. Always proceeding from the most recently identified point on the pn-junction S, the entire course of the pn-junction S can be successively defined in this manner. Fields F.sub.0 through F.sub.8 shown in FIG. 7 are explained later.

The size of the fields of points over which averaging is respectively carried out is varied until the addition or omission of a point from such a field no longer changes the localization of the pn-junction S.

FIG. 8 explains a further method for that case in which the width of the space charge region SCR is great in comparison to the diameter of the scatter volume G. The EBIC profile I.sub.i exhibits a plateau in this case. The constancy of the plateau value of the EBIC signal I.sub.i within the space charge region SCR is first monitored over a certain region of the pn-junction by means of a number of line scans. The measured EBIC signals I.sub.i are logarithmized at every line scan. The exponential portions of the EBIC signal I.sub.i thereby become approximately straight lines. The points F on the logarithmized signal I.sub.i at which the logarithmized signal I.sub.i exhibits just about linear edges are identified for every line scan on the rising and on the falling edge of the EBIC signal I.sub.i. The slope of the logarithmized signal I.sub.i at said points F is then identified in two-dimensional fashion. With these identified slopes, straight lines are placed through the points F on the logarithmized signal I.sub.i, said straight lines exhibiting two intersections T with a further horizontal line proceeding through the maximum of the logarithmized signal I.sub.i. Via their spacing b, these intersections T define the width b of the pn-junction.

The foregoing has always been based on the assumption that the maximum of the signal I.sub.i coincides with the pn-junction. This assumption is met as long as the electrical field in the space charge region SCR is symmetrical and as long as the diffusion lengths of the respective minority charge carriers are of comparable size. When, however, the dopant concentrations in the n-region and p-region differ greatly, as is usually the case, the electrical field distribution is asymmetrical and the diffusion lengths are different. The maximum of the EBIC signal I.sub.i and the pn-junction then do not exactly coincide.

In the example from FIGS. 2, 4, and 5 (where the emitter dopant concentration is 10.sup.20 cm.sup.-3 and the base dopant concentration is 10.sup.18 cm.sup.-3, a deviation .DELTA.x of the pn-junction from the maximum of the EBIC signal I.sub.i is about 10 nm. This means that the true emitter/base junction is negligibly shifted toward the emitter. In order to keep this error .DELTA.x as small as possible, the scatter volume G should be as small as possible. For this purpose, one must work with the lowest possible acceleration voltage for the primary electrons PE.

Fields F0 through F8 are indicated in FIG. 7. Each of the fields F0 through F8 supplies a value of the EBIC signal I.sub.i that has been averaged over the scanning points contained in the respective field. The values for the fields F0 and F8 are not taken into consideration because they have already been identified as belonging to the barrier region. Of the remaining fields F1 through F7, the field F4 exhibits the highest mean value for the EBIC signal I.sub.i. Thus, the field F4 also belongs to the electrical barrier region.

FIG. 9 shows a block diagram of an arrangement for the execution of the method according to the invention. In a scanning electron microscope SEM (Cambridge Stereo Scan S150), the primary electrons PE penetrate into the sample D. The sample D is shown as a diode in FIG. 9. Given a closed external circuit, a "charge separation current" EBIC flows. This EBIC signal is converted into a voltage signal in a current/voltage converter CVC (Keithley 427). This voltage signal is supplied over an analog-to-digital converter ADC to a multi-programmer MPR (HP 6940B). These test signals are processed in a computer COM (HP 9826A). The processing of the test signals in the computer COM is controlled, for example, via the program attached in the appendix A to the application and fully incorporated herein. Test results can be displayed via a plotter PLT (HB 9872A). The deflection of the primary electrons PE is likewise controlled by the computer COM. For this purpose, signals are communicated to the deflection coils SC of the scanning electron microscope SEM via digital-to-analog converters DAC which is associated with the multi-programmer MPR. Test results can also be represented on the picture screen CRT which, for example, is associated with the scanning electron microscope SEM, being represented thereon via a further digital-to-analog converter DAC that likewise is associated with the multi-programmer MPR.

Additional details concerning the construction of the measuring arrangement according to FIG. 9 are contained in the publication "Mapping Of PN-Junctions And Depletion Layers By Computer Processed EBIC Signals" published by the inventors after the filing date of the corresponding German application No. P 33 13 597.5 filed Apr. 14, 1983 and incorporated herein by reference. Of particular interest is page 2, second paragraph and page 3 the last two paragraphs.

Although various minor changes and modifications might be proposed by those skilled in the art, it will be understood that we wish to include within the claims of the patent warranted hereon all such changes and modifications as reasonably come within our contribution to the art. ##SPC1##