JPH06273479A - Method and device for evaluating reliability of wiring - Google Patents

Abstract

PURPOSE:To evaluate conduction life without depending on wiring length by completing the measurement, of a sample wiring length at a point when the shortest conduction life is obtained out of the conduction life of a plurality of samples. CONSTITUTION:A sample 21 and Si substrate 21 are placed inside a constant- temperature bath 30. A constant current is applied from current terminals 23 and 24 through a power supply 31. At the same time, the voltage among voltage terminals 25-29 which are spaced equally is measured 32 and then the resistance between the voltage terminals is calculated according to the relationship between current and voltage. The distance between the voltage terminals is changed from 100 to 100mum for evaluation. The point when wiring resistance increases by 20 % in reference to the point where the resistance becomes stable and reaches a constant value is set as the conduction life of the sample. As the wiring length is reduced, the distribution of the conduction life is increased but the conduction life cannot be reduced without any limitation and a shortest value Lt0 exists. Namely, the conduction life is longer than Lt0 regardless of the wiring length. By estimating the wiring life with Lt0, a highly reliable evaluation can be obtained. Also, it is sufficient if a life which is approximately 20% of the entire part is made known and evaluation can be made quickly.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a reliability evaluation method for wiring, particularly semiconductor device wiring, and a thin film wiring whose reliability is ensured by the evaluation method. In particular, a thin film suitable for ULSI having a laminated structure. The present invention relates to an energization life evaluation method for wiring, an evaluation apparatus, and wiring whose reliability is guaranteed by the evaluation method.

[0002]

2. Description of the Related Art In the past, as a method for evaluating the energization life of a semiconductor device wiring, a take-wow "effect of metal line geometry on electromigration lifetime in aluminum kappa submicron interconnections" I.E.
k, “EFFECT OF METAL LINE GEOMETRY ON ELECTROMIGRA
TION LIFETIME IN Al-Cu SUBMICRON INTERCONNECTION
S ”, IEEE IRPS, p. 185 (1988). Fig. 2), using a 4-terminal sample, applying a high-density constant current above the actual operating conditions, electro Accelerated life tests have been used to accelerate migration.In this evaluation method, the time until the sample breaks is defined as the electrical life of the sample, but the Al alloy / refractory metal laminated structure currently used is used. In terms of wiring, JAC Ondolsec Etoal “Effective Kinetic Variations with Stress Duration”
Four Multi-Layered Metallization (JC Ondrusek et al., "EFFECTIVE KINET
IC VARIATIONS WITH STRESS DURATIONFOR MULTILAYERED
METALLIZATIONS ", IEEE IRPS, p.179 (1988).), Even if the wiring layer made of an Al alloy is broken, the remaining high melting point metal layer maintains conduction, and the sample resistance is somewhat Although it rises, it does not reach the disconnection.
The energization life is measured by the time until the sample resistance rises above a certain ratio compared to when the energization starts. In this way, the energization life of a plurality of samples was obtained, and the time until the energization life of a certain ratio or more of all samples reached the energization life was used as the energization life of the wiring. In addition to direct current, the current to be applied is not limited to DC current, and JS Suehle and HA Schafft, "CU
RRENT DENSITY DEPENDENCE OF ELECTROMIGRATION t ₅₀ E
NHANCEMENT DUE TO PULSED OPERATION ", IEEE IRPS, p.
106 (1990).), Pulsed current is also used.

[0003]

In the above prior art, even if a part of the wiring is damaged, it is a small part of the entire wiring and the wiring functions normally while the resistance increase is small. Stand on the assumption. However, for example, in a semiconductor integrated circuit device, the length of the wiring connecting the individual devices varies. A life evaluation wiring sample with a length of several millimeters produces a small resistance increase that is negligible compared to the resistance of the entire wiring.If damage occurs in a short wiring, a relatively large resistance increase and the semiconductor integrated circuit device malfunctions. Will end up. Therefore, according to the related art, it is not possible to accurately evaluate the reliability of the laminated wiring in which the occurrence of damage does not directly lead to the disconnection.

The object of the present invention is also to evaluate the energization life in a laminated structure wiring without depending on the wiring length or to compare the energization life depending on the wiring length, and to evaluate the energization life by the evaluation method. To provide guaranteed high reliability wiring.

In order to detect a small damage, it is necessary to shorten the wiring length and evaluate the wiring life.
In this case, even if the number of samples is the same, the total length of the wiring to be evaluated becomes short. Therefore, it is necessary to increase the number of samples to be measured for highly reliable evaluation. Another object of the present invention is to provide an energization life evaluation device capable of efficiently collecting the wire life of a large number of samples having a short wire length.

[0006]

Among the above objects, the evaluation of the energization life without depending on the wiring length can be achieved by two different methods. One of them is to apply different currents A and B (A <B, one energization time Ta at current A> one energization time Tb at current B) to the sample at least once in the order of A and B. The voltage is applied and the service life Ls of the sample is defined as the total time during which the current A is applied until the sample resistance increases by a certain ratio or more as a result of energization.
In this case, the current C (A <C,
The reliability of the test result is further enhanced by energizing the energizing time Ta at the current A> the energizing time Tc at the current C), and excluding the sample that is disconnected at the first energization of the current C from the result as an initial defective product. Becomes Further, a pulse current having a constant cycle can be used as the current B.
As a modification of this method, the application of the currents A and B is performed only once, and the sample resistance is measured after the application of the current B. When the sample resistance increases by a certain ratio or more compared to the time when the test is started, the life Ls of the sample is determined. There is also a method in which the life Ls of the sample is set to be longer than the energization time Ta at the current A when the increase in the sample resistance is less than a certain ratio at the energization time Ta at the current A or less.
The lifespans of a plurality of samples are obtained by these methods, and the time until the lifespan of the samples is generated at a certain ratio or more is defined as the lifespan L of the wirings forming the samples.

Another method is to measure the change over time in the sample resistance while applying a constant current, and in an energization life test in which the sample life is the time Ls until the sample resistance rises above a certain ratio by energization, The shortest of the energization lives Ls of a plurality of samples having the same sample wiring length m, Lt (m)
Is measured by changing Lt (m) in the direction of shortening m until it takes a constant value Lt ₀ which does not depend on m, and Lt ₀ is a life L ₀ that does not depend on the wire length of this wire. Is to evaluate.

In addition, in an energization life test in which the life of the sample is the time Ls until the sample resistance rises above a certain ratio by energization, the wiring having different energization life distribution depending on the length of the sample is evaluated. The number n of samples reaching the energization life when the total length of the wiring is constant, and the wiring length m of the sample is m ₁
A constant value n ₀ is taken within the range of <m ≦ m ₂ , and this n ₀ is n (m)
If it is larger than (m> m ₂ ), the wiring length of this wiring is m ₁
By defining the defect occurrence rate per unit length of the wiring evaluated by <m <m ₂ as n _MAX (m ₁ , m ₂ ), different materials or structures can be used even when the energization life changes depending on the wiring length. It becomes possible to compare the energization life between the wirings having the.

Further, a plurality of voltage measuring terminals are provided between both current applying terminals of the wiring sample, each terminal is treated as one sample, and a change with time of the resistance between the terminals is measured, and the terminals are connected in parallel with the sample. A short circuit with an open / close switch is installed on each of the terminals, and the resistance increases between each terminal treated as a sample. It is possible to efficiently collect the wiring life of many small samples.

[0010]

In the former method of evaluating the energization life without depending on the wiring length, in the former method, the application of the current A corresponds to the conventional energization test and occurs in the wiring due to the application of the current B having a higher density than the current A. It is what makes the damage manifest. If no damage is generated in the wiring due to the application of the current A, even if the current B is applied for a short time, the wiring is not so affected. On the other hand, when damage such as half-breakage occurs in the wiring during energization, the growth rate of damage increases rapidly as the applied current density increases, so that the growth of the damaged portion accelerates when the current B is applied. The resistance increases sharply. Therefore, even a small damage in a long wiring, which cannot be found by the conventional energization life evaluation method, can be detected, and it is possible to evaluate the energization life not affected by the wiring length even in the wiring of the laminated structure. On the other hand, in the latter method, it is assumed that the sample wiring is used and the number of samples is increased to evaluate the life. Normal,
If damage such as half-breakage occurs in the wiring while it is energized, the resistance of the damaged part is larger than that of the normal part, so the wiring resistance increases, but when the wiring length is large, the ratio of the increase to the total resistance of the wiring Cannot be detected because is small. Therefore, the length m of the sample wiring is shortened and the sensitivity to damage occurrence is increased to perform the measurement. Since this method requires measurement for different m, the total number of samples increases, but the measurement is not performed until all the samples reach the end of their lives, but the shortest one Lt (m Since the measurement for the sample having the length m of the sample wiring is completed at the time when) is obtained, the measurement can be completed in a short time. If Lt (m) does not become shorter even if m is shortened, this Lt (m) can be considered as the time until damage occurs due to energization. According to this method, it is possible to efficiently evaluate the energization life not depending on the wiring length in a short time even in the wiring of the laminated structure.

Further, the shorter the wiring used for evaluation is, the larger the increase in resistance is detected even if the damage causes only a small increase in resistance. Therefore, even if the defect occurrence rate per fixed number of samples is decreased, the length is fixed. The number of defective occurrences per hit tends to increase as the wiring becomes shorter. Therefore, the total number of samples to be evaluated is n
The sample wiring length m takes a constant value n ₀ in the range of m ₁ <m ≦ m ₂ , and when this n ₀ is larger than n (m) (m> m ₂ ), n _MAX (m ₁ , m ₂ ) = N ₀ is the wiring length of this wiring m ₁ <m
By defining the defect occurrence rate per unit length of the wiring evaluated by <m ₂ , it is possible to compare the energization life that varies depending on the wiring length between the wirings having different materials or structures.

Further, by using the energization life evaluation device, one long wiring sample can be subdivided and treated as a plurality of short wiring samples, and the wiring lives of many samples can be efficiently collected. Also, when the wiring sample is continuously energized, the wiring resistance increases due to damage, which leads to disconnection in extreme cases and overload of the power supply.However, a short-circuit switch installed in parallel between each terminal By using the closed short circuit as the current path, the sample resistance can be apparently reduced, and the power supply can be prevented from being overloaded. In particular, by controlling the opening and closing of the short-circuit switch by a computer, the wiring life evaluation in this method can be automated, which is particularly effective in improving the efficiency of measurement.

[0013]

【Example】

(Embodiment 1) An embodiment of the present invention will be described below with reference to the drawings. FIG. 1 is a schematic diagram showing the configuration of a wiring sample and an evaluation device used in the present invention. Current application terminals 3, 4 and voltage measurement terminals 5, 6 are provided at both ends of the wiring sample 2 formed on the Si substrate 1. During the energization life evaluation test,
The sample 2 is installed in the thermostat 7 together with the Si substrate 1, and a test is performed by applying a current from the programmable power source 8 from the current applying terminals 3 and 4. At the same time, the voltage between the voltage measuring terminals 5 and 6 is measured by the DC voltmeter 9, and the sample resistance is calculated from the relationship between the applied current and the voltage.

FIG. 2 is a sectional view of the wiring sample 2 taken along the line AB in FIG. Below, FIG. 3 showing the flow of manufacturing the wiring sample
A method of manufacturing the wiring sample 2 will be described below. The Si substrate 1 is thermally oxidized to form a thermally oxidized SiO ₂ film 10 of 5000 nm on the surface (FIG. 3A). A 200% 10% Ti-W alloy film 11 is formed on the substrate by DC-magnetron sputtering.
m, Al-1% Si-1% Cu film 12 is continuously formed at 500 nm without breaking the vacuum (FIG. 3B). This laminated film is patterned into a sample shape by the photolithography technique and the dry etching method (FIG. 3C). The width of the formed wiring is 500 nm. A 1000 nm thick phosphosilicate glass film 14 is formed on the wiring pattern 13 thus formed by an atmospheric pressure CVD method to form a passivation film. Pads of the current application terminals (3, 4) and the voltage measurement terminals (5, 6) are opened by photolithography and dry etching (FIG. 3 (d)). Further, it is annealed in a H ₂ atmosphere at 450 ° C. for 30 minutes to prepare a wiring sample. Although it can be used for an evaluation test at this stage, normally, the sample thus formed is diced and mounted on a ceramic package for wire bonding and then tested. FIG. 4 is a diagram showing a change in resistance when a current-carrying life evaluation by a conventional continuous DC application is performed on a wiring sample. When this energization life evaluation test is performed on the Al alloy single layer wiring (A), the energization life is found by the disconnection, but in the laminated wiring sample, the wiring resistance only gradually increases (B). Therefore, in the present embodiment, the fluctuation of the wiring resistance that does not affect the operation of the Si semiconductor element is 2 at maximum.
It was set to 0%, and the time at which the wiring resistance increased by 20% based on the time when the resistance value became stable after the start of energization and took a constant value was taken as the wiring life. 5 and 6 are views showing the influence of the wiring length of the sample on the energization life evaluation result. When the sample is an Al alloy single layer wiring (FIG. 5), the energization life tends to be shortened overall and the distribution thereof tends to become smaller as the wiring length increases. On the other hand, in the case where the sample is a laminated wiring (FIG. 6), the distribution of energization life also decreases as the wiring length increases,
This is because both short-lived samples and long-lived samples are smaller than when the wiring length is small. The energization life of the wiring is determined by the time until the energization life reaches a certain ratio or more of the sample, and in this embodiment, this ratio is set to 1%. At this time, in the laminated wiring (FIG. 6), the longer the wiring length is, the longer the energization life of the wiring tends to be. Wirings of various lengths are used in actual semiconductor integrated circuit devices, and reliability should be made to the wiring with the shortest life among them, and in the case of FIG. Even if the life is evaluated, the reliability of the wiring cannot be guaranteed.

On the other hand, the present embodiment intends to solve the problem by making the method of applying the current different from that of the prior art. FIG. 7 is a diagram showing a change in applied current with time when a current is applied to a sample according to the present invention in comparison with a conventional technique. Conventionally, a constant current A (10 mA) as shown in FIG.
Was applied. On the other hand, in this embodiment, as shown in FIG. 7B, the same current as A is applied for 2 hours, and then a current B (20 mA) larger than A is applied for 1 minute, and this is repeated thereafter. As another embodiment of the present invention, in FIG. 7C, in addition to the current series shown in FIG.
mA) is applied for 2 minutes. Further, the present invention is
As shown in FIG. 7D, instead of the current B in FIG. 7B, a pulse current D (peak current 60 mA, on-time 1 μsec, cycle 100 μsec) may be applied for 1 minute.

The results of evaluating the energization life of the sample wiring by the method of this embodiment will be described below. FIG. 8 is a diagram showing a resistance change when the energization life test is performed by applying the current series of FIG. 7B to the above laminated wiring sample. Although the wiring resistance gradually increases due to energization and reaches the end of its life (A), the time from when the wiring resistance starts to rise to the end of its life is shorter than that of the prior art (B). FIG. 9 is a view showing the influence of the wiring length of the sample on the evaluation result of the laminated wiring energization life by the current series shown in FIG. 7B. The distribution of energization life tends to decrease as the wiring length increases, but the distribution on the short life side, which determines the wiring life, is not significantly affected. Therefore, unlike the case of the conventional laminated wiring evaluation (FIG. 6), it is possible to increase the total extension of the evaluation sample by using a long wiring sample and improve the reliability of the energization life evaluation.

Here, the main points of this embodiment will be described. In this embodiment, it is important to apply a large current to the sample for a short time. Even if a large current is applied, the wiring is hardly newly damaged for a short time of about 1 minute. On the other hand, on the other hand, the growth rate of a damaged portion such as a half-break is greatly affected by the applied current density and is accelerated by the 4th to 6th power of the current density.
Therefore, when there is damage in the wiring, the growth of the damage is accelerated by applying a large current, and the wiring resistance rapidly increases due to the growth of the damage even if the application time is short. With the conventional constant current application, even if damage occurs due to energization, the growth rate of damage is small.Therefore, until the life of energization is reached, the current is applied for a longer time to increase the resistance increase due to this damaged part or Need to cause damage. On the other hand, in the present invention, when even one damage occurs due to the energization, the damage rapidly grows by the subsequent application of a large current for a short time, the wiring resistance is increased, and the sample can reach the energization life. Therefore, the life of the wiring can be evaluated without being affected by the wiring length of the sample.

As the short-time applied current (current B in FIG. 7B) for accelerating the growth of the damaged portion, the test current (FIG. 7) is used.
It is preferable to use a current which is 1.5 times or more, preferably 2 times or more as large as the current A) in (b), and the application time for one application is 1/10 or less of the application time for one application of the test current. . In addition, the number of times this damage acceleration current is applied is at least about 10 times, and it is preferable to perform a preliminary test by changing the test current application time once from several tens of minutes to several hours, and the energization life is to accelerate damage. Perform this test by selecting a range that is not affected by the number of times the application current is applied. Further, according to the application of the current C in FIG. 7 (c), the sample that had already been damaged before energization can be removed before the application of the test current because the resistance increases at this time, The reliability can be further enhanced. As the damage accelerating current, it is possible to use a pulse current instead of the direct current as the current D in FIG. 7D. The merit of using the pulse current is that it is possible to apply a large current that is impossible with a DC current if the pulse peak current is used. The growth of the damaged part is 4 to the current density.
Since it is accelerated by the sixth power, a large damage acceleration current is more advantageous. However, since the influence of Joule heat generation is large, the maximum current density for direct current is about 1 × 10 ⁷ A / cm ² . On the other hand, according to the pulse current application, it is possible to apply the peak current density up to about 1 × 10 ⁸ A / cm ² while suppressing the Joule heat generation by selecting the pulse period and the pulse on time. FIG. 10 is a diagram showing the resistance change of the sample when conducting the energization life test using various damage acceleration currents. It can be seen that even if the pulse current (cycle 100 μsec) is used for the damage acceleration, the result is almost the same as that when the direct current is used. When a pulse current is used, even if the time-averaged current density is the same, it is better to shorten the pulse-on time and increase the peak current density to shorten the time from the start of resistance increase to the end of life, and damage growth. Is being accelerated.

(Embodiment 2) Another embodiment of the present invention will be described below with reference to the drawings. The structure of the wiring sample and the evaluation apparatus used in this example is the same as that of the first example, and the schematic diagram 1, the cross-sectional view of the wiring 2, and FIG. 3 showing the sample preparation flow are common to the first example.

FIG. 11 is a diagram showing the change over time in the applied current when the current is applied to the sample according to the present invention. Test current A
As a result, after applying a current of 10 mA for a certain period of time (Ta time in FIG. 11), a damage acceleration current B larger than A (20 mA)
Is applied for a certain time (Tb time in FIG. 11: 10 minutes in this embodiment). After that, the resistance of the sample is measured and it is judged whether or not it has reached the end of its life depending on the magnitude of the resistance increase. In the present example, the time point when the sample resistance increased by 20% was defined as the current-carrying life, based on the time point when the resistance value became stable and reached a constant value after the start of current application. According to the present embodiment, it is not necessary to consider generation of new damage due to the damage acceleration current B because the evaluation of the sample to which the damage acceleration current B has been applied once ends at that point. However, as a preliminary examination, it is necessary to confirm that the life is not reached even if the current B is applied to the sample before the test current is applied for Tb time (10 minutes in this embodiment). This is for confirming that the sample having no damage before application of the test current does not reach the end of its life even if the damage acceleration current B is applied for Tb time. Test current A
Even if the damage acceleration current B is applied after applying Ta for
As long as is sufficiently small, no damage occurs due to energization, so most of the samples do not reach the end of their lives. However, when Ta becomes large, a sample appears to be damaged during the application of the test current A, and the resistance increases due to the application of the damage acceleration current B. Here, in this embodiment, the time until 1% of the entire sample reaches the life is defined as the wiring life. In the present invention, since it is necessary to evaluate many samples for one Ta, too many Tas are evaluated.
However, it is difficult to carry out an evaluation test. On the other hand, it is possible to evaluate in a short time because it is not necessary to test all samples until they reach the end of their lives.

Table 1 shows the results of evaluating the wiring life by applying the current shown in FIG.

[0022]

[Table 1]

The test current energizing time Ta was changed to three levels, and 400 samples were evaluated for each Ta, and the ratio of the samples that reached the end of their life was shown. When Ta is 100 hours, only 0.5% of the sample reaches the end of its service life, but when Ta is 200 hours, the life is 4%.
%, The sample has reached the end of its service life, and it can be seen that the wiring service life is longer than 100 hours and shorter than 200 hours.

(Embodiment 3) Another embodiment of the present invention will be described below with reference to the drawings. FIG. 12 is a schematic diagram showing the configuration of the wiring sample and the evaluation device used in this example. Si
Current application terminals 23 and 24 and voltage measurement terminals 25 to 29 are provided at both ends of the wiring sample 22 formed on the substrate 21. During the energization life evaluation test, the sample 22 including the Si substrate 21 is placed in the constant temperature bath 30, and a constant current is applied from the programmable power supply 31 from the current application terminals 23 and 24 to perform the test. At the same time, the voltage between the voltage measuring terminals 25 to 29 provided at equal intervals is sequentially measured by the DC voltmeter 32, and the resistance between the voltage measuring terminals is calculated from the relationship between the applied current and the voltage. Distance between voltage measuring terminals is 100 to 10
It was evaluated by changing the thickness at 00 μm.

The sectional structure of the wiring sample 22 and the procedure for manufacturing the wiring sample are the same as those in the first embodiment, and the sectional view of the wiring 2 and FIG. 3 showing the sample manufacturing flow are the same as those in the first embodiment. Figure 1
FIG. 3 is a diagram showing the influence of the wiring length of the sample on the current-carrying life evaluation result. After the start of energization, the time when the wiring resistance increased by 20% was defined as the energization life of the sample, based on the time when the resistance value became stable and took a constant value. In this embodiment, the wiring length of the sample is shorter than that of the conventional method, and voltage measuring terminals are provided at equal intervals for one wiring sample of about 2000 μm in the conventional method, and one sample is provided between the voltage measuring terminals. Handling. Therefore, compared to the conventional method, one programmable power supply is used to evaluate a large number of samples.

As shown in FIG. 13, when the wiring length is shortened, the distribution of energization life becomes wider, but the energization life is not shortened without limit and the shortest value Lt ₀ exists. That is, the energization life is longer than Lt ₀ regardless of the wiring length. Estimating the wiring life with this Lt ₀ enables highly reliable life evaluation. Further, in the present embodiment, attention is paid to a sample having a short energization life, and it is not always necessary to know the energization life of all the samples, unlike the conventional method of extrapolating the life-cumulative failure rate curve to determine the wiring life. It is sufficient if the life of about 20% of the whole is known. Therefore, the evaluation can be performed in a shorter time than the conventional method.

(Embodiment 4) Another embodiment of the present invention will be described below with reference to the drawings. 14 and 15 are diagrams showing the results of conducting an energization life test on two types of samples in which the Al-1% Si-1% Cu film 12 was formed under different conditions in Example 1. The sample A has a substrate temperature of 50 ° C. when the Al alloy film is formed, and the sample B has a substrate temperature of 300 ° C. According to the result of the current-carrying life test by applying a constant current to the two samples (Fig. 14), the wiring life is 69 hours for the sample A and 58 for the sample B.
Sample A having an Al alloy film formed at a substrate temperature of 50 ° C. has a longer life. On the other hand, in the case of the present example in which the current shown in FIG. 7B of Example 1 was applied and the energization life test was performed (FIG. 15B), the wiring life was 33 hours and 53 hours, respectively. , The magnitude relationship between the two is reversed from the evaluation result of the conventional method. Therefore, a Si semiconductor integrated circuit element is manufactured using these two kinds of wiring, and the ambient temperature is set to 8
The device was continuously operated at 0 ° C. and a reliability evaluation test of the device was performed.
Table 2 shows the number of elements that have malfunctioned due to poor wiring continuity.

[0028]

[Table 2]

It can be seen that the device using the wiring manufactured at the substrate temperature of 300 ° C. has a smaller number of defects and higher reliability. That is, the conventional wiring life evaluation method does not correctly evaluate the life of the wiring, and the reliability of the wiring cannot be guaranteed due to the life of the wiring. On the other hand, the wiring life evaluation method according to this embodiment reflects the reliability evaluation test result of the semiconductor integrated circuit element, and it can be said that the longer the wiring life determined according to this embodiment is, the more reliable the wiring is. Also,
The longer the life of the wiring used according to the present embodiment is, the more reliable the semiconductor integrated circuit device is.

(Embodiment 5) An embodiment of the present invention will be described below with reference to the drawings. FIG. 16 is a schematic diagram showing the configuration of the evaluation device of the present invention. Current applying terminals 102 and 103 are connected to both ends of the wiring sample 101, and a current is supplied from a power source 104. The scanner 1 out of the plurality of voltage measuring terminals 106
08, the voltmeter 105 selects a terminal for performing voltage measurement. Here, the power supply 104, the voltmeter 105, the short-circuit switch 107, and the scanner 108 are the control computer 1
The voltage between the terminals when a constant current is continuously applied to the sample is automatically measured at equal time intervals.
The short-circuit switch 107 is closed between terminals whose resistance increases by a certain amount or more as a result of measurement, and a current is passed through a low-resistance short-circuit path. Here, the test was performed by applying a constant current to the sample, but the test can be performed by applying a current with various energizing patterns that can be controlled by the computer 109, such as applying a pulse current with a constant current and a constant period. When a test at a constant temperature is required, the wiring sample 101 is connected to the current applying terminals 102 and 10.
3 and the voltage measurement terminal 106 are held in a constant temperature bath and a test is performed.

The sectional structure of the wiring sample used in this example is the same as that of the example 1, and the sectional view 2 of the wiring and FIG. 3 showing the sample preparation flow are common to the example 1.

FIG. 17 is a diagram showing the resistance change when the energization life is evaluated by continuously applying a DC constant current. The sample had a wiring width of 2.0 μm and a total wiring length of 2000 μm, and was tested at an ambient temperature of 150 ° C. In this embodiment, the wiring is 200 μ
A voltage measuring terminal was provided every m, and a wiring having a total length of 2000 μm was treated as 10 200 μm wirings arranged in series.
The sample is damaged by energization and the resistance gradually increases. In the conventional method, the wiring of 2000 μm is treated as one sample, so even if the resistance is greatly increased by one damage in the wiring, it is judged that the life has reached the end, and if the current is further continued, the load on the power supply will be finally reached. The upper limit of is exceeded and the test cannot be continued any more. Even if one wire is subdivided and treated as many samples with a short wire length, if the increase in the wire resistance is mainly due to one of the samples, the resistance of the whole wire is exceeded until the upper limit of the load on the power supply is exceeded. However, even if the number of samples increases, the number of samples that have actually reached the service life is limited to one or two, and the purpose of efficiently measuring the service life of many samples cannot be achieved. On the other hand, in the present embodiment, among the samples formed by subdivision, those in which the increase in resistance became a certain value or more were closed by closing the short-circuit switches installed in parallel to make the low-resistance short-circuit path the current path, It is possible to reduce the apparent resistance of the entire wiring and keep the load on the power supply small, and to continue the test until the life of all samples is reached. In FIG. 17, the short-circuit switch is closed when the resistance value of the wiring of 200 μm becomes 1.5 times or more of that at the time of energization. Therefore, the resistance increase of the entire wiring from the beginning of energization is 10% or less, and the test can be continued without worrying about the load on the power supply. FIG. 18 is a diagram showing another pattern of resistance change at the time of energization life evaluation by continuous DC application. In addition to the pattern in which the resistance gradually increases as shown in FIG. 17, the wiring resistance may increase rapidly due to the energization, and as shown in FIG. In this case, the wiring resistance temporarily exceeds the upper limit of the power supply load, and it becomes impossible to apply a constant current.
In this embodiment, when the control computer detects a control error signal from the power source, the energization is temporarily stopped and the resistance of the 10 samples arranged in series is sequentially measured. For samples with remarkably large wiring resistance, close the short-circuit switch and use the low-resistance short circuit as the current path to reduce the resistance of the entire sample. After that, the energization test is restarted and the constant current application is restored. In the present embodiment, the short-circuit switch of the sample, which has a resistance that is more than twice the resistance at the beginning of energization in the wiring resistance measurement when a power supply control error occurs, is closed.
When closing the short-circuiting switch, there are two cases, that is, when the wiring resistance is gradually increased as shown in FIG. 17 and when the resistance is rapidly increased as shown in FIG. It is desirable to set ∑ to be more lenient, that is, the short-circuit switch will not be closed until a greater resistance increase. This is a condition for continuing the measurement for each sample as much as possible. On the other hand, if the judgment condition for a sudden resistance increase is more severe, the short-circuit switch will be closed and the measurement will be interrupted by the wiring resistance measurement when a power supply control error occurs before the sample resistance increases to the set value. The situation may occur.

FIG. 19 shows the situation in which damage is caused to the entire wiring as a result of conducting the test by subdividing the wiring sample according to the present embodiment and continuing the energization even after the resistance of the entire wiring is increased due to the damage caused by the energization. It is the figure which showed typically. In FIG. 19, the test was performed by dividing the wiring having a total length of 3200 μm into 32 equal parts, and the energization time was increased to 10, 20, 30, 40 hours in the order of (a) → (b) → (c) → (d). There is. Three
The wiring resistance at the beginning of energization of each of the two parts is 10 to 1
When the resistance of the sample was within 1 Ω and the resistance of the sample increased by 10% or more from the beginning of the energization, the energization life of the sample was defined. The short-circuit switch was set to close when the wiring resistance was more than double that at the time of energization.

Table 3 is a summary of the results shown in FIG.

[0035]

[Table 3]

Since the resistance of all parts divided into 32 equal parts is measured, it is possible to know the life when the wiring is divided into 16 equal parts, 8 equal parts and so on. In the wiring evaluated in this example, a sample subdivided into eight equal parts or more, that is, 400
The samples with a wiring length of μm or less have the same number of samples that have reached the end of life in each measurement from (a) to (d). It was the same. This number of samples is considered as the number of defects generated by energization, and the defect occurrence rate is (a): 1/3200 μm (100
μm, 800 μm), 4 pieces in (d) / 3200 μm (100
.mu.m, 400 .mu.m) and can be used for comparison of energization life between wirings having different materials or structures.

(Embodiment 6) Another embodiment of the present invention will be described below with reference to the drawings. FIG. 20 is a schematic diagram showing the configuration of the evaluation device of the present invention. The current application terminals 122 and 123 are connected to both ends of the wiring sample 121, and a current is supplied from the power supply 124. From the plurality of voltage measuring terminals 126, the terminal for performing voltage measurement with the voltmeter 125 is selected by the scanner 129. Here, the power supply 124, the voltmeter 125, the short-circuit switch 127, and the scanner 129 are controlled by the control computer 130, and the voltage between terminals when a constant current is continuously applied to the sample is automatically measured at equal time intervals. It The short-circuit switch 127 is closed between terminals whose resistance increases by a certain amount or more as a result of measurement, and a current is passed through a low-resistance short-circuit path.

Reference numeral 128 is a short circuit resistance which can be set to an arbitrary value. By providing the short-circuit resistor 128, the resistance fluctuation of the entire sample when the short-circuit switch 127 is closed is reduced, and the load on the power supply 124 is reduced. Also, when the applied current is not direct current but high frequency or rectangular wave, the applied current waveform is deformed by the impedance of the sample,
It is desirable to minimize the impedance change of the wiring during the test. Therefore, a short-circuit resistor 128 or short-circuit impedance that can be set to an arbitrary resistance value is provided,
It is effective in conducting a highly reliable test to minimize the change in the applied current waveform when the short-circuit switch 127 is closed.

[0039]

In the conventional energization life evaluation method, the required life depends on the wiring length of the sample, and evaluation with a long wiring sample does not lead to improvement in reliability of the evaluation result. Further, in some cases, the evaluation result may be contrary to the reliability evaluation result of the actual device. On the other hand, according to the present invention, it is possible to efficiently obtain the wiring life which is not affected by the wiring length. Therefore, the reliability of the semiconductor integrated circuit device in which wirings of various lengths are used can be guaranteed by the energization life required according to the present invention.
It is possible to provide a highly reliable wiring by selecting a wiring having a long energization life by the evaluation method of the present invention, and it is also possible to provide a highly reliable semiconductor integrated circuit element by using the wiring. .

[Brief description of drawings]

FIG. 1 is an explanatory diagram showing a configuration of a wiring sample and an evaluation device.

FIG. 2 is a cross-sectional view of a wiring sample.

FIG. 3 is a process diagram showing a flow of manufacturing a wiring sample.

FIG. 4 is a characteristic diagram of changes over time in wiring resistance during a DC energization test.

FIG. 5 is a characteristic diagram showing a relationship between energization life and cumulative defective rate in a single-layer wiring.

FIG. 6 is a characteristic diagram showing a relationship between energization life and cumulative defective rate in laminated wiring.

FIG. 7 is a timing chart showing a test current application pattern.

FIG. 8 is an explanatory diagram of changes in wiring resistance with time according to the present invention.

FIG. 9 is a characteristic diagram showing a relationship between energization life and cumulative defective rate in the laminated wiring according to the present invention.

FIG. 10 is a characteristic diagram of changes in wiring resistance over time when various damage acceleration currents are used.

FIG. 11 is an explanatory diagram showing a test current application pattern.

FIG. 12 is an explanatory diagram showing a configuration of a wiring sample and an evaluation device.

FIG. 13 is a characteristic diagram showing the influence of the wiring length of the sample on the energization life-cumulative failure rate curve.

FIG. 14 is a current-life-cumulative fraction defective curve diagram of an Al alloy laminated wiring formed at a substrate temperature of 50 ° C.

FIG. 15 is a current-life-cumulative fraction defective curve diagram of an Al alloy laminated wiring formed at a substrate temperature of 300 ° C.

FIG. 16 is an explanatory diagram showing configurations of an evaluation device and a wiring sample according to the present invention.

FIG. 17 is a characteristic diagram showing an example of a change over time in wiring resistance during a DC energization test according to the present embodiment.

FIG. 18 is a characteristic diagram showing another example of changes over time in wiring resistance during a direct current test according to the present embodiment.

FIG. 19 is an explanatory diagram showing a situation of damage occurrence due to energization.

FIG. 20 is an explanatory diagram showing the configuration of another evaluation device and a wiring sample according to the present invention.

[Explanation of symbols]

1 ... Si substrate, 2 ... Wiring sample, 3, 4 ... Current application terminal, 5, 6 ... Voltage measurement terminal, 7 ... Constant temperature chamber, 8 ... Programmable power supply, 9 ... DC voltmeter, 10 ... Thermal oxidation SiO ₂
Film, 11 ... 10% Ti-W alloy film.

─────────────────────────────────────────────────── ─── Continuation of the front page (51) Int.Cl. ⁵ Identification code Office reference number FI technical display location H01L 21/3205 // H05K 10/00 7128-4E

Claims (16)

[Claims] 1. A voltmeter for measuring a voltage between a power supply terminal for applying a current to the sample and a sample having a current applying terminal and a sample resistance measuring terminal at both ends of the sample as a current-carrying portion between the current applying terminals, While applying different currents A and B to the sample at least once in the order of A and B, the change with time of the sample resistance is measured,
The time until the sample resistance rises above a certain ratio by energization is determined, and the total time Ls during which the current A is passed is taken as the life of the sample, and the time until a life of a plurality of samples reaches a certain ratio or more. Is the life L of the wiring that constitutes the sample. 2. A voltmeter for measuring a voltage between a power source terminal for applying a current to the sample and a sample having a current applying terminal and a sample resistance measuring terminal at both ends of the sample as a current-carrying portion between the current applying terminals, Different currents A and B are applied to the sample in the order of A and B, and the sample resistance after the current B is applied is measured. When the increase of the sample resistance is less than a certain ratio for the energization time Ta at A or less, the life Ls of the sample is assumed to be larger than the energization time Ta at the current A, and a plurality of samples reach the end of their life at a certain ratio or more. A method for evaluating an energization life of a wiring, characterized in that a time until it is set is a life L of a wiring constituting a sample. 3. A voltmeter for measuring a voltage between a power supply terminal for applying a current to the sample and a sample having a current applying terminal and a sample resistance measuring terminal at both ends of the sample as a current-carrying portion between the current applying terminals, In the energization life test of the wiring, the same composition and cross-sectional structure were measured in the energization life test in which the sample life was measured by measuring the change over time in the sample resistance while applying a constant current and increasing the sample resistance to a certain ratio or more by energization. While the wiring has a different energization life depending on the length of the sample and has a tendency that the energization life becomes longer as the sample becomes longer in at least a part of the length region, a plurality of wires having the same sample length m are provided. the Lt (m) the shortest among the current life Ls of samples, a constant value Lt ₀ direction Lt (m) does not depend on m to shorten the m
And Lt ₀ is defined as a life L ₀ that does not depend on the wire length of the wire. 4. The current is measured by a sample having a current applying terminal and a sample resistance measuring terminal at both ends of the sample and a power source for applying a current to the sample and a sample having a current-carrying portion as a current-carrying part. Measure the change over time in the sample resistance due to application,
In the current-carrying life test of the wiring, the time Ls until the resistance of the sample rises to a certain ratio or more due to energization is determined as the life of the sample, and a sample having a resistance value higher than that is judged to be defective. The number n of samples that reach the energization life when the total extension of the wiring to be evaluated is constant for wirings having different energization life distributions depending on the length of the wiring, and the wiring length m of the sample is m ₁ <m ≦ m ₂ takes a constant value n ₀ in the range, the n ₀ is n (m) (where m> m ₂₎ wiring length m ₁ of the wiring is larger than <m <per unit length of the wiring was evaluated in m ₂ The method for evaluating the energization life of wiring is characterized in that the defect occurrence rate is defined as n _MAX (m ₁ , m ₂ ). 5. The electric current A according to claim 1, wherein the electric current A is direct current, and the electric current B.
Is a pulse current having a constant period, and the peak value Bp of the current B> DC A, the energization time Ta at the current A> the energization time Tb at the current B (sum of pulse current energization time and pulse off time) Electrical life evaluation method. 6. The method according to claim 1, wherein the energization test is conducted with a current C (A <C, energization time Ta at the current A> prior to energization with the current A).
A method for evaluating the energization life of a wiring, which starts from energization for an energization time Tc) at a current C, and excludes a sample that breaks when the first current C is energized as an initial defect product from the test results. 7. An evaluation device comprising a pair of terminals for applying a current to the sample, a power supply for energizing between the terminals, a voltmeter for measuring the voltage between the terminals, and a control computer. It shall have a voltage measurement terminal for performing voltage measurement at multiple positions between the application terminals, a scanner for connecting a voltmeter to a specific voltage measurement terminal, and a short circuit with an open / close switch between each voltage measurement terminal. Characteristic wiring reliability evaluation device. 8. The wiring reliability evaluation device according to claim 1, wherein the control computer performs selection of the voltage measurement terminal and opening / closing of each short circuit. 9. The wiring reliability evaluation device according to claim 1, wherein the current applied to the sample is a constant current regardless of whether the short-circuit switch is opened or closed. 10. The reliability evaluation device for wiring according to claim 1, wherein the short-circuit switch is closed when the measured voltage between the terminals is out of a predetermined range, and a current is passed through the short-circuit. 11. The method according to claim 1, wherein the resistance value between each terminal of the sample is measured when the power supply is out of control of the computer due to a change in the resistance value of the sample over time, and the measured value deviates from a preset constant range. A wiring reliability evaluation device that restores the power control function by the computer by adjusting the load resistance value to the power supply by closing the short-circuit switch between terminals and passing a current through the short-circuit. 12. The wiring reliability evaluation device according to claim 1, wherein the resistance value of the short circuit can be set to an arbitrary value for each terminal. 13. The life L according to the current-carrying life evaluation method or the life L _{0 according} to the current-carrying life evaluation method is defined as the current-carrying life, and the current-carrying life is determined in advance at an environmental temperature and a current density that are separately determined. Specified time Lspe
Highly reliable wiring that is at least c. 14. A highly reliable Si integrated circuit device using the wiring according to claim 13. 15. A laminated wiring for an integrated circuit device, comprising at least one Al alloy layer, according to claim 1, 2 or 3, having a life L determined by the energization life evaluation method or a life L _{0 determined} by the energization life evaluation method. High-reliability laminated aluminum wiring whose life is equal to or longer than the predetermined time Lspec at a separately set environmental temperature and current density. 16. The highly reliable Si integrated circuit element according to claim 15, wherein the Si integrated circuit element using the laminated wiring including at least one Al alloy layer is an Al laminated wiring.