Beirut Failure Analysis Zhao Gong Semiconductor Engineer 2022-05-13 10:42
The cleanliness of the wafer surface will affect the subsequent semiconductor process and product qualification rate, and even in all output losses, up to 50% is caused by wafer surface pollution. The cleanliness of the wafer surface of
will have a certain impact on subsequent semiconductor processes and product qualification rates. The most common main pollution includes the residues of metals, organic matters and particulate particles. The results of pollution analysis can be used to reflect the degree and type of pollution encountered in a certain process step, a specific machine or an overall process. Early literature pointed out that during the manufacturing process, the loss caused by failure to effectively remove contamination on the wafer surface may account for more than 50% of all production losses.
Common contamination may have an impact on processes and products. For example, metal contamination can cause leakage current in the p-n structure, which in turn leads to a decrease in the breakdown voltage of the oxide and a decrease in the carrier life cycle. Organic contaminants may cause unintended hydrophobic properties on the wafer surface, increase surface roughness, produce atomized (haze) surfaces, and disrupt the growth of epitaxial layers, and can also affect the cleaning effect of metal contamination without first removing the contaminants. Particle contamination may lead to blocking or masking effects in etching and microfilm processes; pinholes and microvoids are generated during film growth or deposition. If the particle particles are large and conductive, it may even lead to short circuits.
Therefore, how to ensure that there is no pollution residue on the wafer surface has always been an important issue. After the introduction of RCA mixture cleaning solution mainly based on hydrogen peroxide in the 1970s, many different cleaning solutions have been used, such as the traditional two-step cleaning process of 1:1:5 ammonia:hydrogen peroxide:pure water ratio (SC-1) in sequence, and then a 1:1:6 hydrochloric acid:hydrogen peroxide:pure water ratio (SC-2) [1]. The effectiveness of different cleaning processes requires evaluation and testing based on actual needs and purposes, and according to different machines. The main purpose is to clean and remove contamination on the wafer surface through the introduction of different cleaning processes.
Gold is the most commonly discussed issue among the three types of pollution. It may come from the reagents used in cleaning, etching, lithography, deposition, etc., or the machines used in the process, such as ovens, reactors, ion implantation, etc., and may also be caused by careless treatment of wafers.
Many analytical technologies and instruments in the past have been used to determine metal contamination, including Auger Electron Spectroscopy (AES), Secondary Ion Mass Spectrometry (SIMS), Time of Flight-Secondary Ion Mass Spectrometry (TOF-SIMS), Rutherford Backscattering Spectrometry (RBS), and Graphite Furnace Atomic Absorption Spectrometry (Graphite Furnace Atomic Absorption Spectrometry, GF-AAS), etc., but because each machine has its limitations, such as insufficient sensitivity, unable to provide accurate quantitative results, can only conduct extremely shallow analysis, or cannot perform multi-element detection at the same time, and the overall measurement time is long, which will make the above analysis technology only achieve limited analysis capabilities.
The previous requirement for the concentration detection of metals on the wafer surface was 1010 atoms/cm2. With the evolution of the process, the detection limit has been reduced to 108 atoms/cm2. The technologies that can meet this analysis requirements are mainly two types: Total Reflection X-ray Fluorescence Spectrometer (TXRF) and inductively coupled plasma mass spectrometer (ICP-MS). How to judge the timing of use of these two detection instruments and give full play to the best performance of metal pollution analysis?
Total Reflection Fluorescence Spectrometer (TXRF) - Using the principle of X-ray total reflection, the wafer surface is excited with X-rays at a very small angle to obtain a map of the content of metal contaminants on the surface. It is a highly surface-sensitive analysis technology to realize the analysis and detection of trace elements .
TXRF has the advantages of non-destructive and capable of fixed-point detection. In the detection of transition metal elements such as copper and iron, the detection limit is about 109–1010 atoms/cm2 2 , which meets the general detection and monitoring requirements; only low-mass metals such as sodium, magnesium, aluminum, etc. will have a high detection limit; lithium, beryllium, and boron elements cannot be detected. Some documents have also pointed out that TXRF has a low recovery rate for copper elements. TXRF can directly test the samples, or use the sample pretreatment technique of ball method and gas phase decomposition (VPD) to detect wafer surface contamination.
The principle of its determination is to use a monochromatic X-ray source to incident on the sample surface at an angle smaller than the total reflection angle. The excitation sample with an atomic layer thickness of only 3-8nm on the surface will be measured by a detector placed on the vertical sample surface. By analyzing its energy wavelength, qualitative information can be provided. The signal intensity can be obtained after conversion through the detection line.
To improve the instrument detection capability to reach the detection limit of 108–109 atoms/cm2 2 , you can consider using VPD or Vapor Phase Treatment (VPT) to concentrate the sample [2-4], or use Synchrotron Radiation (SR) light source to enhance the incident light intensity to improve the detection capability of the measurement [5]. In series use of VPD or VPT systems, whether it is a commercial machine or a self-designed and assembled system, the wafer is mainly placed in a high cleanliness environment, and the wafer is introduced by hydrofluoric acid vapor to condense on the surface of hydrophilic silica, and the following formulas are used to decompose the silicon oxide.
SiO2 + 6HF → H2SiF6 + 2H2O
H2SiF6 → SiF4 + 2HF
droplets condensed on the wafer surface. In VPD, droplets will be further collected at an inclined wafer angle, or additionally used scanning solution to help collect droplets scattered on the wafer surface; at the same time, the silicon wafer surface that becomes hydrophobic also helps to collect droplets intact. After the collected droplets are heated and dried up on the wafer surface, it can be detected. The components of the scanning solution vary slightly according to demand, and usually, except for hydrofluoric acid, it contains hydrogen peroxide that can help improve Cu recovery [6]. During the VPD process, in addition to the effective increase of the signal, the gaseous SiF4 and H2 with volatility will dissipate due to heating or its own volatility before the measurement, and will also help reduce the interference/suppression effect caused by Si matrix dissolution on the measurement process. The efficacy of
VPD is mainly affected by two factors, namely the dissolution efficiency of the acid solution on contaminants and the recovery efficiency of the acid solution. The concentration ratio can be simply evaluated by testing the wafer surface area relative to the measured area where the droplets later dried up. Since the surface contaminants originally distributed at different locations on the entire wafer are concentrated/concentrated to a certain point for measurement, although the system sensitivity is improved, the original spatial resolution (Spatial Resolution) advantage of fixed-point detection is lost. If the overall metal concentration of the concentrated sample is too high, the Cu measurement recovery rate will be lower than 50%.
considers the demand for spatial resolution. If the dispersed droplets are not collected in scanning solution, the wafer is directly heated, so that the condensed droplets are dried up in their original locations and then the measurement is performed directly. Although the data results show that the detection ability is not as good as that of the VPD system, which is only about 1.5–5.0 times improved, the advantages of spatial resolution can be moderately retained [3]. In the study of SR as the light source, it is shown that using a light source with higher intensity can achieve better detection limits. In addition, by using radiation below Si edge, Al atoms on the surface rather than Si atoms on the substrate can be detected for Al contamination above 1010 atoms/cm2 [7].In addition to having few TXRF test resources in Taiwan, the above-mentioned SR is not a widely-installed light source device. Taiwan's SR light source is mainly located in the Synchronous Radiation Research Center (NSRRC) in Hsinchu, so it also limits the convenience and practicality of increasing TXRF sensitivity.
Inductively Coupled Plasma Mass Spectrometer (ICP-MS) - Destroy the sample matrix components by high-temperature plasma, separate ions with mass spectrometer for quantitative analysis, which is sensitive to scan and can detect almost all elements on the earth, and has strong trace element analysis capabilities. Compared with the rarer TXRF, the Inductively Coupled Plasma Mass Spectrometry (ICP-MS) developed in the 1980s, since the sample matrix components can be destroyed by high-temperature plasma and analyzed and measured by high-resolution and high-sensitivity mass spectrometers, it has been widely used in many fields such as environment, biology, forensic science, material analysis, etc., and then connected with different equipment (hyphenation), many specific analytical and detection needs can be achieved. In addition to halogen, other elements that can be measured by ICP-MS are not suitable for TXRF or cannot be detected, other elements that cannot be used for low-mass numbers, such as sodium, magnesium, aluminum, lithium, beryllium, boron, etc., can all be used to obtain good analysis results by ICP-MS. When detecting metal contamination on the surface of 8-inch and 12-inch silicon wafers, the detection limit of most metals generally falls at the detection limit of 108–109 atoms/cm2 2 1.
commonly includes laser erosion systems (LA), arc systems (Arc), and spark discharge systems (Spark). These systems are connected in series to make direct testing of solid samples (or biological samples). If connected in series with a gas chromatography device (GC), separation and determination of gaseous compounds of organotin or organic lead can be performed. In series with an Electro Thermal Evaporation (ETV), the online removal of the sample matrix can be achieved through the Temperature Program and the Matrix Modifier, and analytical and detection of a small number of samples can be performed. However, when not connected to the above special equipment, the sample needs to be converted/dissolved into liquid form before the detection can be carried out. This is a destructive analysis and cannot provide fixed-point detection information for the sample.
In terms of metal pollution detection requirements on the wafer surface, ICP-MS needs to decompose native or deposited silicon oxide film layer or silicon nitride film layer on the wafer surface by hydrofluoric acid to analyze and detect the collected acid solution, and convert the results into surface concentration units of atoms/cm2 . The decomposition of the surface film layer can be done by manual acid dropping, or by automated machine assisted (VPD), where manual method is also called ball method, droplet check method, direct acid droplet decomposition (DADD), liquid phase decomposition-droplet collection method (LPD-DC), and other different names.
As for VPD systems, as described in the previous VPD-TXRF paragraph, after the saturated hydrofluoric acid vapor is introduced into the hydrophilic silicon oxide layer, the exposed hydrophobic wafer surface will make the additional scanning solution easy to roll and collect surface droplets. The components of the scanning solution will vary according to different experiments, but it will usually contain hydrogen peroxide that can help improve Cu recovery. Excessive hydrofluoric acid vapor and additional oxidants such as nitric acid or hydrogen peroxide can promote the formation of volatile SiF4 gas by the silicon oxide matrix, reducing the matrix interference or composite ion interference that the Si matrix may cause in ICP-MS assays, such as 47Ti (28Si19F), 68Zn (40Ar28Si), and 44Ca (28Si16O). However, since the surface oxide layer is treated in a gaseous manner, the thickness of the film layer will affect the length of the exposure time. Usually, the native oxide layer (1.5–3.0nm) will take 20-30 minutes. If it is an oxide layer above 10nm, it may take 3-12 hours of exposure time; and the manual pretreatment method can directly increase the amount of hydrofluoric acid to reduce the overall film etching time.The
VPD system usually has a robotic arm to process the samples to avoid possible pollution during the manual processing of the samples. The internal cleanliness of the machine are usually Class1 environment. Although it can effectively reduce the background value and detection limit of the measurement, it also makes the machine construction cost relatively high. Although the manual balling method is more likely to be introduced with additional contamination, it is widely used in laboratories because it is simple, cheap, fast and has high elasticity.
Currently, ICP-MS with balling method or VPD has been used in wafer surface metal contamination detection for many years, but some testing needs will still face problems that need to be solved. For example, non-silicon wafers such as gallium arsenide and gallium nitride wafers, or wafers such as ceramic, sapphire, glass and quartz as carriers, or samples of silicon wafers with non-silicon oxide (SiOx) or silicon nitride (SiNx) on the surface, or the applicability of thin film components analysis such as polysilicon (Poly-Si), epitaxial silicon, tungsten silicide (SiW) and titanium (Ti), the optimal conditions for detection requirements of precious metals (Pd, Au, Pt and Ru, etc., and the inapplicability of pattern wafers and wafer margins (Edge) and bevel detection requirements [8]. The pretreatment methods of
silicon wafer samples are mostly made of acid liquid for surface film etching. High concentration of acid liquid can produce violent reactions, but there are also concerns about eroding the underlying wafer; if the concentration of acid liquid is too low, it may not be able to erode enough depth. How to achieve the goal of metal component detection through different thicknesses and characteristics of wafers through different acid solutions and techniques?
Currently, most of the samples that are subject to metal contamination on the surface of silicon wafers are mainly silicon wafers with native or chemical grown and thermally grown silicon oxide films or silicon nitride films. Such silicon wafer samples can be pretreated by manual methods or VPD machines as described above. Although thicker films require increased acid use and reaction time, the hydrophobic properties that appear after the film etching can help determine whether to terminate the etching; for the hydrophilic wafer surface, this hydrophobic phenomenon cannot be used to judge the end point of the film etching. What's more, some such as gallium arsenide or gallium nitride wafers, the substrate will be slowly dissolved/dissolved under an acid liquid system containing hydrofluoric acid. The increase in the pretreatment time will cause an increase in the dissolution of the corresponding substrate. Although the increase in metal contamination by high-purity wafers is limited, the dissolved substrate components still have the opportunity to affect the detection data in the detection equipment with the effects of mass spectral interference or non-mass spectral interference. If dilute acid is considered as the extraction solvent, although the substrate dissolution can be avoided, it may be impossible to effectively dissolve/remove the contaminants from the wafer surface. Through the literature, the cleaning effect of different acid solutions on the surface of GaN wafers under different pH values and redox potentials shows that the cleaning solvent with low pH values and high redox potentials helps to reduce metal contamination on the surface of GaN [9]. The discussion in this part is similar to the discussion of the cleaning efficiency of Cu elements on wafer surface in other literatures. In the future, when cleaning, removal or concentration monitoring of metal elements on non-silicon wafer surfaces, helping to further understand the characteristics of wafer materials and the redox potential of the substance to be tested, as well as the pH value of the cleaning and collection solution, it will help achieve the goal. The experimental design and verification of this part, including understanding of the sample background information and whether the laboratory has suitable testing equipment to cooperate, can be expected to have a certain degree of difficulty. In addition, a few ceramic, glass, sapphire wafers, etc. used as carriers, considering the recycling of wafers, they cannot be sampled with a hydrofluoric acid solution, and can only collect contaminants on the surface with the diluted nitric acid solution.
is a film that cannot be dissolved by simple hydrofluoric acid, such as Poly-Si film. Although it is possible to consider oxidizing and decomposing the film with mixed acid containing nitric acid and hydrofluoric acid, in addition to the violent reaction, higher concentrations of acids also have the concerns of further etching to the underlying wafer; if the concentration of the acid solution is too low, it may not be able to decalize enough depth, so it is difficult to manually perform contamination detection of such films.In VPD machine, it provides bulk etching option, which can simultaneously import ozone (ozone) and hydrofluoric acid, and perform component measurements while etching, achieving the need for metal components detection in Poly-Si. The literature points out that the surface of the sample analyzed in the Bulk etching option is more likely to show obvious roughness and hydrophilicity, making the scanning solution suspended at the front end of the scanning nozzle easily flow out of the result [10]. Due to the use and residue of photoresist during the process, the used and the use of
silicon wafers with pattern may be limited to the pattern. In addition to the possibility of low recovery of acid or unrecoverable, the difficulty of acid liquid to roll smoothly on the wafer surface will also lead to differences in the etching time and depth at different locations. There is currently no better solution for the detection of such wafers. It may be considered soaking in an acid tank, but the crystal back is expected to contribute; or it may be done by covering the entire surface of the test surface with a large amount of acid, but a high dilution ratio will be introduced, thereby increasing the chance of not detectable (ND).
wafer The effect of surface pattern on droplets, and the droplets are restricted by pattern to the diffusion and rolling of their liquid.
The above two different sample detection requirements are difficult to achieve by VPD-ICP-MS or LPD-DC-ICP-MS, and may be evaluated for direct sample analysis with TXRF. Although it still faces the problem of high detection limits, the selection of acid components and the discussion of pretreatment time can be avoided.
mainly considers the detection of precious metals in the detection of special elements to be tested. For pretreatment of precious metals to be tested such as iridium (Ir), ruthenium (Ru), palladium (Pd), gold (Au) and platinum (Pt), the surface of the silicon wafer must be hydrophobic, and then the acid solution containing nitric acid and hydrochloric acid, similar to Aqua Regia formula, can be improved as a scanning solution to improve the recovery rate of the to be tested. However, the optimal concentrations of different precious metal elements for hydrochloric acid and nitric acid are different, and their recovery rates generally fall between 20-80% [10-11]. In references, similar sampling methods can be used to determine Au, Pd, Pt and Ag, 74-98% of the recycling results can be obtained [12].
's determination requirements for wafer edge (Edge) and side (Bevel) are mainly used to evaluate the degree of contamination introduced by the wafer in the area in which the wafer box is contacted. If the contaminated distribution area can be accurately evaluated, the proportion of areas that can be used on the wafer surface can be increased. However, without special tools, it is difficult for ICP-MS to detect these two areas [12]. Although the laboratory can consider drawing a groove of a specific depth on the surface of a PFA material object according to needs, limiting the acid solution that can be soaked into the groove only at a certain distance from the edge to perform sampling at a specific location, even if pre-processing is done with such self-designed tools, the absolute amount of the object to be measured will still be limited due to the small sampling area. For the contamination assessment in these two locations, fixed-point detection with TXRF may be a more direct and simple instrument selection [13].
In semiconductor processes, trace amounts of contaminants may cause corrosion, erosion, electron migration or short circuit of components, wafers or final electronic components. In order to minimize the possible defects caused by wafer surface pollution, in addition to precise control of process steps and environmental conditions, metal pollution detection is an indispensable analysis step to ensure the highest pass rate.
Comprehensive the above, metal pollution detection on the wafer surface can assist in evaluating the metal pollution contribution of a wafer through a certain process, a specific machine or all process history. Currently, VPD-TXRF, VPD-ICP-MS or LPD-DC-ICP-MS have the opportunity to meet the detection needs of 108-1010 atoms/cm2 when testing 8-inch and 12-inch wafers. Some elements can even be close to 107 atoms/cm2 or lower.
. TXRF can meet the needs of fixed-point detection and does not require special sample pretreatment. In addition to the original low-mass element not applicable, the detection limit generally falls to 1010 atoms/cm2 .The non-destructiveness of the sample gives the same sample the opportunity to undergo repeated testing and verification, so it is often used in online monitoring equipment on production lines. If the sample pre-processing is further combined with VPD, although the original fixed-point detection capability will be lost, the detection limit can be increased to 108-109 atoms/cm2, meeting the current metal detection needs of metal surfaces. ICP-MS, which has a high prevalence and better detection capability in
machines, can be measured for low-mass elements that are difficult to detect by TXRF. However, because it cannot directly detect wafer surface pollution like surface analysis technology, it must be dissolved and destroyed by acid liquid before it can be tested. Therefore, when faced with wafer samples of other materials or special film samples in the future, the formulation and applicability of the acid liquid are still issues that need to be discussed continuously.
Source: Hongkang Technology Detection
Semiconductor Engineer
Semiconductor Experience Sharing, Semiconductor Results Exchange, Semiconductor Information Release. Semiconductor industry trends, career planning of semiconductor practitioners, and growth process of chip engineers.
17 original content
public account
published in Beijing
