CN116298750A - A method for non-destructive characterization of semiconductor device microstructure damage - Google Patents

Abstract

The invention discloses a nondestructive characterization method for damage of a micro-region structure of a semiconductor device, which is characterized in that a certain electrical bias condition is applied to the device to be tested, transient curves of leakage current changing along with time before and after the stress action of the device to be tested are respectively collected by a testing instrument, the time constant of the damage position of the device is extracted from the transient curves by using a Bayesian iterative time constant extraction method, the time constant of the damage position of the device is presented in a peak spectrum form by combining a peak spectrum time constant extraction technology, and the spectrum valued characterization is carried out by using an amplitude spectrum technology on the basis, so that the evolution process of the damage of the invisible micro-region structure in the device under the stress action is converted into visual spectral line movement, and the accurate positioning, damage degree and spectral valued quantitative characterization of the evolution process of the damage of the micro-region structure of the device are realized.

Description

A method for non-destructive characterization of semiconductor device microstructure damage

technical field

The invention relates to the field of testing and characterization of semiconductor devices, and is mainly applied to the non-destructive characterization of semiconductor device micro-region structure damage.

Background technique

Wide bandgap semiconductor devices have excellent characteristics such as high breakdown voltage, high output power and high reliability, and show excellent performance in the field of high frequency and high power. However, affected by the heterogeneous structure of the device and the growth conditions of the material, the active region of the device is very easy to induce traps under the action of high temperature, dynamic strong field, and electrical stress. However, the existence of traps limits the further improvement of its performance and its wide application. Therefore, understanding the damage location, damage degree, damage evolution process and law of the device microstructure is the prerequisite for effective analysis of the microstructure damage mechanism and reliable design of the device. Non-destructive measurement, precise location and quantitative characterization of the evolution process of semiconductor device microstructure damage.

Contents of the invention

The technical purpose of the present invention is to propose a non-destructive characterization method for semiconductor device microstructure damage based on the relationship between drain transient current changes and semiconductor device microstructure damage. The Bayesian iterative deconvolution of the method Time constant extraction technology, analysis of time constant and peak spectrum changes, to achieve non-destructive characterization of semiconductor device microstructure damage, precise positioning and quantitative characterization of evolution process.

The defect capture and release electron process caused by the damage of semiconductor device directly affects its current change. There is a mapping relationship between the transient change of the current between the drain and the source electrode and the damage of the micro-region of the device. Based on the mapping relationship, by collecting the transient change curve of the drain current before and after the device damage, the time constant of the current transient curve is extracted by using the Bayesian iterative deconvolution time constant extraction method, and on this basis, the time constant spectrum is extracted Integrate to obtain the contribution of damage at different positions to the change of drain current, and express it in the form of spectral lines. By analyzing the drain transient curve before and after the damage, the evolution process of the invisible microstructure damage inside the device is transformed into a visualized spectral line movement, and the precise positioning of the device microstructure damage, the damage degree and the spectral value of the evolution process are realized Quantitative representation.

The technical solution adopted in the present invention is a method for non-destructive characterization of semiconductor device microstructure damage, which can compare the degree of damage to the device by stress with that before the device is damaged, and accurately characterize the change of the degree of device damage and the location of the damage.

S1 builds a transient current test system, as shown in Figure 2, to collect the transient change curves of device leakage current before and after stress (high temperature, high voltage, etc.). The accuracy of the current test system can reach the order of milliseconds. In the measurement of the transient current response, it is necessary to apply a trap filling stage to the damaged part of the device before the measurement process. In the filling stage, a large electrical stress is applied to make the damaged part of the device, that is, the trap Fully fill, then quickly switch to a small measurement bias (measurement phase), monitor the process of releasing electrons from the trap, and then collect the curve of leakage current I _DS versus time. Select a device under test, place the device under test on a constant temperature platform with a temperature of T0, and apply the electrical bias conditions in the above description to the device to collect the change curve of the leakage current I _DS with time.

S2 After collecting the transient drain current response, in order to extract the core information of the device damage position (trap), the transient drain current response curve is further processed to obtain the time constant spectrum of the trap, based on Bayesian deconvolution The time constant extraction method can accurately distinguish the area where the curve amplitude changes more gently, that is, the area where the peak value is annihilated by the peak value of the strong signal and is difficult to identify, or the area with highly overlapping peaks, which is extremely difficult to distinguish after multiple iterative calculations, and can effectively Highlights time constant peaks at device damage sites (traps).

The trap information exists in the transient current response in the form of e exponential change, as shown in formula (1)

τ _i is the time constant of the i-th trap, and ΔI _i is the magnitude of its influence on the current change. The time constant spectrum is mainly used to extract the characteristic time constant τ _i of the trap, and display it in the form of peak spectrum, and the abscissa of the peak value is the trap time constant.

First, introduce the logarithmic time variable:

z=lnt (2)

Then, the time constant spectrum of the trap is as follows:

ΔI(z) is the time constant spectrum, then the transient current I _ds (t) is described as:

Compared with formula (1), (4) is the integral form of channel transient current. Combined with equation (2), the transient current is converted to:

This is a convolution type integral equation of ΔI(τ). The result of differentiating z on both sides of equation (3.5) is:

Define the function W(z) as follows:

W(z)=exp(z-exp(z)) (7)

的表达式为：but

The expression is:

为卷积运算符，则时间常数谱ΔI(z)为：

is a convolution operator, then the time constant spectrum ΔI(z) is:

So far, the time constant spectrum can be solved by Bayesian deconvolution. The time constants of all device damage locations (trap) are presented in the form of peak spectrum. The area enclosed by the peak spectrum envelope corresponds to the variation of the drain-source current, and the abscissa corresponding to the peak value of the time constant spectrum is the device damage location (trap). ), the ordinate is the relative change in the drain-source current caused by the device damage location (trap).

再将x轴数值与y轴数值互换得到积分幅值谱。S3 accumulates the y value of the time constant spectrum according to the total variation of the peak value and the drain current, so that the y value corresponding to each time point is equal to the cumulative value of the y value of all previous time points (including this time point), that is,

Then exchange the x-axis value with the y-axis value to obtain the integral amplitude spectrum.

S4 calculates the first-order derivative of the integrated amplitude spectrum to obtain the differential amplitude spectrum. The difference between the peaks in the differential amplitude spectrum is the absolute action strength of the trap, that is, the sum of all action strengths is equal to the change in the total transient current response. It can more intuitively and accurately characterize the damage location and damage degree of the device microstructure. In addition, the peak ordinate of the differential amplitude spectrum represents the "degree of independence" of this type of trap, that is, how close the time constant of this type of trap is to other traps.

The entire characterization process is shown in Figure 1. The characterization process includes: transient drain current acquisition and correction, time constant extraction, peak spectrum characterization, and amplitude spectrum characterization.

Description of drawings

Fig. 1 Schematic diagram of the device microstructure damage characterization process.

Figure 2 Device transient current test system.

Figure 3 The hardware architecture and electrical bias timing diagram of the transient leakage current measurement system.

Fig. 4 Schematic diagram of spectral value characterization method for stress-induced device microstructure damage.

Detailed ways

The present invention will be described in detail below in conjunction with the accompanying drawings and embodiments.

Taking GaN HEMT as an example, the device is placed on a constant temperature platform with a temperature of T0, and the curve of the leakage current IDS changing with time is collected. Leakage current transient response characterization during device damage (trap) release of electrons (detrapping): In the filling stage, a DC voltage is connected to the drain, the voltage is V0, the source is grounded, the gate is connected to the reverse bias voltage V1, and the filling time is T1; In the measurement stage, a small excitation signal is immediately applied to the filled device under test, that is, the gate voltage is 0V, and the source is connected to a small voltage V2, so that the electrons captured in the device filling stage can be fully released, and the test time is T2. Since the GaNHEMT used in the example is a depletion-type device, if the gate and drain are both in a forward-biased state, the device will be easily damaged when the leakage current exceeds the maximum allowable channel current. Therefore, two separate timing pulses are required to control the working states of the gate and the drain respectively. The hardware architecture and electrical bias timing diagram of the GaN HEMT transient leakage current measurement system are shown in Figure 3. After applying a certain stress (such as electrical stress, high temperature stress, etc.) to this device, place the stress-applied device on a constant temperature platform with a temperature of T0, apply the same electrical bias conditions as before the stress, and test the leakage current of the device again The transient curve changing with time, the transient curve of leakage current changing with time before and after the stress is shown in Fig. 4(a).

After collecting the transient drain current response before and after the damage of the device under test, the transient drain current response curve is further processed to obtain the time constant spectrum, and the time constant extraction method based on Bayesian deconvolution extracts the device from the time constant spectrum The time constants and relative peaks of the trap locations are presented as a peak spectrum. The area enclosed by the peak spectrum envelope corresponds to the variation of the drain-source current, as shown in Figure 4(b), there are three time constant peak spectra, namely DP1, DP2 and DP3, according to the total peak and drain current We can roughly estimate and compare the position and damage degree of the device DP1 before and after the stress of the device, that is, before and after the damage. The peak position of DP2 moves from about 6s to about 3s. From around 360s to around 240s, the degree of damage also increased significantly.

再将x轴数值与y轴数值互换得到积分幅值谱，可以直观定量分析应力作用前后器件3处损伤位置漏电的变化量，如图4(c)所示。Since the peak spectrum characterization cannot directly and quantitatively observe the quantitative characterization of the damage degree of the device, on this basis, the amplitude spectrum technology can be used to realize the spectral value characterization of the device damage, and the extraction method based on Bayesian deconvolution is used to extract the time Constant spectrum, accumulating the y value of the time constant spectrum, so that the y value corresponding to each time point is equal to the cumulative value of the y value of all previous time points (including this time point), that is,

Then exchange the x-axis value with the y-axis value to obtain the integrated amplitude spectrum, which can intuitively and quantitatively analyze the change of leakage at the three damaged positions of the device before and after stress, as shown in Figure 4(c).

The differential amplitude spectrum is obtained by first deriving the integral amplitude spectrum. This method can more accurately and intuitively characterize the drain current variation caused by the three device damages before and after the stress damage, as well as the evolution process of the device damage degree and damage location, as shown in Figure 4(d).

In the present invention, by collecting the change of the drain current in real time, the transient drain current response curve is further processed, and the time constant extraction method based on Bayesian deconvolution realizes the accurate acquisition of the trap time constant spectrum, and based on this The spectral value characterization of device micro-area damage is realized by using the amplitude spectrum technology. Based on the above characterization process, different stress types are applied to the device, such as high temperature, strong field, etc., and the peak spectrum and amplitude spectrum are extracted respectively before the stress is applied (before damage) and after a period of stress (after damage), and the device can be The evolution process of internal invisible microstructure damage is transformed into intuitive qualitative and visualized spectral line movement. By comparing and analyzing the movement of the peak spectral line and amplitude spectrum before and after stress application, the effects of stress type, applied intensity, and duration on the device microstructure can be obtained. The influence of the damage location and damage degree of the zone structure is shown in Fig. 4.

Claims (3)

1. The method is characterized in that the method is based on a Bayesian iterative deconvolution time constant extraction technology, analyzes time constants and peak spectrum changes, and realizes nondestructive characterization, accurate positioning and quantitative characterization of evolution process of the micro-area structural damage of the semiconductor device;

mapping relation exists between current transient change between drain and source electrodes and damage of a micro-area of the device; based on the mapping relation, the drain current transient change curve before and after the damage of the device is acquired, the current transient curve time constant is extracted by using a Bayesian iterative deconvolution time constant extraction method, and the contribution of the damage at different positions to the drain current change is acquired by integrating a time constant spectrum on the basis, and is shown in a spectral line form; by analyzing the transient curves of the drain electrodes before and after the damage, the evolution process of the invisible micro-region structural damage in the device is converted into visual spectral line movement, and the accurate positioning of the damage of the micro-region structural damage of the device, the damage degree and the spectral value quantitative characterization of the evolution process are realized.

2. The method for nondestructive characterization of micro-area structural damage of a semiconductor device according to claim 1, wherein S1 is used for constructing a transient current test system and collecting transient change curves of leakage current of the device before and after stress action; the accuracy of the current test system reaches millisecond level, a trap filling stage is required to be applied to a damaged part of the device before a measurement process in transient current response is measured, larger electric stress is applied in the filling stage, the damaged part of the device, namely the trap, is fully filled, then the device is switched to a small measurement bias, and the process of releasing electrons from the trap is monitored, so that a change curve of leakage current IDS along with time can be acquired; selecting a tested device, placing the tested device on a constant temperature platform with the temperature of T0, and applying an electrical bias condition to the device so as to acquire a change curve of leakage current IDS along with time;

s2, after transient drain current response is acquired, processing a transient drain current response curve to obtain a time constant spectrum of a trap, and accurately distinguishing a region with a more gentle curve amplitude change through repeated iterative computation by a time constant extraction method based on Bayesian deconvolution, wherein a time constant peak value of a device damage position can be highlighted;

trap information exists in transient current response in an e-exponential variation form as shown in a formula (1)

τ _i Time constant, ΔI, for the ith trap _i For which the amplitude of the current change is affected; time constant spectrum for extracting characteristic time constant tau of trap _i The time constant is displayed in the form of peak value spectrum, and the abscissa of the peak value is the trap time constant;

first, a logarithmic time variable is introduced:

z＝lnt (2)

then, the time constant spectrum of the trap is as follows:

Δi (z) is the time constant spectrum, then the transient current Ids (t) is described as:

compared to equation (1), equation (4) is an integrated version of the channel transient current; in combination with equation (2), the transient current is converted into:

this is the convolution integral equation for Δi (τ); the result of differentiating z on both sides of equation (5) is:

the definition function W (z) is as follows:

W(z)＝exp(z-exp(z)) (7)

then

The expression of (2) is:

as a convolution operator, the time constant spectrum Δi (z) is:

the time constant spectrum is obtained by solving the Bayes deconvolution; the time constant of the damaged position of the device is presented in the form of a peak spectrum, the area enclosed by the envelope curve of the peak spectrum corresponds to the variation of the drain-source current, the abscissa corresponding to the peak value of the time constant spectrum is the time constant of the damaged position of the device, and the ordinate is the relative variation of the drain-source current caused by the damaged position of the device;

s3, accumulating the y value of the time constant spectrum according to the total variation of the peak value and the drain current to enable the y value corresponding to each time point to be equal to the accumulated value of the y values of all the previous time points, namely

Exchanging the x-axis value with the y-axis value to obtain an integral amplitude spectrum;

s4, first-order derivation is carried out on the integral amplitude spectrum to obtain a differential amplitude spectrum; the difference value between peak values in the differential amplitude spectrum is the absolute action intensity of the trap, namely the sum of all action intensities is equal to the change of the total transient current response, and the position and the damage degree of the damage of the micro-area structure of the device are intuitively and accurately represented; the ordinate of the peak of the differential amplitude spectrum represents the degree of independence of this type of trap.

3. The method for nondestructive characterization of semiconductor device micro-area structure damage of claim 2 wherein the characterization of S4 comprises: transient drain current collection and correction, time constant extraction, peak spectrum characterization and amplitude spectrum characterization.

Images