JPH0236546A - Test of semiconductor memory - Google Patents

Abstract

PURPOSE:To make it possible to evaluate the evaluation of the soft error of a semiconductor memory correctly with good precision compared to a conventional test method by a method wherein the market failure rate of the malfunction due to alpha rays of the semiconductor memory is computed from the relation between the variability distribution of the proof strength against alpha rays between memory cells and the distribution of the amounts of collecting charges to the memory cell parts. CONSTITUTION:The amount of a collecting charge to each memory cell part is computed on the basis of the irradiation condition of alpha rays, the sizes of the memory cells, the condition of a process and the condition of a bias and by finding the inversion ratio of each memory cell in every a plurality of kinds of the irradiation conditions of the alpha rays, the variability distribution (a strength distribution) of the proof strength against the alpha rays between the memory cells is found. Moreover, the incidence conditions of alpha-particles emitted from a wiring material into the memory cells are simulated and moreover, by finding the amount of a collecting charge in each incidence condition, the distribution (a stress distribution) of the amount of a charge of noise is found. By finding an error rate from the relation between the strength distribution and the stress distribution, the evaluation of the soft due to the alpha rays of a semiconductor memory can be conducted correctly with good precision.

Description

DETAILED DESCRIPTION OF THE INVENTION [Field of Industrial Application] The present invention relates to a method for testing malfunctions (hereinafter referred to as soft errors) caused by alpha rays in a semiconductor memory.

[Conventional technology]

One of the malfunctions of semiconductor memory is a phenomenon called soft error. The radiation that causes this is alpha rays, and its sources are the M wiring materials that make up semiconductor memory, Si chips, and the plastic used to package them.
Natural radioactive substances such as U (uranium) and 7'A (thorium) that are contained in trace amounts in ceramics and other materials. And U.

Rh etc. emit α particles (H- (helium) atomic nuclei) by spontaneous decay, and when these α particles enter a semiconductor memory, electron-hole pairs are generated along this range, which generates noise charges. , which becomes a noise current and causes soft errors in semiconductor memory. The problem of such soft errors is likely to increase in the future as semiconductor memories become more highly integrated and faster.

Soft error evaluation methods have conventionally been described, for example, in the Journal of the Institute of Electronics and Communication Engineers, Vex, / 65 CA 2 PF, 6
High speed bipolar R with 1-66 (19ao) rα particles
As described in AM's Soft Error J, the alpha ray resistance of semiconductor memory is evaluated by the average time (average operating time) from when the semiconductor memory is irradiated with alpha rays until a 1-hat error occurs. . In addition, in order to find the soft error rate of a semiconductor memory, it is assumed that the noise charge distribution generated by α particles and the inverted charge distribution between memory cells in the semiconductor memory are Gaussian distributions, and the soft error rate is calculated from both distribution shapes. is estimated.

[Problem to be solved by the invention]

The above conventional technology has the following problems.

When evaluating the alpha ray tolerance of a semiconductor memory, the average operating time is usually measured under -m class alpha ray irradiation conditions (α particle energy, incident angle), but the alpha ray tolerance between memory cells within one chip is The variation distribution is 2 as shown in Figure 2 1.2.
When evaluating a semiconductor memory using only one irradiation condition,
There is a possibility of incorrect evaluation. That is, in FIG. 2, there are two types of α that can obtain the inverted charge amounts of Qo and Qt.
Considering the beam irradiation conditions, sample 2 under sQo irradiation conditions
Under the irradiation conditions of 1Q1, the α-ray tolerance of Sample 2 appears to be greater than that of Sample 1, and the evaluation results differ depending on the irradiation conditions.

In addition, when calculating the soft error rate of semiconductor memory, it is assumed that the noise charge distribution and the inversion charge distribution are Gaussian distributions as an approximate model, but both There is no mention of a specific method for determining the distribution shape, and it was not possible to determine the error rate.

The purpose of the present invention is to eliminate the problems of the prior art described above,
The object of the present invention is to provide a method that can accurately and accurately evaluate soft errors caused by alpha rays in semiconductor memories.

[Means to solve the problem]

The above objectives are α-ray irradiation conditions and memory cell size.

By calculating the amount of charge collected in the memory cell section from the process conditions and bias conditions, and determining the reversal ratio of the memory cell for each of multiple irradiation conditions, the variation distribution (strength distribution) of alpha ray tolerance between memory cells can be reduced. In addition, the noise charge distribution (stress distribution) is determined by simulating the conditions for the incidence of α particles emitted from the wiring material into the memory cell, and determining the amount of collected charge under each condition of incidence. This is achieved by determining the error rate from the relationship between strength distribution and stress distribution.

[Effect]

Whether or not a soft error occurs due to alpha rays in a semiconductor memory is determined by the amount of charge collected by alpha particles incident on a memory cell. The amount of collected charge depends on the α particle incident conditions (incident energy, incident angle, incident position) into the memory cell portion, memory cell size, process conditions, and bias conditions.

FIG. 5 is a schematic diagram showing a state in which α particles are incident on the memory cell portion of a bipolar RAM under the conditions of incident energy E and incident angle θ, and electron-hole pairs are generated along the range. It is a diagram. Considering the charge collection model from the same figure,
As shown in Figure 4, the charge collection region is b ((transistor active region) + (depletion layer (between buried layer and substrate)) + (funneling length)) in the depth direction of the chip, and b ((transistor active region) + (depletion layer (between buried layer and substrate)) + (funneling length)) Well then! (width of the buried layer that generates the depletion layer), and the amount of collected charge Q can be expressed as Q=y(xt)−y(xt). Here, y(x) is the cumulative charge amount based on the end point of the α particle, x is the distance from the time the α particle enters the charge collection region to the end point, and xt is the distance from the time the α particle enters the charge collection region to the end point. This is the distance to the end point.

Further, xo in FIG. 4 indicates the penetration distance (range) of α particles in Si at incident energy E.

In order to examine the validity of the above model, we compared it with the energy-dependent characteristics of soft errors. Figure 5 shows the experimental results of energy-dependent characteristics. The figure shows four samples 4,
5 and 6.7, the average operating time of the semiconductor memory was measured for each irradiation energy while keeping the incident angle constant. Each sample showed concave characteristics, and the energy was approx. It can be seen that the minimum value exists near 9 MaV. Next, FIG. 6 shows the relationship between the α particle energy and the amount of collected charge obtained from the charge collection model. From the same figure, the energy value at which the collected charge amount is the maximum value is 5.8 MgV
It can be seen that it is nearby. Comparing with the above experimental results,
Since the minimum value of the average operating time and the energy value indicating the maximum value of the collected charge amount are equal, it can be said that the above charge collection model is appropriate.

From the above, the amount of collected charge can be calculated by determining the α-ray irradiation conditions, memory cell size, process conditions, and bias conditions. For example, as shown in FIG. 7, Qo=Qt,
The strength distribution of a semiconductor memory can be determined by irradiating the semiconductor memory under four types of α-ray irradiation conditions such that a collected charge amount of Qt-Qs is obtained and measuring the inversion ratio of the memory cell for each collected charge amount.

On the other hand, the conditions for the α particles emitted from the wiring material of the semiconductor memory to enter the memory cell are as shown in FIG. , Y, and Z. Here, N(E) is the α-ray energy distribution emitted from the wiring material, and Z is the charge collection region 5.
Represents the distance from the top end to the wiring material. Here, since the conditions for emitting α particles from the wiring material (θ, ψ, We can find y. Then, from each incident condition (ψ, x, y), as shown in FIG. By substituting j into the charge collection model equation, the amount of charge collected by α particles generated from the wiring material can be calculated. FIG. 9 shows the occurrence rate distribution (noise charge amount distribution or stress distribution) of the collected charge amount obtained by the above method.

From the above, the market failure rate λ for soft errors is the first
As shown in Figure 0, K is the overlapping part Q of the stress distribution n(Q> and the strength distribution F(Q).-Qmaz, where K is the amount of α particles released from the wiring material.

〔Example〕

An embodiment of the present invention will be described below with reference to FIG.

The testing method for alpha ray resistance of semiconductor memory is as follows:
Set the α-ray irradiation conditions (irradiation energy, irradiation angle, irradiation time, etc.). In step 2.

α-ray irradiation conditions, memory cell size, process conditions.

The amount of charge collected in the charge collection region of the memory cell is calculated from the bias conditions using a charge collection model formula.

In step 3, irradiation of the semiconductor memory with alpha rays is started, and in step 4, the address that caused the error is detected and counted. In step 5, compare the irradiation time of α rays (the time required for all memory cells to be hit by α particles at least once), and if NO, return to step 4. In step 6, α
Determine whether to change the radiation irradiation conditions, and if Yes, return to step 1. In step 7, a strength distribution representing the relationship between the amount of collected charge and the memory cell reversal rate as shown in FIG. 7 is determined. In step 8, the alpha ray tolerance of the semiconductor memory is determined from the slope and intercept of the strength distribution.

Next, in the method for calculating the market failure rate of soft errors, in step 9, conditions for the incidence of α particles emitted from the wiring material into the memory cell charge collection region are determined by Monte Carlo simulation. In step 10, the amount of charge collected in the charge collection region is calculated from the incident conditions obtained in step 9 using the charge collection model formula. In step 11, the occurrence rate distribution (stress distribution) of the amount of collected charge as shown in FIG. 9 is determined.

In step 12, the failure rate is calculated from the overlapping portion of the strength distribution in step 7 and the stress distribution in step 11.

As described above, according to this embodiment, it is possible to express the variation distribution of alpha ray tolerance between memory cells of a semiconductor memory and the noise charge distribution due to alpha particles emitted from the wiring material by the collected charge amount. By estimating the soft error rate from both tangibles, it is possible to accurately evaluate soft errors in semiconductor memories.

〔Effect of the invention〕

According to the present invention, it is possible to evaluate semiconductor memory soft errors more accurately than ever before, and even when improving and developing a product, it is possible to evaluate stress caused by memory cell size, process conditions, and bias conditions. By estimating the distribution and strength distribution, it is possible to accurately evaluate performance against soft errors.

[Brief explanation of the drawing]

FIG. 1 is a diagram showing a test procedure regarding soft error evaluation of a semiconductor memory according to an embodiment of the present invention, FIG. 2 is a diagram showing an example of an inverted charge amount distribution of a semiconductor memory, and FIG. Figure 4 is a simplified model diagram for determining the amount of collected charge; Figure 5 is a diagram of alpha particle energy dependence characteristics of soft errors in semiconductor memory; Figure 6 is a simplified model. Figure 7 is a diagram showing the relationship between α particle energy and the amount of collected charge according to the formula. Figure 7 is a diagram showing the relationship between the amount of collected charge and memory cell reversal rate, which shows an example of the strength distribution. Figure 9 shows the α particle incidence conditions, Figure 9 is a stress distribution diagram showing the relationship between the amount of charge collected from wiring materials and the occurrence rate, and Figure 10 shows how to determine the market failure rate using strength distribution and stress distribution. This is a diagram.

Claims (1)

[Claims] 1. In a method of irradiating a semiconductor memory under test with alpha rays and testing malfunctions caused by alpha rays in the semiconductor memory, charge collection of the semiconductor memory is performed based on alpha ray irradiation conditions, memory cell size, process conditions, and bias conditions. A means for calculating the amount of charge collected in a region and varying α-ray irradiation conditions to detect a malfunction rate for each amount of collected charge and detect variations in alpha-ray tolerance between memory cells in the semiconductor memory. a simulation means for determining the distribution of the amount of collected charge generated in the charge collection region by the α rays emitted from the semiconductor memory, and a means for calculating the distribution of the amount of collected charge generated in the charge collection region by the α rays emitted from the semiconductor memory, based on the distribution of variation in the amount of α ray withstand between the memory cells and the distribution of the amount of collected charge, A method for testing a semiconductor memory, characterized in that a market failure rate of malfunctions caused by alpha rays of the semiconductor memory can be calculated.