Shaker Verlag Aachen 2013 Schriftenreihe des IME

”Elektroschlackedesoxidation von Titanaluminiden”

Von der Fakultät für Georessourcen und Materialtechnik der Rheinisch-Westfälischen Technischen Hochschule Aachen

zur Erlangung des akademischen Grades eines Doktors der Ingenieurwissenschaften

genehmigte Dissertation vorgelegt von Dipl.-Ing.

Jan Bertil Reitz

aus Friedberg / Hessen

Berichter: Univ.-Prof. Dr.-Ing. Dr. h.c. (UA) Karl Bernhard Friedrich Ao. Univ. Prof. Dipl.-Ing. Dr. mont. Helmut Antrekowitsch Prof. Dr.-Ing. Michael Schütze

Tag der mündlichen Prüfung: 22. April 2013

Shaker Verlag Aachen 2013 Schriftenreihe des IME

Band 32

Jan Bertil Reitz

Elektroschlackedesoxidation von Titanaluminiden

WICHTIG: D 82 überprüfen !!!

Bibliografische Information der Deutschen Nationalbibliothek

Die Deutsche Nationalbibliothek verzeichnet diese Publikation in der Deutschen Nationalbibliografie; detaillierte bibliografische Daten sind im Internet über http://dnb.d-nb.de abrufbar.

Zugl.: D 82 (Diss. RWTH Aachen University, 2013)

Copyright Shaker Verlag 2013

Alle Rechte, auch das des auszugsweisen Nachdruckes, der auszugsweisen oder vollständigen Wiedergabe, der Speicherung in Datenverarbeitungs- anlagen und der Übersetzung, vorbehalten.

Printed in Germany.

ISBN 978-3-8440-2022-9 ISSN 1610-0727

Shaker Verlag GmbH • Postfach 101818 • 52018 Aachen Telefon: 02407 / 95 96 - 0 • Telefax: 02407 / 95 96 - 9 Internet: www.shaker.de • E-Mail: info@shaker.de

Danksagung

Für die herzliche und vielfältige Unterstützung meiner Forschung und meines Werdegangs gilt mein Dank zunächst Professor Friedrich. Bei ihm bedanke ich mich neben der ausgezeichneten wissenschaftlichen Betreuung insbesondere für die Unterstützung meines Interesses an der Arbeit im internationalen Umfeld durch Auswahl und Förderung entsprechender Projekte, Konferenzen und Industriekontakte. Ganz besonderer Dank gilt außerdem Tim Georgi-Maschler, André Dammschröder, Andreas Lützerath, Aybars Güven, Johannes Morscheiser und Marek Bartosinski die während meiner Institutszeit von Kollegen zu Freunden geworden sind. Ich bedanke mich explizit bei Claus Lochbichler für Motivation und Inspiration bei der Betreuung meiner Diplomarbeit, sowie für zahlreiche Fachgespräche auch nach seinem Wechsel in die Industrie, die wesentlich zum Gelingen der Fortführung unserer TiAl-Aktivitäten beitragen konnten.

Mein Dank gilt weiterhin der Europäischen Union für die Förderung meiner Forschung im 6.

Rahmenprogramm, Projekt IMPRESS unter der Vertragsnummer NMP3-CT-2004-500635. In diesem Zusammenhang bedanke ich mich im speziellen beim Projektleiter Dr. David Jarvis für die gewährte Freiheit und vorgelebte Motivation durch Werkstoffforschung etwas bewegen zu können, sowie bei Ulrike Hecht von ACCESS e.V. für zahlreiche gemeinsame Sessions am REM und den tieferen Einblick in die Welt des γ-TiAl.

Bei der Vorbereitung und Durchführung meiner Versuche am IME erhielt ich vor allem von Leo Klinkenberg schlagkräftige Unterstützung. Ich bedanke mich bei ihm für die großartige Zusammenarbeit bei der Reparatur und Wartung der Versuchsanlagen und insbesondere im Rückblick auf unsere erfolgreiche Versuchskampagne auf dem Gebiet der Aluminothermie. Mein Dank geht auch an Horst Leuchter und Amir Khamoushkhoo, sowie an das Team aus der mechanischen Werkstatt des IME, ganz besonders an Herrn Kamp für seinen zeitlich flexiblen Einsatz in der Zeit meiner Diplomarbeit und darüber hinaus. Ich bedanke mich bei Herrn Funk für die täglichen, freundlichen Grüße und die geduldige Präparation der undankbaren TiAl-Proben mit dem Trennschleifer. In meinen Dank schließe ich das Administationsteam des IME, insbesondere Irmi Koren und Martin Köllmann ein.

Auch bei unseren Industriepartnern möchte ich mich herzlich bedanken. Über die finanzielle Unterstützung der Unternehmen hinaus, seien hier explizit genannt Dr. Tewes, ThyssenKrupp VDM für die Inspiration zur Nutzung von R, Harald Scholz, ALD Vacuum Technologies für die interessanten Fachgespräche, Ashish Patel, Carpenter Technologies, USA für das beste Industrieprojekt meiner Zeit am Institut, Gerd Zentis vom Elektrowerk Weißweiler für Unter- stützung meiner ATR-Versuche und Herr Dr. Kramer für die besonders lehrreiche und schöne Zusammenarbeit.

Ich danke ich meiner Familie für den starken Rückhalt vom Beginn meines Studiums an, für die moralische, aber auch fachlich-chemische Unterstützung, insbesondere aber für Flügel, Wurzeln und die regelmäßige Ermunterung meinen eigenen Weg zu gehen.

Wahrscheinlich hätte ich diese Arbeit jedoch nie fertig geschrieben ohne die liebevolle und geduldige Unterstützung meiner Freundin Teresa - Dankeschön!

iii

Abstract

Mit ersten industriellen Anwendungen bei Turboladern im Automobilbereich und der Nieder- druckturbine moderner Flugtriebwerke erfährt die Werkstoffgruppe der γ-Titanaluminide (TiAl) derzeit besondere Aufmerksamkeit. Auf dem aktuellen Stand der Entwicklung werden, insbeson- dere im Hochtemperaturbereich, herausragende spezifische mechanische Eigenschaften erzielt, die für eine weitere Effizienzsteigerung von modernen Antrieben und Kraftwerken essentiell sind. In dieser Arbeit wird das Ziel verfolgt, trotz der hohen metallurgischen Anforderungen an die TiAl-Herstellung eine deutliche Senkung der Werkstoffherstellungskosten zu erreichen sowie ein Recyclingverfahren zu etablieren. Mittels Druck-Elektroschlackeumschmelzen (DESU) wird dabei unter Einsatz „aktiver Schlacken“ eine Raffination des Werkstoffs, insbesondere eine Absenkung des Sauerstoffgehalts erzielt, welcher sensibel die mechanischen Eigenschaften beeinflusst. Für mittels Vakuuminduktionsschmelzen, engl. vacuum induction melting (VIM), in keramischen Tiegeln rezyklierten TiAl-Schrott mit einer Sauerstoffkontamination von bis zu 3000 ppm, konnte eine Reduktion auf deutlich unter 500 ppm nachgewiesen werden. TiAl, das alternativ mittels aluminothermische Reduktion (ATR) gewonnen wurde und mit bis zu 16 000 ppm Sauerstoff verunreinigt ist, stellt eine prozesstechnische und kinetische Herausforde- rung für das Verfahren dar und konnte in der Versuchsreihe nicht unter 4000 ppm desoxidiert werden. Eine technisch-ökonomische Betrachtung zeigt, dass die Prozessroute VIM-DESU-VAR (triple-melt), auch im Vergleich mit alternativen Verfahren, die ideale Kombination aus flexiblem Einschmelzen, chemischer Desoxidation und finaler Raffination für das Recycling von γ-TiAl darstellt.

With technical applications in automotive turbo-chargers and the low-pressure turbine of the most recent aero engines, γ-titanium aluminides (TiAl) receive special attention of the technical community these days. On the present state of development, these materials exhibit outstanding weight-specific mechanical properties, even at high temperature operation - a basic design requirement for light-weight, energy-efficient engines and power plants. This thesis aims to reduce material processing cost significantly by establishing a recycling process that fits to their sensitive metallurgical requirements. The application of Pressure Electroslag Remelting (PESR) using “active slags” leads to a refinement and in particular to a reduction in oxygen content which impacts greatly on the mechanical properties. γ-TiAl-scrap remelted by vacuum induction melting (VIM) with an oxygen contamination of 3000 ppm could be successfully treated to levels below 500 ppm, while γ-TiAl obtained through an alternative raw-material route via aluminothermic reduction (ATR) and therefore contaminated with up to 16 000 ppm oxygen presents kinetic and technological challenges to the process and could only be treated down to 4000 ppm oxygen. In a techno-economical analysis, the interesting economical potential of the processing route VIM-PESR-VAR - a triple-melt process that ideally combines flexible scrap melting, chemical deoxidation and final refining - could be highlighted in comparison to alternative processes.

v

Extended Abstract

An integrated melting concept for low-cost γ-TiAl

Due to their good corrosion resistance, low density and interesting mechanical properties at high temperatures γ-phase titanium aluminides are widely regarded as the next generation material for high temperature structural applications in the aerospace and automotive industry.

However high material and production costs for semi-finished products actually slow down an industrial breakthrough to mass production of γ-TiAl.

From a processing standpoint the material yield in investment casting is similar for γ-TiAl as for standard titanium alloy casting with a loss in the range of 40–60 %. On the other hand γ-TiAl alloys are produced at a tremendously higher cost compared to classic titanium alloys like Ti-6Al-4V due to more strict specifications on composition and homogeneity. This is necessary in order to achieve the desired high temperature strength at an acceptable level of ductility. Thus material losses during casting of γ-TiAl should have a more severe impact on overall product cost than in classical titanium alloy casting and hence recycling becomes even more economically interesting for γ-TiAl than it is potentially for titanium.

Since the year 2000 a research group at IME, Aachen aims on cost-reduced production methods for γ-TiAl. While one path follows the alternative aluminothermic production of TiAl from TiO₂pigment and subsequent refining [61, 139], parallel activities aim for development of an efficient process for recycling TiAl-scrap from casting operations in order to avoid the present downgrading of this valuable material as an alloying element in the steel industry.

In IME’s integrated concept (see fig. 0.1) - which includes both alternative primary production and recycling - the first melting of a scrap charge is done by conventional vacuum induction melting (VIM) using specialized ceramic lining. This step can include preliminary deoxidation by injection of calcium metal into the melt [96]. Full deoxidation and reactive refining follows during pressure electroslag remelting (PESR) using a continuously re-activated calcium-metal- containing slag. During final vacuum arc remelting (VAR) micron size slag inclusions as well as dissolved calcium and hydrogen can be removed. For each processing step special equipment requirements and metallurgical challenges have to be met and the different refining opportunities need to be exploited to obtain a sound recycled TiAl-ingot. This dissertation presents the latest progress of the work with a special focus on oxygen removal, refining and homogenization by PESR. Final VAR melting of the obtained material is presently under investigation and results shall be presented in future works.

PESR as a key process for TiAl deoxidation

ESR is widely applied in the production of speciality steels and superalloys nowadays and leads to ingots with superior cleanliness, minimized segregation and a homogeneous microstructure.

As a basic principle the tip of a consumable electrode, made from the alloy to be refined, is submerged into a molten flux, contained in a water-cooled copper crucible. Electrical currents

vii

TiAl scrap 1000-1500 ppm O

alternative TiO₂ raw materials + Al powder

VIM ATR revert,

15 000 ppm O ATR

VIM electrode, 3000 ppm O

ATR electrode, 15 000 ppm O PESR

scrap route alternative production route

PESR ingot, 500 ppm O, 800 ppm Ca

VAR VAR ingot

500 ppm O

Figure 0.1.: Integrated concept for alternative production and recycling of TiAl

(most commonly AC) are applied between the electrode and the flux and result in joule heating up to the melting point of the metal. A liquid metal film forms on the downside of the electrode and hence metal droplets are released which sink through the flux bath due to the difference in density. The slag bath floats on the liquid metal pool which collects the droplets. Through cooling at the mould surfaces, solidification occurs and the process continuously builds up a refined ingot with a controlled microstructure and smooth outer surface. Detailled information about the process and its industrial practice can be found in the literature in the works of Hoyle [65] and others [101, 103, 85, 90, 91].

The potentials of the application of ESR for the remelting of pure titanium using CaF₂-based fluxes have been investigated by various researchers [99, 15, 23, 102, 33, 107, 142] with promising results concerning the possibility to produce slab-shaped ingots and improvements regarding material yield and energy consumption in comparison to VAR. Results also show interesting possibilities to reduce segregation effects due to the fact that lower melt rates compared to VAR are possible. Mechanical properties of ESR remelted c.p. titanium are reported to be comparable to VAR melted titanium [15], despite an ongoing discussion about the solution of fluorine in the metal [120]. Figure 0.2 illustrates how PESR offers an additional chemical refining potential in comparison to VAR.

Theory and calculations

Thermodynamics of metal-slag reaction in PESR

A range of suitable fluxes for ESR of titanium alloys have been investigated thoroughly by Nafziger et al. [99] with the outcome that CaF₂(fluorspar) offers the best combination of

viii

Figure 0.2.: Refining options that PESR offers for titanium alloys in comparison with the established VAR process

thermochemical stability, stable processing conditions and final ingot quality. Thermochemical calculations in the system Ca-CaO-CaF₂have shown that even the deoxidation of pure titanium should be possible by ESR using "reactive slags" where metallic calcium is added to the flux as a deoxidation agent [142]. Ca is chosen as a reactant, because of its high oxygen affinity, good solubility in the slag and available activity data [146] for the equilibrium of Ca and O in titanium melts. Furthermore Ca-reactive slags have been shown to offer a potential for the challenging removal of nitride inclusions [23]. A major challenge in the application of Ca-reactive slags for ESR however is the evaporation of Ca at the melting temperatures of titanium alloys. This effect can be kinetically suppressed by melting under inert gas overpressure, leading to pressure ESR (PESR) as a method of choice for the deoxidation and refining of recycled TiAl scrap material [142].

Dissolution of calcium from the slag into the metal phase results in a reaction according to eq. (0.1). Precipitated CaO is being suspended and then dissolved in the ESR slag and hence CaO activity should be lower than one.

[TiO]_TiAl+ [Ca]_CaF2−−−−

[Ti]_TiAl+ [CaO]_CaF2 (0.1)

with [A]B- Element A dissolved in phase B

ix

K =a(CaO)flux· a(T i)metal

a(T iO)metal· a(Ca)flux = f(T ) (0.2) with K - equilbrium constant, a(A)B- activity of A in phase B

It becomes obvious from eq. (0.1) and eq. (0.2), that through constant feeding of metal with the initial oxygen level into the metal-flux-boundary and continuous readjustment to chemical equilibrium, the flux enriches in CaO during the melt and is being depleted in Ca as the process proceeds. This has severe implications on the chemistry of the ESR process: In order to ensure a uniform oxygen content with respect to the full length of an ingot, calcium has to be added at the exactly correct rate in order to compensate for the increasing CaO-activity. Intensive thermochemical modeling of the metal-slag reaction was applied by Stoephasius et al. [142] in order to enable a controlled deoxidation by ESR. Based on a target oxygen concentration in the metal of 500 ppm, the necessary Ca-activity in the flux can be calculated for every level of CaO-activity as shown in fig. 0.3.

●

●

●

●

● ● ● ● ● ● ●

●

●

●

●

●

●

0 1 2 3

0 10 20 30

c(CaO) /Wt. %

c(Ca) /Wt. %

Temperature

●1750 °C 1650 °C 1550 °C

Figure 0.3.: Necessary Ca content in the active slag as a function of CaO concentration to reach 480 ppm oxygen in a Ti50Al melt.

Due to the current lack of reliable methods for online-surveillance of a(Ca) and a(CaO) in the slag, a process model was introduced based on mass-balance calculations, which includes dissolution of Ca in the metal, pressure dependent evaporation of Ca from the flux and the described deoxidation equilibrium according to eq. (0.2). The approach was put in to practice as a spreadsheet calculation and can be adjusted to melting parameters, e.g. metal melt rate, initial oxygen level of the metal, operating pressure etc. The sheet calculates feed rates for both calcium compensation and feeding of virgin slag, as fig. 0.4 illustrates, base on the assumption that the system operates at or close to chemical equilibrium. While this might be true for the phase boundary itself, it has to be examined if transport kinetics in the slag or metal phase

x

might limit the deoxidation process for higher oxygen containing TiAl materials.

concentration, ATR electrode − 15,900 ppm O concentration, VIM electrode − 3000 ppm O

feed rate, ATR electrode − 15,900 ppm O feed rate, VIM electrode − 3000 ppm O 0

5 10 15 20 25

2 4 6

40 50 60 70 80 90

10 15 20 25

0 25 50 75 100 0 25 50 75 100

0 25 50 75 100 0 25 50 75 100

melting progress /%

feed rate /gmin−1 − c(Ca, CaO) / Wt.%

c(CaO) c(Ca) rate Ca rate CaF2

Figure 0.4.: Higher feed rates for Ca and virgin slag are expected for electrodes obtained by ATR in comparison to those obtained from scrap by VIM.

Kinetics of metal-slag reaction in PESR

Two sources of γ-TiAl material in the focus of this work show different oxygen contents (VIM-type: 3000 ppm vs. ATR-type: 16.000 ppm). For both materials the target in PESR- deoxidation is at 480 ppm in order to obtain the desired mechanical properties. Assuming that the deoxidation reaction is limited to the contact time between liquid metal and liquid slag and that deoxidation stops upon solidification, this gives rise to an easy estimation: Based on metal melt-rates of 0.65 kg min⁻¹achieved in the experimental set-up, a minimum transfer rate for oxygen from the metal to the slag of 1.63 g min⁻¹ (0.1 mol min⁻¹) or 10.09 g min⁻¹ (0.63 mol min⁻¹) needs to be achieved as a baseline for the VIM-cast electrodes and the ATR-cast electrodes respectively.

Looking closer at the ESR process it is assumed that metallic calcium, being dissolved in the slag is transported to the metal-slag interface and reacts there with oxygen being transported from the bulk of the metal phase to the same phase boundary. In parallel Ca atoms cross the phase boundary and Ca dissolves in the metal. When the solubility product of oxygen and calcium in the metal phase is reached, CaO precipitates. The system should feature diffusion boundary layers on both sides of the metal slag-interface, while mass-transport in the bulk slag and metal phase which stirred by buoyancy and Lorentz forces should be fast enough for assuming homogeneous bulk concentrations. Assuming furthermore that thermochemical

xi



 





 



















  

 

 

 







          

Figure 0.5.: Deoxidation reaction across metal slag boundary takes place at three different reaction sites during PESR

equilibrium is quickly achieved at the phase boundary itself, it consequently has to be considered that diffusion of Ca and CaO through a finite layer in the slag and O through a layer on the metal side respectively could be rate limiting factors. To complicate things further kinetic studies in PESR have to consider three distinguished local zones for metal-slag interaction as illustrated in fig. 0.5

Zone 1: Reaction between the top-side of the slag and the liquid film below the electrode. This zone is characterized by continuous surface renewal through release of droplets and a temperature close to the metal liquidus.

Zone 2: Reaction between the bulk slag and the droplet falling through the slag. This field is characterised by increasing heat-transfer to the droplet and a low superheat can be assumed.

Zone 3: Reaction between the upper surface of metal pool and the lower surface of the slag.

In this phase the metal temperature equilibrates with slag temperature and superheats can reach up to 100 K.

Fraser [50] investigated various mass transfer aspects of A.C. ESR of steels (slagging of Mn and desulphurisation) in a very comprehensive thesis work and the findings of his work can be adapted to the conditions for the presented case. In a first step necessary thermophysical properties of slag and metal phase (density, viscosity, interfacial tension, diffusion coefficients of Ca and O in metal and slag, etc.) were determined as functions of temperature using applicable models and data from literature (e.g. Girifalco-Good [55] for interfacial tension). Further volumes, surface areas, residence times, droplet size, thickness of the liquid film below the

xii

electrode and diffusion rates were computed based on these values. These parameters were put together with equipment specific variables (e.g. ingot diameter, melt rate, slag and metal composition) in an MS Excel^TMspreadsheet to allow for easy scale-up scenario calculations. The governing differential equations were numerically solved independently for the three reaction zones at the respective boundary conditions. Residence time at every zone was split in time-steps of 10 % for recursive recalculation of the oxygen concentration. Because only a fraction of the total melting time is modeled with respect to highly reactive zone below the electrode (i.e. 2 s of a total melt of 1200 s) the composition of the slag was assumed to be constant in a first approach.

It could be shown that for our present case, Highbie’s penetration theory ([64], see also [112, 59]) holds for all three reaction zones for the metal and the slag side respectively and correspondingly transport coefficients obtain a square root dependency from the above diffusion coefficients. It can be observed that rate constants differ significantly from reaction zone to reaction zone mainly due to variations in boundary layer thicknesses (e.g. falling droplet vs.

bulk metal pool) and residence times. The liquid film below the electrode (zone 1) exhibits the largest area to volume ratio i.e. a strong impact of mass transport on volumetric concentration, while the metal pool maintains an aspect ratio two orders of magnitudes lower, however at a significantly larger residence time. The falling droplet, regarded by many authors as the main center of mass transfer in ESR due to the fast counter-current speeds at the interface can be nearly neglected for A/V being an order of magnitude lower than at the electrode film and extremely short residence times.

zone 1 − electrode zone 2 − droplet zone 3 − pool

4000 8000 12000

0 1 2 2.56 2.57 2.58 2.590 200 400 600

time /s

c(O) /ppm

model

ATR 15.000 ppm VIM 3000 ppm

Figure 0.6.: Results of the modelling of kinetic effects on deoxidation TiAl by PESR. From left to right: zone 1 - electrode surface, zone 2 - falling droplet, zone 3 - pool surface

For both cases of an electrode of the ATR type (15.000 ppm initial oxygen) and similarly for an electrode from VIM-casting starting at 3.000 ppm initial oxygen, the model predicts the largest drop in oxygen concentration at the liquid film below the electrode where, even though temperatures are lower, diffusion layers are thin, residence times relatively high and high transport rates are being calculated (compare fig. 0.6). At the metal pool interface transport rates are limited and even at the long residence times expected, the final oxygen in the ATR material should not drop below 4000 ppm, while the VIM-cast material is predicted to be below 700 ppm, which corresponds very well to the experiments conducted.

xiii

Experimental approach

Goals and set-up of experiments

The winning of TiAl by ATR and a first trial for remelting by PESR could be shown experi- mentally by Stoephasius [139], however, the proof of a successful deoxidation of this material by ESR is still open presently. In parallel Lochbichler [80] has experimentally proven melting of titanium and γ-TiAl in VIM with ceramic lining and the subsequent casting to electrodes suitable for ESR. These previous works have shown that electrodes with 2500-3000 ppm oxygen can be expected after VIM while electrodes obtained from ATR contain an order of magnitude more oxygen (> 15.000 ppm), which requires faster deoxidation kinetics as shown above. Adding about the double weight and the potential for slag inclusions, ATR electrodes impose an extreme load to the chemical refining capacity of the slag and therefore represent a more challenging case for PESR deoxidation compared to recycled scrap from VIM (see sections 2.3.2, 3.1.2 and 3.1.3).

Based on this situation the following methodology is chosen: In a first set of trials electrodes from binary γ-TiAl (Ti50Al) are obtained by VIM of spent sputter targets, a very clean starting material, and then deoxidized via PESR as a proof-of-principle for a reproducible deoxidation process. In a second set, electrodes are obtained from real foundry scraps of the alloys Ti46Al8Nb, Ti46Al8Ta and TNB-V5 via VIM for PESR deoxidation with the goal to examine recyclability, observe segregation of Nb and Ta and obtain a sound and homogenous ingot that could be reverted for casting trials. In a third phase, the PESR process is subjected to the challenge of extreme deoxidation of electrodes obtained by ATR. In the course of that work, the reverting of loose ATR material by VIM is examined. The works include the definition of a feasible process window (e.g. controller setting, slag mass, charging rates), validation of the thermochemical and kinetic process model but also the acquisition of basic data (e.g. energy consumption) for a techno-economic evaluation of the proposed recycling scheme.

Alternative winning of raw-TiAl via ATR

The basic principles for a controlled aluminothermic reduction of TiO2 and co-reduction of alloying oxides (e.g. Nb₂O₅) have been elaborated at IME in previous works [61, 139]. According to the published methodology, a charge calculation is recursively conducted in combination with thermochemical calculations which include experimentally determined heat losses for correct adjustment of the final reaction temperature. In a cubic reactor a 185 kg charge of TiO₂, Nb₂O₅, Al and KClO₄is ignited and reacts within a timeframe of approximately 2 min. At the bottom of the reactor an aluminium plate, precisely designed to the expected temperatures, melts when the reaction front has reached the drain and the melt is then being cast through a tap-hole into a horizontal crucible. After solidification a 55 kg electrode of Ti46Al8Nb was obtained (see fig. 0.7). In order to investigate homogeneity, one out of several ATR electrodes was sacrificed and sliced by water-jet cutting for analysis of homogeneity as shown in fig. 0.7.

Melting of TiAl scrap material via VIM

It has been shown by various researchers, that melting of titanium alloys and titanium aluminides in particular is possible in yttria (Y₂O₃) and calcia (CaO) crucibles [73, 146, 121, 129]. This offers best chemical homogenization due to inductive stirring, possibilities for composition adjustment via hot sampling and alloy charging, excellent heat control and heat distribution

xiv

Figure 0.7.: Ti46Al8Nb electrode obtained by ATR (above) and slices for sampling (below)

due to direct heating by induced eddy currents, hence a controlled evaporation of volatile elements and most important for scrap recycling a high flexibility regarding feed material (shape, dimension, number of scrap and primary materials). The application of ceramic crucibles leads to a significantly improved energy efficiency compared to melting in water-cooled, so called

"cold crucibles" made from copper. On an experimental scale of 7 kg charge weight for example, power requirements are reduced to 20-25 kW compared to 300 kW [129]. Y₂O₃and CaO offer the best resistance against the strongly reducing titanium melts among the oxide ceramics technically available. With titanium activity aT isignificantly reduced to 0.2–0.3 as compared to pure titanium melts and melting points about 100^◦C lower for the binary γ-TiAl alloys, Ti-Al-X-melts are less aggressive to the ceramic lining than pure titanium, yet a minimum oxygen pickup by attack of the crucible oxides has to be accepted [96] according to eq. (0.3) and can be determined for calcia crucibles based on experimental data from Tsukihashi et al. [146].

[Ca]_Me+ [O]_Me−−−− < CaO > (0.3) with [Ca]_Me- calcium dissolved in liquid melt, [O]_Me- oxygen dissolved in liquid melt

Due to the substantial vapour pressure of Ca in the TiAl-melt continuous evaporation and correspondingly pick-up of oxygen without ever reaching the final chemical equilibrium is observed experimentally. Limiting oxygen pick-up is controlled by evaporation kinetics, which in return are pressure dependent and competing with the reaction rate between the melt and the crucible. In conclusion, lowest oxygen levels should be obtained at a minimum superheat (i.e.

lowest diffusion coefficients, slow evaporation of calcium due to low vapour pressure), laminar stirring (i.e. thick boundary layers), in a well filled, large, crucible with minimal porosity (i.e.

low ratio of contact area and melt volume). It is obvious that modeling of these kinetic effects requires tremendous effort and thus experimental investigation of the matter proves to be necessary.

A first evaluation of the process was carried out using spent sputter targets of the binary composition Ti50Al as a starting material of high purity, low initial oxygen content and avoiding effects of ternary alloying elements. As a second step, scraps from casting trials within the IMPRESS project [70] (alloys Ti-46Al-8Nb and Ti-46Al-8Ta) were recycled.

xv

Figure 0.8.: Scrap from Ti46Al8Ta and Ti46Al8Nb used in trials

Based on representative sampling of the input material and analysis, the optimum charge for the crucible was calculated including adjustment to target composition by adding tita- nium sponge, aluminium granules (99.95 % purity) and tantalum sheet scraps from stamping operations. In particular Al loss during melting was compensated by charging a higher Al concentration from the start. The evaporation rate of Al at our experimental conditions was estimated based on experimental findings of Guo et al. [57] and extrapolated to the actual crucible diameter of 25 cm at an expected holding time of 4500 s. Composition was adjusted by addition of 100 g Al granules for the treated TiAl-Nb alloy and 310 g Al for the TiAl-Ta alloy respectively. Furthermore 91 g of Ta were added to the latter. The material was charged into the crucible at room temperature, the furnace was then closed, evacuated below 1× 10⁻³mbar and flushed with argon twice in order to minimise residual oxygen in the chamber.

Melting trials for consolidation of scrap and casting of electrodes were carried out using a vacuum induction furnace with a melt capacity of 14 l and a nominal melting power of 150 kW. Because of the high reactivity of TiAl melts towards ceramic crucibles, special high purity CaO crucibles were selected. Chemical analysis of the crucible material reveals very little contamination by other oxides (SiO₂0.14 wt %, Al₂O₃0.07 wt %, Fe₂O₃0.05 wt %, MgO 0.48 wt %). The crucible volume allowed a charge weight of approximately 28 kg depending on the shape of the scrap material and the resulting void in the charge (see fig. 0.9).

Melting took place under argon to reduce evaporation of volatile alloying constituents as compared to vacuum melting. To avoid thermal shocks in the crucible, melting power was increased in discrete steps every half hour. After detecting the first molten metal, power was kept constant until all material was liquid. The casting temperature was aimed at 1650–

1700^◦C depending on the liquidus temperature of the target alloy and measured by immersion thermocouples enclosed in a special casing. The homogeneous melt was cast into a water-cooled copper mould in order to obtain an electrode (diameter: 108 mm, length: 750 mm, see fig. 0.9, right) with minimum segregation.

xvi

Figure 0.9.: Heated scrap charge prior to melting (left) and VIM-cast γ-Ti-46Al-8Nb electrode from VIM in dismantled copper mould (right)

Deoxidation by PESR

In order to prevent excessive calcium evaporation and allow for better control of the reducing conditions, all melts in this work were carried out in argon atmosphere at a pressure of 20 bar in a pilot-scale PESR furnace from Leybold-Heraeus (now ALD Vacuum Technologies GmbH).

The equipment is capable of operating at inert-gas over-pressures up to 50 bar, powered with a 5 kA/66 V supply and features latest monitoring and melt-control systems. Melting takes place in a tapered crucible made from copper (diameter: 178 mm-159 mm, length: 880 mm).

Melts were carried out using an industrially available flux with 97.5 % CaF₂, initial CaO- content was determined to be 1.17 wt % by a titration method. During the trials metallic calcium and virgin slag were charged into the melt through screw feeders attached to the sides of the melting furnace. The exact amount of slag and deoxidation agent charged was double-checked after the trial by mass-balancing the material leftover in the slag bunkers.

Start-up of the process was realised by a specially invented procedure using chunks from spent Ti50Al sputter targets for arc ignition in the slag. After the melting process, the ingot was carefully removed from the crucible.

Sampling and chemical analysis

Metal samples were taken with a diamond grinding blade from both ends of the original electrodes and cross sections were obtained from the PESR ingots at three different positions along the ingot axis (fig. 0.10) allowing to differentiate between the composition in the centre of the ingot and the outer radius. Sacrificed ATR electrodes were sliced to discs by water jet cutting. From every sampled disc a 30 x 30 mm sample was removed at half radius. Metallic elements Ti, Al, Nb, Ta, etc. were analysed by XFA using a semi-quantitative method on a PANAnalytical, Model: Axios. Oxygen was determined by hot-gas extraction on a Ströhlein ON-Mat 8500 while residual Calcium was analysed by ICP.

xvii

Figure 0.10.: Samples obtained at three different levels and two radii

Slag samples were taken from the slag skin at the ingot perimeter and at length positions corresponding to the metal samples, for further investigation of slag chemistry and correlation to metal composition. Analysis of these slags was also obtained via XFA after grinding, sieving to grain sizes < 90 μm, and compression of powder pills. As a drawback of this method only the contained metallic elements can be directly detected while differentiation between oxides, fluorides and residual metallic calcium is not possible. This is due to the fact that the detection of fluorine by XFA does not deliver reliable results given the low signal-to-noise-ratio for low atomic order elements like F and O. Therefore CaO content of the slag was determined by a titration method, while the XFA result of all metallic components apart from Ca was recalculated to the corresponding stable oxides. The remaining Ca was recalculated to be CaF₂which is valid with regard to the very low solubility product of CaF₂in a neutral aqueous solution (1.3× 10⁻⁶for CaO compared to 3.9× 10⁻¹¹for CaF₂). The application of a gas-volumetric method to determine remaining metallic Ca via H₂generation could not deliver reproducible results and was therefore rejected.

Results and discussion

Vacuum induction melting and casting

In our VIM melting campaigns, the special CaO crucibles applied lasted about five melts before they had to be replaced. due to cracks in the sidewalls at melt level and forming of oxide crust. Special crucible care was not applied, so that in production scale a significantly increased lifetime can be expected. The results obtained after melting and casting of two electrodes of Ti46Al8Ta and T46Al8Nb are shown in table 0.1. The anticipated composition presented here is based on the average analysis of the original scrap charge and takes into account Al evaporation as well as composition adjustment. It can be observed that alloying of both Al and Ta could be successfully conducted compared to the initial alloy. The results however show significantly higher Al levels than expected and consequently Ta and Nb concentration in both

xviii

alloys appear to be slightly below target.

Table 0.1.: Results of adjusted alloy composition after VIM

Ti46Al8Ta scrap, in at.% Ti Al Ta

composition of scrap 45.85 46.30 7.65

anticipated composition 45.99 46.01 8.00

obtained analysis 44.34 47.80 7.76

Ti46Al8Nb scrap, in at.% Ti Al Nb

composition of scrap 48.18 43.48 8.24

anticipated composition 47.95 43.85 8.20

obtained analysis 46.62 45.13 7.69

Oxygen levels in the Ti46Al8Ta electrode were determined to be 1874 ppm at the top versus 1454 ppm at the bottom side of the casting and 3620 ppm to 2332 ppm respectively in the Ti46Al8Nb electrode. This goes in hand with the observation that Yttrium was detected at the top of each casting (1450 ppm in the Ta-alloy and 2860 ppm in the Nb-alloy, but was not detectable at the bottom and neither in the later obtained PESR ingots after remelting.

Possibly yttria particles from the face coating applied to the mould walls detached upon casting and floated on the not yet solidified TiAl melt while solidification takes place from bottom to top. It can further be discussed whether there is a segregation mechanism for dissolved oxygen in the melt or if CaO particles precipitate upon cooling and float in analogy to the yttria particles.

The Ti46Al8Ta-alloy shows a 350 ppm contamination by Zr, which was detected before in some of the delivered castings and originates from Zirconia components of the shell mould system. Zr contamination could be traced also to the PESR ingots and it appears that there is no removal of Zr by PESR, as can be expected. A contamination by Fe 1430 ppm, Si 895 ppm and Mo 735 ppm could be detected on the Nb-alloy but not in the Ta-alloy and was most likely obtained from residual mould material on the rough surface of the scrap pieces.

Segregation of Ti could be observed in the cast VIM electrodes with slightly decreasing levels from bottom to top of the electrode, while Al concentrations show the exactly opposite trend as expected. Segregation of Nb and Ta are similar with Nb concentration increasing from 7.60 at % at the electrode bottom to 7.78 at % at the top of the electrode while Ta segregates similarly from 7.67 at % at the bottom to 7.84 at % at the top. In order to check for compensation of these segregations by PESR, electrodes are remelted top-down.

Aluminothermic reduction

The analytical results for electrodes obtained by ATR show limited local deviations for the main elements but no pronounced axial segregation effects (see fig. 0.11), which is a matter of horizontal casting as compared to the vertical mould used for solidification of the melt after VIM. The standard deviation of Al with 5.2 rel −% around the average value corresponds to a deviation of 1.38 wt % and is therefore just slightly out of the valid specs which allow Al deviations of 0.5 at % in the final ingot. These values are expected to be met after radial homogenization during PESR.

However the deviation of oxygen concentration has more relevant impact to the process. With

xix

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

● ●●

●

● ●● ●● ●● ●●

●

● ●● ●● ●● ●● ●● ●● ●●

●

● ●● ●● ●● ●● ●● ●●

●

●

● ●

●

● ●

●

● ●

●

● ●

●

● ●

●

● ●

●

● ●

●

●

●

● ●● ●● ●● ●● ●● ●●● ●●●●●●●●

●

●

● ●●●●● ●●●●● ●●● ●●● ●●● ●●● ●●●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

main elements

side elements 20

30 40 50

0 1 2

1 2 3 4 5 6 7 8

sample position

concentration /Wt. %

Element

●

●

●

●

●

●

●

●

●

●

●

●

Al C N Nb O Ti

Figure 0.11.: Analysis of an ATR electrode with an average oxygen concentration of 1.49 wt % and well homogenized Nb-distribution

a mean of 1.47 wt % oxygen appears to be slightly below expectations from thermochemical equilibrium with the final Al₂O₃-CaO slag of ATR (1.59 wt %) but shows a significant gradient along the electrode axis. Furthermore, local deviations are observed that can be contributed to slag inclusions or erosion residuals from the ceramic lining of the mold. In order to still achieve a stable deoxidation, an oxygen concentration of 1.93 wt % (sum of mean and standard deviation) has been used as the input value for the feed calculations controlling the PESR deoxidation.

Pressure electroslag remelting

The PESR control system has been gradually adjusted throughout the trials to compensate for the special process conditions using strongly reducing slags. fig. 0.12 shows the relevant parameters current, voltage, slag resistance, resistance swing and power for a selected melt (see fig. 2.10). Large voltage deviations with a resistance swing of one order of magnitude larger than values obtained in the industrial practice for superalloy remelting (10 % of absolute resistance) are controlled by fine-tuning PID-controllers and filters for tight response (due to short melting times) and at the same time high failure tolerance regarding over-swing. In the result, the main process conditions for a stable deoxidation - a steady electrode feed (= electrode melt rate) - was achieved as fig. 0.12 shows. Based on the logging of power consumption during melting, a specific energy consumption of 3.61 kW h kg⁻¹TiAl could be determined for melts with more than 80 % time at stable melting rate.

xx

Voltage, Resistance, Power/10, Pressure

0 10 20 30

0102030405060

melting time /min

0123456 Melt rate, Swing, ram position / 100 I /kA

U /V

R /mOhm P /kW

p /bar SR /kg*min^−1

Swing /mOhm Vorschub /mm

Figure 0.12.: A stable melt rate, represented by electrode feed rate, could be achieved and is a prerequisite for a stable deoxidation process.

Ti and Al segregation observed in the cast electrodes was smoothed by the backmixing-effect in PESR down to differences in concentration between center and radius section of the ingot to acceptable levels of approx. 0.5 at %. For both Nb and Ta significant differences between center and radius can be observed in the bottom part of the ingot and are probably due to transient melting conditions at the start-up stage of the process. However, concentration of these ternary elements picks up to electrode level during stationary melting with significantly reduced difference between center and radius.

Low oxygen levels < 500 ppm were achieved for VIM-melted electrodes while this goal could not be reached for the high-initial-oxygen electrodes obtained from ATR. Figure 0.13 shows that the analysed O and Ca concentrations are in general agreement to the model of the solubility products we obtained from the experimental study of Tsukihashi et al. [146] and previous results from Lochbichler [80] for deoxidation via VIM in a CaO-crucible, which have been conducted at a super-heat of the melt of 50-100^◦C above the melting point of Ti50Al at 1550^◦C.

Figure 0.13 further illustrates, that for deoxidation by PESR, lower oxygen values can be expected at the same residual calcium concentrations, so that the fit obtained by least squares method (see the bold black line) reverts to a modeled melt temperature of lower than 1550^◦C even though the melting temperature of Ti46Al8Nb and Ti46Al8Ta is 30 K higher than in the binary alloys. This reflects, that due to dissolution in the slag, the capacity of CaO is

xxi

● ●

●

●

●

●

●

● ●

●

●

●

●

●

0 2000 4000 6000 8000

0 1000 2000 3000 4000

Ca /ppm

O /ppm

test series

●

Deoxidation VIM (Lochbichler) PESR − ATR electrodes PESR − VIM electrodes

1550 1600 1650 1700 model fit, T /°C1750

Figure 0.13.: Low oxygen obtained in PESR melted TiAl-alloys vs. modelled solubility product Lcobtained from literature [146]

much lower than in a CaO-crucible (a(CaO)=1) (see also fig. 2.13 in section 2.3.1) which was a constant in the cited works. It also corresponds to the assumption that the highest deoxidation power is achieved at the tip of the electrode at minimum superheat of the liquid phase and not at the phase boundary of the metal pool for which superheat of up to 100 K is assumed in the literature [50].

Side elements and microstructure

The obtained ingots from recycled scraps show an increase in N and C with processing time in PESR, yet the main pick-up is taking place during VIM and can be controlled by technical measures. However for these elements there is no or only limited refining potential in PESR and due to the risk that nitrides or carbides could impact on the mechanical properties, they have to be monitored in future works. Due to the remelting under CaF₂-slags, there is a theoretical possibility for the pick-up of Fluorine during PESR, however it could be shown in a solubility calculation that F should be limited to < 1 ppm in the liquid melt under the expected reducing conditions. In an analysis by Glow Discharge Mass Spectrometry (GDMS) of the Ti50Al remelted here, no significant pick-up of fluorine could be proven. This contrasts with single SEM-EDX measurements of micron-size-inclusions observed in remelted Ti46Al8Nb which qualitatively contain F.

SEM micrographs were prepared from PESR ingots of recycled Ti50Al and Ti46Al8Nb in order to trace potential inclusions or grain-boundary segregations from the named side elements.

In both materials inclusions in the size-range of 1-10 μm were found, mainly in the Al-rich,

xxii

late solidified γ-TiAl matrix and were hence regarded as precipitates. Two types of these could be identified by EDX and are shown in fig. 0.14, one being a Mg-Ca-Al intermetallic of varying composition the other phase a Si-Al-Oxide. As far as Ca is concerned, no measurable concentrations of Ca could be found by EDX in the solidified matrix material, thus Ca seems to precipitate upon solidification in the observed inclusions.

Figure 0.14.: SEM micrographs remelted TiAl show intermetallic Al₁₂Ca₁₀Mg₇(light gray, left) and oxide Al₆Si₃Ti₂(MgCa)₂O₂₂(dark grey, right) precipitates.

In samples of recycled Ti46Al8Nb another type of inclusion was identified by its shiny appearance on the SEM image. EDX reveals >50 at % O and largely deviating concentrations of Ca and Al, however no Ti in these particles. Due to the low atomic order of fluorine and the resulting low signal to noise ratio, the detection of > 10 at % F in all these particles F must be regarded as a qualitative but distinct indication that fluorine in present in the inclusion.

Assuming a ratio of O to F of roughly 5/1 it can be concluded that the particle can not originate from dispersed slag droplets which show about the inverse ratio (CaF₂80 wt %, CaO 20 wt %).

If these 2 μm sized particles have an influence on material properties and why they appear only in the samples of one ingot could not be finally investigated within the scope of this thesis and should be observed more in detail in future works.

Techno economic analysis

For a techno-economic evaluation of TiAl recycling three variants of processing were compared (see also fig. 4.4), with variant three - melting in VIM, deoxidation in PESR and final refining by VAR (triple melt) - showing the best capabilities for treatment of casting scraps and most interesting final ingot quality for reverting the material back to the casting line. From the experimental trials on the demo-scale equipment at IME, a database was set-up including necessary labour, material consumption, consumption of energy and media, metal yield. From this data total cost per kg recycled TiAl on our demo-scale was calculated on a one-shift-per- day basis which results in a recycling capacity of 5 t yr.⁻¹and also taking depreciation of the equipment and university overhead into account. A scale-up calculation was conducted to a small industrial size plant with a recycling capacity of 100 t yr.⁻¹, which should be sufficient to cope with the presently forecasted worldwide production volumes and could be realised with

xxiii

mid-scale melting equipment readily available on the market.

Taking overhead cost and supporting operations into account a profitable operation could be possible when recycled TiAl would sell at 30e kg⁻¹(see fig. 4.8). A sensitivity analysis could show, that the biggest levers on this result are scrap prices, melting yield, personal cost and lifetime of the CaO lining, in VIM in order of decreasing impact. Surprisingly energy consumption is not playing an economically interesting role at present. Nontheless TiAl-recycling also makes sense from an ecological standpoint: A comparison of the specific energy consumed for recycling by VIM-PESR-VAR to the energy consumed for the production of virgin γ-TiAl alloys, starting from the ores shows a 90 % energy saving for a Ti46Al8Nb-alloy and even 95 % for the more ecologically intensive Ti46Al8Ta-alloy.

Conclusions and Outlook

The recycling of γ-TiAl alloys by consolidative melting via VIM in a CaO crucible and subsequent deoxidation by PESR has been demonstrated to prove the high potential of such a process chain.

Alloy composition was successfully adjusted and the major process mechanisms are understood.

A correct sampling of scrap charges in a statistical meaningful manner and a unified standard for chemical analysis of γ-TiAl are to be improved in order to obtain a more precise control of alloy composition and impurity content as necessary for the sensitive γ-TiAl alloys. In parallel to the recycling route also the alternative winning of γ-TiAl alloys by ATR was successfully applied and stable, remeltable electrodes were cast for challenging deoxidation capabilities of PESR.

In can be concluded that PESR offers efficient deoxidation of γ-TiAl even down to 250 ppm if correctly controlled, when starting from VIM-cast electrodes with an initial oxygen content of up to 4000 ppm. This offers an interesting starting point for reuse of the material and might reveal so far unseen material properties. At present the deoxidation of direct-cast electrodes from ATR 15.900 ppm could not be shown, as the challenging requirements of fast process kinetics and more intense requirements on feeding of deoxidation agents could not be surmounted in our experiments. However a kinetic model has been developed and validated, which should enable correct adjustment of the process conditions and successful deoxidation also of this alternative raw material in the future.

Macrosegregation in the obtained ingots is minimal and segregation effects from the VIM cast electrodes could be homogenised by PESR. After the first two processing steps, the material contains residual Ca in a concentration of 500-1000 ppm which is probably too much for direct reuse in an investment casting process due to strong evaporation tendencies. However calculations indicate that Ca should be completely removed by subsequent VAR and trials on the obtained material will be carried out in the near future for experimental proof. A first investigation of microstructure of the PESR ingot material did not show any show-stopping inclusions, however a re-evaluation has to be carried out after VAR refining trials of this material.

A first techno-economic assessment of the proposed processing chain VIM-PESR-VAR shows an interesting economical potential for both recycling of γ-TiAl casting scraps, as well as the alternative winning of γ-TiAl alloys by ATR, with the option to enrich scraps poor in alloying elements by blending in very homogeneous pre-alloys from ATR into the VIM process.

Based on these findings, the following topics should be researched to realise the full potential of PESR deoxidation of TiAl and Ti-alloys in general: Further development and validation of

xxiv

the kinetic model of PESR deoxidation for prediction on the limits of deoxidation rate and better process control. The model could serve as a basis to also understand other mass transfer aspects in ESR. A more precise kinetic investigation of the evaporation of Ca from the remelting slag and an upgrade of the PESR research equipment by an electrode weight control and more precise feeding mechanism would enable better control of the final oxygen level in the ingots.

Further investigation of the solubility product of Ca and O in TiAl by more precise analytics is a key to experimentally determine CaO-activity in high-CaF₂slags, which is not only interesting for deoxidation but also for e.g. desulphurization a critical detail in the remelting of superalloys.

Further tracking fluorine in PESR remelted TiAl is essential to overcome the ongoing discussion about its detrimental effects in electroslag remelted titanium alloys. A close investigation if F could even be alloyed precisely during PESR by adjusting slag composition with less stable fluorides (e.g. MgF₂) could serve for a cheap and controllable application of the halogen effect as an effective possibility for high-temperature corrosion protection of γ-TiAl. Finally the possibilities of integrating ATR pre-alloys into γ-TiAl recycling should be explored more in detail to lever an interesting economical and ecological potential by cost-effective alloying of supply-critical refractory metals like niobium or tungsten.

xxv

Inhaltsverzeichnis

1. Einleitung 1

1.1. Motivation . . . 1

1.2. Wissenschaftliche Einordnung der Arbeit und Verfahrenskonzept . . . 2

2. Grundlagen und eigene theoretische Vorarbeiten 5 2.1. Die Werkstoffklasse der Titanaluminide . . . 5

2.1.1. Legierungsentwicklung und Mikrostruktur der TiAl-Legierungen . . . . 5

2.1.2. Wirkung von Sauerstoff und anderen nichtmetallischen Elementen in Ti und TiAl . . . 9

2.1.3. Eigenschaften und Anwendungen der Titanaluminide . . . 11

2.2. Prozesstechnik zur Herstellung von TiAl-Legierungen . . . 14

2.2.1. Etablierte Prozessroute über Vakuumlichtbogenschmelzen (VAR) . . . . 15

2.2.2. Verfahren auf der Basis von Vakuuminduktionsschmelzen (VIM) . . . . 16

2.2.3. Arc Skull Melting (ASM) für den Schleuderguss . . . 19

2.2.4. Plasma-Lichtbogen-Schmelzen (PAM) . . . 19

2.2.5. Alternative Gewinnungsprozesse in der Entwicklung . . . 21

2.2.6. Elektroschlackeumschmelzen von TiAl . . . 23

2.3. Thermochemie der Desoxidation über DESU . . . 29

2.3.1. Desoxidation von Titanlegierungen . . . 29

2.3.2. Gleichgewichtsbetrachtung der Desoxidation beim DESU . . . 30

2.3.3. Gleichgewichtsbetrachtung für Aluminium in der Randschlacke . . . 35

2.3.4. Gleichgewichtsbetrachtung für die Löslichkeit von Fluor im Metall . . . 36

2.4. Kinetische Modellierung der Desoxidationsreaktion . . . 38

2.4.1. Ausgangslage . . . 38

2.4.2. Kinetisches Modell . . . 40

2.4.3. Ansatz zur quantitativen Beschreibung der Reaktionskinetik . . . 41

2.4.4. Ermittlung notwendiger Stoffdaten . . . 47

2.4.5. Numerische Lösung und Umsetzung in ein Berechnungs-Tool . . . 49

2.4.6. Kinetik der Calcium-Verdampfung . . . 52

2.5. Zusammenfassung der theoretischen Vorarbeiten . . . 54

3. Experimentelle Untersuchungen 57 3.1. Methodik und Versuchsziele, Planung und Analytik . . . 57

3.1.1. Ausgangssituation, Methodik und Ziele . . . 57

3.1.2. Prozessmodell zur Desoxidation . . . 58

3.1.3. Optimales Verhältnis von CaF₂zu Ca bei der Chargierung . . . 58

3.1.4. Analytik und Probennahme . . . 61

3.2. Anlagentechnik und Versuchsbedingungen . . . 67

3.2.1. Reaktor für aluminothermische Reduktion und in-line Abguss . . . . 67

xxvii

3.2.2. Aluminothermische Reduktion . . . 67

3.2.3. Herstellung von Elektroden über VIM . . . 70

3.2.4. Raffinierendes Druck-Elektroschlackeumschmelzen . . . 75

3.3. Erzielte Ergebnisse . . . 83

3.3.1. Charakterisierung der erhaltenen Elektroden . . . 83

3.3.2. Prozessergebnis DESU . . . 88

3.3.3. Desoxidation . . . 92

3.3.4. Homogenität von Ti, Al, Nb und Ta im umgeschmolzenen TiAl . . . 97

3.3.5. Die Nebenelemente C, N, S und F . . . 100

3.3.6. Zusammensetzung von Flugstaub und Schlacken . . . 106

3.3.7. Auswertung der Prozessgrößen . . . 110

3.3.8. Mikrostruktur des umgeschmolzenen TiAl . . . 117

4. Diskussion der Ergebnisse 125 4.1. Fehlerbetrachtung . . . 125

4.1.1. Genauigkeit der Ca- und O-Analytik . . . 125

4.1.2. Prozesskontrolle . . . 127

4.1.3. Vergleich von experimentellen Ergebnissen und kinetischem Modell . . . 130

4.2. Technisch-Wirtschaftliche Bewertung von TiAl-Recycling mittels DESU . . . . 131

4.2.1. Varianten der Prozessroute zum TiAl-Recycling und Skalierung . . . 131

4.2.2. Spezifische Recyclingkosten und Kostenstruktur . . . 132

4.2.3. Wirtschaftlichkeitsbetrachtung und Sensitvitätsanalyse . . . 134

4.2.4. Energieverbrauch und Effizienz des TiAl-Recyclings . . . 136

4.3. Abschließende Bewertung des Verfahrens . . . 138

4.3.1. Technische Reife des Verfahrens . . . 138

4.3.2. Erweiterung durch Verknüpfung von ATR- und Recycling-Route . . . . 139

4.3.3. Bewertung technischer Alternativen zur DESU-Desoxidation . . . 140

4.3.4. Bewertung der bisher erzielten Werkstoffqualität . . . 142

5. Zusammenfassung und Ausblick 147 5.1. Zusammenfassung der erzielten Ergebnisse . . . 147

5.2. Ausblick auf weiterführende Untersuchungen . . . 150

A. Prozessdatenschriebe der durchgeführten Schmelzen 153 B. Zeitliche Entwicklung des Prozessdrucks 161 C. Konstanten, Parameter und Annahmen für das kinetische Prozessmodell 165 C.1. Ergebnisse aus der Modellierung temperaturabhängiger Parameter . . . 165

C.2. Ansatz zur Modellierung des fallenden Tropfens . . . 165

C.3. Beschreibung des Ca-Transports in Argon durch Diffusion und Konvektion . . . 168

D. Technischer Anhang 171 E. Weitergehende Ergebnisdokumentation 175 E.1. Abschätzung der Pooltiefe . . . 175

E.2. Analytik . . . 176

E.2.1. Metallanalytik einer über ATR gewonnen TiAl-Elektrode . . . 177

xxviii

E.2.2. Flugstaubanalysen . . . 178 E.2.3. Schlackenanalysen . . . 178 E.2.4. Auswertung REM-EDX-Aufnahmen . . . 181

F. Parameter Wirtschaftlichkeitsberechnung 183

xxix

Abbildungsverzeichnis

0.1. Integrated concept for alternative production and recycling of TiAl . . . viii

0.2. Principle and refining potential of PESR . . . ix

0.3. Ca and CaO in active slags to reach target oxygen level . . . x

0.4. Result of the process model for PESR deoxidation . . . xi

0.5. Deoxidation reaction across metal slag boundary . . . xii

0.6. Results of the modelling of kinetic effects on deoxidation . . . xiii

0.7. Sliced ATR electrode for analysis . . . xv

0.8. Scrap from Ti46Al8Ta and Ti46Al8Nb used in trials . . . xvi

0.9. Scrap charge prior to melting and cast electrode . . . xvii

0.10. Samples obtained at three different levels and two radii . . . xviii

0.11. Analysis of sliced ATR electrode . . . xx

0.12. Stable melt rate for stable deoxidation . . . xxi

0.13. Low oxygen obtained in PESR melted TiAl-alloys vs. model . . . xxii

0.14. SEM micrographs of inclusions in remelted TiAl . . . xxiii

1.1. Verfahrenskonzept zur kostengünstigen Gewinnung von Titan . . . 3

2.1. Phasendiagramm Ti-Al . . . 6

2.2. Einfluss der Legierungselemente auf die α-Transustemperatur . . . . 7

2.3. O in Ti: Sauerstofflöslichkeit und α-Stabilisierung . . . . 10

2.4. Schematische Darstellung einer VAR Schmelzanlage . . . 15

2.5. Induktives Einschmelzen von TiAl . . . 17

2.6. Aufbau von Plasma-Brenner und Plasma-Schmelzanlage . . . 20

2.7. IME-Verfahren zur ATR von γ-TiAl . . . . 22

2.8. Prinzip und Raffinationspotentiale beim ESU . . . 24

2.9. Prinzipskizze einer Elektroschlackeumschmelzanlage . . . 25

2.10. Saarschmiede ESU und erstarrende Schlacke . . . 26

2.11. Typischer ESU-Schmelzschrieb . . . 30

2.12. Ellingham-Darstellung möglicher Desoxidationsmittel . . . 31

2.13. Ca und CaO Aktivitäten im DESU Schlackensystem . . . 33

2.14. Ca-Konzentration in der aktiven Schlacke in Abhängigkeit von CaO . . . 34

2.15. Modellierung der Fluor-Konzentration in γ-TiAl und Titanschmelzen . . . . 39

2.16. Reaktionsmechanismen Desoxidation DESU . . . 40

2.17. Reaktionszonen beim ESU . . . 42

2.18. Modellierte Desoxidationsrate und Verlauf der Sauerstoffkonzentration . . . 51

2.19. Ca Dampfdruck und Verdampfungsrate . . . 53

2.20. Spezifische Verdampfungsgeschwindigkeit . . . 54

3.1. Funktionsschema des Prozessmodells zur DESU-Desoxidation . . . 59

3.2. Ergebniss des Prozessmodells für die DESU Desoxidation . . . 60

3.3. Bestimmung des optimalen Verdünnungsverhältnisses CaF₂/Ca . . . 62

xxxi