Investigation into the selectivity of CVD iron from Fe(CO)5 precursor on various metal and dielectric patterned substrates

Investigation into the selectivity of CVD iron from Fe(CO)5 precursor on various metal and dielectric patterned substrates

Surface & Coatings Technology 201 (2007) 8998 – 9002 www.elsevier.com/locate/surfcoat Investigation into the selectivity of CVD iron from Fe(CO)5 pre...

771KB Sizes 0 Downloads 0 Views

Surface & Coatings Technology 201 (2007) 8998 – 9002 www.elsevier.com/locate/surfcoat

Investigation into the selectivity of CVD iron from Fe(CO)5 precursor on various metal and dielectric patterned substrates M.F. Bain ⁎, Y.H. Low, D.C.S. Bien, J.H. Montgomery, B.M. Armstrong, H.S. Gamble School of Electronics, Electrical Engineering and Computer Science, Queen's University Belfast, Belfast BT9 5AH, UK Available online 4 May 2007

Abstract It has been shown that CVD iron from a Fe(CO)5 precursor deposits selectively on dielectric surfaces over tungsten surfaces. No similar selective CVD mechanism for titanium and aluminium to SiO2 surfaces was observed. It was established that the selectivity between the tungsten surface and the SiO2 surface could be enhanced through the oxidation of the tungsten surface. Depositions carried out on oxidised tungsten (WOX) and SiO2 substrates showed that iron layers up to 0.5 μm thick with a resistivity of 18 μΩcm can be deposited with excellent selectivity. The selective mechanism is attributed to the electrochemical properties of the tungsten or WOX layer, which prevents the reduction of the iron precursor. Selectivity loss was attributed to defects or impurities adsorbed to the tungsten surface. © 2007 Elsevier B.V. All rights reserved. Keywords: CVD Iron; Selective deposition

1. Introduction The deposition of iron by CVD from the metalorganic iron pentacarbonyl, Fe(CO)5 precursor has been widely reported. Deposition techniques ranging form direct thermal decomposition [1,2] to assisted deposition techniques using laser [3] or ion [4] have been described. These techniques have typically been used in the fabrication of magnetic nanostructures. Work carried out previously has shown that deposition of iron occurs preferentially on a silicon dioxide (SiO2) or silicon nitride (Si3N4) surface to tungsten surfaces [5]. Depositions carried out on patterned W/SiO2 substrates have shown that layers up to a thickness of 180 nm with a resistivity value of 19 μΩcm can be deposited selectively with only some iron island nucleation on the tungsten surface. A number of possible applications for this novel technique have been investigated. A promising technology has been developed which allows the fabrication of self-aligned sub 100 nm iron nanowires using silicon nitride spacers and a dual selective deposition process [6]. The aim of the work presented is to investigate whether other metals exhibit the same selectivity mechanism when exposed to ⁎ Corresponding author. Tel.: +44 28 90975462; fax: +44 28 90667023. E-mail address: [email protected] (M.F. Bain). 0257-8972/$ - see front matter © 2007 Elsevier B.V. All rights reserved. doi:10.1016/j.surfcoat.2007.04.068

the iron precursor and to attempt to understand the selectivity mechanism such that it may be enhanced. If the selective process can be predicted and enhanced it may lead to novel magnetic devices. Selective deposition is defined as the process whereby deposition is preferentially carried out on one surface to the exclusion of the deposition on another. There are a number of publications which describe the preferential growth of iron on a deposited iron seed layer using Fe(CO)5 precursor [7,8]. No publications, to the knowledge of the authors, have reported selective deposition on a dielectric surface over another metal surface. 2. Experimental The CVD system consisted of a load-locked, low pressure, single wafer process chamber. The iron precursor, iron pentacarbonyl, Fe(CO)5, is stored in a stainless steel bubbler and maintained at 0 °C. The precursor is supplied to the processing chamber by direct draw though a restrictor. During deposition, the precursor delivery line is heated to 50 °C, and the process chamber walls are heated throughout to avoid precursor condensation. Hydrogen was used as the carrier gas as previous results indicate it aids the removal of the reaction by-products from the wafer surface, thus reducing deposited layer resistivity [9,10].

M.F. Bain et al. / Surface & Coatings Technology 201 (2007) 8998–9002

All iron depositions carried out in this work were thermal, non-assisted processes. The deposition conditions were maintained constant for all the depositions described. The temperature was set at 400 °C as previous results indicated that the temperature had no effect on the selectivity mechanism [5]. The deposition temperature however produces a large effect on the resistivity of the iron layer, at low temperatures, T b 250 °C the iron layer is non continuous, as the temperature increases the layer coalesces [2]. As the deposition temperature increases further the resistivity of the layer decreases, this is attributed to grain growth and the reduction of impurities, primarily carbon and oxygen, incorporated in to the layer, this was confirmed by auger analysis [11]. All depositions were carried out on 100 mm b100N oxidised silicon substrates (tox = 500 nm). A layer of tungsten, aluminium or titanium was then deposited onto the oxidised silicon substrate by physical vapour deposition (PVD). The metal layer was then patterned using a “chequer-board” mask with standard photolithographic techniques. The exposed metal was etched back to the oxide under layer by a wet etch of KOH: KH2PO4: K3Fe(CN)6: H2O, 0.67 g: 1.7 g: 1.65 g: 50 ml for tungsten, 50:1 HF for titanium and H3PO4: H2O: CH3COOH, NHO3, 16:2:1:1 for aluminium. Resulting in a substrate as shown in Fig. 1 below. The sheet resistance of the deposited iron layer was determined using four-point probe measurements. The thickness and resistivity of the layer was verified by patterning the iron layer

Fig. 1. (a) Patterned ‘Chequer-board’ PVC metal layer deposited on an oxidised wafer (b) X section SEM of selective CVD iron on patterned tungsten.

8999

by etching using 12% nitric acid, H2O solution. The step produced was measured with a Tencor Alpha Step-100 profilometer. Cross sectional and surface analysis of the layers was performed in a Leo Supra-25 SEM system. Fig. 1(b) shows a typical cross section of a selective deposition obtained on a patterned tungsten silicon dioxide (SiO2) substrate. The thickness of the W, Al, and Ti layer was measured before and after iron deposition to ensure no interaction with the iron precursor, none was observed. 3. Results and discussion 3.1. Effect of metal layer on selectivity Depositions were carried out on the patterned tungsten, titanium and aluminium layers on SiO2 for 5, 10, 15, 20 and 30 min. After patterning the thickness of the deposited CVD iron layer was measured on both the oxide region (tox) and the metal region (tm). The selectivity of the deposition is summarised as the tm/tox ratio. That is to say a value between tm/tox = 0 representing no iron deposition on the metal region and tm/tox = 1 where the deposition is entirely uniform across both surfaces. Fig. 2(a) shows the effect of deposition time on the thickness of the iron deposited on the oxide region for patterned tungsten layers and aluminium layers (titanium results were identical to those for aluminium). From Fig. 2(a) a deposition rate of 7.5 nm/min was obtained on the oxide region of the tungsten patterned substrate and a deposition rate of 4.5 nm/min for the aluminium patterned substrate. Fig. 2(b) summarises the selectivity of the deposition using the tm/tox ratio for tungsten, titanium and aluminium patterned substrates. The deposition rate calculated from Fig. 2 (a) shows a deposition rate on the patterned tungsten layer that is almost twice that on the patterned aluminium. This would suggest that little or no precursor was consumed on the tungsten-coated regions. It also implies that the rate-limiting step of the deposition reaction at the oxide surface is mass transport onto the substrate surface. The fitted line to these thickness points does not appear to go through the origin point suggesting there is an initial ‘fast’ deposition of 30–40 nm. This can be attributed to either a gas surge at the start of the deposition process or a two-stage deposition process. The initial deposition appears the same on all three metals this would suggest this ‘fast’ deposition rate is not controlled by a surface reaction but due to an uncontrolled gas surge. It is clear from Fig. 2(b) that depositions carried out on the aluminium and titanium patterned SiO2 substrates showed no selectivity. Depositions performed on the patterned tungsten layer showed excellent selectivity for the time shown. On closer inspection iron nucleation and small isolated island growth on the tungsten surface was observed after 15 min of deposition as can be seen in Fig. 1(b). The iron layer, however, was not continuous and no thickness value could be obtained. The iron depositions were repeated on tungsten layers deposited on SiO2 by CVD, patterned, as before to ensure the selectivity mechanism was due entirely to the properties of the tungsten and not the deposition method. The results obtained were, within experimental error, identical.

9000

M.F. Bain et al. / Surface & Coatings Technology 201 (2007) 8998–9002

If this is true material such as platinum, gold and selenium, which have electronegativity values of 2.28, 2.54 and 2.5 respectively [9] will also exhibit a selective process. Unfortunately these materials were unavailable or incompatible and could not be tested. The only material available to test this hypothesis was silicon, which has an electronegativity value of 1.9. Prior to iron deposition the silicon/SiO2 substrate was dipped in 10:1 HF for 10 s ensuring a hydrogen terminated silicon surface, H has an electronegativity value of 2.2. The hydrogen terminated silicon showed good selectivity with iron layers up to 100 nm deposited only on the oxide region. Results were difficult to replicate, it is proposed that this is due to the desorption of the hydrogen from the silicon surface under vacuum at a deposition temperature of 400 °C and possible regrowth of the native oxide. A further possible variable in the selectivity mechanism is the presence of a thin native oxide at the tungsten surface. 3.2. Rapid thermal oxidation of tungsten To determine what affect any native tungsten oxide layer has on the selectivity of CVD iron, PVD tungsten layers 300 nm thick were deposited and patterned as before. The tungsten was then subjected to oxidation by rapid thermal annealing at 700 °C. A two-stage process was developed to form the oxide; the tungsten was initially annealed in a nitrogen environment at 750 °C

Fig. 2. (a) Effect of deposition time on iron thickness on oxide for patterned W and Al substrates (b) Selectivity of CVD iron process on patterned Al, W, Ti substrates.

There are a number of possible explanations as to why selective deposition occurs. It is possible that the adsorption of the iron precursor, Fe(CO)5 onto metal surfaces has a low sticking coefficient hence preferential deposition occurs on the oxide region. In this case one would expect an increased nucleation time on the tungsten surface but a continuous layer would eventually form and not the large island growth that is observed. Typically selective deposition by CVD occurs on a metallic surface, this is attributed to the electron exchange between the metallic surface and the gas precursor. The deposition then continues due to preferential conditions at the selective surface. The strength of this initial interaction is determined by the electrochemical properties of the metal seed layer and the chemistry of the precursor. A measure of the ability of an atom or molecule to attract electrons is described by the electronegativity of the layer. The electronegativity values for tungsten, titanium and aluminium using the Pauling scale are 2.36, 1.61 and 1.54 respectively [12]. It is possible that the high electronegative value of the tungsten prevents the decomposition of the iron precursor. It is proposed the selective deposition process is controlled by two mechanisms. The properties of the tungsten layer prevents the reduction of the iron precursor, whilst at the oxide surface dissociation of the reactive gas occurs at the growth surface enabling the reaction to proceed at a greater rate.

Fig. 3. (a) Cross sectional SEM of tungsten oxide formed by Rapid Thermal Oxidation (b) X-ray diffraction analysis of oxidised tungsten layer.

M.F. Bain et al. / Surface & Coatings Technology 201 (2007) 8998–9002

for 10 s, the temperature was then reduced to 700 °C and oxygen was introduced into the chamber for 30 s. The high temperature anneal prior to oxidation was performed to set grain growth before oxygen was introduced. Oxidation timing was crucial; if all the tungsten was consumed during the reaction catastrophic delamination of the layer from the SiO2 occurred, a cross section SEM of the as formed tungsten oxide is presented in Fig. 3(a). The SEM of the tungsten layer after oxidation shows the layer thickness is ∼0.65 um and a thin layer of tungsten (0.1 um) remained below the WOX layer to ensure adhesion. X-ray diffraction (XRD) analysis as shown in, Fig. 3(b) indicated the oxidised layer to be made up of WO3 and WOX, a thin W layer was also detected. XRD analysis was performed on an ix Panalytical X'Pert Pro, CuKα, λ = 1.54 Å, step size 2 theta = 0.0167°. 3.3. Selective CVD iron on oxidised tungsten CVD iron depositions were carried out on patterned WOX/SiO2 and W/SiO2 substrates, again the deposition conditions were maintained constant as before. No continuous iron layer was observed on the tungsten or the tungsten oxide region after 30 min. Some nucleation and island growth had occurred on both. The deposition time was increased to 60 min, an almost continuous layer was observed on the tungsten regions Fig. 4(a) and (c), however on the oxidised tungsten surface only small island growth was observed, Fig. 4(b) and (d). The cross section of the oxidised

9001

tungsten layer, Fig. 4(d) shows a curling at the edge of the WOX layer. This is attributed to the oxidation of the exposed sidewall as the tungsten layer was patterned prior to oxidation. The over oxidation causes delamination from the SiO2 substrate; this feature has no effect on the selectivity. From Fig. 4(c) and (d) the deposition rate on the oxide has altered, with an iron layer ∼ 300 nm deposited on the substrate patterned with tungsten and a 460 nm layer deposited on the substrate patterned with WOX. This is attributed to the loss of selectivity on the tungsten surface. A highly selective CVD iron layer of thickness ∼ 0.5 μm was successfully deposited on a silicon dioxide substrate patterned with tungsten oxide. The resistivity of the deposited iron layer was found to be 18 μΩcm. From Fig. 4 it is clear that the selectivity mechanism is enhanced through the oxidation of the tungsten. It is proposed the oxidation of the tungsten surface enhances the mechanism, which prevents iron nucleation, unfortunately data on the electrochemical properties of WOX could not be found. Selectivity loss can sometimes be initiated through the readsorption of reaction by-products from the reactive surface (SiO2) to the non-reactive surface (W/WOX), this is typically characterised by selectivity loss initiating in the area next to the reactive region. Iron nucleation was observed equally across the entire tungsten surface. It is proposed that selectivity loss is initiated at defects or adsorbed impurities on the tungsten/tungsten oxide surface that allow the initial nucleation of the iron layer. It is

Fig. 4. Micrographs of the surface and cross sections of tungsten and tungsten oxide layer after a 60 min iron deposition at 400 °C (a) surface scan of tungsten, (b) surface scan of WOX, (c) cross section of tungsten, (d) cross section of WOX.

9002

M.F. Bain et al. / Surface & Coatings Technology 201 (2007) 8998–9002

possible that the oxidation process removes the adsorbed impurities from the tungsten surface enhancing selectivity. 4. Conclusion Selective deposition of CVD iron has been demonstrated on SiO2 surface over tungsten surfaces no selectivity was observed on Al and Ti surfaces. The selectivity can be enhance through the oxidation of the tungsten layer (WOX) with iron layers up to 0.5 μm thick with a resistivity of 18 μΩcm deposited with excellent selectivity. The selective mechanism is attributed to the electrochemical properties of the tungsten or WOX layer, which prevents the reduction of the iron precursor. At the oxide surface the rate-limiting step of the deposition is the mass transport to the substrates surface. Selectivity loss was attributed to defects or impurities on the tungsten surface. Acknowledgments The assistance of John Meneely at QUB School of Geography with XRD analysis was greatly appreciated.

References [1] J.V. Armstrong, M. Enrech, C. Decrouez, J.G. Lunney, J.M.D. Coey, IEEE Trans. Magn. 26 (1990) 1629. [2] Y.H. Low, D.C.S. Bien, J.H. Montgomery, H.S. Gamble, 10th Joint Vacuum Conf. Slovenia, 2004, p. 30. [3] R. Cirease, R. Alexandrescu, I. Voicu, I. Morjan, G. Punga, Surf. Coat. Technol. 80 (1996) 229. [4] M. Takeguchi, M. Shimojo, K. Furuya, Nanothechnology 16 (2005) 1321. [5] Y.H. Low, M.F. Bain, D.C.S. Bien, J.H. Montgomery, B.M. Armstrong, H.S. Gamble, Microelectron. Eng. 83 (no.11-12) (2006) 2229 Nov/Dec. [6] Y.H. Low, M.F. Bain, D.C.S. Bien, S.J.N. Mitchell, H.S. Gamble, Electrochem. Solid-State Lett. 9 (no.12) (2006) G340 Dec. [7] P.M. George, J.L. Beauchamp, Thin Solid Films 67 (1980) L25. [8] D.P. Adams, L.L. Tedder, T.M. Mayer, B.S. Swartzentruber and E. Chason, Phys. Rev. Lett. Vol.74, No.25, pg.5088–5092. [9] R. Kaplan, N. Bottka, Appl. Phys. Lett. 41 (10) (1982) 972 15 November. [10] M.J. Hampden-Smith, T.T. Kodas, Chem. Vap. Depos. 1 (No.1) (1995) 8. [11] Y.H. Low, Thesis ‘Chemical Vapour Deposition of Iron and Cobalt layers’ December 2006. [12] CHEMIX School Interactive Periodic table of Elements - Electronegativity Chart.