硅片金属杂质分析

AuthorJunichi TakahashiKouichi YounoAgilent Technologies9-1 Takakura-Cho, Hachioji-ShiTokyo, 192-0033JapanAbstractA newly designed, high-sensitivity reaction cell induc-tively coupled plasma mass spectrometer (ICP-MS) wasused to determine trace metals in the presence of a highconcentration of silicon (2000 ppm).A number of silicon-based interferences prevent straightforward analysis oftitanium, nickel, copper, and zinc by conventional ICP-MSso the test also served to evaluate the magnitude of anyphysical or spectral interferences arising from the matrix.The 7500cs ICP-MS incorporates an Octopole ReactionSystem for interference removal, allowing the analysis ofthese elements directly at the analyte mass. Detectionlevels for all SemiconductorEquipment and MaterialsInternational (SEMI)required elements determined usingthis technique are in the range 0.20 – 40 ppt. Characterization of Trace Impurities in Silicon Wafers by High Sensitivity ReactionCell ICP-MSApplication IntroductionSemiconductor manufacturers are constantly striv-ing to improve devices and lower manufacturingcosts. Device improvements demand increasedminiaturization, faster operational speeds, andgreater integration. Lowering manufacturing costsrequires higher yields and decreased devicedefects. As device features are minimized to meetthese goals, the need to characterize trace metalcontamination in silicon wafers at lower concen-trations becomes more important. Various surfacecontamination concentration sampling techniquesare used to determine the purity of both the nativeand thermal oxide layer on silicon wafers, includ-ing surface metal extraction (SME) and liquid dropdecomposition (LDD). The silicon concentration ofa sample solution obtained from the naturally oxi-dized layer is less than 10 ppm, while it is up to2000 ppm for the thermally oxidized layer. Samplesolutions for bulk silicon wafer analysis containsilicon up to the 2% level. Applications relating tosilicon wafer analysis are particularly challengingdue to the small sample volumes, for example a200– 250 μ L sample volume is typical of the SMEtechnique; the silicon-rich matrix and aggressivereagents used to prepare silicon wafer samples forSemiconductoranalysis; and the ultratrace levels at which metalcontaminants need to be measured. The instru-ment used in this application was an Agilent7500cs ICP-MS equipped with a low flow nebulizer.The use of a MicroFlow nebulizer (MFN-100),Peltier cooled spray chamber, wide diameter torchinjector, and highly efficient 27 MHz plasma gener-ator operating at 1600 W (normal power) ensuredcomplete ionization of the sample matrix, therebyeliminating concerns regarding analytical stability. High Silicon Matrix - Interference RemovalPolyatomic overlaps on some semiconductor-critical elements present specific analytical prob-lems for conventional quadrupole ICP-MS for manysemiconductor applications. Plasma based interferences Ar, ArH, and ArO limit the determi-nation of Ca, K, and Fe respectively, while silicon-based interferences (shown in Table 1) interferewith important elements such as Ti, Ni, Cu, andZn. With the 7500cs reaction cell ICP-MS, analystscan attenuate interferences using a controlledenvironment within the Octopole Reaction System(ORS) cell, which is pressurized with the simplecell gases hydrogen (reaction mode) and helium (collision mode). 2Table 1. Silicon-Based Polyatomic Interferences on Ti, Ni, Cu,and ZnPolyatomic ions Mass Analyte30SiO, 29SiOH 46 Ti28SiF, 30SiOH 47 Ti29SiF, 28SiFH 48 Ti30SiF, 29SiFH 49 Ti28SiO2 60 Ni28SiOF, 30SiO2H 63 Cu29SiOF, 28SiOFH 64 Zn30SiOF, 29SiOFH 65 Cu28SiF2 , 30SiOFH 66 Zn30SiF2, 29SiF2H 68 ZnTable 2. ICP-MS Operating ConditionsParameter ICP-MS ConditionsRF Power 1600 WSampling depth 8 mmCarrier gas flow 0.45 mL/minMakeup gas flow 0.68 mL/minExtraction lens 1 – 152 VExtraction lens 2 – 22 VMethodologySample PreparationPieces of silicon wafers were soaked in 1:3 HF solu-tion for 10 minutes to remove any surface deposits.The wafer surfaces were then rinsed and dried in astream of argon gas. Bulk silicon pieces (2 g) weredigested in high-purity grade HF (25 g) (Tama PureChemicals, Japan) and HNO 3 (15 g) in a sealedvessel, which was heated to 60 ° C on a hot plate.Vigorous shaking was necessary to dissolve any by-products, for example, ammonium fluorosilicate.On cooling, ultrapure water was added to make upa 2% silicon solution, and these were furtherdiluted by a factor of 10 using 3.8% HF and 6.8%HNO3 to produce a final concentration of 0.2%(2000 ppm).ICP-MS AnalysisAn Agilent 7500cs ICP-MS equipped with an ORS,ShieldTorch System (STS) and an MFN-100 wasused for this analysis. The sample was aspirated atan uptake rate of 68 μ L/min. A platinum 2.0-mminjector torch, perfluoroalkoxyalkane (PFA) spraychamber and PFA endcap were also used. The7500cs is fitted with a platinum interface as standard. Instrument operating conditions aregiven in Table 2. 3Method of QuantificationCalibrations were performed using matrix-matched 2000 ppm silicon standards. One of thedigested Si samples was spiked to final concentra-tions of 0, 20, 60, and 100 ppt with multi-elementstandards to create an external calibration. Allother Si sample concentrations were determinedagainst this external calibration and were cor-rected to account for the dilution factors. No inter-nal standards were added to avoid the risk ofcontamination. The external calibration avoidedthe need for time consuming standard additions,which require the need to spike every sample.The effectiveness of the external, matrix-matchedcalibration was gauged by spiking samples with amulti-element standard at a concentration of 50 ppt (ng/L) and calculating recoveries. Figure 1illustrates a representative calibration curve, inthis case Cu in the matrix matched solution.Figure 1. Copper calibration curve in 2000 ppm silicon solution.4Table 3. DL (3 sigma), BEC and Spike Recoveries of SEMI Specified Elements in 2000 ppm Silicon SampleElement Plasma H2 Gas flow He Gas flow DL 3 sigma, Spike recovery (%)(mass) power (W) (mL/min) (mL/min) n = 10 (ppt) BEC (ppt) 50 ppt spikeLi (7) 1600 - - 0.64 1.2 81Be (9) 1600 - - 1.1 1.4 82B (11) 1600 - - 6.6 22 79Na (23) 1600 5.0 - 16 18 95Mg (24) 1600 - - 0.68 2.9 83Al (27) 1600 5.0 - 6.9 7.7 113K (39) 1600 5.0 - 12 36 79Ca (40) 1600 5.0 - 9.8 45 91Ti (48) 1600 - 5.0 40 76 94V (51) 1600 - 5.0 0.72 0.17 93Cr (52) 1600 5.0 - 7.8 7.5 97Mn (55) 1600 5.0 - 1.3 2.2 91Fe (56) 1600 5.0 - 2.7 31 87Co (59) 1600 - 5.0 1.6 2.6 98Ni (60) 1600 - 5.0 6.8 6.9 94Cu (63) 1600 - 5.0 2.6 15 95Zn (64) 1600 - 5.0 11 14 94Ga (71) 1600 - 5.0 15 51 95Ge (72) 1600 - 5.0 8.1 3.3 90As (75) 1600 - 5.0 18 19 94Sr (88) 1600 - - 0.91 1.6 82Zr (90) 1600 - - 1.5 2.3 84Nb (93) 1600 - - 0.31 0.30 83Mo (95) 1600 - - 2.1 2.1 85Ag (107) 1600 - - 2.7 4.5 80Cd (111) 1600 - - 1.4 1.5 88Sn (118) 1600 - - 0.84 2.8 85Sb (121) 1600 - - 0.29 0.80 84Ba (137) 1600 - - 1.3 0.90 85Ta (181) 1600 - - 0.18 0.17 85Au (197) 1600 - - 2.5 2.0 86Tl (205) 1600 - - 0.40 0.23 82Pb (208) 1600 - - 0.81 1.2 85Bi (209) 1600 - - 0.81 1.3 84Th (232) 1600 - - 0.20 0.19 80U (238) 1600 - - 0.28 0.18 81Results and DiscussionDetection limits (DL) and background equivalentconcentrations (BEC) for a full suite of semicon-ductor elements in the 2000 ppm silicon sampleare summarized in Table 3. As the data in the tabledemonstrates, all elements in the high siliconmatrix return DL and BEC at ppt levels, even forthe more difficult elements K, Ca, Ti, Fe, Ni, Cu,and Zn. These results highlight the effectiveness ofthe ORS cell for removing plasma and matrix-based polyatomic interferences in the high siliconmatrix. 5The 50-ppt spike recoveries for all elements are inthe SEMI specified range of 75%-125%. The quanti-tative recovery also indicates the absence of anynebulization or transport interferences. Note: allrecoveries were determined without the use of aninternal standard, therefore simplifying samplepreparation and eliminating a potential source ofcontamination.A short-term stability study was performed byadding a 100-ppt standard into a 1000-ppm siliconsolution and analyzing the spiked sample over a 2-hour period. Instrument stability over thisperiod was excellent with %RSD values typically5%. A stability plot of representative elements isshown in Figure 2.Li (7) Be (9)B (11) Na (23Mg (24) Al (27)K (39) Ca (40Ti (48) V (51)Cr (52) Mn (55Fe (56) Co (59Ni (60) Cu (63Zn (64) Ga (71Ge (72) As (75Sr (88) Zr (90)Nb (93) Mo (95Ag (107) Cd (11Sn (118) Sb (12Ba (137) Ta (181Au (197) Tl (205Pb (208) Bi (209Th (232) U (238Normalizedioncount00.20.40.60.81.01.21.40 20 40 60 80 100 120Time /minFigure 2. Analysis of 100-ppt multi-element standard spiked into 1000-ppm silicon solution measured repeatedly over a 2-hourperiod. Analysis time per sample was 340 s including 60 s sample uptake time and three replicate measurements.Agilent shall not be liable for errors contained herein or for incidental or consequentialdamages in connection with the furnishing, performance, or use of this material.Information, descriptions, and specifications in this publication are subject to changewithout notice.? Agilent Technologies, Inc. 2003Printed in the USAMay 13, 20035988-9529ENwww.agilent.com/chemFor More InformationFor more information on our products and services,visit our Web site at www.agilent.com/chem. For more information about semiconductor mea-surement capabilities, go towww.agilent.com/chem/semicon.ConclusionsThe 7500cs ORS with a MicroFlow nebulizer (MFN-100)was used to determine 36 elements in a 2000 ppm Si matrix sample, with a full analysis taking onlyabout 5 minutes for a sample volume of about 350 μ L. A simple sample preparation method isdescribed, which is applicable to the analysis ofbulk silicon wafer samples. All analytes were mea-sured directly on mass, in a single analytical runwith automatic switching of operating parameters(all data was acquired under normal plasma oper-ating conditions, that is, 1600 W RF forwardpower) and with the results being combined auto-matically into a single report. The results highlightthe effectiveness of the ORS for removing plasmaand matrix-based polyatomic interferences on K, Ca, Fe, Ti, Ni, Cu, and Zn in the presence of ahigh concentration of silicon. All potential interfer-ences are attenuated using the ORS with simplegases H2 or He, and the analysis is fast and robust.The 50 ppt spike recovery data demonstrate theeffectiveness of operating at high RF power withnegligible plasma ionization suppression from thehigh silicon sample matrix. The quantitative recov-ery also indicates the absence of any nebulizationor transport interferences.