Final Publishable Summary Report of EPT300

PDF download: Final Publishable Summary Report of EPT300

 

EPT300 Final report 2015

 

 

 

 

 

ENIAC JOINT UNDERTAKING

“Final Publishable Summary Report”

 

EPT300

Enabling Power Technologies on 300mm wafers

 

ENIAC Call 2011/2

 

 

 

 

 

 

 

 

Coordinator

Infineon Technologies Austria AG

Johann Massoner

Cristina De Luca (P.M.)

 

Management Support

 

EUTEMA

Erich Prem

Jörg Irran (P.M.)


 

 

Enabling Power Technologies on 300 mm wafers

EPT300 Project - ECSEL JU

Introduction

EPT300 (Enabling Power Technologies on 300mm wafers) is a 36 month project that “aims to be a decisive step forward to strengthen Europe’s leading position in power semiconductor technologies and More-than-Moore manufacturing capabilities relating to energy efficient electronic solutions. Power semiconductor devices fabricated in a European leading pilot line for 300mm wafer production are the scope of the project, for which manufacturing excellence, cost competitiveness and challenging applications are critical boundary conditions”. EPT300 project started as planned by first of April and the consortia achieved already many intermediate deliverables and milestones, according to the technical annex plan. The partners collaborate to the innovative approach in a professional and active manner. EPT300 on one hand addresses challenges in research on process and production technologies on the other hand on handling and automation for advanced power technologies based on 300mm wafers. The major axes are the following:

·        The investigation of the future performance requirements and reliability targets for demonstrator power technologies including application.

·        The development and research of enhanced substrate materials and ultra-thin wafer technologies

·        The investigation and implementation of enhanced equipment and new power processes for 300 mm

·        Automation challenges for 300 mm thin wafers are covered by developing new concepts and requirements, researching new possibilities in factory automation and material handling in a mixed wafer environment and also in wafer transportation carrier, automated wafer substrate tracking and transportation.

The final challenge of EPT300 project is: the setup of the first pilot line and high volume power 300mm production with the first demonstrators in CoolMOSTM, IGBT and SFET technologies.

 

 

 

Consortium

EPT300 is a focused 36 month project with 23 partners including large industry, small and medium enterprises and research institutes from Austria, Germany, Italy and the Netherlands.

The very broad know how across consortium members associated with high competences of a semiconductor company like Infineon can act as engine for “R&D in Europe micro- and nanotechnologies”.

AUSTRIA

Infineon Technologies Austria AG is a legally independent subsidiary, 100% owned by Infineon Technologies AG in Germany. It is one of the globally acting manufacturing and research &development centres of Infineon Technologies AG.

Infineon Technologies IT-Services GmbH is a global center of competence delivering and developing IT services and solutions for approximately 25.000 IFX employees at more than 80 IFX sites on three continents (Asia Pacific, Europe, North America).

Lam Research Corporation has been a major supplier of wafer fabrication equipment and services to the worldwide semiconductor industry for more than 30 years.

mechatronic systemtechnik gmbh is a successful, fast-growing high-tech company headquartered in Villach, Austria. mechatronic has several years of experience in the semiconductor industry providing fully automated handling solutions for handling thin or warped (dome-, bowl-, or potato-shaped) wafers.

SICO Technology GmbH is an SME employing 45 people. SICO manufactures and processes fused silica, fused quartz, various types of silicon and ceramic products for various industries that include semiconductor, optical, automotive and aerospace.

Artesyn Austria GmbH&CoKG 100% owned by Emerson who is a world leading supplier of power supply equipment, from 1 Watt mobile phone chargers, several kW telecom rectifiers and solar inverters up to MW UPS solutions for data centers and wind mills. Emerson serves the entire industry, from Telcom, Datacom, Computing and Storage, Medical, Lighting, Solar, Electric Drives, Wind Power, Defense and Arospace.

CTR is a non-universitary, ISO 9001:2008-certified R&D centre focussing mainly on Advanced Sensor Systems and Technologies. As a professional, industry-oriented R&D service provider CTR conducts both fundamental research and application-oriented R&D projects for and with its customers.

eutema is a strategic research and technology consultant company based in Vienna, Austria with a focus on ICT. It is skilled in designing and managing research programmes and projects as well as RTD policies.

Universität Klagenfurt The Department of Statistics is part of the Faculty of Technical Sciences of the University of Klagenfurt.

ITALY

LPE was founded in 1972 as a producer of chemical wet benches and small cvd reactors. Starting from 1987 LPE focused his activity to silicon epitaxial reactor for power devices. From then LPE has established various satellite companies in the USA, in Czech Republic, in France, in Japan and in 2004 in China.

GERMANY

HAP was founded in 1991 by two scientists of Carl Zeiss Jena. HAP is a SME and develops and manufactures handling solutions, robotics for wafer, carrier, masks, pellicle and for PCBs and complex automatization for semiconductor industry and solar industry.

Infineon Technology Dresden GmbH is one of the largest production sites of Infineon Technologies AG. Infineon Technologies AG, headquartered in Munich, Germany.

Infineon Technologies AG, headquartered in Munich, Germany, focuses on three central challenges facing modern society: Energy Efficiency, Mobility and Security. Infineon offers semiconductors and system solutions for automotive and industrial electronics, chip card and security applications.

For over 10 years, Roth & Rau - Ortner GmbH (former Ortner c.l.s. GmbH) has participated in complex production logistics with a broad portfolio of customised services for the semiconductor industry. The focus of its work is on the development and production of tailor-made automation solutions and customer specific service concepts for the improvement of company internal material flow for semiconductor manufacturers.

Semikron is an internationally leading power semiconductor manufacturer. 2011 marks the 60th anniversary of the German-based family enterprise which employs 3600 people worldwide

Siltronic AG is a global leader in the market for ultra-pure silicon wafers and the partner of numerous leading chip manufacturers. Siltronic AGdevelops and manufactures wafers with diameters of up to 300 mm at production sites in Europe, Asia and the United States.

Technische Universität Dresden is a leading region of semiconductor engineering and manufacturing in Europe (Globalfoundries, Infineon, X-FAB, ZMD) with large second and third-level delivery industry and a strongly evolving software industry (Deutsche Telekom, SAP-SI).

NETHERLANDS

Philips Healthcare is an expanding business that plays a significant role in the vision of Philips that is built around healthcare and lifestyle.

ASM is a leading manufacturer of wafer-fab and packaging equipment and materials. ASM’s Front-end Operations has a 40-year track record in the industry, particularly in all types of CVD technologies such as LPCVD, ALD, RTCVD, PECVD and Epitaxy.

Bruco Integrated Circuits B.V. is specialized in design of custom made Integrated Circuits. Bruco Integrated Circuits offers design services that focus on the development of analog, mixed-signal, RF and high voltage custom made integrated circuits.

Heliox is specialised in the research, design and manufacturing of high performance power conversion products. The R&D department of Heliox is mainly located in The Netherlands.

Eindhoven University of Technology Electromechanics and Power Electronics (EPE) group has an extensive expertise in the fascinating field of advanced methods and tools to enhance the analysis, design and multi-objective optimization of innovative electromagnetic structures and cyclically switched networks.


 

Technical achievements and exploitation potential

EPT300 project is based on the concept of 300mm wafers in a 1:1 transfer approach from 200mm to fully prove compliance with application requirements. So the old device in 200mm should have the same performance of the transferred device in 300mm. The challenges in EPT300 are related to the processes, equipment, automation and handling, substrates that are needed for such a “challenging” transfer to a higher area of wafers. It is obvious that all partners in the projects contributed in a professional and constructive way to these remarkable results. All contributions finally lead into a released technology platform. It’s also worth to mention that for the partner Infineon the overall cost especially to mitigate the risk by running alternative paths have been much higher as claimed by the EPT300 project, which is only a part of a global activity for 300mm.

 

PERFORMANCE indicators concerning EPT300:

Demonstrators for all technologies (CoolMOS™, SFET4  and IGBT 1200V) available.

Demonstrators

Technologies

Planned

Achieved demonstrators and qualification for production

CoolMOS

C3,C6

C3,C6,C9

SFET

4

4, 5

IGBT

1200V

1200V, RCH3

Overall improvements:

Yield Frontend

  8” yield performance baseline exceeded by >0.5%

Yield Backend

YB 1,2% better than planned

Costs reduction

Actual cost performance improvement for 300mm Wafers compared to 200mm Wafers is in the expected range of  22%-30% @ pilot line Villach.

 

Technology readmap

Technology roadmap for 300mm power technologies is defined for the sites Villach and Dresden.

Line stability

Line stability cpk à better than 8”

Thin wafer manufacturing processes specified and defined

Qualified and base line volume established in Infineon Villach; also Dresden ready to run initial volume

Proof of the optimized logistic

 

A 300mm logistic flow over the whole process line including front end to back end and a secure transport of 300mm thin wafers is now established and qualified at Infineon (IFAT, IFD and IFAG) ant its partners in the supply chain

Demonstrating a thickness of about ½ of the current state of the art

 

300mm thin wafer technology is demonstrated and qualified for CoolMOS™ (200µm & 60µm), IGBT (120µm) and SFET technology (40µm). 1st results to enabling the next generation of ultra-thin wafer technology are available. Further investigation by concept studies to be finished to explore the limits

Raw material costs efficiency in comparison to 200mm

For the unified power substrate and lightly doped n type wafers an increase in cost of 5 - 10% is expected on the long term. [JI1] Therefore the price difference per area will go up. These disadvantages of the 300mm substrates in cost could be  compensated by measures taken in 300mm wafer production.

Uniformity optimization @ Si spin etch

Improvement of the uniformity from 2,7% to 2,1% was achieved with hardware (equipment) and process modifications.

 

Future performance requirements and reliability targets for demonstrator power technologies including application

In the EPT300 project, new base materials and new equipment were explored for power semiconductor technologies. Therefore we used this as opportunity to design new product characteristics. A “try and error” approach would not be feasible because of the enormous costs. This activity was mainly conducted in the 1st and 2nd year of the project and was successfully closed; the results were used as input for the definition of future application power converter designs in the different application domains.

All application demonstrator designs have been defined and simulated by the different teams. The goal of the defined demonstrators was to proof the 1:1 capabilities of the EPT300 devices. Targeted applications in this project were:

·           Next generation power converter for Medical Equipment

·           Low EMI power converter

·           Highly efficient telecom rectifier

In the last year of the project the functional models has been converted to the final demonstrators by implementing necessary updates of the design or convert the lab model to a first production prototype model. The updated demonstrators have been successfully finalized and used to conduct the 1:1 capability measurements of the transferred products which are already available on a 200mm.

 

IMPROVEMENT per application:

Next generation power converter for medical equipment

Demonstrator was developed and will be available for Medical X-ray products with the following advantages:

1)    Excellent EMI behavior by using soft switching techniques (cost advantages because of less measures needed to stay in the requested limits for the medical application, this could be housing bus also additional suppression components. At the end this will save costs);

2)    Higher efficiency – loss reduced

3)    Size reduction and costs reduction achieved

4)    Efficiency of isolated concept increased

Low EMI power converter:

The demonstrator is a “PI-AVR” (PI for an alternator voltage regulator) with low EMI active 3 phase automotive rectifier. The target of loss reduction was achieved and the efficiency gain was in the required range.

Highly efficient telecom rectifier:

The prototype development resulted in a highly efficient telecom rectifier. Increase the efficiency was a must to enable convection cooling. This could be achieved by using the latest power electronic components.

Burn–in reduction:

Industrial and automotive applications show a steadily increasing demand in terms of quality and reliability. Various qualification tests are implemented in order to guarantee the products long term stability. Semiconductor devices show an increased failure rate at the beginning of their life. Different procedures are applied for keeping this early life failure rate (ELFR) level at a minimum. However, in case of novel technologies, new production lines etc., the actual ELFR needs to be evaluated. This is done by stressing an independently selected number of devices after production – referred to as burn-in study. A burn-in study is successfully completed if the early life failure probability p is proven to meet a target ppm-level at a certain confidence level. In general, zero failures are required. Once a fail occurs, the burn-in study actually has to be restarted. The total savings of new burn-in methodologies exceed the expected results and lead to a faster closure of BI studies and release from 100% BI.

 

AUTOMATION challenges

Within the EPT300 project a lot of automation challenges were addressed already at the beginning of the project establishing the first worldwide semiconductor fabs using 300 mm power manufacturing processes. Many of the new processes required new materials, or equipment which was not available at the market supporting the classic semi 300 standards. Therefore, high efforts were spent for equipment and process automation for a lot of new non-standard tools. Classic automation systems using robotic tool loading are using conveyor based tool loading system like robotics (see Fig.1b) or overhead (OHT) based  automated material handling system, see Figure 1 c below. Within the EPT300 project all different automation concepts were investigated for their optimized use within the new 300 mm wafer facilities and implemented within the pilot lines.

Another focus area was the development of automated handling procedures of 300 mm thin wafer handling. Those substrates are very sensitive in terms of wafer breakage and all end effector systems within the special thin wafer process chain must be optimized to the individual requirements of each handling procedure. Figure 2 shows the handling platform used at IFAT and IFD using typical end effectors, shown in Fig. 2b. Handling of thin wafers was optimized to the lowest possible contact regarding particle and defect density level (see example of a typical defect density map within Fig.2 c) as well as lowest level of trace element contaminations on the wafers.

 

Very complex is the use of many different substrate types including orientation of the wafers. A lot of new automation- as well as new identification concepts, and sorting steps were worked out fulfilling the requirements of the power 300 business.

As a new method, a tool box for analysing and simulation of new automation concepts as well as capacity evaluation for existing automated material handling system was developed which is the base for many new performance monitoring before implementation of cost intensive transportation hardware. Figure 3 shows an example for the virtual fab analysis which is used now within the ramp up of the fabs.

Figure 3: Example for fab simulation (demonstrator was worked out within EPT300)

 

 


 

ABSTRACTS from EPT300 - INNODay 2015:

EASYLOAD 300-200: Automation with Focus on Mixed-Wafer Fab

HAP GmbH Dresden
Gostritzer Str. 63

01217 Dresden

Presenter:      Burkhard Stegemann
E-mail:                       
burkhard.stegemann@hap.de

Project: EPT300

-.-.-.-.-.-.-.-.-.-.-.-.-.-

Motivation

Automation in a wafer fab with multiple wafer sizes has special challenges. Especially the mixed processing of 200mm and 300mm for power devices, with optional thin wafer and relevant substrate capability, is of importance.

Description

The automation of parallel loading/ handling of 200mm and 300mm wafer carriers onto production systems needed a solution.

Further, a sorting tool with the following features was looked at:

·         Batch sorting

·         Load/ unload of different wafer sizes in standard boxes

·         Load/ unload of different wafer sizes in special, non-standardized, boxes, as for example metal carriers for furnaces (200mm/ 300mm)

Innovation

A complex simulation has been carried out how an automation interface for a parallel production line of 200mm and 300mm wafers could be designed.

Additional, a realisation for the product EASYLOAD includes:

·         Improved range of flexible wafer pitch of batch endeffector (6m-15mm pitch)

·         Wafer mapper now works at 3 different positions/ distances to cope with different cassettes and wafer sizes: Centre, 25mm forward, 50mm forward

Results

A concept for an automation interface for a parallel production line of 200mm and 300mm wafers was created.

The product EASYLOAD 300-200 was developed. The new needed functionality is implemented.

 

Outlook

HAP continuously develops new systems for the semiconductor industry and improves existing ones. The latest product family, the HAP-HERO® FAB, is a fully free travelling automated guided vehicle (AGV) for both, local transport and sophisticated handling/ loading tasks. The HAP-HERO® FAB is capable to handle 200mm wafers in cassettes or transport boxes as well as 300mm wafers in FOUPs, FOSBs or cassettes and other substrates.

Automation concept for parallel 300/200mm line

EASYLOAD 300-200 for mixed wafer size

New endeffector with variable pitch

 

 

                            


 

Automated Planning and Creation of Simulation Experiments for Semiconductor Manufacturing AMHS

Thomas Wagner1, Germar Schneider2

1TU Dresden
2Infineon Dresden GmbH

Presenter: Thomas Wagner
E-mail:           
thomas.wagner2@tu-dresden.de, germar.schneider@infineon.de

Project: EPT300

-.-.-.-.-.-.-.-.-.-.-.-.-.-

Motivation

Especially for power semiconductor production, new manufacturing and automation concepts are required, e.g. regarding the use of other substrates than silicon wafers. To allow a judgment on how an existing automated material handling system (AMHS) can cope with the new challenges or which alterations are required, a material flow simulation is essential. However, the planning and creation of such simulation experiments is difficult because of the systems complexity, the large amount of boundary conditions and the effort of manually modifying and testing many different variants. In order to assist in this process, a method for rapidly creating valid simulation experiments by using a knowledge base that allows for the reuse of experiments and system experts’ knowledge is suggested.

Description

The premise to the work flow depicted in Figure 1 is the availability of formal conceptual knowledge about manufacturing systems and their simulation. Much of this can be acquired from existing ontologies, e.g., OntoCAPE. Some important concepts needed for simulation are not included in OntoCAPE such as and transport systems. Consequently, these concepts are added in separated ontology modules. As a result, by the integration of existing ontologies and following their concepts, the introduced knowledge base is easily extensible. In a first step, individual fab specific manufacturing data must be mapped to abstract concepts described in the various ontologies mentioned above. Since the amounts and formats, e.g., CSV-Files and databases, of core data present in different manufacturing sites is not always fully known upfront, in most cases proprietary software modules are needed for this task. The demonstrator software tool uses a plugin-based approach to easily exchange or augment the pool of software modules for data acquisition. The simulation generator composes the simulations from small templates by querying the domain models stored in the ontology. For this purpose, a user friendly modeling front end was implemented. Using this front end, the experimenters define the variants they want to compare. For instance they will, e.g., change the product mix or rearrange tools. The possibilities are automatically restricted by executing queries on the ontology as depicted in Figure 2. Querying as well as reasoning over conflicts is carried out hidden from the experimenters. As a result, valid simulation models of the new factory variants are created. Using the wizard shown in Figure 3, the experimenters may also define a series of experiments by selecting the parameters they want to modify as well as the parameter ranges they are interested in. The demonstrator will then automatically compile and mass-execute the set of appropriate simulation experiments.

 

 

Innovation

A methodology is presented to assess the effects to a fab’s AMHS that result from modifications, e.g., integrating new equipment or changing the product mix. The core idea is to automatically generate detailed AMHS simulations from a knowledge base. This eases the time consuming human task of repeatedly creating and executing necessary simulation experiments. A plugin based software tool was introduced that populates the knowledge base and allows the user to freely design valid. An extensible set of plausibility checks is performed during the design by accessing the underlying knowledge to derive restrictions, e.g., space limitations, media requirements or necessary environmental conditions. When generating the simulations, the ontology is used again to produce robust material flow simulation models.

Results

The demonstrator has been applied to exemplary industrial use cases, for example experiments for testing different production scenarios and product mixes. In one use case, a fab production bay should be tested if it can cope with smaller lot sizes. For this purpose, the parameters “wafer starts per week” (several amounts from low to high volume), “vehicle count” (3-18, steps of 3) and “lot size” (1, 3, 6, 12 and 25) were tested by parametrizing the wizard mentioned above. As a result, a total of 150 experiments were automatically created and executed, allowing estimations about the impacts of producing with smaller lot sizes, see Figure 4.

Figure1: Workflow

Figure 2:Validation of tool positions

 

Figure 3: Experiment creation wizard

 

Figure 4: Results

 


 

Advanced Statistical Models for Semiconductor Burn-in Studies

Daniel Kurz1, Horst Lewitschnig2, Jürgen Pilz1

1 Department of Statistics, Alpen-Adria-Universität Klagenfurt
2 Infineon Technologies Austria AG

Presenter: Daniel Kurz, Horst Lewitschnig
E-mail:            daniel.kurz@aau.at, horst.lewitschnig@infineon.com

Project: EPT300

-.-.-.-.-.-.-.-.-.-.-.-.-.-

Motivation

A burn-in (BI) study is applied to release full BI testing of some (newly developed) products by demonstrating a target failure probability for the produced devices. However, from a statistical point of view, there is substantial potential for improvement with respect to the handling of BI studies by considering further available information (e.g. countermeasures introduced to the production process, synergies between different chip technologies, differently reliable chip subsets, etc.) within the estimation of a product’s failure probability.

Description

The failure probability of the devices is estimated on the basis of the observed number of BI relevant failures in the BI study. If this estimate is below the predefined target failure probability, full BI testing can be released. Nevertheless, the classical handling of BI studies suffers from various drawbacks, e.g. restart of BI study if failures occur, no consideration of synergies between different chip technologies, too conservative handling of failures on multiple reference products, no consideration of differently reliable chip subsets, etc.

Innovation

We developed a series of novel statistical models for the estimation of the failure probability of a chip, which contribute to a more efficient handling of semiconductor BI studies. These models involve (i) an estimation method for BI failures, which are tackled by countermeasures implemented in the chip production process, (ii) an estimation concept for the failure probability of a chip, which takes advantage of synergies (e.g. comparable chip layers) among the different chip technologies, (iii) an approach for handling BI studies on multiple reference products with different chip sizes and (iv) a novel area scaling model, which is capable of scaling differently reliable chip subsets separately from each other.

Results

All of these models contribute to a reduction of the efforts of BI testing (e.g. less burnt devices, less BI equipment, less engineering resources, etc.), and lead to a faster closure of the BI study. These models can save up to 1/3 of the burn-in costs.

 

 

 


Figure 1: Benefit due to countermeasure (CM) model.



Figure 2: Benefit due to model for synergies between different chip technologies.

 

References

D. Kurz, H. Lewitschnig and J. Pilz: Decision-Theoretical Model for Failures Tackled by Countermeasures, IEEE Transactions on Reliability, vol. 63, no. 2, pp. 583-592, 2014.

D. Kurz, H. Lewitschnig and J. Pilz: Modeling of Chip Synergies for Failure Probability Estimation in Semiconductor Manufacturing, Submitted to: Journal of Applied Statistics, 2014.

D. Kurz, H. Lewitschnig and J. Pilz: Failure Probability Estimation With Differently Sized Reference Products For Semiconductor Burn-in Studies, Applied Stochastic Models in Business and Industry, 2015. DOI: 10.1002/asmb.2100

D. Kurz, H. Lewitschnig and J. Pilz: An Advanced Area Scaling Approach for Semiconductor Burn-in, Microelectronics Reliability, vol. 55, no. 1, pp. 129-137, 2015.

D. Kurz, H. Lewitschnig and J. Pilz: Survey of Recent Advanced Statistical Models for Early Life Failure Probability Assessment in Semiconductor Manufacturing, Proceedings of Winter Simulation Conference, pp. 2600-2608, 2014.


 

Application of Combined Multi-Parameter Simulation for the Investigation and Optimization of Epitaxial Silicon Deposition

J. Kaczynski1, J. Baumgartl2 and M. Kraft1

1 CTR Carinthian Tech Research AG, Europastr. 4, 9524 Villach/St. Magdalen, Austria
2 Infineon Technologies Austria AG, Siemensstr.
2, 9500 Villach, Austria

Contact E-mail:         martin.kraft@ctr.at

Project: EPT300

-.-.-.-.-.-.-.-.-.-.-.-.-.-

Motivation

In modern semiconductor manufacturing, two - contradicting – requirements emerge: i) possibly high homogeneity, e.g. of layer thickness and composition and ii) high throughput. As both are equally essential, the processes must be optimised to account for both. Using 300 mm substrates makes this even more of a challenge, since larger areas have to be e.g. held at constant temperature and supplied with closely controlled media streams. Traditionally, such optimisations are done experimentally, which is a proven but – even when supported by DoE approaches – lengthy and disruptive method that does not always provide unambiguous results. To thus to better understand the underlying parameters and processes governing system behaviour, a joint team of CTR and IFAT hence decided to supplement these activities by using combined multi-parameter simulations of a core processes of 300 mm power semiconductor manufacturing: single-wafer epitaxial deposition,

Description

For the simulations, a detailed, precise 3-D geometry model of the epitaxy reactor was established that included the impact of system heat-up on the model geometry itself (Fig. 1). Fully parameterised material models of the fluids and solids in question were combined with multi-parameter models of the physical processes in the system, i.e. fluid dynamics, heat transfer and the chemical equilibrium reaction yielding the desired silicon layer. The latter was introduced as a special user-defined function on the finite volume level, which proved essential for achieving a realistic and reliable prediction.

Results

In the course of the project, dedicated models were developed and validated against data from actual measurements. While the first models correctly predicted the basic shape, the actual deposition rates were over-predicted by ~ 60%. Only by using a specially developed, refined model that eliminates a number of common simplifying assumptions it became possible to predict the deposition rates with an accuracy better 5% (Fig. 2). This model could then be used to investigate several of the more pressing problems, trace them back to their physical origins and develop, assess and compare a range of counter-strategy approaches. For instance, “cold spots” in epitaxial deposition could be linked to a system purge flow; with an optimised design, the gradient is reduced to almost zero (Fig. 3).

Fig. 1: 3-D geometry of the single-wafer epitaxy reactor chamber, scaled to operating temperatures

Fig. 2: Comparison of simulated and measured Si deposition rates; the deviation of simulation results (for standard settings) and experimental rates (for an optimised parameter set) is < 5%

 

Fig. 3: Temperature gradients in the wafer susceptor before (left) and after (right) optimisation; Boundary condition: perfectly homogenous temperature distribution at the wafer surface; shading scaling 2 K

 


 

 

FEM Simulation and Experimental Validation of Warping of Bare Large-Diameter Si Thin Wafers

J. Schicker, A. Iravany, T. Arnold and C. Hirschl

CTR Carinthian Tech Research AG, Europastr. 4, 9524 Villach/St. Magdalen, Austria

Contact E-mail:         johannes.schicker@ctr.at

Project: EPT300

-.-.-.-.-.-.-.-.-.-.-.-.-.-

Motivation

For both i) optimising wafer handling in semiconductor manufacturing and ii) facilitate understanding and eventually modelling the mechanisms of slip-line development, the ability to accurately and efficiently simulate the wafer behaviour, i.e. its deformation and stress development, under bending conditions is a vital prerequisite. Still, this task is as yet unresolved and highly challenging, mainly due to i) an extreme aspect ratio of the geometry, leading to ill-conditioned matrices in the governing equations, ii) the anisotropic nature of monocrystalline silicon, which prohibits usual simplifications and leads to disproportionately large systems with high calculation efforts, and finally iii) the problem of contact modelling, which evolves to be a computing time-consumer and a source of uncertainties.

Methodology and Scope

Lacking a significant experimental database, a combined strategy was chosen. First, a series of successively thinner and larger wafers was calculated for a ball-on-ring test arrangement, beginning with a moderately compact wafer with known deformation values and a minor degree of difficulty, and then evolved with a focus on result plausibility. The simulation models were refined towards Finite Elements with faster convergence behaviour, more finely adjusted numerical convergence criteria, and faster converging contact algorithms. For validation, the deflection of suitable thin wafers was measured by THz imaging and compared to values calculated using the researched methods. However, already minor intrinsic stresses, like tensile membrane stresses introduced with wafer grinding, may cause significant additional warping (Fig. 1); to gain results for model validation, wafers without intrinsic stresses were needed (Fig. 2).

Results

A key result is that higher order elements converge faster and with fewer elements to a final solution. Very slender wafers can only be calculated efficiently by using shell elements, due to the extremely increased calculation effort when using solid elements. Although such shell elements lack accuracy, specifically in the vicinity of concentrated loads, suitable stress estimates can be obtained by employing sub-models for these regions. The thus predicted warpage maximum vs. successively increased wafer slenderness is shown in Fig. 3. Fig. 4 validates this by showing the calculated shape of a wafer centrally loaded by a weight in comparison to the THz measurement. It can be seen that the agreement is not perfect, but the error is significantly less than 10% of the experimental values.

 

Fig. 1: Experimentally determined deformation of an uncoated 200 mm / 300 µm wafer on a ring with Di=194 mm. Left: although the wafer is loaded by its dead weight it is still slightly (about 30 µm) bulged up. Right: the same wafer turned upside down in the same ring shows an extended sagging of about 200 µm; the different bulging can only be explained by tensile membrane pre-stressing e.g. due to single-sided wafer grinding.

Fig. 2: Experimentally determined deformation of an uncoated, stress-relief 200 mm / 100 µm wafer on a ring with Di=194 mm. Top side up (green) and bottom side up (blue) show nearly identical deflections from dead load, characterising a nearly stress-free wafer. Left: 3D plot; middle and right: projection of the measured 3D values onto the the y-z-plane

 

Fig. 3: Finite Element results for various wafers in a ball-on-ring test arrangement: wafer deflection vs. the diameter / thickness quotient, a fixed force on the ball (21.5 N)

 

Fig. 4: Comparison of Simulation and Measurement of an approximately 114 µm thick wafer on a ring with Di=194 mm, centrally loaded by a 100 g weight: the solid red line shows the calculated wafer warping whereas the blue and green dots depict the values obtained by THz measurements. Since no measurements can be done where the weight is located the points in the centre are missing. The blue dots were obtained using a weight with a hole to obtain a rough estimate of the deflection beneath the weight.


Optimizing the Performance of Wet Chemical Silicon Etching by Optimizing Dispense Type

 Stefan Detterbeck1, Heinz Cramer1

1Lam Research AG
Presenter:      Stefan Detterbeck
E-mail:                       
Stefan.detterbeck@lamrc.com

Project: EPT300

-.-.-.-.-.-.-.-.-.-.-.-.-.-

Motivation

For future technology nodes a clear need for improvement of the performance of the silicon etch process is projected. Main focus is on etch rate uniformity and the stability of the etch rate uniformity. Today values of 5% (3sigma) for both of these parameter are achievable, the future need for these values however is less than 3% (3 sigma). Etch rates as of today are acceptable.

Description

This paper shows variations of key process parameter, spin speed, medial flow and dispenser profile, and their response to the wet etch results. The stability of the results was not deemed to be sufficient for optimized parameter, so a new dispense nozzle was developed to improve this stability. The new dispense nozzle was tested for the response by variation of the above mentioned key parameter.

Innovation

A new nozzle will be shown, which enables a curtain shaped dispense pattern, traditional dispense patterns are either single point dispense or spray dispense. Both of these dispense approaches were did not show sufficient promise. The innovation is a special type of nozzle enabling this media curtain, as well as enabling the usage of this media curtain by guarantying sufficient performance for switching this media curtain On and Off

Results

This paper will show the response for variation of the key process parameter spin speed, media flow and boom swing profile for a point type dispense and also for a curtain type dispense.

The differences in the response as well as the potential for creating more stable and repeatable results will be discussed.

The overall variation and specially the variation between multiple wafers could be significantly reduced.


 

Modeling Supporting Tools to Design Pilot Scale CVD Reactors

Filippo Rossi1, Maurizio Masi1 , Danilo Crippa2

1 Dipartimento di Chimica, Materiali e Ingegneria Chimica, Politecnico di Milano
2  LPE Epitaxial Technology, Baranzate, Milano

Presenter: Danilo Crippa
E-mail:           
danilo.crippa@lpe-epi.com

Project: EPT300

-.-.-.-.-.-.-.-.-.-.-.-.-.-

Motivation

It is our intention here to describe the modelling procedure used to support the design of the chamber of the CVD reactor for the epitaxial silicon deposition on 300 mm wafers. The goal was to design a reactor that allows the deposition of thick films at high deposition rate while keeping the crystallographic quality of the films grown on 200 mm wafers. Other goals were the maximum utilization of the reactants and the possibility of a fast switching the dopant concentration to realize sharp junctions.

Description

The reactor design was here performed entirely through a modelling procedure aimed at obtaining a reasonable prediction of its performance. At the end, the final design that satisfactorily answers to the industrial priorities can be obtained only through full 3D models embedding a realistic chemical kinetics. These requirements are somewhat in contrast with the industrial necessity of “fast solutions”. These fast solutions are particularly required when different design ideas need to be tested and compared in order to decide which one should be adopted.  Accordingly, the full 3D model should be adopted only to verify solutions previously identified through simplified models addressing only partial aspects of the problem. This hierarchical procedure [1] was already adopted for the analysis of different types of CVD reactors (barrel & horizontal ones) thus; it will be not addressed here in detail, being the model equations described in [17–20].

Innovation

The innovation here presented is related to the first application of the above procedure to simulate a CVD chamber for the epitaxial silicon deposition on a 300 mm wafer. The final simulated system is a 300mm single wafer horizontal reactor. During the first screenings, a configuration of a very large chamber with the possibility to load 3x300 mm wafers was also considered. About the chemistry side, the innovation consists in considering also the presence of the etching reaction due to the presence of HCl in large extent (large size reactors present a greater utilization of the TCS thus gases in the tail part of the reactor are richer in etchants.


 

Results

Examined reactor configurations through scaling analysis and simplified models were:

·         cold-wall 1x12” scaled in width & length, while keeping same height;

·         cold-wall 1x12” scaled in width & length, while keeping same aspect ratio;

·         hot-wall 1x12” scaled in width & length, while keeping same height;

·         hot-wall 1x12” scaled in width & length, while keeping same aspect ratio;

·         cold-wall 3x12” scaled in width & length, while keeping same height;

The scaling analysis was performed imposing the same performances in terms of growth rate & uniformity (identified by the Damkholer dimensionless number) as the Pe3061. Accordingly, the comparison was performed with respect to the running consumptions (H2, TCS, energy).

Main conclusion confirms that hot-wall reactors requires the consumption of higher chemicals quantity because the double deposition surface (susceptor + walls). However, the almost absence of temperature gradients should improve significantly the crystallographic quality of the deposited film.

In a reasonable forecast, the cold-wall 1x12” reduced height could be operated at higher growth rates (4 and 6 micron/min), while the cold-wall 3x12” probably will present a growth rate lower than Pe3061 (i.e., 2.5 micron/min) to reduce consumptions and to match the crystallographic quality.

Finally, an example of a 3D simulation, comparing two different shapes of the reactor chamber is presented to highlight the overall capability of the whole procedure here adopted (Fig.1).

1.1              

Figure 1. Example of a distributed inlet-jet system on a horizontal CVD reactor for silicon epitaxy using SiHCl3/H2. (a) Gas velocity, (b) gas temperature, (c) precursor mass fraction and growth rate over the susceptor. Half of the reactor is shown.


 

300 mm Silicon Wafers for Power Devices

Thai Do, Erich Gmailbauer, Thomas Schröck, Frank Mümmler

1  Siltronic AG, Burghausen

Presenter:      Frank Mümmler
E-mail:            frank.muemmler@siltronic.com

Project: EPT 300

-.-.-.-.-.-.-.-.-.-.-.-.-.-

Motivation

The basis for all semiconductor devices is the availability of semiconductor substrates. Without substrates, there will be no devices.  In order to enable a 300mm Power Device Pilot production for the first time in the world, advanced Silicon substrates had to be developed within the EPT300 project

Description

Requirements for Power Device substrates are quite different to CMOS substrates. When Power Devices moved to 300 mm in the EPT300 Project, also new substrates had to be developed.

300mm arsenic doped silicon crystals were grown for the first time with specific challenges to handle the high amount of melt and arsenic dopant for 300mm crystals with very high dopant concentration.

Another challenge was the development of substrates for IGBTs, as Float-zone wafers were not available for 300 mm.

Innovation

300mm arsenic doped substrates in the resistivity range <3 mOhmcm were developed during the project. Also 300mm ultra low Oi substrates for IGBT devices were developed.

Results

Power devices could be successfully produced on the new substrates and thus a pilot production on 300mm wafers for Power Devices could be started for the first time in the world.


 

Dissemination

A public website to represent EPT300 and its progress was designed, implemented and launched by Eutema in co-operation with Infineon Technologies Austria AG within the first seven months of the project. The design and content went through an internal agreement process to gain the consent of all partners of the EPT300 consortium. The website is updates regarding news, activities and dissemination. The website is hosted by eutema www.EPT300.eu. An eRoom has been setup to enhance the collaboration within the consortium (hosting of files and providing notification mechanisms as described in deliverable 6.1). The portal is hosted by Infineon. The maintenance was performed by IFAT supported by Eutema which is a continuous activity during the duration of the project. http://www.infineon.com/cms/en/product/index.html

To disseminate not only the results, but also the benefits of the co-operation between research / academia and industrial partners as well as the role of research / academia in a large pilot-line project such as EPT300 was one of the endeavours successfully demonstrated by activities such as:

·        Dedicating the second day of the year two meeting of the consortium to the demonstration of work and progress of the academic partners and research organisations to invited participants

·        Contributing as a demonstrator in the multi key enabling technologies pilot line research project mKPL. http://www.mkpl.eu/home/

·        Participation in ECSEL Austria Events (eg. 16th and 17th September, 2014)

·        Conducting an EPT300 / EPPL day – the Innovation Day 2015 (6th of May 2015)

Based on the dissemination plan, close co-operation of the consortium and the efforts of all partners a large number of dissemination activities were presented to the steering board via the dissemination approval procedure. Managed by eutema and IFAT this resulted in a total number of ~130 dissemination activities.

These activities are reaching from small activities as newsletter distributions and website contributions to large activities as the Innovation Day 2015 that was held at Infineon Villach addressing the international semiconductor equipment and process community (scientific and industrial) and funding experts.

The dissemination activities further include scientific publications and contributions to high impact conferences as well as journals well received within the industrial and scientific community. These include:

·        APEC Conference

·        PRECEDE Conference

·        IEEE Transactions on Emerging and Selected Topics in Power Electronics  

·        2015 International Symposium on Semiconductor Manufacturing Intelligence (ISMI2015)

·        European Network on Business and Industrial Statistics (ENBIS) 2015

·        Eighth International Workshop on Simulation (IWS)

·        IEEE EuroSimE 2015, Budapest/HU

·        AEC Reliability Workshop 2015, April 28-30

·        IEEE Transactions on Power Electronics

·        EPE 2015 ECCE Europe, Geneva

·        WinterSim Conference 2013 + 2014

·        ECS, May 2015, Chicago

·        IECON 2014

·        Microelectronics Reliability (MR)

·        ENBIS (European Network for Business and Industrial Statistics) 2014 Conference, Linz (Austria)

·        Conference on Modelling and Analysis in Semiconductor Manufacturing (MASM) 2014

·        Journal of Applied Statistics (JAS)

·        Applied Stochastic Models in Business and Industry (ASMBI)

·        Nordstat 2014 (Turku, Finland)

·        ENBIS 2013 (Ankara, TUR)

·        Third Symposium on Games and Decisions in Reliability and Risk (Kinsale, Ireland)

·        CASI 2013 (Kildare, Ireland)

·        ECCE down-under 2013

·        IEEE Transactions on Reliability

·        Quality and Reliability Engineering International (QREI)

And finally - as a further dissemination and exploitation highlight to be mentioned - 7 diploma and PhD theses were completed during the course of EPT300 conducted at IFD, UNIKL, CTR and TUD.

Economical and social impact

The Project is fully in line with the Europe 2020 initiative where the European Commission has set ambitious targets for reduction of greenhouse gas emissions, energy efficiency and electro-mobility. Striving towards these goals, power semiconductors designed and manufactured at low cost and in sufficient quantities in Europe and enabled by equipment and materials supplied by European companies. With the help of EPT300, the European semiconductor manufacturing and the European equipment & materials industries is the first worldwide with a released 300mm power semiconductor processing line dedicated to power device production. By this, EPT300 is an important step towards realizing leading edge high volume manufacturing capabilities in Europe. The 300 mm pilot line activities and researches has a huge social and economic impact, first EPT300 enabled the EPPL – eRamp and Power Base Projects, second as social impact we address the press release were Infineon recently informed the public regarding the expansion plans foreseen investments and research costs amounting to a total of € 290 million, crating approximately 200 new jobs in the period from 2014 and 2017, primary in R&D. All EPT300 goals have to be considered as a highly challenging aspect, which deserved high attention and much effort by Infineon and the Consortium well beyond the project itself. With EPT300 project ECSEL JU, Europe is reinforcing its important role in innovation and competence for power electronics. We're making an important contribution to the success of the European semiconductors by coupling the innovation factory in Villach with volume production in Dresden using the achievements of EPT300 Project.

 


 [JI1]