JPH0527974B2 - - Google Patents

Description

[Detailed description of the invention]

The present invention will be explained in the following order. A. Industrial field of application B. Overview of the disclosure C. Prior art D. Problem to be solved by the invention E. Means for solving the problem F. Example G. Effect of the invention A. Industrial field of application. It relates to wafer manufacturing methods, and in particular to processing methods for standardizing and stabilizing semiconductor wafers, typically made of silicon, for use on both unipolar (FET) and bipolar device manufacturing lines. be. B. SUMMARY OF THE DISCLOSURE This invention discloses the standardization and stabilization of semiconductor wafers supplied by different vendors (A, B,...) so that they can be used on both unipolar and bipolar device manufacturing lines. This processing method includes: (a) classifying wafers into classes (C1-C3) for each wafer having the same ^initial cluster distribution; All wafers in a given class are subjected to the same thermally matched processing cycle to form a defect-free zone of 10-30 μm thickness, corresponding to cluster densities in the range of ¹⁰⁹ clusters/ ^cm3 . (c) cleaning the wafer. C. Prior Art Conventionally, it has been known that contamination cannot be completely removed during the manufacture of semiconductor integrated circuits. To overcome this difficulty, various techniques have been developed to date to getter impurities and other contaminants, such as internal gettering, which relies on the presence of oxygen atoms in a supersaturated solid solution in silicon. The technology is known. These atoms accumulate during the subsequent heat treatment and form small aggregates or aggregates. This phenomenon is accompanied by the occurrence of secondary defects such as stacking faults and dislocations. These aggregates and secondary defects constitute minute defects that serve as nucleation sites for gettering impurities and contaminants. The efficiency of this type of gettering depends on the motion of the collection of oxygen atoms, which itself depends on the initial concentration of oxygen atoms and the distribution of nucleation sites.
Although these phenomena are not fully understood at present, the literature indicates that three conditions must be satisfied to achieve the desired internal gettering effect: (1) The concentration of oxygen must be high enough to cause accumulation. (2) The existence of a defect-free zone in the plane of the active face of the wafer. (3) nucleation sites;
(small clusters or agglomerates of oxygen atoms) pre-existing in the wafer. This first condition can be easily met.
This is because most commercially available wafers are typically Czochralski
It has an oxygen concentration in the range of 26 to 40 ppmA, which is found to meet the first condition above. In addition, in the case of bipolar devices, the oxygen concentration is 26 to 36 ppmA.
It is known that it is necessary to use wafers in the range of (see US Pat. No. 4,344,815 for details of this). Next, regarding the second condition above, the wafer
It is known that when subjected to high temperature heat treatment of 1000° C. or higher, the tendency of oxygen atoms to diffuse outward (exodiffusion) allows the formation of defect-free zones. That is, during such heat treatment, the concentration of oxygen atoms approaches the solubility limit within a region several microns thick below the surface of the wafer.
This solid solution then stabilizes as it is no longer supersaturated. Then, during the next heat treatment step, the oxygen atoms aggregate in the bulk of the wafer rather than at the surface, thereby forming defect-free zones on the surface. Next, regarding the third condition, it is known that nucleation sites or clusters are generated during heat treatment performed at a relatively low temperature (900° C. or lower). This process tends to reduce supersaturation by forming small clusters of oxygen and relies on random fluctuations in oxygen concentration. Let's represent these things graphically. The processing steps corresponding to the second and third conditions above are the first
It is represented by the curve in the figure. That is, FIG. 1 is a graph showing changes in the nucleation rate and the thickness of the defect-free zone versus the temperature T of the heat treatment process. Conventionally, various attempts have been made to establish manufacturing methods that combine the above two processes. In experiments by the inventors of the present invention, it was recognized that the order in which operations are performed is important. A method intended to counteract the effects of previous heat treatments therefore consists of the following treatments: (a) When a silicon rod grown by the Czyochralski method is cooled (or the resistivity Process the wafer at high temperature (e.g. 1250 °C) to destroy any clusters formed (during the annealing process performed) to stabilize the wafer. (b) Next, exodiffusion of oxygen atoms occurs. (c) This is followed by a heat treatment at temperatures in the range of 500-900°C to bond and redistribute the clusters. The disadvantage of this thermal cycle is that in (a) above, approximately
This involves a high temperature of 1200℃. The second method relies on the existing distribution of clusters, in which the above operations are performed in reverse order. That is, the first heat treatment at 800-900 °C serves to bond some of the existing clusters, while the second heat treatment in the temperature range 1000-1100 °C (this temperature is higher than the initial heat treatment in the first method above). (lower than the cycle temperature) destroys the clusters near the surface of the wafer, leading to out-diffusion of oxygen into this region. However, this second method (which was previously proposed by the inventor) has also been found to be insufficient. In this method, if the number of clusters generated on each wafer is plotted on the horizontal axis and the number of wafers is plotted on the vertical axis, the number of clusters is calculated at both tails of the normal distribution curve (2 × 10 ⁸ and 2 × 10 ⁹ It has been found that this spreads in the vicinity of ³ cm3). Number of clusters is 2×10 ⁸ /
cm ³ or less is unsuitable as a wafer for FET. This is because in FET manufacturing lines, subsequent heat treatment steps are not performed at very high temperatures, which are insufficient to grow the clusters as large as necessary for proper gettering action. On the other hand, the number of clusters is 2× ¹⁰⁹ /
cm ³ or more is unsuitable for bipolar wafers. This is because bipolar manufacturing lines involve subsequent high-temperature processing (e.g., high-temperature processing such as growth of epitaxial layers and diffusion formation of subcollectors), which causes excessive cluster growth, resulting in excessive gettering effects. Undesirable. Thus, it does not seem appropriate to subject all wafers to the same thermal cycle.
Therefore, it may be necessary to perform a thermal treatment that is specific to each wafer and that is compatible with a particular application, but then a considerable number of different wafers must be managed for each application. It will stop happening. D Problems to be Solved by the Invention The purpose of the invention is to provide a method for standardizing and stabilizing semiconductor wafers so that they can be used in both unipolar device manufacturing lines and bipolar device manufacturing lines. It is about providing. Another object of this invention is to provide a controlled and reproducible method for internally gettering contaminants in a wafer, regardless of the internal nature of the wafer. Yet another object of the invention is to provide a method applicable to semiconductor wafers sourced from different vendors and having initial oxygen concentrations over a very wide range, thereby greatly facilitating the availability of wafers. It is in. E Means for Solving the Problems According to the present invention, a standardized processing method for semiconductor wafers that can be used commonly for manufacturing unipolar and bipolar devices measures the condensation rate of oxygen atoms dissolved in semiconductor wafers in multiple lots. and classifying the semiconductor wafers into a plurality of classes according to a plurality of ranges of their condensation rates. In order to generate predetermined clusters, a cluster generation step is performed in which a specific heat treatment is applied to all semiconductor wafers in the same class, which differs for each class, but is the same for all semiconductor wafers in the same class, and a defect-free zone of a predetermined thickness is formed on the surface. The steps include applying substantially the same heat treatment to all classes of wafers to stabilize the wafers. The configuration of the present invention is as follows. a step of grouping semiconductor wafers in a plurality of lots according to a plurality of initial oxygen content ranges, extracting at least one semiconductor wafer from each group as a sample and heating the sample at a temperature substantially corresponding to a cluster formation temperature; A process of testing the oxygen condensation rate; a process of grading each group of semiconductor wafers into a plurality of classes according to the above oxygen condensation rate; heating semiconductor wafers of all classes to produce clusters in each semiconductor wafer corresponding to a predetermined cluster density; A standardized processing method for semiconductor wafers that can be used commonly for the manufacture of unipolar and bipolar devices, comprising the steps of heating in a heating cycle. The heating cycle for the sample will be described later in connection with FIG. 3. A plurality of ranges, for example three ranges, for the amount of condensed oxygen are discussed below in connection with FIG. F. Embodiment According to the present invention, wafers are first classified into a plurality of classes. Each of these classes includes wafers with approximately similar precipitation rates, or similar initial distributions of clusters. This is essential for standardizing wafers. This is because by applying the same thermal treatment to all wafers, even if the effect of the previous thermal treatment can be canceled out, the oxygen concentration will be in the range of 26 to 40 ppmA. This is because it would be impossible to achieve the same cluster distribution across the wafer. Now, the method of the invention described below is possible because tools are currently available that can classify wafers as a function of their initial oxygen concentration without damaging the wafer's surface. FIG. 2 is a block diagram illustrating the steps of the method for standardizing and stabilizing wafers according to the present invention. In this example, the wafers to be processed are silicon wafers supplied by two vendors A and B, whose oxygen content is approximately
It ranges from 26 to 40 ppmA. Note that ppmA is an abbreviation for “parts per million atomic” and corresponds to 0.5×10 ¹⁷ oxygen atoms per 1 cm ³ of silicon. In step 1, wafers with proper manufacturer identification are stored prior to processing. As mentioned before, the condensation motion of oxygen atoms is 2
two parameters: the initial oxygen content of the wafer and the initial number of nucleation sites. Measurement of these parameters and classification of the wafers takes place in step 2, which will be described in detail below. The first parameter is measured directly on every wafer by infrared absorption spectroscopy using the well-known Baker Standard:

[0]pppmA=9.63αcm−1 Here, α is the absorption coefficient for infrared light of 9 μm. This measurement must be done very quickly so as not to damage the wafer surface. This ensures that the initial oxygen content of wafers supplied from vendor A is in the range of 30 to 39 ppmA, and that of wafers supplied from vendor B.
It was found to be in the range of 27-36ppmA. All wafers from the same vendor and exhibiting approximately similar initial oxygen content by infrared absorption spectroscopy are then formed in one lot. Such lots are formed, for example, based on the following ranges: 27-30 ppmA, 33-36 ppmA, 36-39 ppmA. However, the second parameter cannot be determined by direct measurement. Therefore, a heat treatment called a condensation test is performed in order to condense the oxygen atoms onto the existing clusters. For this purpose, let us use the amount of condensed oxygen as an approximation to the number of clusters. However, the number of clusters may not actually be calculated from that amount. Note that this condensation test is performed only on sample wafers in each slot. Also, the temperature at which the setting test is conducted must be high enough to result in little or no cluster formation, and high enough to destroy relatively small clusters. FIG. 3 is a graph of curves representing the amount of condensed oxygen versus the temperature at which the heat treatment is performed for wafers having different amounts of oxygen. From this graph, the maximum value of the amount of condensed oxygen is approximately 800
℃, and that there is a minimum value between 800℃ and 1000℃ (furthermore, from Figure 1 and its explanation,
It can be seen that relatively small clusters are generated at temperatures above about 900°C, and the number of clusters formed is drastically reduced in the temperature range of 800°C to 900°C). If based on a condensation test at a temperature of 800° C., the number of clusters would be too large and the amount of condensed oxygen would be so excessive that there would no longer be a defect-free zone in the wafer after final processing. It is appropriate to condense at a temperature of approximately 900°C in order to have clusters, but not in excessive numbers, but still large enough to be safe (i.e. not collapse) during subsequent high temperature processing. One thing can be seen from the curve in Figure 3. In this way, the wafer
One possible solution is treatment for at least 20 hours in a nitrogen atmosphere at 900°C. However, although satisfactory results can be obtained by this method, a time of 20 hours or more is too long for practical use. Therefore, a faster heat treatment was set. This heat treatment method first heats the wafer to 900°C to ensure sufficient cluster growth.
and then treating the wafer at 1100° C. to cause the condensation of oxygen atoms on these clusters. This treatment method has the advantage of facilitating the detection of changes in oxygen content. Since the absorption band of condensed oxygen is different from the absorption band of oxygen in the molten state, we can compare the levels of the absorption band of oxygen in the molten state before and after the condensation test. amount can be determined. Therefore, the preferred condensation test is in a nitrogen atmosphere.
The heat treatment cycle consists of 2 hours at 900°C and then 8 hours at 1100°C. Figure 4 shows the three classes C1− before the condensed oxygen pair.
This is a graph showing the correspondence between the initial oxygen content of sample wafers classified as C3: Class condensation rate Equivalent amount of condensed oxygen C1 Low ~0~10ppmA C2 Intermediate ~10~20ppmA C3 High >20ppmA Note that a finer distribution than this distribution may be used, or a slightly different distribution may be used. The sample used for the coagulation test is then rejected. The lot formed as described above is the fourth
distributed within these three classes using the information given in the figure. The table thus obtained is as follows.

[Table] Thus, upon completion of step 2 in FIG. distributed within two classes C1-C3. Thus, although any one class may include wafers with different amounts of oxygen, the wafers have approximately similar initial cluster distributions so that their condensation characteristics at 900° C. are substantially the same. Wafers supplied by vendors other than Vendors A and B are also processed as described above;
A condensation test is performed on the selected sample and the wafers are sorted into the appropriate class. This test also takes into account changes during manufacturing that can affect the movement of oxygen atoms to condensate.
It needs to be done periodically on the sample. In practice, the corresponding class (C1, C2 or
C3) can be easily determined as long as the wafer vendor (eg A or B) and the initial oxygen content are known. The sole purpose of the condensation test performed on the sample is to guard against possible changes in the crystal pulling process introduced by the vendor. Now, according to the main objective of this invention, which is to supply the production line with stabilized (i.e. exhibiting the same properties) and standardized wafers:
A user implementing the method of the invention must first determine the thickness of the defect-free zone, ie the number of clusters that are identical for the wafer. Considering the current state of the art, it seems reasonable to set the thickness equal to at least 10 μm (for example in the range 10-30 μm). Then, the number of clusters will be 2×10 ⁸ to 2×10 ⁹ clusters/cm ³ , preferably approximately equal to 5×10 ⁸ clusters/cm ³ where one cluster exists every 10 μm. . The invention provides a thermal cycle, which is referred to as an adaptation cycle. This is because the cycles are tailored to each class of wafers to ensure that all wafers meet the user's specifications before they are used on the production line. This adaptation cycle is shown at 3 and 4 in FIG.
It consists of two steps shown in . That is, step 3
In step 4, clusters are generated, and in step 4, defect-free bands are formed. The table describes in detail how step 3 is executed. Note that in each of the three classes C1-C3, the initial temperature is between 800 and 900°C.
Furthermore, although the periods shown in the table are suitable values, they may be changed as desired. Step 3 can be performed in a programmable furnace. Three classes as shown below
C1-C3 are processed individually.

[Table] Now, unlike step 3, step 4 is the same for all wafers regardless of class. For the formation of defect-free zones, in a suitable gas atmosphere,
It is necessary to carry out the heat treatment at a temperature in the range of 1050-1200°C (preferably 1100°C) for 90 minutes. This process is
Corresponds to a diffusion length (√) on the order of 10 μm.
The temperature increase from 900° C. to 1100° C. (in the second half of step 3) is intended to combine the number of clusters so that a stabilized wafer is obtained later. The temperature increase is carried out in an atmosphere of dry oxygen containing 1% HCl in such a way that the upper temperature limit is reached in 30 minutes. The formation of defect-free zones is formed by outdiffusion of oxygen at 1100° C. in two different atmospheres.
The different atmospheres are: the first in dry oxygen containing 1% HCl (to grow enough oxide on the wafer) for 30 minutes; the _second in 97.5% N2;
60 minutes in a gas mixture of 1.5% _O2 and 1% HCl. The furnace temperature was then increased over 20 minutes using the N ₂ /O ₂ /
The wafer is removed from the 900°C furnace by ramping down to 1000°C in an HCl atmosphere, and then ramping from 1000°C to 900°C in an _N2- only atmosphere over a further 20 minutes.
Note that this temperature change from 1100°C to 900°C needs to be done carefully so as not to cause distortion of the wafer. In this connection, considering that the risk of contamination at 1100°C is almost non-negligible and that the occurrence of stacking faults reaches its maximum value at that temperature, the above-mentioned gas atmosphere determination is I must say that it was extremely difficult. The various atmospheres mentioned above, especially N ₂ /
The O ₂ /HCl gas mixture was selected after numerous tests. It has been found that almost no stacking faults occur on the surfaces of wafers exposed to this particular gas mixture. The stacking faults were small in size and were limited to a band less than 1 mm wide around the wafer. The table below summarizes the heat treatments applied to all wafers, regardless of class, in step 4.

TABLE FIG. 5 is a diagram schematically representing the thermal adaptation cycle for each of the three classes. Finally, in step 5 of FIG. 2, the wafer is cleaned. This cleaning involves first immersing the wafer in a diluted HF solution, then in deionized water, then in chlorinated water (to make the wafer hydrophilic), and then again in deionized water. Soak in ionized water, then Huang
This is done by immersion in a solution. The wafer thus obtained can be used in both bipolar and unipolar device production lines. These wafers contain the desired number of connected clusters and are therefore stabilized, making them sensitive to possible variations in the manufacturing process when used in both unipolar and bipolar manufacturing lines. It has the advantage of not being One advantage of the method of the present invention is that it is applicable to individual wafers regardless of vendor or wafer nature, and as a result, the acquisition of new wafers from various vendors is more flexible. can be done in a highly economical way,
That's what it means. Note that while the above method is intended to handle wafers whose initial oxygen content is in the standard range of 20 to 40 ppmA, the method is intended to handle wafers whose initial oxygen content is outside that range, as described below. It is also applicable to certain types of wafers. That is, if the value exceeds 40 ppm, the initial number of clusters is large enough to omit step 3 (to generate clusters) (to form defect-free bands and combine clusters). Step 4) is carried out at 900° C. as described above. On the other hand, if the initial oxygen content of the wafer is less than 26 ppmA,
must be extended from 26 minutes to 60 minutes, as shown in step 3(a) of FIG. 5, for example. At this time, Step 4 is not changed. Now, in order to judge whether the object of the invention was achieved, the following items were checked: Surface conditions: (i) The thickness of the oxide layer and its deviation Δe were measured using ellipsometry. And the measured thickness is
From 109.7 nm to 116.4 nm, Δe ranged from 0.7 to 1.7 nm depending on the position of the wafer in the boat. (ii) After the deoxidation process, the surface was exposed using Wright Etch, and with the exception of the few stacking faults mentioned above,
No abnormal conditions were observed. (iii) The wafers were visually inspected before and after the deoxidation step and were found to meet the most stringent specifications for surface condition. (iv) Curvature measurements were taken at the beginning and end of the process.
As a result, no distortion of the wafer due to thermal cycling was observed. (v) X-ray topography was performed to detect crystalline dislocations, but no dislocations were found. Stability of oxygen solution: If the oxygen solid solution is unstable, heat treatment at 750°C should result in a change in resistivity due to donor formation. However (as mentioned above)
No such change in resistivity is observed upon heat treatment at 750° C. for 20 hours, thus demonstrating that the oxygen solid solution is stabilized. Number of clusters generated: Upon completion of the process according to the invention, it was found that the changes in the oxygen content of the wafer were minimized. Therefore, a heat treatment at 1050 °C for 24 h was performed to cause condensation on the stabilized clusters. From the amount of condensed oxygen, Ham's law (FSHam, JAP 30, 1518,
The number of condensation clusters was calculated using the following formula derived from Np = 2.3 × 10 ^-4 × D ^-3/2 × τ ^-3/2 × C ^-1/2 . is the oxygen diffusion coefficient at 1050℃,
C is the residual oxygen concentration and τ is the incubation time of the condensation process given by: τ=t/Log(C _O −C _L /C−C _L ) where C _O is the initial oxygen The concentration, C _L is the solubility limit at the applied temperature (here C _L =8 ppmA), and t is the duration of the treatment. The number of clusters generated in all three classes of wafers subjected to the entire thermal cycle was calculated. The table shown below as an example shows the results obtained for several samples.

【table】

[Table] Note that these theoretical values give a good approximation by calculating the actual number of etch pits after wrapping and chemical surface exposure. Please note that this has been confirmed. Defect-free zone thickness: The bebel technique was used to measure the defect-free zone thickness in wafers processed at 1050° C. for 24 hours. This is because defects formed immediately after the pre-annealing operation are too small to be imaged. The curve in FIG. 6 represents the change in number of clusters generated versus defect-free band. As shown in this figure, a defect-free zone of at least 10 μm thickness is obtained for this cluster. Electrical properties of the defect-free zone (MOS test): The electrical properties of the defect-free zone within a few microns thick region below the surface of the wafer form a MOS capacitor, rapidly changing from the accumulation mode to the inversion mode. It was evaluated by examining the transition response when switching to . No particular problems were found with the pretreated wafers. And by testing such structures on beveled wafers, we found that the lifetime of the carrier decreases as it moves away from the wafer surface;
It was found that this indicates the existence of a defect gradient. G. Effect of the invention As described above, by applying the method of this invention, the wafer can be stabilized regardless of the initial oxygen concentration of the wafer, and it can be used in both bipolar device and unipolar device manufacturing lines. This has the advantage that it can be adapted as desired.

[Brief explanation of the drawing]

Figure 1 is a graph of temperature versus nucleation rate;
The figure is a block diagram of each process according to the present invention, Figure 3 is a graph of temperature versus amount of condensed oxygen, and Figure 4 is
FIG. 5 is a graph representing the thermal adaptation cycle according to the invention; FIG. 6 is a graph of the number of clusters versus the thickness of the defect-free zone. be.

Claims (1)

[Claims] 1. A step of grouping semiconductor wafers in a plurality of lots according to a plurality of initial oxygen content ranges, extracting at least one semiconductor wafer from each group as a sample, and subjecting the sample to a cluster generation temperature. a step of testing the oxygen condensation rate at substantially corresponding temperatures; a step of grading each group of semiconductor wafers into a plurality of classes according to the above oxygen condensation rate; and a step of grading and classifying each group of semiconductor wafers into a plurality of classes under substantially the same oxygen atmosphere for each class. heating each class of semiconductor wafers with different heating cycles to produce clusters in each semiconductor wafer corresponding to a predetermined cluster density; heating all classes of semiconductor wafers above under an atmosphere containing oxygen and nitrogen; A standardized processing method for semiconductor wafers that can be used commonly for the production of unipolar and bipolar devices, comprising: heating with the same heating cycle higher than the heating cycle.