JPH03257843A - Evaluation of semiconductor substrate - Google Patents

Abstract

PURPOSE:To simultaneously evaluate a natural oxide film, etc., which cannot be evaluated by a PBS method with machining damage evaluation by measuring the attenuation coefficient of a semiconductor substrate using ellipsometry. CONSTITUTION:In ellipsometry, He-Ne gas laser having a wavelength of 6328Angstrom is used and the intrinsic attenuation coefficient k0 of a complete GaAs crystal becomes the lowest of 0.211. When the surface of the crystal substrate becomes incomplete by machining damage, etc., as superfluous levels are created in the forbidden band, the symmetric property of Bloch function is destroyed and the attenuation coefficient increases according to the destruction. Therefore, the increase of (k-k0), which shows the attenuation coefficient increase, becomes the barometer of the substrate surface crystal incompleteness and the increase of the machining damage. In ellipsometry, a wafer 1 has a large attenuation coefficient (k) by a natural oxide film formed on the surface. Therefore, a wafer having a small value of (k) has a less natural oxide film, surface defects, machining damage, etc.

Description

DETAILED DESCRIPTION OF THE INVENTION Field of Industrial Application The present invention relates to a method for evaluating characteristics on the surface of a semiconductor substrate by ellipsometry, also called ellipsometry.

<Prior Art> In order to achieve high integration and high yield of semiconductor devices with good characteristics, it is necessary to use a uniform semiconductor substrate with good characteristics.

For example, in order to achieve high speed, high integration, and high yield in a GaAs digital 1vIC, it is necessary to
It is necessary to set the threshold voltage (vth) of the As MESFET to a central design value and to make it uniform across the substrates used. In manufacturing the above GaAs MESFET, the ion implantation method with good controllability is widely used to form the active layer that determines the vthO value. Even if manufactured using this ion implantation method, the v
The non-uniformity of th can be caused by the manufacturing process, such as variations in the gate length Lg or the thickness distribution of the insulating film, but here we will discuss dislocations in the semiconductor substrate, deep levels EL
2. An evaluation method regarding variations caused by damage caused by processing, etc. will be explained. This variation in the semiconductor substrate results in variation in the active layer in the MESFET of the above example. Many studies have been conducted on the influence of dislocations and EL2 on this variation in the active layer, and various theories have been proposed. Also, in recent years, attention has been paid to the effects related to the distribution of machining damage during polishing.

In general, semiconductor substrates are produced by slicing an ingot grown into a single crystal by pulling it into wafers, smoothing it by mechanical polishing, and removing processing damage caused by the polishing by chemical etching. Processing damage to the substrate is that processing damage that cannot be completely removed by etching and remains on the surface of the substrate.

The remaining machining damage causes lattice defects such as F dislocations and distortions in the lattice constants, so the fabricated MESFE
The vth of T became uneven. Conventionally, this machining damage is PB
S method (Photon Back Scatterin)
g) was detected and evaluated. (Evaluation by this PBS method, for example, R, C, C1arket aL,
”5 surface damage in SEM
i-insulatingGaAs 5tibStra
teS”, GaAs ICSyml)-, p, 4
1. 1987 Nato VC explanation Les “Cile.) Above P
The BS method is a method in which the surface of a semiconductor substrate is scanned with a laser beam, and the degree of scattering due to defects (including surface irregularities, foreign matter, etc.) caused by residual processing damage is detected and evaluated.

<Problems to be Solved by the Invention> The PBS method described above is widely used as a method for evaluating defects due to residual machining damage.

However, although this (2) PBS method can detect roughness and defects on the surface of a semiconductor substrate with high sensitivity, it can detect other defects such as a uniform native oxide film (native-oxide) on the surface of a semiconductor substrate.
There is a problem in that if a semiconductor substrate is covered with an oxide layer (which naturally forms on the surface when it is left in the atmosphere), it cannot be evaluated.

SUMMARY OF THE INVENTION An object of the present invention is to provide a semiconductor substrate evaluation method that can be used as a conventional processing damage evaluation method.

<Means for solving the problem> The present invention is based on the measurement of the attenuation coefficient of a semiconductor substrate using ellipsometry, and the evaluation of a semiconductor substrate by ellipsometry, which is the same as the PBS method, is generally used to evaluate etching damage and ion implantation damage. It is used in (For example, K. Watanabee et al., “El
lipometric 5tudy ofsil
icon 1nplanted with Bo
Ron Ionsin low doses? ,
Appl+Phys+Lett+, 84°p518.1
979 etc.).

In the evaluation by ellipsometry, it is also possible to evaluate semiconductor substrates.

Here, if the emissivity of the semiconductor substrate to be evaluated by ellipsometry is R1, the absorption coefficient is α, and the transmittance is T, then the following relationship exists.

R = [2 to (n-1)2+]/[2 to (n+1)2+]
...(1) α=4π/λ ...(2) T=(1-R2)exp(-αyD, 4l-R2exp
(-2αX)] - (3) where λ is the wavelength of the incident laser beam, n is the refractive index of the semiconductor, k is the attenuation coefficient, and X is the penetration depth.

From the above equation (2), the absorption coefficient σ is a linear correlation with the attenuation coefficient, and since n and k are found in ellipsometry, k
It is easier to evaluate using Therefore, k will be used in the following explanation. (Of course, conversion is easy, so n may be used.) Generally, in ellipsometry, H with a wavelength of 6328A is used.
An e-Ne Gannulas laser is used, and the inventive one is 0.211
When the surface of the crystal substrate is disturbed due to processing damage, etc., extra levels are created in the forbidden band, destroying the symmetry of the Bloch relation, and the attenuation coefficient increases depending on the degree of the disturbance.

(For example, M.H., Brodsky et al.
.

5tructura1. optical and e
electrical properties of
amorphous siliconfi1ms',
, Phys, Rev, B, t p2682.1970)
Therefore, an increase in (k-ko), which indicates an increase in the attenuation coefficient, is an indicator of an increase in crystal imperfections or processing damage on the substrate surface.

In order to confirm that the ellipsometry described above can also evaluate processing damage on the substrate surface in the same way as the PBS method, three types of GaAs wafers (■, ■, and ■) were examined using ellipsometry. Figure 1 shows the distribution of bIX attenuation coefficients measured at 5 m intervals. The numbers shown in FIG. 1 are the values obtained by subtracting the average value of k on the substrate 1cave from the measured kO value and multiplying the result by 1000. Tsumashi 1000 X (k
kave).

In the distribution diagram of k above, the solid line indicates the average value of the attenuation coefficient k over the wafer. It connects the points where the thumbnail is 0, and the - dotted chain line indicates the boundary where the number in the figure becomes 10 or more, and the dotted line indicates the boundary where the number in the figure becomes 110 or less.

By looking at the ranges indicated by the eight types of lines in FIG. 1, the uniformity of the distribution of machining damage, etc. on the wafer can be clearly understood.

In Figure 1, there is no noticeable difference in the distribution of k between wafer I and Ru.

The following figure 2 shows wafer 1. The respective PBS scattering intensity distributions and ellipsometry attenuation coefficients are shown by the line 1 line scanned horizontally to the orientation flatness of each wafer and the dotted line line scanned vertically through the center of ■ and ■. This is a graph for comparing the distribution. In FIG. 2, for wafer (3) with large processing damage, the trends of the evaluation values by the two evaluation means described above are in very good agreement. This is considered to be because, from a macroscopic perspective, defects on the wafer surface that scatter the laser beam also cause an increase in the attenuation rate k. It is difficult to find correspondence in minute changes between the measurement curves of the two evaluation procedures for wafers I and ■.The reason for this is that the ellipsometry equipment used for this measurement is similar to the PBS method equipment. Because it was not possible to increase the density of measurement points (due to manual movement),
This is thought to be due to the inability to respond to minute changes on the wafer surface measured by the BS method.

The following Table 1 shows the values and PB of the respective attenuation coefficients obtained by horizontal and vertical scanning of each wafer shown in Figures 1 and 2.
The average value of the S scattering signal value and the specifications of each wafer are shown together.

In Table 1, wafers are processed using the standard lapping process (
After mechanical and chemical polishing processes), the wafers were packaged and stored in a normal air atmosphere, wafers were subjected to a standard lapping process and stored in a dry N2 atmosphere, and wafers were subjected to only mechanical polishing and wafer cutting. The wafers are from the same storage.

From Table 1 above, it can be seen that in ellipsometry, wafer I has a larger attenuation coefficient k than wafer II due to the natural oxide film formed on its surface, but there is a clear difference between the measured values using the PBS method. There is no difference. The cause of this is thought to be the particles formed on the substrate surface.

<Function> The non-destructive semiconductor substrate evaluation method using ellipsometry of the present invention can be performed in conjunction with the evaluation of processing damage by measuring the attenuation coefficient of the substrate and the PBS method, so it is a practical and efficient semiconductor evaluation method. This is a method for evaluating the board.

<Example> Embodiments of the present invention will be described below with reference to the drawings.

The semiconductor substrate used in this example has a diameter of 2 inches and has a high resistance G
aAs wafers, and the source of these wafers,
Table 2 categorizes the wafers by type of wafer, such as semiconductor crystal growth method and amount of impurities, and assigns classification numbers to each type.

Figure 8 shows the distribution of wafer attenuation coefficients measured by ellipsometry using the same method as explained in Figure 1 for the various wafers listed above, and the correspondence with the EVA classification numbers is shown in Figure 8. , (a) of this Figure 3.

...(g) corresponds to this. Furthermore, t(1), ID, . . . in this figure are determined for several wafers of the same classification, and the respective results are shown. Ave and devi added to each figure indicate the average value and variance of the measured attenuation coefficient for each wafer. The average value in Table 2 is the average value of all measured values for wafers with that classification number.

From Figure 8, which shows the above ellipsometry values and the distribution of k on a semiconductor substrate, we have shown the characteristics of each acquisition route, and in particular, in the distribution of k, pay attention to the solid line showing the average value. When you look at it, you can see that a distinctive pattern is formed. These differences in patterns may be due to the distribution of the natural oxide film, as mentioned in the explanation for Figure 1 above, but this alone does not mean that patterns with characteristics are formed for each evaporator manufacturer, which differs depending on the route of acquisition. FIG. 4 shows the results of HCI processing to examine the relationship between the patterns. Figure 4 (a) shows the k and its distribution of the wafer before the natural oxide film is removed by HCI, and (b), (C) and (d) of the same figure show the k of the wafer before removing the native oxide film by HCI, and Figure 4 (b), (C) and (d) show the k of the wafer after HCI treatment. shows k and its distribution after 24 hours, 48 hours, and 72 hours. Looking at the progress in FIG. 4, even though the MCI processing pattern changes somewhat after the HCI processing, it approaches the pattern before the MCI processing as time elapses. From the above, even if the evaluation of k by ellipsometry of the present invention mainly targets the natural oxide film formed on the semiconductor surface, the growth of that film may also cause damage to the underlying GaAs substrate or its processing damage, etc. It is thought that the fact that they are different from each other is a factor that makes the distribution patterns similar depending on the acquisition route.

Therefore, we thought it would be meaningful to summarize the characteristics of evaluation by ellipsometry for each wafer acquisition route, and summarized them as follows.

1) A-1 Average k is slightly high. However, the average value line is uniform at the periphery of the wafer and at the center of the reed, which can be said to be the best distribution.

The distribution of k is almost the same as flA-1, but the distribution pattern of k is different due to the difference in dopants.

Moreover, the average [is] getting lower.

8) The average value for B-1 is the lowest. Although the average value line is band-like toward the center of the wafer, the uniformity is good.

4) C-1 kC) The distribution is complicated (four target patterns, the average value is high, so it cannot be said that it is a good wafer.

The distribution pattern of k is almost the same as that of A), and the average value of k is close to A-1. This low elementary concentration is C~2.
4 X 1014cx, high carbon concentration is C~2.
It is about 4×1015os”.

6) D-1 Above +2) It is characterized by the average of B and C, and there is no particular tendency, and the variation in kC) distribution is quite large. However, the average value is on the low side.

Taking all the above into account, we can see that the distribution of k is uniform because A
and C-2,8, and the one with the lowest average value is B.
I can say that.

Subsequently, an FET was manufactured using the substrate evaluated by the ellipsometry of the present invention as described above, and a comparison was made with the characteristic distribution of the FET. The FET was fabricated using wafers selected from the various wafers shown in Table 2.

The cross-sectional view of the manufactured FET is as shown in Figure 5.The FET manufacturing process and dimensions of the main parts are shown in Figure 3.
It is as shown in the table.

8g 8 Table In other words, n-type impurities are ion-implanted at a low concentration into a semi-insulating GaAs substrate, the impurities are activated by rapid annealing (RTA), and the gate, source, and drain metal electrodes are removed by a lift-off process. It is being produced.

Regarding the fabricated FET, the substrate sheet resistance, FETclVth, K (=μ
Table 4 summarizes the average values and variances of ξ/2aLg) and I dss . Furthermore, the n value and variant φb representing Schottky characteristics corresponding to various wafers are
Table 5 shows this and its variance.

Table 5: Comparison of Schottky characteristics of various substrates Further, FIG. 6 shows the correlation between the attenuation coefficient and the Vth O dispersion of the FET, and it shows a very good correlation.

The following FIG. 7(a) shows the relationship between the Schottky characteristic On and the value, and FIG. 7(b) shows the relationship between the value and the dispersion of n (σn). However, in this figure f, no clear correlation can be found. In FIG. 8(a), barrier hide φb,! : I am comparing the values, but I can't find a clear correlation.

≠1, and it can be seen that there is a fairly clear first-order correlation between the value and the variance of φb in FIG.

From what is shown in each of the above figures, the k value is σvth and σφ
It can be seen that there is a good correlation with b. This may be due to the fact that these values are related in the following relationship.

Vth=φb −Vp ” (6Vth)” = At Co $b)2+A2
('Vp)2 (Al and A2 are correlation coefficients) The above is a study of the correlation between the measured value of 2 and the characteristics of devices fabricated using the wafer, and depending on the correlation, it may be even more useful. relationships can be found. When the value of the wafer carried in before the device fabrication process is small, the variation in the barrier hide φb is small, and the variance σvth of the threshold voltage is also small.
In addition, good Schottky characteristics are easily obtained. The reason for the above correlation is not clear, but wafers with smaller values indicate fewer natural oxide films, surface defects, processing damage, etc., and this is thought to be the cause of reducing the variation in φb (D). .

In the above embodiment of the evaluation of a semiconductor substrate for measuring the attenuation coefficient k of ellipsometry according to the present invention, a GaAs substrate was explained.
It is not limited to As substrates, but semiconductors such as Si and InP can also be evaluated in the same way, and each substrate can be evaluated non-destructively, and predetermined selection can be performed, and the selected wafers can be By using it, it is possible to increase the production efficiency of semiconductor devices.

<Effects of the Invention> According to the evaluation method of semiconductor substrates using ellipsometry according to the present invention, wafers before processing can be non-destructively inspected, and information regarding processing damage and surface defects such as natural oxide films on the semiconductor substrates can be obtained. can be obtained as an attenuation coefficient, and from this (Z)k, it has become possible to estimate variations in characteristic values when devices are manufactured using wafers. Therefore, it is possible to select a wafer suitable for the device to be manufactured before the process.

Productivity in manufacturing semiconductor devices can be improved.

[Brief explanation of drawings]

FIG. 1 is a diagram showing the distribution of Dk of the present invention for various surface-treated wafers, FIG. 2 is a diagram comparing the measured values of the present invention and the PBS method, and FIG. (Figure 4 shows the time-dependent change in value after hydrochloric acid treatment, Figure 5 shows the configuration of Example 0 FET, Figure 6 shows k and σV
FIG. 7 is a diagram showing the relationship between n and σn with respect to k, and FIG. 8 is a diagram showing the relationship between k, φb, and σφb.

Claims (1)

[Claims] 1. A semiconductor substrate characterized in that the state of processing damage, oxidation, etc. on the semiconductor substrate surface is estimated and selected from the measurement of the attenuation coefficient or the distribution of the attenuation coefficient on the semiconductor substrate surface by ellipsometry. evaluation method.