A comparison of statistically-based and neural network models of plasma etch behavior

Himmel, C.D.; Kim, B.; May, G.S.

doi:10.1109/ismss.1992.197650

Cited by 18 publications

(2 citation statements)

References 9 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The problem of modeling the CVD process has been approached using different techniques, both linear (such as ordinary least square (OLS) and partial least squares) and nonlinear (such as artificial neural networks (NNs)). It has been shown that NNs guarantee better performance in modeling semiconductor manufacturing processes than other linear approaches. NNs are flexible computing frameworks and universal approximators that can be applied to a wide range of learning problems with a high degree of accuracy .…”

Section: Introduction and Related Workmentioning

confidence: 99%

A virtual metrology system based on least angle regression and statistical clustering

Susto

Beghi

2012

Appl Stoch Models Bus & Ind

View full text Add to dashboard Cite

In semiconductor manufacturing plants, monitoring physical properties of all wafers is crucial to maintain good yield and high quality standards. However, such an approach is too costly, and in practice, only few wafers in a lot are actually monitored. Virtual metrology (VM) systems allow to partly overcome the lack of physical metrology. In a VM scheme, tool data are used to predict, for every wafer, metrology measurements. In this paper, we present a VM system for a chemical vapor deposition (CVD) process. On the basis of the available metrology results and of the knowledge, for every wafer, of equipment variables, it is possible to predict CVD thickness. In this work, we propose a VM module based on least angle regression to overcome the problem of high dimensionality and model interpretability. We also present a statistical distance-based clustering approach for the modeling of the whole tool production. The proposed VM models have been tested on industrial production data sets.

show abstract

Section: Introduction and Related Workmentioning

confidence: 99%

A virtual metrology system based on least angle regression and statistical clustering

Susto

Beghi

2012

Appl Stoch Models Bus & Ind

View full text Add to dashboard Cite

show abstract

“…Furthermore, these neural process models require fewer training experiments. 5 In order to characterize the PECVD ofSi02 films deposited under varying conditions, we have performed a 251 fractional factorial experiment with three center-point replications.6 Data from these 19 experiments was used to develop neural process models describing seven output responses: deposition rate, refractive index, permittivity, film stress, wet etch rate, silanol concentration, and water concentration. Neural networks were trained using the feed-forward error back-propagation (FFEBP) algorithm.…”

mentioning

confidence: 99%

Modeling the properties of PECVD silicon dioxide films using neural networks

et al. 1994

View full text Add to dashboard Cite

Silicon dioxide films deposited by plasmaenhanced chemical vapor deposition (PECVD) are useful as interlayer dielectric for metal-insulator structures such as MOS integrated circuits and multichip modules. The PECVD of Si02 in a SiH4/N20 gas mixture yields films with excellent physical properties. However, due to the complex nature of particle dynamics within the plasma, it is difficult to determine the exact nature of the relationship between film properties and controllable deposition conditions. Previous modeling techniques such as first principles or statistical response surface methods are limited in either efficiency or accuracy. In this study, PECVD modeling using neural networks has been introduced. Neural networks have been shown to exhibit superior performance in both accuracy and prediction capability compared to statistical models.The PECVD of Si02 was characterized via a 251 fractional factorial experiment, and data from this experiment was used to train neural networks using feed-forward error back-propagation (FFEBP) algorithm. From these neural process models, the effect of deposition condition on film properties has been studied, and sensitivity analysis has been performed to determine the impact of individual parameters. The deposition experiments were carried out in a Plasma Therm 700 series PECVD system. The models obtained will ultimately be used for several other manufacturing applications, including recipe generation and optimization, and process control. ThJTRODUCTIONChemical vapor deposition (CVD) is an important processing technique in the manufacture of integrated circuits. The CVD process consists of the formation of a non-volatile solid film on a substrate by the reaction of vapor phase chemicals (reactants) that contain the required constituents. While atmospheric CVD (APCVD) and low pressure C\'D (LPCVD) are currently the dominant methods of employing CVD, some processes require a lower thermal budget than these two techniques allow. As opposed to thermally grown Si02, CVD Si02 is often used to maintain shallow diffusions or to prevent changes in profiles. Both APCVD and LPCVD Si02 yield films with good electrical properties, and the deposition processes are relatively straightforward. However, when these layers are deposited on silicon wafers, both materials are in tension. This not only reduces mechanical strength, but also limits the film thickness.Plasma-deposited oxide, on the other hand, is usually in compression, allowing film thickness up to 5 im to be obtained without cracking. Rather than relying solely on thermal energy to initiate and sustain chemical reactions, plasmaenhanced CVD (PECVD) uses a radio frequency (RF) induced glow discharge to transfer energy into the reactant gases, allowing the substrate to remain at a lower temperature than in APCVD or LPCVD. In the presence of the plasma, the reactant gases are transformed to various active species (such as neutral radicals and ions) due to collisions with energetic electrons. The highly reactive nature of these specie...

show abstract

Modeling and control of a semiconductor manufacturing process with an automata network: An example in plasma etch processing

Rietman

Patel²,

Lory

1996

Computers & Operations Research

View full text Add to dashboard Cite

A comparison of statistically-based and neural network models of plasma etch behavior

Cited by 18 publications

References 9 publications

A virtual metrology system based on least angle regression and statistical clustering

A virtual metrology system based on least angle regression and statistical clustering

Modeling the properties of PECVD silicon dioxide films using neural networks

Modeling and control of a semiconductor manufacturing process with an automata network: An example in plasma etch processing

Contact Info

Product

Resources

About