Matrix Assisted Laser Desorption/Ionization Orthogonal Acceleration Time-Of-Flight Mass Spectrometry: Development and Characterization of a New Instrument

Home , Daly detector, Sector mass spectrometer

Matrix Assisted Laser Desorption/Ionization Orthogonal Acceleration Time-of-Flight Mass Spectrometry: Development and Characterization of a New Instrument

By David Sean Selby

A candidate for the degree of Doctor of Philosophy

Submitted to the University of New South Wales December 2002 Certificate of Originality

I hereby declare that this submission is my own work and to the best of my knowledge it contains no material previously published or written by another person, nor material which to a substantial extent has been accepted for the award of any other degree or diploma at UNSW or any other educational institution, except where due acknowledgment is made in the thesis. Any contribution made to the research by others, with whom I have worked at UNSW or elsewhere, is explicitly acknowledged in the thesis.

I also declare that the intellectual content of this thesis is the product of my own work, except to the extent that assistance from others in the project’s design and conception or in style, presentation and linguistic expression is acknowledged.

David Selby Abstract

The performance of a linear matrix assisted laser desorption/ionization mass spectrometer (MALDI-oa-TOFMS) was improved with more reproducible sample preparation methods, a higher rate digitiser for integrating signals and customisable computer control, data acquisition and analysis in the LabVIEW programming environment. This resulted in a ~20% improvement in resolution (up to 4,400) and enabled measurement of desorption velocities of 1,000 - 1,800Êms-1 for analytes with m/z 615 Ð 1,350ÊDa, with matrix ion velocities being 4,000 Ð 4,800ÊmsÐ1. Detector limitations and restrictions on source axis energy (and hence velocity) required for the analysis of ions prevented detection of other species with this instrument.

A 20 kV reflecting geometry MALDI-oa-TOFMS was constructed to overcome these limitations and extend the mass range. This mass spectrometer was able to analyse ions desorbed with a wide range of energies (32 Ð 197ÊeV). The resolution was found to be 8,000 -10,000. Best mass accuracy was 15-80Êppm (internal standards ). External calibration gave larger mass errors, mostly due to timing jitter, but the mass axis was stable for >2 weeks. Mass accuracy was independent of the analyte and matrix used. Ions with m/z of ~10,000 - 20,000ÊDa were observable with the use of a pulsed lens in the target region. This lens increased signal approximately 20 times, but degraded resolution. The detection limit of the instrument (sample consumed) was estimated to vary from 10ÊÐÊ90Êfmol, by extrapolation, with more moles required at higher m/z.

The microsphere plate (MSP) electron multiplier used in the reflecting instrument was found to have a temporal response of <Ê1Êns FWHM, but with a low secondary electron conversion efficiency, making it unsuitable for high m/z species. Experiments were also performed with a novel rectangular mesh grid, which (in correct orientation) provided similar resolution to conventional square mesh grids, but with significantly improved transmission and hence sensitivity.

i Acknowledgments

Before commencing this project I knew little about the practicalities of automated instrument control, design and characterization, while by its completion I had learned enough to become an important member of a team. I would like to thank a number of individuals who contributed to this process.

Firstly I would like to thank the two scientists who made this project possible: my supervisor, Associate Professor Michael Guilhaus, who in addition to his thorough supervision often made one feel more like a colleague than a student, for which I especially thank him; my co-supervisor, Dr Victor Mlynski, who was responsible for modifying the linear oa-TOF instrument into MALDI mode and largely responsible for the physical design and construction of the reflecting geometry MALDI-oa-TOF instrument.

Secondly I would like to thank others who made a significant contribution to the project: Paul Hallohan and Mitchell Davis from the School of Chemical ScienceÕs Mechanical Workshop, who fabricated parts to our exacting standards; John Morgan of the School of Chemical ScienceÕs Electronics Workshop, who provided assistance on innumerable occasions concerning computers and electronics; Dr Len Cherkson of the School of Chemical ScienceÕs Electronics Workshop, who designed our push out pulse generator; Dr Joseph Brophy, operator of the School of ChemistryÕs Mass Spectrometry Unit, who provided samples and advice; Dr Keith Fischer and Rui Zang, who provided samples; and Associate Professor Mark Duncan formerly of the Ray Williams Biomedical Mass Spectrometry Facility and Martin Bucknall and Ann Poljack of the Ray Williams Biomedical Mass Spectrometry Facility, who provided advice, samples and access to a commercial MALDI instrument.

Thirdly I would like to thank those who helped to make the stay enjoyable: My fellow students in the Analytical Mass Spectrometry group, Patricia Calderon, Brett Ireland, Mark Lewin (who was also an excellent proof reader), Nirand Pongpun and Nageeb Sousou, and Dr Robert Goldsack of Tim-Tam fame.

Finally (and perhaps most importantly) I would like to thank my parents, Janice Selby and Luc Szulevicz, without whose upbringing I would not have turned out the way I did or have had the persistence to complete this thesis.

179 Contents

Chapter 1: Introduction

1.1 Overview of Thesis 1

1.2 Time-of-Flight Mass Spectrometry 2

1.2.1 Introduction to Mass Spectrometry 2

1.2.2 The Time-of-Flight Analyser 3

1.2.3 Resolution and TOF Mass Analyser Performance 5

1.2.4 Other Measures of TOFMS Performance 6

1.2.5 Improving Resolution: Ion Optics, Time-Lag Focusing and the Ion 8 Mirror

1.2.6 Coupling with Continuous Ionization Sources 10

1.3 Orthogonal Acceleration Time-of-Flight Mass Spectrometry 11

1.3.1 Introduction to Orthogonal Acceleration Time-of-Flight Mass 11 Spectrometry

1.3.2 Principles Specific to oa-TOFMS 12

1.3.3 Types of oa-TOFMS Instruments and Applications 15

1.4 Matrix Assisted Laser Desorption/Ionization 18

1.4.1 The Development of MALDI 18

1.4.2 Theory of Desorption/Ionization 20

1.4.3 Performance Limiting Factors: Initial Kinetic Energy, Detector 20 Efficiency and Reproducibility

1.4.4 Improving TOFMS Performance with Sample Preparation 22

1.4.5 Improving Performance with Instrumentation: Ion Mirrors and Delayed 23 Extraction

ii 1.4.6 Range of Applications: Types of Analyte and Obtaining Structural 24 Information

1.5 MALDI-oa-TOF Combination 26

1.5.1 MALDI-oa-TOF Instrumentation and Applications 26

1.5.2 MALDI with Hybrid Sector/oa-TOF Analysers 27

1.5.3 Why Combine MALDI with non-tandem oa-TOF Analysis? 29

1.6 Computer Simulations of oa-TOFMS 31

1.6.1 The Importance of Simulations in the Design and Characterization of 31 oa-TOF Instruments

1.6.2 I-Opt 31

1.6.3 SimTOF 33

1.7 Aims 35

Chapter 2: Instrumentation and General Methods

2.1 Overview of Instruments and Methods 36

2.2 Prototype Linear MALDI-oa-TOF 36

2.3 MALDI-oa-TOF with an Ion Mirror 37

2.4 A Commercial Instrument: The Voyager 39

2.5 Sample Preparation 39

2.5.1 Matrices, Analytes and Solvents 40

2.5.2 Cleaning of Sample Slides 40

2.5.3 Dried Droplet Sample Preparation 40

2.5.4 Electrospray Sample Deposition 41

iii Chapter 3: Computerised Instrument Control and Analysis

3.1 A Brief Introduction to LabVIEW and Virtual Instruments 43

3.2 Computer Hardware 44

3.3 Instrument Control 45

3.3.1 Computer Controlled Laser Triggering and oa Pulse 45

3.3.2 Control and Monitoring of Power Supplies 46

3.4 Data Acquisition and Analysis 48

3.4.1 Data Acquisition 48

3.4.2 Analysis 50

3.5 Other Programs and Macros 51

Chapter 4: Improving Performance of the Linear MALDI-oa-TOF

4.1 Importance of Improving Performance 53

4.2 Optimising Dried Droplet Sample Preparation 53

4.2.1 Dried Droplet Sample Preparation Methods 53

4.2.2 Appearance of Target and Spectral Quality for Dried Droplet Samples 54

4.3 Electrospray Deposition Sample Preparation 55

4.3.1 Electrospray Deposition Procedures 55

4.3.2 Appearance of Target and Quality of Spectra Obtained with 56 Electrospray Deposition

4.4 Using a Faster Digitiser and the Effects of Laser Power 58

4.4.1 Procedures Used in Assessing Effects of Digitiser and Laser Power 58

4.4.2 Measured Effect of Digitiser Rate 59

4.4.3 Effect of Laser Power 60

4.5 Summary of Early Experiments and Sample Preparation 61

iv Chapter 5: Fundamental Studies Ð Desorption Velocity/Energy Experiments on the Linear Instrument

5.1 Importance of Desorption Velocity/Energy 62

5.2 Velocity Distribution in the Desorption Axis 63

5.2.1 Introduction: Published MALDI Desorption Velocities and Relevance to 63 oa-TOF Instrument Geometry

5.2.2 Experimental: Methods used to Measure Source Axis Velocity 65 Distribution

5.2.3 Results and Discussion: Source Axis Velocity and its Effect on 66 Instrument Performance

5.3 Correlation of Velocity and Position in the TOF Axis 67

5.3.1 Relevance of Correlation or Lack of Correlation of Position and 67 Velocity to Instrument Performance

5.3.2 Experiments and Simulations used to Determine the Degree of 68 Correlation

5.3.3 Comparison of Experimental Results with Simulations 70

5.4 Extension to Other Samples: Higher m/z and Matrix Effects 71

5.4.1 Methods of Sample Preparation and Analysis 72

5.4.2 Analysis of High m/z and HCCA Matrix Results 73

5.4.3 Conclusions Ð Limits of the Instrument 75

Chapter 6:Testing and Improving MALDI-oa-TOF with an Ion Mirror

6.1 Introduction 76

6.2 Initial Experiments with Fullerene Standard, Gramicidin S and Insulin in 77 DHB Matrix

v 6.2.1 Techniques Used in Initial Experiments for Fullerene Standard, 77 Gramicidin S and Insulin

6.2.2 Signals and Mass Accuracy for Fullerene Standard, Gramicidin S and 78 Insulin at -3 and -15ÊkV

6.3 Solving Field Leakage Problems in the Target and Fill Up Regions 80

6.3.1 Target Region 81

6.3.2 Grid 1 and Field Penetration into the Fill Up Region 82

6.4 Establishing Resolution at Ð17.5 kV 84

6.4.1 Methods Used in Experiments Establishing Resolution at Ð17.5 kV 84

6.4.2 Resolution Obtained at Ð17.5 kV 86

6.5 Improving Sensitivity 88

6.5.1 S/N and Approaches to Improving Sensitivity 88

6.5.2 Sensitivity Improvement with an Einzel Lens 90

6.5.3 Sensitivity Improvement with a Pulsed Lens 91

6.5.4 Signal Preamplifier 93

6.6 Overall Performance at Ð3ÊkV, Ð15ÊkV and Ð17.5ÊkV and Initial 95 Performance at -20ÊkV: Conclusions

Chapter 7: Improving Mass Accuracy and Characterizing the Analyser of the Reflecting MALDI-oa-TOF

7.1 Ion Optics at Ð20ÊkV 96

7.2 Optimising the Source and Analyser Potentials 98

7.2.1 Push Out Pulse Shape and Potential 98

7.2.2 Mirror Backplate Potential 100

7.2.3 Explaining Focusing Criteria: Push Out Pulse and Mirror Potentials 101

7.2.4 Delay Time and Probe Potential Ranges 102

vi 7.3 Mass Accuracy Instability (External Calibration) 103

7.3.1 Contributions to Mass Axis Instability 103

7.3.2 Determining Drift on the Power Supplies 104

7.3.3 Long Term Mass Error 107

7.3.4 Drift in Start Triggers: ∆ t Between the Application of the Push Out 108 Pulse and Oscilloscope Zero Time

7.4 Characterizing the Analyser 111

7.4.1 Mass Accuracy Experiments 111

7.4.2 Determining the Limiting Resolution and Quantifying Contributions to 115 Peak Width

7.4.3 Spectra of Species Used to Find Mass Accuracy and Limiting 119 Resolution

7.5 Extension of the Instrument to other Samples: TPP, Myoglobin, DNA and 119 Polyethylene Glycol

7.6 Estimating Instrument Sensitivity 121

7.7 Conclusions on Analyser Performance 123

Chapter 8: Evaluation of the Detector

8.1 Why Characterise the Detector? 125

8.1.1 Importance of the Detection System 125

8.1.2 Microsphere Plate Theory 126

8.2 Design of the Detection System 127

8.3 Evaluation of Detection System Performance 128

8.3.1 Measuring the Background Count Rate and Dark Current 128

8.3.2 The Effect of Bias Potential on Gain 130

8.3.3 Temporal Response 131

vii 8.3.4 Mass Dependence of Gain and Detector Conversion Efficiency 132

8.4 Overall Analysis of Detection System 136

Chapter 9: Assessing the Effects of Grid Geometry

9.1 The Special Importance of Grid Geometry in oa-TOFMS 138

9.2 Simulation of Grid Effects 140

9.2.1 Simulation Techniques 140

9.2.2 Simulation Results 142

9.3 Experimental Determination of Grid Effects 143

9.3.1 Physical Techniques 143

9.3.2 Physical Results 144

9.4 Comparative Analysis 145

9.5 Conclusion and Postscript 146

Chapter 10: Conclusions 148

References 152

Appendices

Appendix 1 Glossary of Abbreviations and Defined Terms 171

Appendix 2 IGOR Pro Macros and Functions 175

Appendix 3 List of Publications and Conference Presentations 177

Acknowledgments 179

viii Chapter 1: Introduction

“I have endeavoured in this book to give some account of the experiments on Positive Rays … I feel sure that there are many problems in Chemistry which could be solved with far greater ease by this than by any other method.” Sir J.J. Thomson, 1913 [1]

1 . 1 Overview of Thesis This thesis describes the development and characterisation of matrix assisted laser desorption/ionisation - orthogonal acceleration time-of-flight (MALDI-oa-TOF) mass spectrometers, with the joint intentions of (i) using these instruments to further understanding of the chemical and physical processes involved in both MALDI and oaÐTOF and (ii) demonstrating the potential of the MALDI-oa-TOF combination for the analysis of samples. The work performed has been divided into a number of chapters and appendices, with this chapter providing the introduction.

Chapter 2 is the general methods chapter, with descriptions and standard operating conditions for the mass spectrometers used, together with standard sample preparation methods utilised for most experiments. The software written during the course of the project, which provided computer control of custom made instruments and computer assisted data analysis is outlined in chapter 3. Chapter 4 details the experiments performed with a prototype linear MALDI-oa-TOF instrument, in order to optimise instrument performance and sample preparation for the oa-TOF instruments. The velocities gained by ions in the desorption/ionization process with the linear MALDI-oa- TOF instrument are investigated in chapter 5. The commissioning of the reflecting MALDI-oa-TOF mass spectrometer and a number of improvements made to its design during the commissioning phase are reported in chapter 6. Chapter 7 contains a detailed account of issues related to mass accuracy in the reflecting MALDI-oa-TOF and provides characterisation of the overall performance of the analyser in that instrument. The performance of the microsphere plate (MSP) detector installed in the reflecting oa-TOF instrument, in particular the mass dependence of its gain, is the subject of chapter 8, while chapter 9 investigates the effects of analyser grids upon resolution. The conclusions of this thesis are provided in chapter 10. Three appendices are also included in this thesis: appendix 1, a glossary of abbreviations and defined terms; appendix 2, the full text of the macros written for the program Igor Pro, used to assist in data analysis and presentation; and appendix 3, a list of publications and conference presentations that resulted from the research presented in this dissertation.

1 The remainder of this chapter is divided into a number of sections, which introduce the analyser (oa-TOF), ionization technique (MALDI), combination of the two (MALDI-oa- TOF), the custom software used to simulate the instruments and the aims of this thesis. Section 1.2 provides an introduction to mass spectrometry and the principles applicable to time-of-flight (TOF) analysers and it is followed by section 1.3, which explains the importance of oa-TOF analysers. The next major section, 1.4, details the development, limitations and importance of MALDI ion generation, with section 1.5 detailing the reasons for combining MALDI with oa-TOF and reviewing the reported literature. The software used in simulating the oa-TOF mass spectrometers is described in section 1.6. Finally, the aims of this thesis are listed at section 1.7.

1 . 2 Time-of-Flight Mass Spectrometry 1.2.1 Introduction to Mass Spectrometry Mass spectroscopy was developed in the early part of the 20th Century by the physicists who were investigating positive rays [2]. Mass spectroscopy is a technique whereby ions are separated according to their mass to charge ratios (m/z), with the instruments used described as mass spectroscopes or mass spectrometers. In a mass spectroscope ions of the different mass to charge ratios are dispersed and detected in parallel, usually with a photographic plate; while in a mass spectrometer ions are detected sequentially (for a scanning analyser) with an electronic detector [3].

The forerunner to all mass spectroscopes is Thompson’s positive-ray parabola instrument, a version of which was used in the identification of two isotopes of neon, with atomic weights of 20 and 22ÊDa [1]. The first mass spectroscope was constructed by Aston and reported in 1919 [4], while the first true mass spectrometer was developed by Dempster in 1918 [5]. All three of these instruments utilised electric and magnetic fields to separate ions according to their m/z.

Most modern instruments are mass spectrometers, containing at least (i) a method of producing ions (an ion source), (ii) an analyser able to separate the ions on the basis of their m/z and (iii) a means of detecting the ions. Analysis, detection and sometimes formation of ions typically occur under vacuum conditions, since the principles used to separate the ions require an absence of collisions with residual gas molecules. Spectra are usually recorded and analysed with a computer, an approach that was first reported in the 1960s [6]. Figure 1.1 provides a block diagram illustrating the components of a modern mass spectrometer, and table 1.1 lists common ion sources, mass analysers and detectors.

mass sample ion source detector analyser

VACUUM

computer

Figure 1.1: Block diagram of the major components common to all typical modern mass spectrometers. Note that in some cases the ion source is located outside the vacuum chamber or at a higher pressure than the mass analyser and detector.

diode deflection plates (source) 317 cm lens

amplifier focusing/accelerating vacuum vacuum electrodes vacuum sample gas oscilloscope

Figure 1.2: Diagram of the EI-TOFMS reported by Cameron and Eggers in 1948. Ions were given a constant final energy of ~500 eV and the ion beam was interrupted by applying 200 V to one of the deflection plates. Table 1.1: Examples of the basic components of mass spectrometers

Ion sources Mass analysers Detectors thermal ionization magnetic (B) = photographic plate spark source double focussing (EB) faraday cup electron impact (EI)= reversed geometry (BE) electron multiplier photoionization (PI) ion cyclotron resonance (ICR) magnetic electron multiplier chemical ionization (CI) quadrupole (Q) continuous dynode electron multiplier field ionization (FI) quadrupole ion trap (ITMS) dual (or triple) channelplate field desorption (FD) radio frequency (RF) daly detector multiphoton ionization (MPI) time-of-flight (TOF) diode array detector fast atom bombardment (FAB) fourier transform (FTMS) channelplate array detector plasma desorption mass triple quadrupole (QqQ) == image currents spectrometry (PDMS) secondary ion mass four sector (EBEB) inductive detector spectrometry (SIMS) thermospray (TS) hybrid (EBqQ) infrared laser desorption (IRLD) hybrid (EB/TOF) matrix assisted laser Hybrid (Qq/TOF) desorption/ionization (MALDI) electrospray ionization (ESI) tandem TOF/TOF = Standard abbreviations are provided in parentheses, where applicable. == An upper case letter indicates the analyser component is used for m/z analysis, while a lower case letter indicates that the component is used to transmit all ions. (based upon Table 1.1 in [23], with modification) 1.2.2 The Time-of-Flight Analyser As indicated above, early mass spectrometers required the use of both electric and magnetic fields. In 1946 Stephens suggested that a mass spectrometer could be made with only electric fields that would be able to separate packets of ions, on the basis that ion flight time down a vacuum tube would be related to ion m/z [7]. This type of instrument is called a time-of-flight (TOF) mass spectrometer. The first TOF instrument was constructed by Cameron and Eggers in 1948 [8] and a similar instrument was reported by Keller in 1949 [9], both instruments using electron impact (EI) to generate ions. Figure 1.2 shows a simplified diagram of the EI-TOFMS constructed by Cameron and Eggers. In both early TOF instruments ions were accelerated through a single electric field to a constant energy, of the order of hundreds to thousands of electron volts. In this energy regime ions travel below relativistic speeds, allowing the behaviour of the ions to be described with newtonian equations [10]. Table 1.2 lists the quantities and units corresponding to the symbols in the equations given below, together with a number of important quantities discussed later in this chapter.

The ions experience an accelerating force from the electric field: FE=q and Fa=m , so Eq a = (1.1) m Acceleration is also defined to be the rate of change of velocity, du dt, so in the accelerating region: Eq uu−=∫dt and 0 m Eq uu=+ t (1.2) 0 m The time to traverse the accelerating region is given by: uu− t =m 0 (1.3) a Eq and the displacement the ions travel during this time is: −= ss0 ∫ udt Eq (1.4) =+ut t2 0 2m

After being accelerated, the ions will receive an energy of UD, which is related to the total potential (V) that accelerated the ions:

2 = ()=−1() UEsuuDaDqV = q2 m 0 This allows calculation of the drift velocity: 3 Table 1.2: Important quantities in deriving time-of-flight equations

quantity symbol description value/unitst mass m mass in kg kg m mass in daltons, with 1 dalton defined to be 1/12 the Da= or amu mass of a carbon-12 atom charge q charge C e charge of a proton (unit charge) 1.60218 x 10-19ÊC z number of unit charges (q/e) time t time-of-flight s

t0 time before or after t = 0 that ion begins to accelerate s

ta time that ion is in accelerating field s

tD time ion spends in field free drift region (linear TOF). s

td response time of the detection system s displacement s distance m

s0 initial position of ion, relative to zero position m

sa distance over which accelerating field acts in a single m stage accelerator

sa1 distance over which the first accelerating field acts in m a dual stage accelerator

sa2 distance over which second accelerating field acts in m a dual stage accelerator D drift region, distance between end of accelerating m region and detector (linear TOFMS)

Dsf location of the spatial focus plane, relative to the end m of the accelerating region

D1 first drift region, distance between accelerating m region and front of ion mirror

D2 second drift region, distance between front of ion m mirror and detector

d distance an ion that started with ‘s0’ and‘u0’ of zero m would penetrate into the ion mirror energy U translational (kinetic) energy of ions, either in joules J or eV== (SI units) or electron volts

U0 initial translational energy of ions J

UD energy of ion in D (linear TOFMS) or D1 and D2 J (TOFMS with ion mirror) velocity u velocity ms-1 -1 u0 initial velocity ms -1 uD drift velocity ms electric field E electric field strength Vm-1 -1 E1 electric field in the first accelerating region of a dual Vm stage accelerator -1 E2 electric field in the second accelerating region of a Vm dual stage accelerator t Values given for non SI units. All values taken from the appendices of Halliday and Resnick [188]. = 1ÊDa = 1.661 x 10-27 kg == 1ÊeV = 1.602 x 10-19 J A conversion factor from m/z to m/q (SI units) can be calculated from the above to be: m/q = (1.037 x x10-8/n)m/z, where ‘n’ is the number of unit charges. 2qEs uu=+ a (1.5) D 0 m and hence time spent in the drift region: = D tD uD (1.6(a)) m 1 =+D  2qEsa u0   =+m 1 or tD D  (1.6(b)) 2qV u0 The total time of flight is obtained by summing the time spent in each region and for a linear TOF instrument with single-stage acceleration it is: =+ TOF ttaD (1.7) In practice, the measured TOF will include an offset between the commencement of timing and acceleration or ion formation (t0) and the detector response time (td), so the following is a more realistic measure: =+++ TOF tttt0aDd (1.8) >> TOF instruments are usually designed to ensure that uuD0 and that t0 and td provide a relatively small contribution to overall TOF, so it follows from equations 1.4 and 1.6 that: TOF∝ mz / (1.9) which can be expressed as: mz/.=a TOF + b (1.10) where ‘a’ is the constant of proportionality and ‘b’ is a constant principally due to t0 and td. Equation 1.8 can be extended with additional terms, if necessary, for instruments with multiple accelerating, ion mirror or drift regions and the relationship between TOF and m/z expressed in expression 1.9 and equation 1.10 applies to extended versions of equation 1.8, provided all accelerating or decelerating fields are constant on the timescale of the measurements. This enables equationÊ1.10 to provide a useful calibration relationship for a TOF mass spectrometer, with the coefficients ‘a’ and ‘b’ calculated from the times of flight of two or more known m/z values.

TOF mass analysers have a number of advantages, with the most significant being: (a) a (theoretically) unlimited mass range; (b) the ability to record an entire spectrum from a single packet of ions, unlike scanning analysers, and hence: (i) removing variation in relative signal intensities between m/z due to source variations; and

4 (ii) providing the ability to acquire spectra rapidly, based upon the TOF of the highest m/z ion (typically in the 10Êµs to c.100Êµs range); and (c) a relatively simple instrument (in theory) since only electric fields are required. There are, however, a number of limits on simple TOF mass analysers, with the most significant relating to resolution and coupling with continuous ion sources. These limits and the methods used to reduce their impact are best understood when considered in association with the measures of mass analyser performance.

1.2.3 Resolution and TOF Mass Analyser Performance The instruments described above, together with other early TOF instruments [11-13], all provided comparatively poor instrument performance. The main limitation was a very low resolving power. Resolving power is the ability of the instrument to separate ions of similar m/z. Resolving power is conventionally measured with the ratio of m/∆m, where

‘m’ is the m/z value of interest and ‘∆m’ is the width of the mass peak at that m/z value.

The value of m/∆m is defined to be the resolution, with ∆m usually measured by the full width at half maximum height (FWHM) for a TOF analyser [14].

In TOFMS it is often convenient to assess data presented in the time domain, so it is useful to be able to determine resolution from time data. For a constant energy TOFMS: mt∝2 or mt=A2 where ‘A’ is a constant of proportionality, assuming a negligible zero offset. Differentiating this expression with respect to time provides: dm = 2At dt which becomes, for a finite interval, after rearranging: ∆∆mAtt=2 and hence: m t = (1.11) ∆∆m 2t From this expression it can be seen that the limit on mass resolving power is due to differences in the measured flight times for ions of the same m/z. A number of factors contribute to the distribution in flight times, as can be seen from an examination of equations 1.4 to 1.8, including: (a) the initial distribution (prior to acceleration) in:

(i) ion kinetic energy and hence velocity (u0) in the TOF direction;

(ii) ion position (s0); and

(iii) time of formation (t0); 5 (b) non-ideal acceleration fields (variations in UD);

(d) glancing collisions between ions and background gas species (affects uD); and (e) metastable decay, where ions fragment in the drift region, with the energy released slightly altering the velocity (uD) of the fragment ions. The most important factors limiting resolution in the early instruments were the initial dispersions, listed at (a).

1.2.4 Other Measures of TOFMS Performance In addition to resolution, analyser performance is measured with a number of other factors, the most important of which relate to mass accuracy, sensitivity, mass range and repetition rate. These measures of performance will be discussed briefly in the following paragraphs. Other performance measures, such as dynamic range, are important in instruments used routinely for quantitative analyses. These quantitative measures were not, however, investigated in the course of the experiments described in this thesis and will not be discussed further.

Mass accuracy is the deviation of measured mass to charge ratios from the true mass to charge value. The level of mass accuracy is determined by reference to the mass error, reported as either an absolute error (measured m/z less true m/z) or a relative error (absolute mass error divided by true m/z). The errors are typically reported in millidalton (mDa) units for absolute errors and as parts per million (ppm) values for relative errors for the current generation of TOF analysers. A mass error can be determined either for a monoisotopic mass or an average mass. The monoisotopic mass refers to the m/z for a peak arising from (typically) the most abundant isotopes of the elements, while the average mass is calculated from the average masses of the elements for a single chemical species, weighted for isotopic abundance, representing the mass centroid [15]. Clearly, monoisotipic masses can only be accurately measured with sufficient resolution (ideally × RFWHMÊ>Ê2Ê Êm/z). Resolving power is also important when determining average masses, since accurate mass determination requires sufficient resolution to ensure that the measured m/z only contains the species for which the mass is being determined. It should be noted that difficulties in determining the centroid of an unresolved mass cluster, due to chemical interferences and baseline noise, ensure that it is usually preferable to measure monoisotopic masses, rather than average masses, where resolution permits [16].

The detection limit is a measure of the smallest amount of sample that can be detected or quantified by an instrument or technique. The detection limit for a mass spectrometer is determined by the efficiency of: (i) ionization, (ii) the mass analyser and (iii) the detection

6 system, in respectively ionizing, analysing and detecting the species of interest. The mass analyser efficiency itself involves two factors, the “duty cycle” and the “transmission efficiency”. The duty cycle is a measure of the fraction of the total time of ion production that ions of a particular m/z are sampled for mass analysis, while the transmission efficiency is a measure of the proportion of ions entering the analyser that reach the detector. The duty cycle of conventional TOF instruments with continuous ion sources (such as EI or CI) is often quite low, since in-line gating is required. In an in-line gating instrument the TOF analyser receives ions from the source for a small proportion of the ion production time [17]. Duty cycle efficiency is, however, very high (often 100%) for a TOF analyser utilised with a pulsed ion source, since all ions from each pulsed ionization event can (in theory) be sampled, in contrast to a scanning analyser which will only sample a small mass range from each pulsed ion packet [18]. Transmission efficiency in TOFMS is largely independent of the type of ion source, and instead is directly related to the transparency and number of grids used in the analyser. TOF analyser transmission efficiency is always less than 100%, unless a gridless analyser is used [19].

The mass range represents the range of masses that can be analysed with an instrument. A TOF analyser has, in theory, an unlimited mass range, with the only limit being that it takes longer for ions of higher m/z to reach the detector. In practice the mass range of TOF analysers is usually limited by the ability of the ion source to produce ions of high m/z, the upper mass limits of the detector, recording system limits and repetition rate considerations.

The repetition rate is a measure of how quickly the instrument can record more than one spectrum. The repetition rates of TOF analysers are, in theory, governed by the time of flight required for the highest m/z being analysed. In practice, the upper limit for the repetition rate is also governed by data system processing time, since data can be acquired very rapidly. For instance, a TOF instrument may acquire spectra over a m/z range of hundreds to thousands, with each peak preferably containing at least 10 data points, to satisfy Nyquist sampling requirements [20] and accurately reproduce the signal, typically generating single sweeps with thousands or tens of thousands of data points in tens to hundreds of microseconds. Further, representative spectra are usually obtained by averaging a number of sweeps, requiring rapid data acquisition and averaging. In combination, flight times and data processing requirements usually permit repetition rates in up to the kilohertz range [21].

7 1.2.5 Improving Resolution: Ion Optics, Time-Lag Focusing and the Ion Mirror The most significant limitations to the resolution of TOF instruments are the initial ∆ ∆ ∆ dispersions of isobaric ions ( so, uo and to) so improved resolution requires either minimisation of the initial dispersions or correction of the initial dispersions by the analyser. In general, the effects of initial temporal dispersion are reduced by either (i) ∆ minimising t0 by using a pulsed source or pulsed accelerator; or (ii) increasing the TOF by using a longer drift region or lower accelerating potential, to decrease the relative ∆ contribution of t0 to TOF. While the variations in initial position and initial energy are also directly minimised, their effects are able to be at least partially corrected by focusing with appropriate analyser ion optics.

Spatial focusing, illustrated as a space-time diagram1 for a single stage accelerator in ∆ figure 1.3, is used to minimise the effects of the distribution in initial position ( so) on total time of flight for isobaric ions. In spatial focusing, the ion initially closer to the detector plane experiences less of the accelerating field and hence has a lower final velocity than the ion initially further from the detector. The spatial focus plane is located where the ion initially further from the detector has caught up with ion initially closer. For a TOFMS with a single stage accelerator (illustrated in figure 1.3) the spatial focus plane is located at a distance (Dsf) from the accelerating grid, which is twice the average distance ions travel in the accelerating region (sa) [22]. In a TOFMS with a two stage accelerator, appropriate setting of the electric fields in the first and second accelerating regions enables the (first order) spatial focus plane to be located wherever desired in the drift region, according to the following equation taken from Cotter [23] at 32:

2()ss+E1 a1E2 a2 D=sf (1.12) sa1 where ‘sa1’ is the average distance ions travel in the first accelerating region, ‘sa2’ is the length of the second accelerating region, and ‘E1’ and ‘E2’ are the fields experienced in the first and second regions respectively. In practice, ions often have a range of initial positions, and spatial focusing is used to create a small spread in arrival time for isobaric ions. Importantly, the location of the space focus plane is independent of mass, so all masses have the same space focus plane [22].

∆ The effects of initial velocity spreads ( uo) (and hence energy spreads) on time of flight are greatest for a pair of isobaric ions with the largest magnitude initial velocity in

1 All space-time diagrams provided in this chapter are based upon figures from [10]. 8

detector plane

D=2sa drift region

acceleration region

Figure 1.3: Space-time trajectories showing spatial focusing in a TOFMS with a one stage accelerator.

drift region ∆t

acceleration +u0 region -u0

(u0 ,s0 ,t0 )

Figure 1.4: Space-time trajectories illustrating the origin of the turn around time, labelled ∆t. opposite directions, +u0 and Ðu0, illustrated in figureÊ1.4. The difference in total flight time, as can be seen, is the time it takes for the ion with initial velocity Ðu0 to be decelerated and re-accelerated to a velocity of +u0. This is called the “turn around time,” the effects of which can be minimised by using a strong initial accelerating field, or by increasing the length of the drift region. Unfortunately, minimising the effects of turn around time conflicts with the requirements of spatial focusing. Increasing the length of the drift region ensures that the detector is further than 2sa from the accelerating grid, preventing efficient spatial focusing for a single stage accelerator. Increasing the strength of the initial accelerating field increases the ratio E1/E2 in a two stage accelerator, resulting in a detector location that is likely to be further from the accelerating region than the distance required for spatial focusing according to equation 1.12. Thus it proved very difficult to simultaneously minimise the effects of initial spatial and energy spreads in the earliest TOF instruments.

Wiley and McLaren developed an ion source with two stage acceleration and a method for correcting both spatial and energy dispersions in 1955 [22], called time-lag focusing, illustrated on a distance-time diagram in figureÊ1.5. The time-lag method compensates for initial dispersions by forming the ions in a field free region, then applying an extraction pulse (relatively small since only the first accelerating stage was pulsed) after a suitable delay. The delay allows ions with differing energies to distribute spatially within the accelerating region, permitting initial kinetic energy to be corrected with spatial focusing fields. Unfortunately, the optimal length of the delay time was mass (but not velocity) dependent, so only a small portion of the spectrum could be obtained at high resolution with any particular delay time. Instruments employing this technique can only obtain full spectra with a box car method, with repetitive ionization cycles that utilise successively longer delay times to resolve successively higher masses.

Another method used to correct energy dispersions is based upon the use of an ion mirror, a technique developed by Mamyrin and co-workers in the 1970’s, who called the overall instrument a “mass reflectron” [24, 25]. An ion mirror consists of a region with a field that reflects the ions at the end of the drift region, with ions traversing through a second drift region before being detected, illustrated in figureÊ1.6 as a space time diagram. An ion mirror utilises the property that faster ions penetrate further into and spend longer in an ion mirror than slower ions of the same m/z., which can be used to offset the shorter time faster ions will spend in the other regions of the instrument. Optimal resolution occurs when the additional time spent by faster ions in the ion mirror is the same as the additional time the slower ions spend in the drift regions and accelerator. For a TOFMS incorporating a single stage ion mirror, it can be shown (to

detector plane

drift region

grid 2 acceleration region 2 grid 1

A +u0 B acceleration region 1 -u0

delay

Figure 1.5: Space-time trajectories showing time-lag focusing in a TOFMS with a two stage accelerator. ± The delay allows the initial energy spread (illustrated with u 0 at time A) to be converted to a large spatial spread at time B, with the delay chosen to provide a sharp focus at the detector plane for ions of a particular m/z.

ion mirror

grid 2

drift region

detector plane m/z m/z1 2

B B spatial focus

grid

A accelerating A region

Figure 1.6: Space-time trajectory diagram showing the effect of a single stage ion mirror in providing focusing for two masses starting at A in a single acquisition. Ions are spatially focused (but defocused for energy) at the spatial focus (B) and are focused for both spatial and energy spreads at the detector plane. first order and ignoring any difference in ion flight time in the accelerator) that this optimal resolution occurs when [23, 24]: += DD124d (1.13) with ‘D1’ and ‘D2’ representing the length of the two drift regions, while ‘d’ is the distance an ion with both u0 and s0 of zero will penetrate into the mirror. Thus the energy focusing effect of an ion mirror is independent of m/z. An ion mirror can be combined with spatial focusing to simultaneously correct initial position and energy, as indicated in the trajectories given in figure 1.6. This method uses a relatively strong initial accelerating field to produce a spatial focus a short distance into the first drift region. Ions of the same m/z have a small spatial distribution at this spatial focus, but a large energy spread. The spatial focus acts as a virtual source with respect to the ion mirror, which is then able to correct the energy spread at the detector plane. Importantly, however, it should be remembered that the ion mirror is unable to correct for turn around time, since ion mirror energy correction is based upon differences in final velocity.

1.2.6 Coupling with Continuous Ionization Sources TOF mass spectrometers require ions in temporally small packets. This is compatible with pulsed ion sources, which generate ions in small time windows, allowing pulsed sources to be used to generate ions from samples located at the beginning of the accelerating field. It is more difficult to couple TOF analysers with continuous ion sources, as continuous sources do not create ions in temporally small packets. In fact, this difficulty is one of the reasons that the earliest TOF instruments had poor resolution, since they contained continuous EI sources, contributing a significant t0.

One method used to create temporally small ion packets with continuous sources is beam deflection, with precision beam deflection first reported in a TOF mass spectrometer in the 1970s [26]. In beam deflection methods the source continuously creates a beam of ions, but ions are only analysed and detected when the beam is modulated, directing ions to pass through a sampling slit [17]. The idea behind beam deflection is that the temporally and spatially small packet of ions sampled can provide a time domain spectrum with acceptable resolution. Unfortunately, there is an inherent conflict between resolution and duty cycle in this method, since optimal resolution requires minimal initial dispersions and this can only be achieved by sampling for the smallest possible time at the start of each TOF cycle. For instance, an instrument with beam deflection may provide a packet width of 10Êns to generate the desired resolution and operate at a repetition rate of 10ÊkHz [17], providing a duty cycle of 0.01%. Duty cycles as low as this are clearly not desirable, since they result in poor sensitivities.

10 Ion storage devices have been used to improve the duty cycle in continuous source TOF instruments [27]. In this approach ions are produced continuously, but stored in a potential well located at the beginning of a Wiley-McLaren type source, with the packet injected and detected at the spatial focus, providing a high duty cycle. Unfortunately, simple ion storage methods do not provide an ideal solution. Ions can not be trapped indefinitely, as ions are lost due to collisions with trapping elements [27]. Further, trapping methods provide significant initial space and energy dispersions, related to storage device geometry and trapping potentials. More recently, a hybrid instrument has been constructed by Quian and Lubman that uses a quadrupole ion trap to store ions prior to analysis by a TOF instrument incorporating a reflectron [28]. This approach appears to provide improved results when compared to simple trapping devices. This can not, however, be considered a conventional method for coupling continuous ion sources to a TOF analyser, since it is actually a more complex tandem instrument, an ITMS/TOF, with the ion trap able to act as a first stage mass analyser, rather than only as a simple ion storage device.

1 . 3 Orthogonal Acceleration Time-of-Flight Mass Spectrometry 1.3.1 Introduction to Orthogonal Acceleration Time-of-Flight Mass Spectrometry Orthogonal acceleration time-of-flight mass spectrometry (oa-TOFMS) provides a simple solution, in theory, to the conflict between resolution and duty cycle when using continuous ion sources with TOF mass analysis, without the use of trapping potentials. The main innovation of the oa-TOFMS approach is to conduct the TOF analysis in a direction orthogonal to the ion source axis, illustrated in figure 1.7 for both (a) linear and (b) reflecting instruments. In an analysis of oa-TOFMS it is important to define two key (orthogonal) directions, the TOF direction and the source direction, as indicated in figure 1.7. This is because an ion’s measured TOF is only related to its velocity component in the TOF direction, while its velocity component in the source direction determines the time it takes to fill the orthogonal accelerator, and is important in determining the exact location of an ion when it reaches the detector plane.

Independence of the TOF and source directions provides two significant advantages: (a) the reduction of the initial dispersions with appropriate design, providing a narrow ion beam (for a continuous source) with:

(i) a small zero averaged spread in initial uTOF,;

(ii) a small known spread in initial sTOF,; and

(iii) a t0 independent of ion formation; and

l a lb b b lp lp

beam optics* orthogonal beam optics* accelerator V (+) po Vpo(+) ion 0 ion 0 source source V (-) G1 V0 bias(+) G1 po G2 G2 orthogonal detector 0 accelerator focus 2 V (+) V (+) (energy) beam G3 beam G3 V (-) θ θ tof Vtof(-)

isobaric ion packet

focus 1 (spatial) axis conventions

source axis

isobaric ion packet

G4 TOF axis

ion mirror spatial focus detector

Vm(+)

Figure 1.7: Layout of typical oaTOF systems utilising continuous sources for (a) linear and (b) reflecting geometries. Ions enter from the ion source and the voltage polarities are those applicable for positive ions. The axis conventions used are indicated in the centre. Focusing in the TOF axis is demonstrated by narrowing of the isobaric ion packet at the focal planes, while dispersion in the source axis is illustrated by gradual broadening in that dimension. (b) a duty cycle of almost 100% for a continuous ion source2, assuming the ratio of source velocity to TOF drift velocity is controlled to provide that the time taken to fill the orthogonal accelerator is less than or equal to the time taken for ions to be accelerated and drift to the detector. The first of these advantages results in improved resolution, while the second increases sensitivity.

Although not reported in the general scientific literature, experiments using oa-TOFMS were first conducted in the 1960s, on a Bendix instrument with a plasma source [30]. The instrument was, in effect, a linear TOF with the output of an atmospheric pressure plasma gun sampled perpendicular to the accelerating region. The orthogonal acceleration approach was primarily used because it permitted easier sampling. The use of steering plates and limits of technology meant that the instrument had a resolving power just sufficient to separate ions of N2, O2, Ar and CO2. Unfortunately, this research by the Bendix Corporation was published as a technical report and was not widely known or followed up. The concept of oa-TOFMS was rediscovered by the groups of Guilhaus and Dodonov in the 1980s and 1990s [31-35].

The principles applicable to conventional TOF instrumentation also apply to the specific case of oa-TOF, and thus, for instance, calibration is performed with equation 1.10, resolution can be determined with equation 1.11 and ion mirrors can be incorporated into instruments to improve focusing. There are, however, a number of important areas where oa-TOFMS differs from conventional TOFMS, including drift trajectory, ion gating considerations, grid influences, resolution and calibration expectations and (for continuous sources) duty cycle. An analysis of how these areas apply to oa-TOF instruments will be provided below, followed by a brief review of the reported types of instrument and some applications.

1.3.2 Principles Specific to oa-TOFMS As illustrated in figureÊ1.7, the spontaneous drift trajectory of ions in oa-TOF instruments is not perpendicular to the source axis (and hence not parallel to the TOF axis as may be expected), and instead is inclined at an angle of θ to the source axis. This angle arises as a result of an ion’s non-zero velocity in the source direction (usa). Importantly, usa is unaffected by the field (if ideal) experienced in the oa, which only accelerates an ion

2 The duty cycle can never be 100% for a continuous source without an (additional) ion storage element, such as that used by Campbell, Collings and Douglas [29] since ions are unable to enter the oa and are thus lost while the push out pulse is on. 12 θ in the TOF direction, to a final TOF velocity component of uTOF. The value of can be readily determined with the following equation:

−  u  θ = tan 1 TOF  (1.14)   usa This spontaneous drift trajectory can prove useful in oa-TOF instruments employing ion mirrors, allowing the detector to be offset from the accelerator, observed in even the earliest reflecting geometry oa-TOFMS [32]. The spontaneous trajectory can, however, interfere with the design of linear instruments. In an effort to remove the effects of usa, the earliest, and a number of other linear oa-TOF instruments, employed parallel ion steering plates in the drift region, to steer the ions into a transverse trajectory (ensure θ is 90¡) presumably for engineering convenience [30, 36-38]. It has, however, been shown with trajectory simulations that steering of ion packets to create a transverse trajectory in oa- TOFMS results in a substantial loss of resolving power [39]. More recently, experiments have confirmed that steering the ion packet in the drift region reduces resolution [29]. This means that θ can never be 90¡ in an oa-TOF instrument that maintains optimal resolution. Further, since the value for uTOF is linked to ion focusing and hence resolution, it is clear from equation 1.14 that the only way to adjust θ without affecting resolution is by modifying usa prior to orthogonal acceleration.

Drift trajectory considerations are relatively straightforward in instruments that employ constant energy sources, such as EI, since in those instruments all ions, regardless of mass, experience effectively the same ratios of kinetic energy in both the source and TOF directions, resulting in the same small range of θ for all masses [34]. Thus for constant energy sources, appropriate θ values are readily obtained by adjusting the source acceleration potential. Drift trajectory considerations are not as straightforward for ion sources that generate ions with usa that remains (relatively) constant with mass, such as ions produced in supersonic beams. In this instance θ decreases with increasing mass and detection of ions over a large mass range requires either (i) a detector significantly longer

(in the source direction) than the orthogonal accelerator or (ii) correction of usa. Most early instruments employing supersonic beams utilised deflection plates in the drift region, to allow the instruments to be tuned for specific mass ranges [30, 33, 36-38]. Unfortunately, this results in a loss of resolution and an inability to obtain a full spectrum. It is possible to give the ions additional energy in the source direction that is large compared to that for the highest m/z, prior to orthogonal acceleration, but this can decrease sampling efficiency, since if usa is too large the oa will refill prior to the

13 collection of the preceding TOF spectrum. A better approach uses collisional cooling in a mulitpole device, which is able to reduce the average ion energy to that of the bath gas θ [40]. This ensures that a smaller range of can be used for all masses. It also allows usa to be set to a value that provides a high duty cycle, if possible, where the time taken to fill the oa is approximately the same as the TOF of the highest m/z ion. It should, however, be noted that using a delay time to maximise duty cycle for the highest m/z species, results in a mass discrimination effect that favours high mass ions, since low mass ions will often have higher usa than the highest mass species and hence a portion of the low mass ions will have exited the fill up region of the oa prior to the application of the push out pulse. Thus, unlike most other types of mass analyser, an oa-TOFMS employing a constant energy source can actually transmit a larger proportion of high mass species than low mass species.

While duty cycle and the effects of θ are very important when considering the behaviour in the source axis prior to ions entering the oa, the oa itself is designed primarily to maximise resolution, with duty cycle an important secondary consideration. The Wiley- McLaren accelerator [22] provides one of the best approaches to correcting initial dispersions and providing relatively high resolution [41]. Thus it is not surprising that most orthogonal accelerators are, in effect, specialised two stage Wiley McLaren ion accelerators. As such, the main requirements of the orthogonal accelerator are that: (a) its first stage is able to act as a field-free fill up region, allowing ions to enter without being deflected in the TOF direction; (b) an electric field is able to be very rapidly generated in the fill up region, with the potential provided by a push out pulse (POP), with the POP persisting until after the heaviest ions have left the region, to avoid compromising the spatial focusing optics; and (c) no ions are able to enter the fill up region and hence orthogonal accelerator while the POP is applied, since these species would be accelerated later and generate “false” signals at higher m/z values. The magnitude of the POP is sometimes reduced by dividing the first region into two stages, illustrated in the potential diagrams for two typical oa arrangements given in figureÊ1.8. Application of a POP introduces relatively little temporal dispersion [42], provided POP rise time and amplitude are applied reproducibly. Finally, the oa can be designed to ensure that ions that would enter the oa during application of the POP are deflected, for instance with bevelled edges at the beginning of the fill up region [31].

Other considerations in oa-TOFMS include the effects of grids on sensitivity and resolution. Grids act to reduce sensitivity, since the presence of conducting wires ensures

A POE beam G1 G2 G3 B POE beam G1 G2 G3

Vpo(+) Vpo(+) drift region drift region 0 0

V V

Vtof(-) Vtof(-) distance from POE distance from POE

Figure 1.8: Potential diagrams for typical orthogonal accelerator configurations where (A) a single pulse is applied to the push out plate electrode (POE) and (B) where simultaneous positive and negative pulses are applied to the POE and G1 to keep the potential near the beam path close to zero. that grids have less than 100% transmission. Grids also create electric field inhomogeneities, since there is always some field penetration between the mesh wires. In fact, avoidance of field penetration can require a small offset potential when creating field free regions in a TOF accelerator source [22]. The effects of field penetration increase in magnitude as (i) the spaces between conductors increase or (ii) the differences in field strengths between the regions separated by the grid increase. Grid effects have been treated analytically for ion mirrors [43] and oa-TOFMS [44], but these treatments do not include the effects of an ion’s angle of approach to the grid. An assessment of grid effects in oa-TOFMS, particularly those linked to angle of approach, was an important part of this thesis and they are addressed in detail in chapter 9.

Resolution in oa-TOF instruments is typically fairly high, when compared to conventional TOF instruments of similar size, since, as mentioned in section 1.3.1, initial dispersions are quite low. Further, the narrow TOF axis dimensions of the fill up region of the orthogonal accelerator can be used to generate averaged initial TOF direction ion velocities close to zero, and a mean initial position lying on the source axis. This ensures that while the initial spreads affect resolution, they should not significantly affect mass accuracy, so oa-TOF instruments have near ideal calibration relationships [18]. The small deviations from ideality relate to the rise time and stability (amplitude) of the POP. All ions travel during the rise time of the POP, but lighter ions travel further, resulting in a slightly smaller accelerating energy from the POP and hence slightly longer than expected flight times, while any other temporal variation in POP amplitude will also have a mass dependent effect on energy.

1.3.3 Types of oa-TOFMS Instruments and Applications The largest area of activity in oa-TOFMS involves electrospray ionization (ESI). Electrospray ionization was developed by Fenn and coworkers [45] several years prior to the rediscovery of oa-TOF, and the first ESI-oa-TOF combination was reported by Dodonov’s group [32], with a detailed description of this instrument provided in a subsequent report [33]. The use of oa-TOF analysers with ESI offers a number of advantages: (a) duty cycle related sensitivity; (b) compatibility with liquid chromatography; and (c) both high m/z capability and excellent mass accuracy, when compared with the quadrupole analysers typically used in ESI analyses. A number of groups recognised at least some of these advantages, with Fenn’s colleagues showing the duty cycle advantages of the ESI-oa-TOF approach, shortly after the release

15 of Dodonov’s early work [37, 46]. The most important developments in ESI-oa-TOF will be given below.

The early ESI-oa-TOF instruments had limited resolution, in part because they generally used steering plates to correct source direction velocities, since ESI produces ions in a supersonic beam. This difficulty was overcome by Standing’s group in 1995, by the use of a collisional focusing device to transfer the ions from the source to the oa fill up region [47], with a detailed report provided in 1998 [40]. Other important instrument developments in ESI-oa-TOF include the use of an ion guide to improve transfer efficiency between the source and oa [48] and the development of commercial tandem instruments incorporating an oa-TOF as the final stage, with the first type of instrument being a magnetic sector oa-TOF hybrid [49], proposed in 1992 [50], and the second and more widely distributed type being a quadrupole oa-TOF (Qq-oa-TOF) hybrid [51].

The high m/z capabilities of ESI-oa-TOF were demonstrated on the instrument developed by Standing’s group, in studies showing that higher m/z and non-covalent species could be observed in an ESI source operated at physiological pH [52, 53], with others performing high m/z experiments with commercial instruments [54]. ESI-oa-TOFMS has been coupled to on-line chromatographic separations, with the high mass accuracy and resolution provided by oa-TOF assisting in identification of species separated by liquid chromatography [55], while the high speed and mass accuracy of oa-TOF has proved important in exact mass measurements of species separated by capillary electrophoresis [56]. Tandem mass spectrometry, with ESI and oa-TOF as the second analyser has been shown to be very sensitive, with either a magnetic sector [57] or a quadrupole [58] used as the first mass analyser. The ESI-Qq-oa-TOF approach appears particularly promising, with one study concluding that it provides a simpler and superior approach for high speed LC/MS/MS than sector mass spectrometry or fourier transform mass spectrometry [59]. ESI-oa-TOFMS has been applied in many other analyses, which are detailed in several reviews [18, 60, 61].

Another important ion source used in combination with oa-TOF analysers is EI. The EI- oa-TOF combination was first suggested in 1989 by Dawson and Guilhaus [31], with a working instrument reported in 1992 and 1993 [34, 62]. The most significant advantage of this instrument was that it was able to obtain spectra with sufficient resolving power, sensitivity and ion abundance fidelity at high acquisition rates to enable its use as a detector for fast gas chromatography (GC) [63, 64]. Mass accuracy and library matches were significantly better than those reported with an axial reflecting TOFMS system [65]. Fast GC/MS with an oa-TOF analyser has also been reported with two separate instruments, one using an EI source and another with a hyperthermal surface ionization 16 (HSI) source, which demonstrated that coeluting components could be readily deconvoluted at the high acquisition rates possible [66]. The HSI source creates ions through the interaction between a supersonic molecular beam and a heated metal interface, and has been shown to be very sensitive when used to ionize polynuclear aromatic hydrocarbons [67].

Inductively coupled plasma (ICP) sources have also been used with oa-TOF analysers, to take advantage of speed benefits and full element sensitivity, facilitating coupling of ICP- MS with transient inlet systems (for instance electrothermal vapourization) and in theory providing better sensitivity than scanning analysers, although the large range of energies in the source direction has the potential to cause mass discrimination. Initial work on ICP- oa-TOFMS was conducted by Hieftje and colleagues in the early 1990s [38]. The first version of the instrument used cylindrical optics to transfer ions to the oa [68], while a later version used direct current (DC) quadrupole optics, which provided superior sensitivity [69]. The multi-element speed advantage was demonstrated in this work, but overall sensitivity was lower than that of scanning analysers [70]. A major limitation of this instrument was the recovery time of the detection system, which was thought to be the main reason why accurate isotope ratio abundance measurements for adjacent masses was limited to cases where the low abundance isotope was lighter than the high abundance isotope [71]. A commercial ICP-oa-TOFMS incorporating a reflectron was developed in parallel with the work of the Hieftje group [72]. This commercial instrument was able: (i) to decrease the effects of mass discrimination by using a large oa; (ii) to improve the recovery time of the detector by using a discrete dynode multiplier instead of a microchannel plate; (iii) to improve the dynamic range with a hybrid system, combining a time to digital convert (TDC) for very low to low range signals and an integrating transient recorder (ITR) for mid to high range signals; and (iv) to use an ion quenching device, located at the first spatial focus, to eliminate most of the signal from background ions, for instance Ar+.

None of the above applications uses a pulsed ionization source. This is not surprising, since the most significant advantage of oa-TOF, when compared to the conventional TOF approach, is to improve duty cycle for continuous sources, an advantage that is not applicable to pulsed sources. The oa-TOF analyser has, however, been extensively used with a pulsed ionization source, matrix assisted laser desorption/ionization (MALDI). The advantages and disadvantages inherent in the MALDI-oa-TOF approach will be discussed after MALDI is introduced.

17 1 . 4 Matrix Assisted Laser Desorption/Ionization 1.4.1 The Development of MALDI The earliest ion source commonly used in mass spectrometry was the EI source, first reported in Dempster’s instrument [5]. The EI source has not changed significantly since it was improved by Neir in the 1940s [14]. An EI source requires its sample to be in the gas phase prior to ion formation, and imparts a large amount of energy to the analyte during ionization. Attempts at gas phase volatilzation result in thermal decomposition of some species, while excessive ionization energy often causes extensive fragmentation of molecular species. The combined effects of decomposition and fragmentation mean that it is often very difficult (if not impossible) to obtain useful spectra of large, non-volatile or labile species by EI. This has encouraged the development of milder forms of ionization [73]. Many milder ionization techniques have been developed, but possibly the most relevant to the development of MALDI are fast atom bombardment (FAB), plasma- desorption mass spectrometry (PDMS) and laser desorption mass spectrometry (LDMS).

FAB, developed by Barber and colleagues [74] involves focusing a fast moving (keV) beam of argon atoms on a sample containing an analyte dissolved in a neutral matrix of (typically) glycerol. The argon atoms are often replaced with an ion beam, in which case the technique is termed liquid secondary ion mass spectrometry (LSIMS) [75]. The main role of the matrix was to replenish the sample and to ensure that relatively small amounts of sample would produce spectra for a reasonable length of time, allowing the FAB source to be used with scanning analysers. Importantly, the experiments by Barber’s group appear to represent the first mass spectrometry experiments where a matrix was intentionally used to disperse the analyte [76].

PDMS, introduced by Torgerson, Mcfarlane and Skowronski [77], involves bombarding a sample deposited on a thin foil (which acts as both a support and ionization matrix) with (MeV) 252Cf fission fragments. PDMS is a pulsed source, with a spectrum of the ions generated from the sample by each fission event recorded in a TOF spectrometer. Molecular ions with m/z of over 20,000ÊDa have been recorded with PDMS [78], setting the record for the heaviest M+ ions detected by mass spectrometry [79], prior to the development of MALDI.

LDMS has been used to produce ions from organic molecules for mass analysis since the 1960s [80]. LDMS involves irradiating a sample with a laser pulse, with laser wavelengths from the UV to IR used for a wide variety of irradiation times. When short pulses are used, laser desorption is compatible with TOF instrumentation and laser

18 desorption TOFMS has been used in the analysis of industrial polymers and oligosaccharides, amongst other species [81].

MALDI mass spectrometry was invented by combining the ideas of LDMS with the use of a matrix to assist in ion production, a function performed by matrices in both FAB/LSIMS and PDMS. MALDI mass spectrometry was first reported in the 1980s by the groups of Hillenkamp [82] and Tanaka [83]. Hillenkamp’s group was able to obtain molecular weight information for peptides and other species of up to m/z 2,845, contained in a molar excess of UV absorbing solid or liquid matrix materials. The desorption/ionization energy was provided by a 266Ênm frequency quadrupuled Nd- YAG laser, with the energy predominantly absorbed by the matrix. Tanaka’s group was able to obtain molecular ions and multiplets of up to 100,000ÊDa, for proteins and polymers with molecular weights of up to 25,000ÊDa from a glycerol matrix containing cobalt powder, with ion production by 337Ênm radiation from a N2 laser. After the high mass work of Tanaka and publication by Karas and Hillenkamp [84] of high mass spectra of proteins (up to m/z 66,750ÊDa), obtained with 266Ênm laser radiation and nicotinic acid matrix, many other groups commenced MALDI-TOF studies. Interest in MALDI- TOF was sufficiently high for several companies to release dedicated commercial MALDI-TOF instruments by 1992 [85] and sub-femtomole (<10-15 mole) sensitivities were achieved before 2000 [86]. The high sensitivity and applicability of the technique to high mass analysis, in particular to biological compounds, has resulted in the publication of thousands of papers involving application of the technique.

A diagram of a typical linear MALDI-TOF instrument is provided in figure 1.9. The main difference between MALDI-TOF instruments and other TOF mass spectrometers, is the use of a laser to generate ions on a sample slide, which is typically used as the first electrode in the accelerator. The most widely used lasers in MALDI are reliable and inexpensive 337Ênm N2 lasers [87]. MALDI spectra have, however, also been produced with UV lasers of wavelengths 266Ênm [82], and 355Ênm (frequency tripled Nd-YAG) µ µ [88]; IR lasers of wavelengths of 2.94Ê m (ER-YAG) [89], and 10Ê m (TEA-CO2) [90]; tunable free electron laser arrangements [91]; and OPO systems [92, 93].

Different matrices provide useful spectra for each wavelength, and investigation of new matrix materials occurred in parallel with experiments with different lasers. Most MALDI applications have involved the use of solid chemical matrices, which were found to generally perform better than liquid matrices [85]. The importance of the matrix has resulted in many studies that have investigated new matrices [94-98] and important matrix parameters such as UV absorption properties [99] and proton affinities [100]. Table 1.3

chevron MCP drift tube detector

detector signal

N laser (337 nm) ±20 kV 2

power supply

computer

sample

±10 kV grid

expanded ions, neutrals B

0V grid ±20 kV

337nm laser

Figure 1.9: Schematic of a typical linear MALDI-TOF instrument with two stage acceleration (A), showing an expanded view of the ion source and acceleration region (B). The ion source, accelerator, drift tube and MCP detector are all under vacuum. Table 1.3: A small selection of MALDI matrices

matrix usable wavelengths applications and references 2,5-dihydroxybenzoic acid 266 nm, 337 nm, 355 proteins [129], lipids [189, 190], (DHB) nm, 2.79 µm, 2.94 oligosaccharides [191] µm, 10.6 µm 3,5-dimethoxy-4- 266 nm, 337 nm, 355 proteins [94] hydroxycinnamic acid nm, 2.79 µm, 2.94 (sinapinic acid, SA) µm, 10.6 µm α-cyano-4-hydroxycinnamic 337 nm, 355 nm peptides [192], lipids [190] acid (CHCA) 3-hydroxypicolinic acid 337 nm, 355 nm oligonucleotides [144] (HPA) nicotinic acid 266 nm, 2.94 µm, proteins [84] 10.6 µm lists some of the more commonly used matrix materials, with information on useful laser wavelengths and analytes.

1.4.2 Theory of Desorption/Ionization Some aspects of the MALDI process are straightforward, but a complete theoretical understanding of the desorption/ionization process is still lacking [101, 102]. Analyte molecules are deposited in a large molar excess, typically several hundred to thousand fold, of matrix crystals. The matrix is chosen so that it readily absorbs the laser radiation and is relatively easily sublimed. The matrix thus functions as the energy absorbing medium, and acts to dilute and isolate the analyte molecules from each other [82, 84, 103]. If the energy density achieved upon laser irradiation is above a threshold level, an explosive phase transition occurs [104], forming a gas jet containing analyte molecules [105]. Thus the velocity of species in this gas jet (or plasma) is an important characteristic of the MALDI process, and these velocities have been measured in a number of experiments, as is discussed below in section 1.4.3 and in chapter 5 of this thesis.

While the desorption process is fairly well understood, at least qualitatively, there is no unified model describing the ionization process [106]. Early MALDI publications assumed that, at least in UV MALDI, analyte ions were produced by photochemical processes that resulted in the transfer of protons from matrix to analyte molecules in the desorption plume [107]. While it appears correct that many ions are created by photochemical processes, ionization in MALDI is a complex phenomenon. The species formed include protonated, deprotonated and cationized ions, together with neutral molecules, fragments and radicals, with the types of ions formed relatively independently (in a qualitative sense) of factors such as the matrix, solvent composition, solution pH and analyte acid base properties [106]. It is likely that no single mechanism can explain all the ions formed, even in a single MALDI event [108]. Importantly, whatever the exact mechanism, large analyte molecules are able to be desorbed and form (predominantly) singly charged ions, which are then available for mass analysis.

1.4.3 Performance Limiting Factors: Initial Kinetic Energy, Detector Efficiency and Reproducibility Early MALDI-TOF instruments suffered from a number of limitations, the most important of which were: (a) lower than expected resolution (only several hundred); (b) lower than expected mass accuracy; (c) poor shot-to-shot reproducibility; and (d) inability to detect ions of very high m/z.

20 Low resolution was related to the effects of initial kinetic energy and poor mass accuracy resulted from the initial kinetic energy and the presence of unresolved adduct species. Poor shot-to-shot reproducibility was related to sample inhomogeneity, while the inability to detect high mass ions was related to the efficiency of the detection system.

The inability to detect high mass ions was relatively easy to overcome, to a certain extent. Detection efficiency for various species is based upon the yield of secondary electrons from the conversion electrodes for incident ions (the “conversion efficiency”). The conversion efficiency is a function of the velocity of the incident ions, with higher velocities resulting in the emission of more electrons [109, 110]. It is clear from equation 1.5 that ion velocity in TOF is inversely proportional to the square root of m/z, with the result that ions of high m/z were not detected owing to velocities below those required for efficient production of secondary electrons. For instance, it is known that the micro channel plate (MCP) detectors commonly used in TOF instruments do not efficiently detect ions with velocity below 104Êms-1 [110, 111]. Fortunately, large ions generated by MALDI are believed to produce low mass secondary ions on collision with a conversion electrode [112] and these ions are able to start the secondary electron cascade, albeit less efficiently than low mass primary ions. Thus singly charged ions with a velocity as low as ~2.6×103Êms-1 have been observed with m/z of approximately 1×106ÊDa, although signal to noise was very low [113]. More recently, Faraday charge collection detectors [114], ion to photon detectors (IPD) [115] and cryogenic detectors [116, 117] have been tested in an attempt to improve high mass detection efficiency. The Faraday charge collector based detector was found to generate a signal that was less mass dependent than an MCP, with reasonable response at high mass, but it provided lower sensitivity at low mass than an MCP and had a rise time of 25Êns, much larger than an MCP. It was hoped that an IPD would provide improvements in detection efficiencies for high mass, but experiments indicated that the relative detection efficiency of an IPD compared to an MCP decreased with increasing m/z [118]. Cryogenic detectors offer the potential of detection efficiencies approaching 100% at all masses, but in their current state of development require very low temperatures (<Ê2ÊK), are too small for general use as TOF detectors (<Ê1Êmm2), and have a poor temporal response [119]. Thus neither Faraday charge collection, ion to photon, nor cryogenic detectors currently provide a suitable alternative to the MCP.

Lower than expected resolution was, to a large extent, the result of the spread in kinetic energy obtained by analyte ions during the desorption/ionization process, combining with the reduction in observed peak resolution caused by unresolved isotope peaks in clusters. Poor mass accuracy was the result of two effects: (i) increased analyte ion velocity, due to

21 kinetic energy obtained in the desorption/ionization process, resulting in lower apparent mass measurements; and (ii) low resolution preventing separation of [M+H]+ analyte species from matrix adduct species, resulting in broad quasi-molecular ion peaks with higher than expected average mass measurements [120]. Thus it was recognised during early MALDI experiments that improving resolution and mass accuracy would require an understanding of ion desorption velocities. In addition, the velocity of the desorption gas jet is regarded as a fundamental physical property of the MALDI process [121] providing further impetus to measure it.

The velocity of desorbed species has been measured with a number of different methods, including: effects on calibration in delayed extraction TOFMS [121, 122], postionization of desorbed neutrals with a second (delayed) laser pulse [123], delay times in oa-TOF instruments [124, 125], a purpose built instrument providing field free desorption/ionization and measurement of desorption angle [126]; and via high speed imaging of the desorption plume [127]. The results from these experiments, which will be analysed in detail in chapter 5, indicated that desorption velocity depends upon sample preparation, the analyte and matrix used, amongst other factors. Experiments with all geometries showed matrix ions had higher velocity than high mass analyte ions. Above c. 3,000ÊDa velocity was essentially constant for analyte molecules, at least within each reported method of measurement, with velocity measurements varying from several hundred to over one thousand meters per second, depending upon conditions. Constant velocity at high mass has been confirmed in studies for multiply charged MALDI ions from m/z of 160,000 to 480,000ÊDa [128]. Thus the initial kinetic energy in MALDI is proportional to mass.

Two types of approach have been used to improve the performance in MALDI-TOFMS. The first type of approach has involved modifications to sample preparation procedures, since varying matrices and methods of mixing/crystallising the analyte and matrix can affect the number of ions produced, level and types of adduct and desorption energies. The other type of approach involves modifications to instrumentation. Key instrument modifications used to improve high mass detection have been discussed above. The most important modifications used to improve resolution and mass accuracy, delayed extraction and ion mirrors, will be discussed below, after sample preparation.

1.4.4 Improving TOFMS Performance with Sample Preparation Some improvements in MALDI-TOF performance have been provided by the identification of new matrices, such as those that provide fewer adducts ions [94, 129]. Other improvements have resulted from the use of co-matrix materials, such as

22 nitrocellulose, which was found to increase the signal from peptide ions and improve reproducibility, assisting in quantitation experiments [130]. A significant amount of research has also been devoted to the development of methods of sample deposition, with the main aims of improving consistency of samples, resolution, mass accuracy and sensitivity. Two methods of sample deposition that have resulted in better MALDI results are fast evaporation techniques [131, 132] and electrospray deposition (ESD) of samples [133, 134], which are discussed briefly below.

In the fast evaporation method matrix is deposited in a fast drying solvent, such as acetone with a trace of water, to produce a fairly homogeneous surface of small matrix crystals. Analyte solution is then layered on top of the matrix, with rinsing of the resulting sample spot with a drop of acidified water used to remove impurities, after the sample has dried. The fast evaporation technique was found to provide attomole range sensitivity, improved spot to spot reproducibility of spectra and increased resolution for peptides [131]. In a subsequent study the authors observed an order of magnitude improvement in mass accuracy for peptide samples (better than 50Êppm in some analyses) when compared with earlier experiments using conventional sample preparation [132].

The ESD method involves the application of an electric field of several kV to a hypodermic needle, with the applied potential causing a solution passing through the needle to be deposited on a grounded electrode. When preparing samples for MALDI analysis, the sprayed solution contains the matrix and/or analyte and a MALDI target is used as the grounded electrode [133, 134]. Axelsson and colleagues determined that ESD of synthetic polymers, a peptide and protein provided improved sample homogeneity and spot to spot reproducibility when compared to conventional air dried droplet sample preparation techniques, with the best reproducibility for samples prepared by spraying a single solution containing both analyte and matrix [133]. Hensel and co-workers investigated the application of ESD to the quantitative analysis of peptides. They found that peptides were distributed more homogeneously when prepared by ESD than conventional air drying methods, providing more reproducible results and enabling quantitation with internal standards over a limited concentration range (2-10Êµg/mL) [134].

1.4.5 Improving Performance with Instrumentation: Ion Mirrors and Delayed Extraction The reflecting analyser, first developed by Mamyrin and coworkers [24] is able to improve resolution in most TOF analysers, with time spent by ions in the ion mirror compensating for initial positive kinetic energy distributions. Thus, while conventional

23 linear MALDI instruments have provided resolutions of no higher than 1,000 (fwhm) [135], RE-TOF resolutions have been reported of ~6,000 (fwhm) for peptides of up to m/z 3,000ÊDa [136]. The distribution in initial ion kinetic energy occurring during MALDI is, unfortunately, too large to be fully corrected by a mirror [137] and thus further methods are required to correct or minimise the effects initial energy. The simplest method for minimising the effects of this distribution, used since early experiments, involves the application of large accelerating potentials to reduce the relative contribution of the initial energy spread to drift velocities and hence resolution. The high velocities that result from large potentials also assist in the detection of high mass ions. A more effective method of improving resolution involved cooling MALDI ions in an ion trap prior to RE- TOF analysis, reducing the initial energy [138], but this approach was not widely adopted, no doubt due to the higher cost and complexity of the instrument, when compared to conventional TOFMS.

The most successful and widely used approach in correcting the initial energy dispersion has been the adaptation of Wiley-McLaren time-lag focusing [22] to MALDI-TOF instruments, with the earliest reports appearing in 1994 and 1995 [135, 137, 139, 140]. The technique applied to MALDI instruments is often termed “delayed extraction” and it is able to correct the dependence of flight time on initial energy, to first order, by delaying the extraction of ions from the source, with the resulting resolution then increasing almost in proportion to the ion path length [137]. Importantly, since MALDI provides (approximately) an initial point source, there is no need to provide spatial focusing in conventional instruments, so the delay time can be optimised for energy focusing only, permitting greatly enhanced resolution. Delayed extraction provides further benefits, including less ion fragmentation in accelerating fields, reduced chemical noise and matrix background and a reduction in the dependence of flight time on laser intensity [137]. Importantly, delayed extraction provides significant improvements in mass accuracy, with reported mass accuracy of ~15Êppm (external calibration) and ~5Êppm (internal calibration) for peptides of up to 4ÊkDa [141]. The main disadvantage of delayed extraction is that the delay time for optimal resolution is mass dependent, so recording a spectrum for a wide range of masses requires sequential recording of spectra at different mass ranges with subsequent combination to provide a high resolution spectrum.

1.4.6 Range of Applications: Types of Analyte and Obtaining Structural Information MALDI mass spectrometry has been applied principally to the analysis of polymers, with most activity in the areas of biopolymers such as proteins and peptides [82, 84], oligosaccharides and other carbohydrates [142, 143] and oligonucleotides [144, 145], although there have also been many investigations into synthetic polymers [83, 146,

24 147]. A particularly important area of application is proteomics, the large scale study of proteins, a rapidly growing area of science in the post-genomic era, where MALDI analysis plays a critical role in the protein identification procedure [148].

As discussed above, an important feature of MALDI is that it generates predominantly quasi-molecular ions, providing molecular weight information for many large and labile species, often with little or no fragmentation. While molecular weight information is very important, it is difficult to unequivocally identify species without the presence of fragment ions that provide structural information. For some analytes, such as oligonucleotides, useful fragment ion spectra can be obtained by careful selection of the matrix [149]. In other instances, fragments have been obtained with three main approaches: (i) analysis of metastable ions, (ii) digestion of samples prior to introduction into the mass spectrometer and (iii) collision induced dissociation (CID).

Metastable ions have been analysed by post source decay (PSD) in reflecting TOF instruments [150], providing sequence information for peptides [151, 152]. In linear TOF instruments, ions that fragment in the drift region (metastable ions) retain the same velocity as the parent ion and thus arrive at the detector at the same time as the parent ion. In a reflecting instrument the ion mirror, however, is able to temporally disperse ions with different kinetic energies, with smaller fragments having less energy and thus reaching the detector earlier than parent ions. Difficulties of the technique include complex calibration [152] and the requirement of different ion mirror voltages for focusing of different PSD masses [153]. Mass dependent focusing has been reduced with an instrument incorporating a curved field reflectron [23, 154], allowing a larger m/z range PSD spectrum to be observed at one setting.

Partial enzymatic digests of proteins have been used in the analysis and identification of proteins, an approach called ‘peptide mass mapping’, with the resulting digest giving a MALDI-TOF spectrum containing primarily quasimolecular ions of peptides providing a ‘peptide mass fingerprint’, used to identify the original protein by comparison with library spectra [155]. Peptide mass mapping has proven important in proteomics and a number of groups have conducted research into automation of the process [156, 157].

CID has been used to provide detailed structural information for various MALDI ionized species, either from tandem instruments incorporating hybrid analysers with an oa-TOF second stage analyser [49, 158] and TOF/TOF analysers [159]; or in MSn trapping instruments employing ion traps [160] or fourier transform ion cyclotron resonance [161, 162].

25 1 . 5 MALDI-oa-TOF Combination 1.5.1 MALDI-oa-TOF Instrumentation and Applications The first MALDI-oa-TOFMS experiments were conducted by Cotter and co-workers [124, 163, 164]. The authors built what they termed a delayed extraction linear MALDI instrument. Ions were generated transversely to the TOF analyser and the probe was set 8Êmm back from the TOF axis to achieve the required delay. Ions were detected for proteins of over 100,000ÊDa [163] and small oligodeoxyribonucleotides [164]. Resolution in these experiments was no higher than 300 (fwhm) for the oligonucleotides and even lower for the larger proteins, where broad unresolved isotope distributions prevailed. The signal-to-noise ratio of the signals was low, as would be expected at high mass with very low resolution. The use of deflection plates to give trajectories transverse to the desorption axis probably contributed to the low resolution, together with the lack of a beam collimation device. The MALDI-oa-TOF approach was also investigated by Standing’s group, who were able to achieve a resolutions of no higher than 1,400 for proteins of up to ~17,000ÊDa, with unresolved isotope distributions [165].

MALDI-oa-TOFMS allows measurement of ion desorption velocity and energy, and it thus has the potential to contribute to an understanding of the MALDI process. Both Cotter’s group [124, 163] and the Manitoba group [125] have determined the desorption velocity distribution from the delay time between laser firing and triggering the push-out- pulse. Cotter and Pan found that nicotinic acid matrix and vasopressin analyte ions had the same velocity distribution. Dworschak, Ens and Standing determined that above m/z 1,000ÊDa analytes in DHB had the same velocity.

The UNSW electron ionization oa-TOF was converted to a linear MALDI oa-TOF, by the replacement of the electron ionization source with a MALDI sample probe [166]. The desorbed ion-plume was collimated before entering the orthogonal accelerator, unlike in the instrument of Cotter’s group, by a 2 mm wide slit located 35 mm from the probe. In initial experiments with 5,10,15,20-tetraphenylporphine, gramicidin S, substance P and

C60, with 2,4-dihydroxybenzoic acid matrix, resolutions of between 2,900 and 3,300 were obtained. Average mass accuracy with external calibration was 56Êppm. Resolution was later found to be limited by the sampling rate of the digitiser, 400ÊMs/s, as is explained in chapter 4, section 4.4.2. This instrument is described in detail in chapter 2 and further experiments performed with it are detailed in chapters 4 and 5.

While this thesis was completed, Krutchinsky et al. [167] adapted an electrospray oa- TOF mass spectrometer with a collisional focusing interface into a MALDI instrument. Resolution of 4,000 to 5,000 (FWHM) was obtained at masses of less than 3,000. At

26 higher masses the isotopes were not sufficiently resolved to allow direct estimation of the instrument resolution, although this would not be expected to decrease. These results were similar to those observed with axial injection MALDI techniques, but the authors were able to use both electrospray and MALDI ion sources on the same analyser. Additionally, mass accuracy was claimed to be approximately 30Êppm with external calibration below m/z of 6,000ÊDa.

More recently, Ruotolo and colleagues have constructed a novel instrument incorporating MALDI with on-line separation and what appears to be oa-TOF, called a MALDI ion mobility orthogonal TOF mass spectrometer [168]. MALDI ions are generated in a region of moderate pressure (1 - 10ÊTorr) and separated over a 4Êcm drift region by a weak (10ÊV/cm) electric field according to ion mobility, with the ions then mass analysed by an orthogonally arranged TOFMS, with a short drift tube. The prototype instrument provided very low resolution in both ion mobility separation (~25, FWHM) and oa-TOF analysis (~400). The ability to provide on-line separation appears promising, however, permitting on-line analysis of protein digest mixtures, so it will be interesting to see how it performs in a higher resolution setup.

1.5.2 MALDI with Hybrid Sector/oa-TOF Analysers MALDI is very good at providing molecular weight information for quasi-molecular ions, since little analyte fragmentation occurs. PSD provides some fragment ions, as discussed in section 1.4, but it relies upon metastable ions and, in contrast to collision induced dissociation (CID) with a collision cell, there is little control of the degree of dissociation and the reaction pathways. Thus tandem instruments incorporating a collision cell for CID provide a more generally applicable method for producing fragment ions. It is impractical, in MALDI applications, to use tandem instruments that rely solely on scanning mass analysers. This is because of the low duty cycle of scanning analysers coupled with the lack of signal reproducibility between laser shots. Utilising an oa-TOF mass analyser after the collision cell increases the duty cycle, making the tandem approach more practical, which has been the main impetus for the development of MALDI sector/oa-TOF instruments.

Bateman et al. have described and built a combined magnetic sector and oa-TOF instrument in which a complete MS/MS spectrum can be recorded for each laser pulse [49]. They were able to obtain a high energy collision induced dissociation (CID) spectrum of renin substrate (m/z 1,758.9ÊDa) by MALDI, using 800ÊeV and xenon as the collision gas. It had the same sequence ions as those found in high energy CID of renin substrate using a liquid secondary ionization mass spectrometry (SIMS) on a four

27 sector instrument. Later developments with this instrument included the installation of an innovative magnet bypass so that ions beyond the m/z limits of the magnetic sector could be transferred to the collision cell and oa-TOFMS.

Fragmentation of underivatised N-linked oligosaccharides ionized by MALDI was investigated using three different types of analyser by Harvey et al. [169]. In spontaneous fragmentation on a magnetic sector mass spectrometer and PSD-MALDI with a reflecting TOFMS most product ions were due to glycosidic cleavage. High energy CID on a hybrid magnetic sector / oa-TOFMS allowed the researchers to control the fragmentation energy (800ÊeV) and gave product ions corresponding to both glycosidic cleavage and cross ring fragments, similar to that in high energy CID for SIMS (secondary ion mass spectrometry). This generated additional sequence and branching information. The only significant disadvantage of the hybrid instrument was that 10 to 20 minutes were required to obtain good signal to noise, since few ions were analysed from each laser shot.

Peptides have been sequenced by MALDI CID on an AutoSpec oa-TOF instrument using 800ÊeV of energy and xenon as the collision gas [170]. The authors found that the quality of sequence information was similar to that obtained from high energy CID with a SIMS ion source. Sensitivity was 1Êpmol for standard peptides.

The structure of a peptide from a single neuron has been determined using an AutoSpec oa-TOF [171]. The study used a conventional MALDI mass spectrometer to obtain the mass spectrum of a neuron from the anterior lobe of the right cerebral ganglion of a freshwater snail. The protonated and cationized ions of the peptide APGWamide were then analysed by MALDI-CID on the hybrid instrument. The CID spectra confirmed the identity of the peptide, demonstrating the feasibility of analysing such small samples by MS/MS with oa-TOF as the second stage.

Kuster et al. [172] investigated the CID MALDI mass spectra of oligosaccharides with different moieties attached to the reducing terminus. The same hybrid analyser configuration used in the preceding reports was used for this series of experiments. Fragmentation behaviour was strongly dependent on the nature of the attached substituent. Adding asparagine to the reducing terminus was seen as the most suitable modification, as it yielded good MALDI sensitivity and a high level of CID information. In further work with complex oligosaccharides, Harvey et al. investigated the high- energy CID spectra of [M+Na]+ ions from 17 underivatized oligosaccharides often found attached to asparagine in glycoproteins [173]. Three different types of cleavage ions occurred, which assisted in determining the structure of the compounds.

28 Synthetic polymers were investigated by Scrivens and coworkers. In one study [158], oligomers of polymethylmethacrylate and polystyrene of mass 1,000 to 3,000ÊDa were investigated by MALDI-CID with cation attachment on an AutoSpec TOF. PSD-MALDI was investigated on a second instrument. MALDI-CID on the hybrid instrument gave better reproducibility and signal to noise than PSD-MALDI. The MALDI-CID spectra with different cations varied more for precursor ions of lower m/z than for those of higher m/z. In another investigation, Jackson et al. [174] conducted further MALDI-CID on polystyrene with the AutoSpec TOF, using silver and copper for cation attachment. The signal to noise ratios for fragment ion peaks was much higher than those obtained by the authors using LSI-MS/MS in an earlier publication [175] and allowed the authors to propose mechanisms for the formation of some of the fragment ion series. On the basis of the work conducted, the authors concluded that MALDI-CID on the AutoSpec oa-TOF is a very useful structural probe for the polymers analysed.

More recently, commercial quadrupole-oa-TOF (Qq-oa-TOF) instruments have been developed, employing collisional cooling prior to the mass analysing quadrupoles (like the MALDI-oa-TOF analyser reported in 1998 [167]), providing higher sensitivity, incorporating reflecting oa-TOF analysers for improved resolution and providing instruments with interchangeable MALDI and electrospray sources [176, 177]. Resolutions of approximately 10,000 and mass accuracies of 10Êppm have been reported in both MS and MS/MS modes, although sensitivity is limited above m/z 6,000 [176]. In experiments by Baldwin et al., a MALDI-Qq-oa-TOFMS was found to provide CID spectra comparable in quality to those obtained with similar instruments employing an electrospray source, with sequencing data from peptides at subpicomole sensitivities [178]. Importantly, precursor ion selection with a quadrupole analyser permitted selection of a narrower precursor ion window than would be possible with a PSD instrument [178]. A MALDI-Qq-oa-TOF instrument has also been used in pseudo-MS3 experiments, with selection of ‘hot’ matrix materials and high laser fluence settings providing non- selective fragmentation in the collisional cooling region prior to MS/MS analysis, assisting in structural studies of peptides that provide predominantly single product ions in MS/MS analysis [179].

1.5.3 Why Combine MALDI with non-tandem oa-TOF Analysis? It is clear that oa-TOF is a very useful analyser for MALDI when used as the second stage in a tandem analyser, providing more generally useful product ion spectra than PSD instruments. The current utility of simple MALDI-oa-TOFMS is not, however, as clear. Prior to the development and widespread application of DE-TOF, the MALDI-oa-TOF approach was used in an attempts to improve resolution [163, 165] and it has also proved

29 useful in the measurement of desorption velocities [124, 125]. DE-TOF has proven very successful at improving resolution and mass accuracy by correcting the energy spread caused by the MALDI process, so it may be thought that MALDI-oa-TOF is no longer as useful a technique, since: (a) a high resolution oa-TOF instrument, with a sampling slit (for collimation of ions) may not sample all MALDI ions, reducing sensitivity when compared to DE- MALDI-TOF, where all ions can be sampled; and (b) each spectrum may only contain ions with a small range of m/z values, depending upon the length of the fill up region of the orthogonal accelerator in the source axis and ion velocity components in the source direction.

These disadvantages are not as great as they appear. Firstly, high mass MALDI analyte ions have most of their desorption velocity expressed axially (perpendicular) to the sample surface [127], permitting a significant proportion to be sampled through a slit. Secondly, a MALDI-oa-TOFMS can be designed to provide for an oa that is long enough in the source dimension to sample a large range of m/z, providing desorption velocities are known. Further, the MALDI-oa-TOF approach offers a number of advantages when compared to DE-MALDI-TOF, due principally to the decoupling of velocity in the source direction, where most desorption/ionization energy is expressed, from TOF analysis. The most significant advantages are: (a) mass independent calibration, allowing (in principle) the use of two point external calibration for accurate mass measurements over a large range of m/z, a simpler and more stable procedure than calibration in DE-MALDI-TOF; (b) optimal resolution that is likely to be mass independent, since oa-TOF provides spatial and energy focusing for ions with small initial dispersions, unlike DE- MALDI, which requires a mass dependent delay for optimal resolution; (c) no transmission of neutral species to the detector, since only ions are injected by the orthogonal accelerator (an advantage for linear analysers) and highly reduced transmission of low mass matrix ions to the detector, since higher desorption velocities provide that most low mass ions will not be in the accelerator at the same time as high mass analyte ions. This avoids low mass detector saturation and prolongs detector life; and (d) the ability to determine the (extended) field-free desorption velocity of ions in the source direction, whereas DE-MALDI measures desorption velocities after a short period of time. This is useful, since determining the final desorption velocity and comparing it to desorption velocities obtained with DE-MALDI may assist in understanding the MALDI process. Thus MALDI-oa-TOFMS is a technique that should be investigated further.

30 1 . 6 Computer Simulations of oa-TOFMS 1.6.1 The Importance of Simulations in the Design and Characterization of oa-TOF Instruments In common with many other research instruments, a modern oa-TOFMS is a high precision machine and small alterations in its design can have a drastic effect on performance. It would be very difficult to design a close to optimal mass spectrometer without either numerical simulation before commencing construction, or construction of a large number of prototypes. Simulation can also be used to predict and confirm experimental results, to provide an enhanced theoretical understanding of the operation of existing mass spectrometers. Thus simulations fulfil the following functions: (a) Simulations enable us to design a close to optimal mass spectrometer, without requiring the construction of a large number of prototype instruments, saving time and resources. Instead, the proposed prototype can be simulated and the instrument parameters adjusted in the simulation until appropriate results are obtained. In oa-TOFMS the simulation is used to provide approximate times of flight and resolution for a certain instrument geometry and accelerating potentials. The geometry and potentials can be adjusted, within allowed limits, to provide the approximately optimal solution, allowing such an instrument to be built. (b) Simulations can be used to predict results of specific experiments with existing instruments. The results of the experiments can be compared with the simulations, to determine how well the theoretical model fits the experiment. If the model does not fit, this allows the researchers to refine the model, improving the researchers’ understanding of the principles governing instrument performance, potentially allowing improved instruments to be designed in the future.

Most important simulations for the work presented in this thesis were conducted with I-Opt and SimTOF. Both of these programs were written by current and former members of the University of New South Wales Analytical Mass Spectrometry Group. These two programs have not been reported in detail in the literature and so will be explained below.

1.6.2 I-Opt I-Opt is an ion optics program, written by Dawson and Guilhaus. It is based upon classical finite-difference numerical calculations [180]. It fulfils a similar function to commercial programs such as SIMION 3D (Scientific Instrument Services, Ringoes, NJ, USA) [181] that can determine the electric and magnetic fields in the regions surrounding and adjacent to magnets and charged conductors with an iterative method, then calculate 31 the trajectories of ions that pass through the electric field regions and field free space that define the instrument. I-Opt itself was developed in the late 1980s and, due to the processor speed and RAM limits of the then available personal computers (PCs), it was only designed to model 2 dimensional arrays, much like the version of SIMION available then, SIMION PC/PS2 [182].

I-Opt predicts the electric fields, after boundary conditions are defined, by ensuring all array points satisfy the Laplace equation [183] for a potential function in two dimensions: d 2φφd 2 +=0 (1.15a) dx2 dy2 In a numeric method, equation 1.15a holds where: φφφ++ + φ φ =01,, 10 0 ,−− 1 10 , (1.15b) 00, 4 where the various φ values correspond to the electric potentials at array points located adjacent to each other in the following pattern: φ 01, φφφ −10,,, 00 10 φ 01,− In each iteration, the optimisation algorithm adjusts the value of the electric potential (φ ) at each point, until either the sum of the absolute values of the residuals in equation 1.15b for all points is zero, or the user stops the optimisation process.

I-Opt contains many features, of which the following were important in this project: (a) a graphical interface, providing a user friendly environment to define electrodes (including transparent grid elements); (b) a choice of several symmetries: (i) planar with one 2 dimensional plane; (ii) planar with two orthogonal planes, providing slices of a 3 dimensional region; (iii) cylindrical, providing accurate simulation of a slice through a 3 dimensional cylindrical region; and (iv) repeat symmetry for repeating elements with offset voltages in 1 dimension, which involves a ‘wrapping’ technique where the last column of elements is mapped onto the first column, with either zero offset (for instance in grid simulations) or a voltage offset (for instance in discrete dynode multiplier simulations);

32 (c) TOF optimised trajectory algorithms, able to calculate trajectories with a “look ahead” method that is able to determine the forces experience by ions midway between trajectory steps; and (d) the ability to check the quality of the simulated results in trajectory calculations with a conservation of energy check, which compares the kinetic energy change ∆ calculated from the velocity of simulated ion trajectories( KEtraj) with the energy change expected for ions based upon the ion charge and overall potential ∆ ∆ difference ( KEvolt). If the electric fields were not accurately simulated, KEtraj ∆ will not equal KEvolt, unless the errors cancel.

1.6.3 SimTOF Ion optics programs like SIMION 6.0 can be used to calculate times of flight for ions in simulated TOF instruments [184]. Such an approach is, however, highly inefficient, since the user is required to define all electrodes and potentials in 2 or 3 dimensions, wait for the resulting arrays to optimise and then have the program calculate the required ion trajectories. When simulating an entire instrument the arrays will typically be fairly large, to provide the necessary level of accuracy. Further, any change in electrode location will require re-optimising of the array, adding additional time. It would be much more efficient to calculate ion times of flight with appropriate equations, such as those given in section 1.2.2, provided the method used is also able to calculate the effects of the factors that lead to a spread in flight times for isobars. Once the appropriate flight times have been calculated, the application of probability density theory [185] can be used to generate realistic simulated spectra for proposed TOF instruments.

SimTOF is a PC based TOFMS simulator, written by Guilhaus, Mlynski and Lewin, that is able to generate realistic spectra in simulated instruments, by utilising a probability density function (PDF) to generate the signals for two m/z species at a time [186]. The core PDF equation of SimTOF is: −−τ 2 −− 2 −− 2 {(,)}Ts0 u {}ss0 {}uu0 σ2 σ2 σ2 f()τ=×× C∫∫ e2 t e2 s e2 u dsdu (1.16) where ‘f(τ)’ is the probability density function, ‘τ’ is the TOF, ‘s’ is the displacement, ‘u’ σ σ σ is the velocity and ‘ t’, ‘ s’ and ‘ u’ are the standard deviations of the initial values in time, displacement and velocity respectively. The PDF is used by the SimTOF algorithm to calculate an accurate peak shape.

33 The SimTOF algorithm is illustrated diagramatically in figure 1.103. The combined temporal contributions of grid dispersions (‘d’), the detector (‘e’) and HV ripple (‘f’), labelled ‘a’ in the figure, correspond to the first exponential term in the PDF integral, while the initial velocity (‘b’) and initial position (‘c’) correspond to the second and third terms respectively in the PDF integral. The user is required to enter values for the following, prior to calculation of the TOF mass spectrum: (i) the expected spatial spread and velocity (or energy) spread in the ion source for a conventional TOF instrument, or the values for spatial and velocity spread in the fill-up region of an orthogonal accelerator for an oa-TOF instrument; (ii) the estimated or known temporal contribution of the detector (‘e’) and high voltage supplies (‘f’); (iii) the estimated energy spread contributed by electric field inhomogeneities near each of the accelerating or mirror grids (‘d’); (iv) the m/z and relative intensities of the two species for which peaks will be plotted (‘g’); and (v) the displacements between elements and potentials required to define the TOF analyser (‘h’).

Once this data has been entered, the SimTOF algorithm is able to generate the appropriate spectrum for the two m/z values. The spectrum can be generated in either one of two formats. The first format is a smooth spectrum, as shown at ‘i’ in figure 1.10, generated by the program smoothly stepping through the velocity and space intervals of the double integral in the PDF equation. The second format involves the use of a Monte Carlo method [187], which simulates the signal from randomly generated single ions, with the user stopping the simulation once the summed signal from the desired number of ions has been plotted. When a sufficient number of ions have been simulated, a spectrum obtained with the Monte Carlo method approximates the smooth mass spectrum provided by the first format. The smooth spectrum represents the expected spectrum when a large number of ions are analysed, while Monte Carlo simulation results can be usefully compared to experimental results where only a small number of ions have been detected.

Finally, SimTOF was also able to generate simulated spectra for entire m/z clusters, if required, from the smooth spectrum for two m/z values. This feature relied upon the observation that the performance of the analyser does not usually change significantly over a small mass range, corresponding to the range of masses observed in clusters due to isotopic contributions at higher mass. Thus SimTOF could generate a simulated

3 Figure 1.10 was based upon the figure by M. Guilhaus and M. Lewin available at the URL http://www.bmsf.unsw.edu.au/research/instrumentation/SimTOFcalc.html on 18 October 2002.

d ef

Figure 1.10: Flowchart outlining the process used by SimTOF to generate a spectrum. spectrum (‘k’ in figure 1.10) for an entire cluster, based upon the peak shapes obtained for the two simulated m/z values, providing the user enters the mass and relative intensity information for the cluster (‘j’ in figure 1.10).

1.7 Aims This project had a number of aims. The first aim was to assist in the development of and then optimise and characterize MALDI-oa-TOF instrumentation, including a prototype 3ÊkV linear instrument and purpose-built 20ÊkV reflecting instrument. The other aims of the project, which could only be fulfilled once instruments had been developed, were: • to investigate the desorption velocities (in both TOF and source directions, if possible) obtained by matrix and analyte ions in the MALDI ion source after

desorption/ionization by the 337Ênm N2 laser; • to investigate the feasibility of using a MSP as an ion detector in MALDI-oa-TOFMS; and • to investigate the effect of accelerator grid wires on oa-TOF resolution and sensitivity.

35 Chapter 2: Instrumentation and General Methods

2 . 1 Overview of Instruments and Methods Three mass spectrometers were used in the experiments described in this thesis. Most of the experiments were conducted with two prototype oa-TOF mass spectrometers constructed for this project, one instrument with a single linear drift region and the other incorporating an ion mirror. A commercial MALDI-TOF instrument, the Voyager DE- STR, was also used for some experiments.

The layout and operating principles of the three mass spectrometers used in this project are given in this chapter, together with the main sample preparation techniques that were utilised. Detailed descriptions of the computer software and hardware used to control the experiments will be provided in the next chapter, including details of mass calibration. A number of specialised experiments required methods tailored to the specific circumstances. These specialised methods will not be discussed in this chapter and will instead be addressed later in the appropriate chapters.

2 . 2 Prototype Linear MALDI-oa-TOF A linear MALDI-oa-TOFMS was created by modification of the existing prototype GC/MS (gas chromatography/mass spectrometry) oa-TOF apparatus [166]. This involved removing the electron impact ion source and installing the MALDI probe and laser optics. The layout of the analyser’s main components is given in figure 2.1. Potentials and distances between important elements are summarised in table 2.1 and described in more detail below. The instrument was configured to analyse positive ions only.

The ion source region consisted of a sample probe, source isolation valve and roughing pump. New samples were inserted into the vacuum system according to the following method. The sample probe was extracted from the analyser vacuum chamber into the vacuum lock and the valve was closed, isolating the lock from the analyser. The probe was then removed from the lock and a new sample slide (constructed from aluminium or stainless steel) was fitted. The probe was reattached to the lock, and the lock was pumped down by a roughing pump. The source isolation valve was then opened, and the sample was inserted into the analyser vacuum chamber, ready for analysis.

A pulsed 337 nm UV nitrogen laser (model VSL-337-NDS, by Laser Science International, Newton, MA, USA) was used to generate the ions in the source region. A

60 Ls-1 turbomolecular pump orthogonal acceleration pulse driver source isolation lens valve

source shield optics mirror laser window

open analyser housing

-6 MALDI 10 Torr source probe sample slide axis

attenuator orthogonal accelerator

fixed-length bellows

θ = -1 utof tan outer drift-tube (0 V) usource

inner drift-tube (-3000 V)

detector plane optics table laser

detector housing 2 x 10-7 mbar

detector angle adjustment 100 mm

Figure 2.1: Scale diagram of the linear MALDI-oaTOF instrument used in this project, showing the relationship between drift angle ( θ ) and velocity components in the TOF and source axes. Table 2.1: Linear MALDI - oaTOF Specifications

component distance potential (V) Aperture source: sample slide 1.0 mm from sampling orifice 0 sampling orifice 26.0 mm from sampling slit 0 2.0 mm ∅ sampling silt 13 mm from centre of oa 0 1 mm slit accelerator: push-out plate -1.2 mm from ion beam centre 88 grid 1 5.2 mm from push-out plate 0 grid 2 5.5 mm from grid 1 -100 grid 3 12.0 mm from grid 2 -2898 drift region and detector: flight tube grid 3 to detector -2898 20.0 mm ∅ detector 1500 mm from grid 3 -2898

push-out pulse plate

from source axis source G1

to detector

Figure 2.2: Cross section of the orthogonal accelerator of the linear MALDI-oaTOF instrument (not to scale). 5 ns laser pulse was applied perpendicular to the surface of the sample probe. Laser intensity was controlled with the attenuator, whilst beam direction and focusing were controlled by the optics mirror and lens illustrated in figure 2.1.

Ions and neutrals formed in the desorption plume were collimated by a slit located 27 mm from the probe. The sampled species then travelled to the fill up region of the orthogonal accelerator, located between the push out plate and first grid. The orthogonal accelerator consisted of a push out plate, three grids and field spacer rings, illustrated in figureÊ2.2, arranged parallel to the ion beam. The grids were made from 1,000ÊxÊ1,000 lines per inch nickel mesh. When ions of interest were in the fill up region the push out pulse generator was triggered, applying a potential to the push out plate that ejected the ions into the accelerating field, while neutral species were undeflected. After being accelerated the ions travelled down a 20Êmm (diameter) x 1,500Êmm (length) TOF drift tube towards a 25Êmm diameter chevron mounted microchannel plate (MCP) detector (Galileo Electro- Optics Corporation, Sturbridge, MA, USA), with its surface arranged parallel to the accelerating meshes and push out electrode. The angle between the drift tube and orthogonal accelerator could be adjusted between (approximately) 83 and 88¡, since the drift tube was attached to the chamber by fixed-length bellows (figure 2.1).

The push out pulse generator also triggered the TOF transient recorder, which recorded the output from the MCP detector. Two different integrating transient recorders were used to record the time spectra, either a LeCroy 9450 or LeCroy 9384 Oscilloscope, both from the LeCroy Corporation (Chestnut Ridge, NY, USA). The LeCroy 9450 is a 350ÊMHz, 400 Msamples/s 2 channel digital oscilloscope, whilst the LeCroy 9384 is a 1ÊGHz, 1 Gsample/s 4 channel digital oscilloscope, capable of recording up to 4 Gsamples/s when operated in 1 channel mode. Initial experiments were performed with the LeCroy 9450. When the LeCroy 9384 became available, it was used instead, since it had a greater bandwidth and faster sampling rate, giving a better representation of nano second scale peaks in the TOF spectrum. The overall triggering sequence involved in generating ions and obtaining a spectrum is given in figure 2.3.

2 . 3 MALDI oa-TOF with an Ion Mirror After the success of the first prototype instrument a purpose-built chamber was designed for MALDI-oa-TOFMS with an ion mirror. This oa-TOF could operate at accelerating potentials of up to 20ÊkV and required five power supplies. Like the linear instrument, it was configured for positive ion analysis only. The overall layout of the instrument is given in figure 2.4 and the values of key parameters are summarised in table 2.2. A photograph of the mass spectrometer is shown in figure 2.5.

variable delay

Laser 1

Push out pulse generator 2

Oscilloscope 3

time (µs)

Figure 2.3: TTL trigger sequence for MALDI oa-TOF experiments. TTLs 1 and 2 are provided by the computer and are of approximately 10 µs duration. TTL 3 originates from the push out pulse generator. All devices are triggered by the rising edge of a TTL.

pulse generator

1.0 kV target MSP detector

G1 source G2 region G3 delay orthogonal θ attenuator accelerator

UV laser -20 kV 337 nm -6

signal start G4 computer

+ LabView¨ ITR The pressure instrument. diagram of the 20 kV MALDI-oaTOF A operation. during Torr was approximately 10

ion

mirror Figure 2.4: IEEE 1.3 kV Table 2.2: Reflectron MALDI - oaTOF specifications, showing typical potentials when the instrument was operated at Ð 20 kV.

component distance potential (V) Aperture source: sample slide 1 mm from sampling orifice 50 - 200 sampling orifice 25 mm from oa entrance slit 0 1.0 mm ∅ source lens 0 - 180 oa- entrance slit 50 mm from centre of oa 0 2.0 x 14.0 mm accelerator: push-out plate 1.25 mm behind ion beam 1005.4 centre grid 1 5.36 mm from push Ð out plate 2.50 50.0 mm ∅ grid 2 5.5 mm from grid 1 -1334.6 50.0 mm ∅ grid 3 40.0 mm from grid 2 -20000 50.0 mm ∅ mirror: mirror grid 550.0 mm from grid 3 -20000 mirror backplate 158.0 mm from mirror grid 1300 detector: detector (front) 619.3 mm from mirror grid -20000 63.0 mm ∅ collection plate 60.7 mm from front of detector 0

Figure 2.5: Labelled photograph of the 20 kV MALDI-oaTOF, with a superimposed ray (cyan) indicating the path followed by laser emitted photons. The source region (probe and lens assemblies) was powered by a TC945 high voltage power supply (Tennelec, now part of Canberra Industries Inc., Meriden, CT, USA). The push out plate and mirror backplate utilised separate +1.5ÊkV MK series supplies (Glassman High Voltage, Salem Industrial Park, NJ, USA). Grid 1 was connected to a custom built power supply. All other grids and the detector were connected to -25ÊkV MK series supply, from Glassman High Voltage. A more detailed discussion of the components follows.

The basic ion source was similar to that of the prototype, with two important modifications. Firstly, a potential was applied to the probe in some experiments, ranging from 0ÊV to 200ÊV, to adjust the desorption energy to allow more ions to reach the detector. Secondly, an Einzel lens was installed in the source region (figure 2.6(a)) to direct a larger proportion of the ions into the orthogonal accelerator and thereby increase sensitivity. A positive potential was applied to the lens (40 Ð 150ÊV), when required, to direct more ions through the 2Êmm wide x 14Êmm high slit, located 25Êmm after the entrance to the lens assembly. Since a potential could be applied to the probe, its position was adjusted to keep it 1Êmm back from the sampling orifice. Samples were inserted and removed utilising the procedure outlined for the first instrument. Eventually, the Einzel lens was replaced with a pulsed lens (figure 2.6(b)), which could operate without a biased probe. It provided focussing of the ions’ position in the TOF axis and velocity in the desorption axis, by applying a potential (40 Ð 200ÊV) after a short delay (2 Ð 20Êµs,

0.1Êµs resolution).

No significant changes were made to the laser or UV optics between the linear and reflectron instruments, as can be seen from a comparison of figuresÊ2.1 and 2.4. The orthogonal accelerator in the new instrument was designed in a similar fashion to that of the old instrument, consisting of three grids located as indicated in tableÊ2.2. The most important difference was the construction of the grids, which consisted of 120ÊxÊ12 lines per centimetre (lpc) nickel mesh made by electodeposition (Buckbee Mears, St Paul, MN, USA), to improve transmission. The grids were oriented with the 120 lpc wires parallel to the source axis.

The mirror consisted of a 40ÊxÊ4Êlpc Towne Technologies (Somerville, NJ, USA) grid, located 550 mm from the third accelerating grid and a backplate, 158Êmm behind it. Eleven spacer frames electrically separated by resistors were used to create a constant field between the mirror mesh, held at the accelerating potential, and the backplate, which was set to a higher potential than the push out pulse. 38

lens electrical contact grounded element teflon insulator

sample probe

source axis A to orthogonal accelerator

lens element

teflon insulator

grounded elements lens electrical contact

teflon insulator

sample probe

source axis B to orthogonal accelerator

lens element

Figure 2.6: Cross-section diagrams of the source region in the 20 kV instrument (a) with the Einzel lens installed and (b) with the pulsed lens installed. Conducting elements are shaded dark grey, insulating parts are shaded light grey and free space (vacuum) is white.

ions

drift region potential (in contact with analyser liner)

spring contact PEEK

MSP 22 MΩ

60 MΩ

PEEK 60 MΩ steel support spacer

rings 41 mm 60 MΩ

60 MΩ

22 MΩ

ground

detector signal collector plate (to oscilloscope)

Figure 2.7: Diagram of the detector assembly, with resistors listed. The 22 MΩ resistances represent single resistors, while each 60 MΩ resistance represents a chain of 1x15, 1x12 and 1x33 MΩ resistor, connected in series. A further resistance, connected in parallel with the MSP and located outside the vacuum system, is not included in this figure.

back plate supply (+1.3 kV)

10 -6 Torr 100 kOhm

back plate

396 MOhm accelerator (mirror) mirror grid supply (-20 kV)

drift region lining -20 kV

grid 3

0.5 MOhm MSP 228 MOhm (oa)

15 MOhm grid 2 284 MOhm grid 1 POP

+2.5 V detection system (oscilloscope)

push out pulse 2.2 MOhm driver (+1 kV)

12.5 MOhm

1 pole 2 position gain selector switch 1.2 MOhm

Figure 2.8: A schematic diagram of the circuit for the oaTOF analyser and detection system. Series of resistors used between key elements (such as grids) or between components and ground have been combined into effective resistances, to provide greater clarity. The detector assembly (figure 2.7) was installed parallel to the oa on the same mounting plate (see figures 2.4 and 2.5). This avoided the necessity of adjusting the detector angle, which had been required for the linear instrument. It consisted of a 70Êmm diameter El- Mul (Yavne, Israel) single thickness microsphere plate (MSP) electron multiplier, followed by 5 field spacer rings (to evenly step the potential down to ground) and a collector plate for the emitted electrons. Charge arriving on the collector plate was measured as a transient voltage with the LeCroy 9384. Spacer rings were required to step the voltage down to ground, because the MSP had a maximum limiting voltage of 3.5ÊkV and operating accelerator voltages were over 15ÊkV. The bias potential across the MSP was adjusted with an external resistor box and could be set to 2.94 or 3.14ÊkV, when the accelerating potential was 20ÊkV, with proportionately lower bias at lower accelerating potentials. A more detailed discussion of the detector assembly is provided in chapter 8. The overall electrical diagram of the instrument is given in figure 2.8.

2 . 4 A Commercial Instrument: The Voyager A small number of experiments were performed on the Voyager DE Ð STR (Perseptive Biosystems, Framingham, MA, USA). This was a delayed extraction MALDI-TOF mass spectrometer, capable of operating in either linear or reflectron mode. The flight path was 2.0Êm in linear mode and 3.0Êm in reflectron mode. A schematic diagram of the instrument, based upon figure 1-19 from the instrument’s user manual [193], is given in figure 2.9. The sample was desorbed and ionised with a 337Ênm nitrogen laser, and after a delay of the order of hundreds of ns, the ions were accelerated down the flight tube by a 25ÊkV pulse applied to the sample holder. The ions then passed through a tuning grid, which was held at a small percentage of the total pulsed potential, and then a grounded grid at the entrance to the drift tube. In linear mode the ions were transferred directly to a detector, while in reflectron mode they passed through an ion mirror prior to registering on a detector. TOF transients were recorded on an integrating recorder, the Tektronix TDS540C (Beaverton, Oregon, USA), a 500ÊMHz bandwidth, 2ÊGS/s oscilloscope. The entire process was controlled by a Pentium computer running under Windows NT 4.0 with proprietary software developed by Perseptive Biosystems.

2 . 5 Sample Preparation Sample preparation is critical in MALDI mass spectrometry. Well prepared samples give improved spot to spot reproducibility and larger analyte ion yields, whilst poorly prepared samples give little or no analyte ion signal. In addition, accurate characterisation of a new instrument requires consistent sample preparation techniques, for results from samples prepared on different days to be usefully compared. The main sample preparation

variable voltage grid attenuator reflector laser prism detector ion mirror sample plate guide wire main source chamber

video camera grounded grid linear sample timed ion aperture (grounded) detector loading selector chamber collision cell (optional)

ion path in ion mirror laser path

Figure 2.9: Schematic of the Voyager-DE STR mass spectrometer. techniques used will be discussed in this section and the development of these techniques will be discussed in chapter 4. Once developed, the same sample preparation techniques were used to prepare samples for both mass spectrometers, since they utilised the same sample probe and laser.

2.5.1 Matrices, Analytes and Solvents Matrix and analyte samples were principally chosen to test the two instruments constructed during the course of the project. A number of factors were used when choosing the samples, including: (a) proven ability to generate positive ions, typically by appearing in published spectra, (b) ease of availability, (c) ensuring a wide mass range was covered, and (d) cost. A list of the samples and solvents used is given in table 2.3, together with details of the suppliers.

All substances were used as obtained, without further purification, except for water. Purified water (“MilliQ water”) was obtained by purifying municipal water with a MilliQ system (Waters Corporation, Milford, MA, USA). The MilliQ system used provided a very low level of impurities, by passing the water through cartridges containing fine filters, ion exchange resins and activated carbon.

2.5.2 Cleaning of Sample Slides Before samples were deposited the sample slides were cleaned. Cleaning was accomplished in the following manner: first the soiled slides were washed with MilliQ water and wiped dry; then the sample slides were cleaned with an abrasive metal polish, Silvo (Reckitt and Coleman, Ermington, NSW, Australia) to remove any contaminated surface; and finally the slides were rinsed first with water, then with isopropanol, to remove any residue and wiped dry with lint free tissues.

2.5.3 Dried Droplet Sample Preparation Most samples used in this project were prepared by the dried droplet method. Solid samples had to be solvated, as an initial step. Peptide samples and their matrix components were dissolved in 1:1 (v/v) acetonitrile and 0.5Ê% aqueous trifluoroacetic acid. DNA samples and matrix compounds were dissolved in 1:1 (v/v) acetonitrile and water. 5,10,15,20-tetraphenyl-21H,23H-porphine (TPP) was dissolved in 1,2-

40 Table 2.3: List of chemical substances used during the project substance molecular weight supplier matrices: α-cyano-4-hydroxycinnamic acid 189.17 Aldrich 2,5-Dihydroxybenzoic acid (gentisic acid) 154.12 Sigma 3,5-dimethoxy-4-hydroxycinnamic acid 224.21 Aldrich 3-hydroxypicolinic acid 139.11 Aldrich ammonium citrate 226.19 donated picolinic acid 123.11 Aldrich analytes: 5,10,15,20-tetraphenyl-21H,23H-porphine 614.75 donated DNA oligomers (synthetic) varied GeneWorks fullerene standard (mixture of C60/C70) 720.67/840.78 donated gramicidin S 1214.4 donated insulin (bovine pancreas) 5733.6 Sigma insulin oxidised chain A (bovine pancreas) 2531.6 Sigma insulin oxidised chain B (bovine pancreas) 3496.0 Fluka melittin (bee venom) 2846.5 Sigma myoglobin (horse heart) 16951.5 Sigma polyethylene glycol 1000 varied donated ribonuclease (bovine) 13682.3 Biotech substance P 1348.7 Sigma synthetic peptides varied donated ubiquitin (bovine red blood cell) 8564.9 Sigma solvents: 1,2-dichloroethane 98.95 May & Baker acetonitrile 41.05 Rhône diethylether 74.12 Rhône isopropanol 66.10 APS MilliQ purified water 18.02 Waters toluene 92.14 May & Baker trifluoroacetic acid 114.02 Aldrich

Supplier key: Aldrich Aldrich Chemical Co., P.O. Box 355, Milwakee, WI, USA APS APS Ajax Finechem, 9 Short St, Auburn, NSW, Australia donated The sample was given to us for free by other researchers. Fluka Fluka Chime AG CH-9471, Buchs. Geneworks GeneWorks Pty Ltd, P.O. Box 11 Rundle Mall, Adelaide, SA, Australia May & Baker May & Baker, West Footscray, VIC, Australia Rhône Rhône-Poulenc Chemicals, Clayton South, VIC, Australia Sigma Sigma Chemical Co, P.O. Box 14508, St Louis, MO, USA Waters Waters Corporation, Milford, MA, USA dichloroethane and C60/C70 fullerene mixture was dissolved in toluene. Polyethylene glycol samples were dissolved in 1:1 (v/v) isopropanol and water. Matrix samples were made at a concentration of 10 or 20Êmg/mL. Analytes were made to a concentration of between 1 and 4Êmg/mL, with higher concentrations used for more massive species, to provide a smaller spread of mole ratios.

Samples were deposited on the probe in a layered fashion. First, an appropriate volume of matrix was applied to the sample slide with a micropipette (1 to 10ÊµL), and evaporated to dryness under an infra-red heat lamp. The required volume of matrix was calculated to ensure the molar ratio of matrix to analyte was 1000:1 for most experiments. If the required volume was over 10ÊµL it was divided into aliquots smaller than 10ÊµL. Each aliquot was layered on top after the previous one had been evaporated. This was done to prevent excessive spreading of sample spots. After this, 2ÊµL of the analyte was placed over the matrix spot and left to evaporate under the infra red lamp. Once this procedure was completed the sample could be introduced into the mass spectrometer, or stored elsewhere for later analysis.

Up to eight spots were routinely placed upon a sample slide using this method. Multiple spots on a single slide were prepared simultaneously, to minimise exposure of sample to the infra red lamp.

2.5.4 Electrospray Sample Deposition Electrospray sample deposition apparatus was constructed, then used to prepare a number of low molecular weight samples. Figure 2.10 is a schematic diagram of the spray region and figure 2.11 is a photograph of the entire apparatus. It consisted of a Harvard Apparatus 22 Syringe Pump (South Natick, MA, USA) which drove a 250ÊµL gastight syringe (Hamilton, Reno, Nevada, USA) at flow rates from 0.5 to 20.0ÊµL a minute. The solution flowed through a PEEK transfer line into the grounded box, and through the spray needle, as illustrated in figures 2.10 and 2.11. Spray needles were obtained from SGE (Ringwood, Vic, Australia), and included both bevelled and flat tipped removable needles for 25ÊµL syringes. Solution was sprayed directly onto a sample slide, located underneath the spray needle on a turntable that rotated at 60 revolutions per minute. The distance between the spray needle and sample could be adjusted between 4 and 15Êmm, by raising or lowering the spray needle.

sample feedline

high voltage contact spray needle

deposited spray sample insulating box grounded (spring) rotation contact

shelf

motor and gearbox

Figure 2.10: Schematic of the electrospray region of the spray deposition apparatus.

Figure 2.11: Photograph of the electrospray deposition apparatus. The potential necessary for the electrospray process was provided by a 10ÊkV Wallis (Worthing, UK) power supply. Its output was applied to the spray needle, and the sample slide was connected to ground. During spraying experiments the voltage was adjusted until a stable Taylor cone was observed, providing the necessary droplet size. Occasionally experiments were performed at a higher potential. Varying amounts of sample were deposited by this method.

42 Chapter 3: Computerised Instrument Control and

Analysis

Time-of-flight mass spectrometers require precise control of timing and voltages. They are also able to rapidly generate large amounts of data, requiring fast data acquisition and making efficient data processing desirable. Additionally, when building prototype instruments, a flexible system is needed. A customisable computer based virtual instrumentation system, such as that provided by a combination of the LabVIEW¨ programming environment with suitable hardware, is able to satisfy these requirements. This chapter details the LabVIEW based system created for this project. Fully functioning versions of the programs developed for use with the instruments described below are included on a CDROM, attached inside the rear cover of this thesis. They can be loaded and viewed with LabVIEW version 5, although use of some functions requires installation of appropriate computer boards.

3 . 1 A Brief Introduction to LabVIEW and Virtual Instruments LabVIEW is a program development environment produced by National Instruments (Austin, Texas, USA) [194]. It utilises a graphical programming language, G, where the programmer creates compilable programs by ‘wiring’ icons together, in contrast with conventional programming languages, where the software is generated from typed lines of text.

The accessible portion of a LabVIEW program consists of two main parts, a front panel and a block diagram. Users interact with a program through the front panel. Its appearance is similar to the panel of a real instrument, containing screen representations of controls (for instance knobs, sliders and buttons) used to input data and indicators (for instance LEDs, charts and graphs or thermometers) for presenting results. The block diagram contains various icons that are wired together to create the source code, which is hidden when programs are being executed, like the components and connections covered by the case of a real instrument. The programs written with LabVIEW emulate real instruments in appearance and function, so they are called virtual instruments (VIs). An example of a simple VI used to multiply two numbers is given in figure 3.1.

Two additional features of LabVIEW make it particularly suitable for instrument development. Firstly, it contains specialist libraries for data acquisition, instrument

Figure 3.1: The front panel (a) and block diagram (b) of a simple program that multiplies two numbers and displays the result. Data is entered on the front panel and the result is displayed there. The wiring and icons on the block diagram are used to create the code. control and data manipulation, in addition to general programming tools, which can be pasted directly into VIs. Secondly, it adheres to the concept of modular programming [195], allowing VIs to be placed within each other, in a hierarchical structure. A VI with its front panel visible to the user is known as a top level VI, while a VI placed within another is known as a sub VI. This greatly simplifies VI development by allowing commonly used features to be created once as a sub VI and then pasted into higher level VIs whenever required, reducing the complexity of block diagrams and decreasing overall program size. Combining this modular programming environment with appropriate hardware allows the creation of flexible, customised instrument control and data acquisition systems for prototype instruments, in a relatively small amount of time.

3 . 2 Computer Hardware It was important to know about the computer hardware used for instrument control and analysis, since this limited what could be done with the software. Computer hardware for each of the two oa-TOF instruments was different, and is briefly explained in the following paragraphs and summarised in tableÊ3.1.

The linear instrument was run from a Mac IIfx computer (Apple Computer, Cupertino, CA, USA), with 20ÊMB of RAM and LabVIEW version 3.1 installed. Two Nu-Bus boards, a multifunction board and a GPIB (IEEE-488) board, were obtained from National Instruments and installed inside the computer. The multifunction NB-MIO-16H- 9 board, incorporating analogue to digital conversion (ADC), digital to analogue conversion (DAC) and timing circuits, was used to control and monitor the power supply and timing requirements, whilst the GPIB board (NB-DMA-2800) was used to download TOF spectra to the computer.

The reflecting geometry instrument required a more powerful computer, to run LabVIEW 5.0 and allow additional real-time instrument control. A Power Macintosh 8100/110AV was used for this task. It was upgraded with a 216 MHz G3 CPU accelerator board (MAXpowr G3 PDS, Newer Technologies, Wichita, Kansas, USA) and had 72ÊMB of RAM installed. A similar multifunction MIO board (a NB-MIO-16XH-42) to that installed in the Mac IIfx was used for timing and power supply control. Important specifications (in this project) for both the MIO boards are contained in table 3.1. The same GPIB board was used as in the Mac IIfx.

44 Table 3.1: Computer hardware used for each instrument, including multifunction data acquisition board specifications.

linear oa-TOF reflectron oa-TOF computer used Mac IIfx Power Macintosh 8100

GPIB board NB-DMA-2800 NB-DMA-2800 multi function board NB-MIO-16H-9 NB-MIO-16XH-42

Analogue output channels 2 2 resolution 12 bits 12 bits range 0 Ð 10 V 0 Ð 10 V

Analogue input channels 8 differential 8 differential resolution 12 bits 16 bits range 0 Ð 10 V 0 Ð 10 V rate 100 kbyte/s 24 kbyte/s

Timing channels 3 3 resolution 16 bit 16 bit clock 1 MHz 1 MHz 3 . 3 Instrument Control Two main types of instrument control VIs have been created. Firstly, there was the program used to control timing of the laser and push-out-pulse generator, and secondly, there were the power supplies’ VIs.

3.3.1 Computer Controlled Laser Triggering and oa Pulse The main role of the Fire Laser VI was to co-ordinate the timing of laser firing and the application of the push out pulse. Exact timing of the delay between these two events was critical, since the push out pulse had to be triggered when the masses of interest had reached the oa. This was especially the case for the linear instrument, which had a relatively long desorption region and narrow oa. In addition to providing an accurate push out pulse delay, the VI allowed the laser to be conveniently triggered from the computer, and ensured that there was enough time between laser shots for even the heaviest ions to reach the detector, before the push out pulse generator was triggered again. The front panel of the fire laser program is given in figure 3.2, with the same VI used for both linear and reflecting instruments.

Operating the Fire Laser VI consisted of a number of steps. First the ‘Prefiring Sequence Controls’ had to be set, whilst the VI was running (the running arrow was displayed, as in the top left corner of figure 3.2). This involved user selection of the number of shots required with the ‘No. shots’ control (allowed range 1 – 1,000), whether a user delay or a four shot sequence was desired and the required pulse delay (allowed 3 to 200Êµs). The ‘FIRE’ button was then pressed in the ‘Laser Firing Controls.’ The program checked once a second to see if the ‘FIRE’ button had been pressed and, if it had, the sequence was allowed to complete, with the ‘Shot’ indicator counting the number of shots fired and the blue light above it flashing each time the laser was triggered. For each count a TTL (transistor-transistor logic) pulse was sent to trigger the laser and another to trigger the push out pulse generator after the set delay. If the firing sequence had to be stopped before the full number of shots had been fired the ‘ABORT’ button was pressed. This stopped the laser firing after the laser and push out pulse had been triggered one more time, a helpful feature, used to stop acquisition if the sample was fully depleted in mid sequence.

When user delay was selected, the delay for every shot was as set with the ‘User OA- pulse delay’ control. If a four shot sequence was selected, the first delay time was that set by the ‘User OA-pulse delay’ control, followed by 1.5, 2.0 and 2.5 times the delay listed in that control. After the first four shot sequence it was repeated until the ‘shot’ counter

Figure 3.2: Front panel of the Fire Laser Vi (running), used to trigger the laser, followed by the push out pulse generator after the set delay. reached the nearest multiple of four, rounding down, to the number set in the ‘No. shots’ control.

3.3.2 Control and Monitoring of Power Supplies In contrast with the Fire Laser VI, concerned with accurate timing of pulsed signals to ensure good instrument performance, power supply VI’s were concerned with ensuring and monitoring the stability of the power supplies of the TOF. Accurate and precise results in a TOF mass spectrometer require stable power supplies, since every ppm error in the total accelerating potential translates to an equivalent ppm mass error in the resulting spectrum. A detailed explanation of this relationship is given in chapter 7, in the sections concerning mass accuracy.

For the linear oa-TOF instrument, only the power supply for the push out pulse was connected to the computer. The accelerator potential was provided by an older model supply, unable to be interfaced to our computer. In the reflectron instrument all three power supplies capable of affecting mass accuracy and resolution were interfaced to the computer. The base program used in both cases was similar, since the top level VI utilised the modular capabilities of LabVIEW. A sub VI, called High Voltage, was created to control and monitor each individual supply. It had to be given the maximum voltage and current ratings of the power supply, together with the desired potential setting. Once this was done, it could set the required potential using a 0 Ð 10ÊV DAC (digital to analogue converter) output and monitor actual potential and current with two 0 Ð 10ÊV ADC (analogue to digital converter) input channels on the MIO board.4 The 12 bit DACs, used to set the potential, allowed the actual mirror or push out pulse voltage to be varied in increments of 0.37ÊV and accelerator potential in increments of 6.1ÊV. When the set and actual voltages differed by more than 0.05% of the full range (slightly larger than two least significant bits for a 12 bit DAC or ADC) an error lamp was triggered to alert the user to the discrepancy. Where only monitoring of potential and current was required a different sub VI, Meas High Voltage, was used. This VI had to be informed of the maximum voltage and current settings and was then able to monitor it through two ADC channels, in the same manner as the High Voltage VI.

The top level VI used for control and monitoring of power supplies was the Set Potentials VI. Figure 3.3 represents its front panel for the reflectron instrument, when switched off. The linear machine used the same program, without the sections for the accelerator and

4 All computer controlled power supplies could be set with a 0 Ð 10 V signal to give a high voltage output scaling from 0 to their full range potential (+1.50 kV or Ð25.0 kV for the supplies used). Their monitor lines would output signals from 0 Ð 10 V, reflecting where the actual output potential and current were within the power supplies’ full range. 46

Figure 3.3: Front panel of the Set Potentials VI (not running) used to control and/or monitor the three mass analyser power units. mirror backplate supplies. Two copies of the High Voltage sub-VI were used, one for the push out pulse and one for the accelerator supply. Only a single copy of the Meas High Voltage sub VI was required, for the mirror backplate. It would have been preferable to have set all three supplies from the computer, but since only two analogue outputs were provided on the MIO board, the supply that needed to be changed the least in our experiments, the mirror, was monitored with the computer and adjusted manually. The two power supplies that were both set and monitored were treated slightly differently. The push out pulse supply could be almost instantly set to any potential within its allowed range (0 to 1,200ÊV) with resolution determined by the 12 bit DAC (to 0.37ÊV). A different approach had to be taken for the accelerator, since allowing it to be rapidly set to any potential within its range (0 to Ð20.0ÊkV) could have resulted in damage to the detector or other sensitive components. Instead the program allowed users to set it to any potential in the allowed range, with 100ÊV resolution, and ramped the potential by 100ÊV every 0.5Ês, until the desired potential was reached. When experiments were finished the program was switched off with the power switch, ensuring all power supply potentials were set to zero, with the accelerator supply being ramped to zero, if required.

In the Set Potentials VI, error lamps, actual voltage and current monitors were used to determine whether the instrument was performing as designed in the short term. However, they gave no indication of the long term stability of the power supplies, essential for mass accuracy in TOF mass spectrometers and specifically required in some experiments. Capability to do this was attained by modifying the Set Potentials VI into the Plot Potentials VI, shown in figure 3.4. The same sub VIs were used to control and monitor the power supplies. The main difference was that it was able to plot the behaviour of the voltages over time, and allowed the user to save the results at the end of the acquisition, by interacting with a dialogue box. It also did not set the power supplies to zero when switched off, or ramp the accelerator supply during changing, actions which interfered with its monitoring function. Due to these modifications the Plot Potentials VI was only used for long term monitoring experiments and not routine operations, since it did not include safety features to prevent rapid changes to the accelerator supply.

In figure 3.4 the monitored voltages for the three power supplies are plotted over a 320Ês interval, in a chart that plots potential against scan. The top trace (POP) is for the push out pulse supply, the second (Accel) gives the result for the accelerator and the third (BP) represents the mirror backplate potential. The final trace (bkgrnd) gives the result for a blank (grounded) trace, scaled to the same range as the push out pulse and mirror backplate (0 to 1500ÊV at the high voltage output represents 0 to 10 V at the measuring device) used to determine the accuracy and precision of the ADC obtaining the measurements. The user was able to select the acquisition rate in readings/minute, from 1 47

Figure 3.4: Front panel of the Plot Potentials Vi (running), showing plots of potentials obtained after acquiring readings from three supplies and a control channel for 5 minutes and 20 seconds. to 60. If the file was saved this information was included, together with the start time for the acquisitions and the raw data, in delimited text format, allowing the data to be plotted in other packages as a function of time. The maximum amount of data collected was limited by the memory of the computer. This did not interfere with our experiments, where we acquired up to 72,000 data points without difficulty.

3 . 4 Data Acquisition and Analysis LabVIEW VIs were an essential part of the data acquisition and analysis process for the instruments created during this project. Every mass spectrum acquired required the use of at least one these programs. The data acquisition VIs were carefully designed to ensure that original time spectra were saved without any further processing. Any additional processing was saved as a separate file with the analysis VI.

3.4.1 Data Acquisition Time of flight data was downloaded from the digital oscilloscopes to the computer at the end of each acquisition, via a GPIB interface between the computer and the oscilloscope. One VI was created to co-ordinate this function for each oscilloscope, Spectra Control 9450 for the LeCroy 9450 and Spectra Control for the LeCroy 9384. Since both programs performed a similar function their front panels were almost identical, with the main difference between them related to the sub VIs used to perform GPIB operations, provided by National Instruments with the LabVIEW program. These sub VIs were pasted in without modification, and one such sub VI was required for each data acquisition VI, to read the TOF trace from the oscilloscope.

The front panel of the Spectra Control VI is given in figure 3.5. The rest of this section will explain its operation, since it was more extensively used than the program for the other oscilloscope. The VI had six main functions, executable by clicking on the appropriate button in the ‘Program Controls’ section of the panel. Once the VI started running, it would continue to run until the power button was switched off. Controls were set to their desired settings and the appropriate function button, labelled F1 to F6, was selected, either by clicking with the mouse or pressing the corresponding function key on the keyboard, to obtain the desired results.

When the ‘Read’ button was selected, the trace listed in the ‘Trace to read/save’ control was downloaded from the oscilloscope and displayed on the graph display, as either a time or mass spectrum. It was necessary to choose whether to download as a mass or time spectrum before pressing the read button, using the mass/time button located at the top left of the graph. In figureÊ3.5 a mass spectrum has been downloaded. If a time

Figure 3.5: Front panel of the Spectra Control VI (running) used to download data from the LeCroy 9384 oscilloscope and provide preliminary analysis. A mass calibrated melittin spectrum is shown in the graph display, together with calibration information in the 'Mass Spectrometer Constants' box. spectrum had been displayed the title would have read ‘Time Spectrum’ and the axis label ‘time (s),’ instead of ‘Mass Spectrum’ and ‘mass (m/z)’ respectively. The number of points in the downloaded spectrum was displayed in the ‘Number of points in spectrum’ indicator.

When ‘Save’ was pressed the selected trace was downloaded from the oscilloscope and saved to disk as a time spectrum, with the user prompted to select a folder and name the file in a dialogue box. Files were saved with the start time (six significant figures) and time between data points (three significant figures) in ASCII format, placed before the amplitude values in signed word binary format (16 bit). This required the minimum amount of storage space. The largest number of points that could be saved per file in this manner was 100,000. The files were not allowed to be saved directly from the oscilloscope to disk as calibrated mass spectra, since incorrect calibration would render such files useless. Saved files could be reloaded by pressing ‘Load’ and displayed as either mass or time spectra, depending on the setting of the ‘mass/time’ button.

Some simple processing facilities were included in the VI, to allow rapid analysis of experimental data. The first of these was the ability to display mass spectra, when the ‘mass/time’ button was set to ‘mass’, by converting time data to mass data with the calibration relationship given in equation 1.10, utilising the flight times for two known m/z species to calculate the calibration constants. The user was required to provide correct time information in seconds for the calibrating species, placing time data in ‘P1-t(s)’ and ‘P2-t(s)’ controls, and the corresponding m/z in ‘P1-m/z’ and ‘P2-m/z’ controls. If desired the required time values could be obtained from a displayed time spectrum, by placing the two cursors on the centre of the peaks corresponding to the calibration masses and pressing the ‘Select’ button. This transferred the time values of the cursor locations into the appropriate controls, ‘P1–t(s)’ and ‘P2–t(s)’, so only the corresponding masses had to be entered manually.

The next set of data processing features were linked to the ‘Integrate’ button and location of the two graph cursors. When the ‘Integrate’ button was selected the ‘Centroid,’ ‘Peak area,’ difference in amplitude values (‘ y’), difference in mass or time values (‘ x’) and number of data points between the cursors (‘No. data points’) was calculated. This allowed the calculation of important peak parameters, depending on the location of the cursors, essential in analysing a TOF spectrum.

The final feature included in the program was filtering. When the ‘Filter’ button was selected a low pass Chebyshev or Butterworth filter was applied, in order to reduce the level of high frequency noise in the spectrum. Sampling frequency, order and high cutoff 49 were typically set as displayed in the ‘Filter and Zero’ box to obtain an optimal result for isotopically resolved data in the spectra typically obtained.

3.4.2 Analysis It became clear that adding further processing functions to the data acquisition VIs was an inefficient procedure, since (i) each additional feature reduced the computer memory available for downloading and saving the raw data, the primary role of the VIs; and (ii) new features had to be separately added to each data acquisition program, requiring additional programmer time. Instead, it was more efficient to create a separate VI for analysis, that could reload data files saved with either TOF data acquisition VI, since they both saved files with the same format. This also meant that if a new acquisition device was used, for instance a time to digital recorder, the same analysis VI could be used, providing files were saved in the same format.

The ToFAn VI was created to fulfil this role. Its front panel (figure 3.6) appears similar to that of the data acquisition VIs, with some different features. Data analysis was simplified for the user by grouping activation buttons with relevant controls and indicators. Functions available included loading files, saving to a different format, scaling amplitude values, obtaining peak parameters, filtering and reducing the number of data points. Like all the other VIs created for this project, the program would continue to run until the virtual power button was switched to the off position.

Spectra were loaded as either time or mass spectra with the ‘LoadTOF’ or ‘LoadMS’ buttons. When loaded as a mass spectrum the numbers displayed in the ‘Mass Spectrometer Constants’ box were used to calibrate the data, in the same manner as the data acquisition VIs. The spectrum displayed could then be saved in delimited text format, with a column of mass or time values and another with corresponding amplitude values, by selecting ‘Save,’ then choosing a location and file name in the dialogue box that appeared. This saved the trace as displayed, including any processing performed in this program. The suffix ‘txt’ was placed on the end of file names by the VI, to differentiate from binary files saved directly from the oscilloscope. The file path of the original file was appended to the end of the file for auditing purposes and additional comments could be included by entering them in the ‘notes’ box before saving. Data files saved in this manner took approximately ten times the disk space of the original binary files, but could be exported to many other applications.

Besides mass calibration, four additional processing options were included. The ‘Filter’ function was identical to that described above under data acquisition and will not be

Figure 3.6: Front panel of the ToFAn VI (running) used to process, analyse and export saved data files. A mass spectrum of TPP is displayed, after being zeroed, filtered with a 100 MHz Chebyshev lowpass filter and recomputed to the original 0.2V/div verticle scale on the oscilloscope. discussed further here. The ‘Integrate’ function was very similar to the equivalent feature in the data acquisition VIs, the only difference being that it also calculated the sampling rate of the spectrum (‘Sampling freq’) required for filtering, if a time spectrum was displayed. ‘Scale’ and ‘Decimate’ were new features.

The scaling function was used to either zero the baseline of the trace, or rescale the y Ð axis to correspond to the original display on the oscilloscope, depending how the DSO/zero button was set. When rescaling the y axis, the correct oscilloscope vertical scale setting (in volts per division) had to be selected with the ‘DSO scale’ control and the probe attenuator value set with the ‘Probe Atten.’ control. The zero baseline function offset the trace by subtracting the average of the first twenty points’ amplitude value from each data point.

Decimate was a simple function, which removed either every second or nine out of ten data points, depending on the setting of the 50%/90% button when ‘Decimate’ was selected. It was used to reduce the number of data points in over sampled data, where each mass peak contained hundreds of data points. Whether data was over sampled could be determined from the number of data points indicator (‘No. data points’) in the integration box. The decimate function could be executed a number of times, until the required level of data reduction was obtained.

The combination of the data acquisition and analysis VIs provided the rapid acquisition, analysis and spectral exporting required for our data. Other programs were used to fulfil further processing and analysis requirements.

3 . 5 Other Programs and Macros A number of other useful computer based tools were created for data analysis and instrument simulation, utilizing either IGOR Pro Version 3 (Wavemetrics, Lake Oswego, OR, USA) or Microsoft Excel 98 (Microsoft, Redmond, WA, USA). The most important of these were the IGOR macros used to load and display saved mass spectra and Excel files written to simulate the instrument.

Before the ToFAn VI was written, spectra were processed in IGOR Pro, an integrated program for visualising, analysing and presenting scientific data. Loading and processing TOF files in IGOR required a number of steps, so macros were written to automate the process. One macro loaded the files and generated TOF spectra and a second converted

51 TOF spectra into mass spectra, after prompting the user for mass and flight times for two calibration points. The text of these macros is included in appendix 2. 5

For some experiments the results had to be fitted to a Boltzmann-Maxwell function. This feature was not present in any of the graphing packages available in our laboratory. IGOR allowed fitting to user defined functions, so a procedure was written to provide Boltzmann-Maxwell curve fitting and a copy of the text of this function is included in the Macros and Functions appendix.

Construction of our instruments and thorough evaluation relied upon simulations in addition to experimental results. Many simulations were performed with I-Opt and SIMTOF software, written by other members of the research group and described in Chapter 1. Others utilised simpler Microsoft Excel spreadsheets, which allowed simulation of individual parameters. These are included on the CDROM in the folder ‘Excel Simulations’ and will be referred to in later chapters where their results were used.

5 After the ToFAn VI was created these macros were little used, since it was simpler to create mass spectra using the the VI and load the text files directly with IGOR. IGOR was still used to export graphics and provide other forms of processing. 52 Chapter 4: Improving Performance of the Linear

MALDI-oa-TOF

4 . 1 Importance of Improving Performance A linear MALDI-oa-TOF instrument had been constructed by Mlynski and Guilhaus just prior to the commencement of this project, with early experiments determining its resolution (2,800 - 3,600) and mass accuracy (90Êppm) for m/z 614 to 1351 [166]. A detailed description of this instrument was given in Chapter 2. The role of experiments described in this chapter was to improve the performance of this mass spectrometer, to allow important desorption velocity experiments to be performed, which will be described in Chapter 5.

Previous work by Mlynski had indicated that the reproducibility of samples was poor, both within a single sample spot and between different sample spots. A more detailed characterization of the instrument required better sample preparation techniques, since many experiments would have to be performed to assess the effect on ion signals of varying independent instrument parameters. These experiments could not be successfully carried out while sample inhomogeneity caused large variations in ion signals, as the resulting spread in measurements would introduce significant uncertainties, making it difficult to measure small changes in signals due the parameters under investigation. Poorly prepared samples also reduced the apparent sensitivity of the instrument. Thus many early experiments were designed to improve sample preparation. Other parameters impacting on instrument performance, such as the effects of laser power and digitiser rate, were also investigated, to optimise results from the mass spectrometer.

4 . 2 Optimising Dried Droplet Sample Preparation 4.2.1 Dried Droplet Sample Preparation Methods Samples of TPP and gramicidin S were prepared using (i) the dried droplet method discussed in chapter 2, (ii) a modified version of the method of Mlynski and Guilhaus [166] utilizing ambient drying conditions and (iii) a variant of the acetone based rapid drying technique of Vorm et al [131, 132]. In (i) the matrix was present in a 1,000 - 2,000 fold molar excess and matrix was “layered” on the probe before the analyte, with rapid drying under an IR lamp between applications. Solutions were made with solvents as described in chapter 2 (section 2.5). The same solutions and mole ratios were used for method (ii), but deposited and dried in a different manner. Gramicidin S and DHB matrix

53 were deposited together on the sample slide and left to dry under ambient conditions. The solvent used for TPP was not miscible with the matrix solution, so for those compounds the analyte was deposited first and left to dry, following which the matrix was layered on top. Drying was performed under ambient conditions.

For method (iii) DHB matrix and gramicidin S were made up in separate solutions of acetone and 2% purified water. TPP was made in 1,2-dichloroethane. A saturated solution of the matrix was made. Analytes were made at an approximate concentration of 1 mg / mL. 0.5 µL of the matrix solution was applied to the probe. After it had dried 1 µL of analyte was layered on top and dried rapidly, under the IR lamp if required.

Samples were analysed in the mass spectrometer with the same instrument conditions for all samples. Laser attenuator was set to 0.4, the push out pulse to 88ÊV, grid 2 to Ð100ÊV and the accelerator to 2,898ÊV. Spectra were recorded with the 400ÊMS/s oscilloscope (LeCroy 9450).

4.2.2 Appearance of Target and Spectral Quality for Dried Droplet Samples The deposited samples prepared with these three methods were visibly different, as can be seen in figure 4.1. The rapid drying technique gave a surface largely covered by thick clusters of DHB crystals, the dried droplet technique gave a more uniform coverage of smaller crystals and the ambient drying technique gave large spindle shaped crystals characteristic of DHB around the edges of the sample spot, with few crystals elsewhere. TPP crystals were quite dark, so they were readily distinguished from the matrix. In the more rapid drying techniques they were fairly evenly distributed, while ambient drying left them mainly around the edges of the large DHB crystals.

The dried droplet technique gave fairly reproducible results. Reasonable spectra were obtained for both analytes. Figure 4.2 shows a sample spectrum of TPP obtained from this sample preparation technique. Gramicidin S spectra were not quite as good.

The best signals for gramicidin S were obtained with ambient drying, with spectra such as that given in figure 4.3. These results were obtained from the large crystals. The remainder of the sample spot gave poor results, with the signal obtained (if any) diminishing rapidly after less than five shots. Unfortunately, it was difficult to ensure that the large crystals would coincide with locations on the slide where the laser would strike, since sites of crystal nucleation and growth were not controlled. Thus only some samples gave good signals. For TPP, however, ambient drying provided the poorest spectra, when compared to the other two sample preparation methods.

(i)

Dried droplet, Gramicidin S Dried droplet, TPP

(ii)

Ambient, Gramicidin S Ambient, TPP

(iii)

Acetone, Gramicidin S Acetone, TPP

5 mm

Figure 4.1: TPP and gramicidin S sample spots prepared by methods (i) to (iii), as explained in the text. TPP spots were brown. In black and white images there was little contrast between this and the sample slides. Contrast was better for gramicidin S, which gave white crystals.

[M+H]+

[M]+

-6 52.7 52.8 52.9 53.0x10 TOF (s)

Figure 4.2: A spectrum of TPP in DHB. The sample was prepared with the dried droplet technique, where the solvent was evaporated under a heat lamp.

Figure 4.3: A spectrum of gramicidin S, based upon figure 4 of Mlynski and Guilhaus [166]. The sample was prepared with the ambient drying technique and this spectrum was obtained from one of the large crystals near the edge of the spot. Results obtained with the acetone based rapid drying technique were disappointing. No spectra were obtained for gramicidin S with reasonable signal to noise. Results were a little better for TPP, where signal to noise improved, but the same results were obtained for locations on the sample that had no matrix for this compound. This is because TPP absorbs the 337Ênm radiation, in effect acting as its own matrix. Thus it is able to give a strong signal when the matrix is present in a large excess (MALDI) and when present without a matrix (LDMS).

The overall results have been qualitatively summarised in table 4.1. Method (i), the dried droplet approach with drying under a heat lamp, was deemed to be the most generally useful of these techniques for our experiments, since both samples gave reasonable spectra over most of the accessible sample spot. This was not achieved with either of the other two techniques. This improved reproducibility was probably associated with the appearance of the spots, since the other techniques gave samples which appeared less homogeneous on the scale of laser irradiance, approximately 0.1mm x 0.1mm. Thus the dried droplet method outlined in section 2.5.3 was used for all other experiments where small sample spots were required.

4 . 3 Electrospray Deposition Sample Preparation Some experiments performed with this instrument required the acquisition of fifty or more separate spectra to understand certain aspects of the instrument. It was preferable to perform these experiments with a single fairly homogeneous sample, to reduce the effects of differences in sample preparation on the results. Samples prepared by the dried droplet approach did not provide enough sampling points for this, so an electrospray deposition (ESD) method was developed. Electrospray deposition had been successfully used to prepare samples for plasma desorption mass spectrometry [196]. Additionally, while these experiments were performed a paper by Axelsson and colleagues was published showing that ESD provided enhanced reproducibility and signal intensity in another instrument [133].

4.3.1 Electrospray Deposition Procedures A detailed description of the ESD apparatus was given in section 2.5.4 and it will not be repeated here. Several different solutions were made for these investigations, with both aqueous and more volatile solvents. The analyte and matrix materials were dissolved in the same solvent and deposited together. The composition of the solutions used is given in table 4.2.

55 Table 4.1: Summary of the quality and reproducibility of spectra obtained with different sample drying techniques.

Compound Dried Droplet Ambient Drying Acetone Rapid Drying (method (i)) (method (ii)) (method (iii)) quality reprod. quality reprod. quality reprod. TPP good yes fair no fair yes gramicidin S good yes very good no poor not relevant

Table 4.2: Samples used for ESD

sample analyte DHB (matrix) solvent name conc (mg/mL) conc (mg/mL) (I) TPP 0.5 10 1:1 1,2- dichloroethane:diethylether (II) gramicidin S 0.5 20 1:1 acetonitrile: water + 0.1% TFA (III) gramicidin S 0.5 20 ethanol Initial experiments were performed to investigate the shape of the sample spot. For these experiments the turntable was not switched on and solution (I) was sprayed directly onto a piece of grounded stainless steel shim located 5.3mm from the spray needle. The potential applied to the needle was varied and the spraying process observed. A flow rate of 1µL/min was used. Solution (I) was used since TPP gave coloured crystals that were easy to see. The spray pattern on the shim was photographed with a digital camera, the Dage CCD 100 (Dage-MTI Inc., Michigan City, IN, USA).

Additional experiments involved depositing solutions (II) and (III) onto sample slides. In these experiments the turntable was rotated during deposition, to ensure samples were deposited in an annulus, covering all of the surface available to be analysed. Spray needle potential was set to the lowest value that gave a stable spray. The samples were then analysed with the mass spectrometer, operated as described in chapter 2, to determine if spectra could be obtained from all accessible locations on the sample slide.

4.3.2 Appearance of Target and Quality of Spectra Obtained with Electrospray Deposition Figure 4.4 shows the results of depositing TPP in DHB on the stainless steel shim at two critical potentials. Crystals obtained were very small and not resolved by the CCD camera, which had a pixel resolution of approximately 10Êµm. The small spot at the top of the figure (2.4 x 2.0Êmm) was obtained when a potential of 2.0ÊkV was applied to the spray needle. This was the lowest potential at which stable electrospraying occurred. The solution formed a single Taylor cone, with a fine mist (the electrosprayed droplets) emitted from its tip, depositing as indicated. Below this potential large drops formed and fell off the end of the needle. When potential was increased to 2.9ÊkV droplets were leaving from the rim of the needle tip and several small cones were observed. At 3.5ÊkV droplets were leaving so rapidly, presumably due to rim emission, that there was not enough solution to form any cones. The spray pattern at the bottom of figure 4.4 (6.3 x 6.3Êmm) was obtained at 3.5ÊkV. It was decided to operate with a spray potential that created a single cone, since this caused electrospraying at a single predictable location. If multiple cone or rim emission occurred during deposition onto a rotating sample slide the sample’s location could not be controlled and crystals were often not in the correct location for MALDI experiments.

Gramicidin S in solution (II) failed to give consistent results by elecrospray deposition. On one occasion an excellent sample was obtained, where every point on the sample gave reasonable mass spectra for over 100 laser shots. On some other days the potential which

(a) 2.0kV 2.0mm

2.4mm

(a) 3.5kV 6.3mm

6.3mm

Figure 4.4: Spray patterns obtained on a piece of metal shim with ESD at (a) 2.0kV and (b) 3.5kV. The spray needle was located 5.3mm above the grounded shim. gave stable spray (over 4ÊkV) also resulted in an electrical discharge, preventing ESD. Sometimes spraying would work, but the sample would fail to dry quickly enough to generate the small crystals that were desired. In short, it appeared that successful ESD of samples dissolved in aqueous acetonitrile was too dependent upon ambient weather conditions, especially humidity and temperature. If humidity was too high the solvent did not evaporate fast enough and the spray potential could cause electrical discharging. It would have been possible to reduce the weather dependence by controlling the atmosphere inside the chamber. This would, however, require changes to the apparatus, so instead it was decided to switch to a more volatile non-aqueous solvent system with a lower surface tension. A more volatile non-aqueous solvent would evaporate quickly, at a rate independent of the humidity, generating the very small crystals desired. The lower surface tension would also allow spraying to occur at a correspondingly lower potential, since formation of cones and droplet emission are based upon the competing effects of surface tension and electrostatic forces exerted on dissolved charge carriers [197]. This provided the added benefit of reducing the risk of discharging.

Solution (III) utilised an appropriate non-aqueous solvent system, ethanol, and gave good samples that avoided the problems encountered with solution (II). A spray potential of 3.4 to 3.6ÊkV was required for stable spraying. Figure 4.5 shows a typical sample obtained with a deposition rate of 2ÊµL/min, deposition time 40 to 50 minutes and an approximately 5mm separation between spray needle and target. Samples deposited by this method gave consistent results all over the target, with upwards of 70 laser shots required to deplete the sample at each location. Solution (I) gave similarly good results for TPP by this method.

Unfortunately, this method of sample preparation had two major disadvantages when compared to dried droplet sample preparation. Firstly, good results required a non- aqueous solvent system in which both matrix and analyte were soluble, limiting the samples that could be prepared. Secondly, far more sample was required. A typical spray deposition experiment deposited at least 100ÊµL of analyte on a single sample slide, at a concentration of 0.5Êmg/mL with an appropriate level of matrix. A further 50ÊµL was required to rinse out and fill the delivery syringe, spray needle and PEEK tubing. The total amount of sample consumed was (approximately) 75Êµg. For dried droplet samples a single spot only required 2ÊµL of 1 to 4Êmg/mL analyte, corresponding to 2 to 8Êµg of analyte. No sample was wasted on rinsing the delivery system, since disposable pipette tips were used. Thus approximately 10 to 40 times as much analyte was consumed to

Figure 4.5: A picture of a sample prepared by ESD, after being analysed in the mass spectrometer. A section of the sample has been expanded to reveal the finer structure. Note the small holes in the ring formed by matrix and analyte crystals, due to ablation by the laser. They are approximately 0.1 x 0.1 mm.

Absorbance

Figure 4.6: The proportion of incident radiation transmitted (T) as a function of absorbance and incident radiation (log plot). prepare a single sample by spray deposition. While it would have been possible to modify the ESD apparatus to require less sample for rinsing, the design of our ion source would still require the deposition of significantly more sample than that required in dried droplet preparations. Because of this, spray deposition was only used to prepare samples for experiments that required a large number of spectra.

4 . 4 Using a Faster Digitiser and the Effects of Laser Power Early experiments were conducted with a 400ÊMS/s oscilloscope. In those experiments each peak in the mass spectrum usually had less than 10 data points, potentially undersampling the signal. This could give rise to poor peakshapes, with ramifications for mass accuracy and resolution. Experiments were conducted with a 4ÊGS/s oscilloscope, to investigate the effects of increasing the sampling rate.

In conventional MALDI-TOF systems it had been observed that reducing laser power to near threshold levels improved resolution [94]. In MALDI oa-TOF the effects of the ion source were largely decoupled from the analyser and thus laser power was expected to have a smaller (if any) effect on resolution. Accordingly, it was useful to investigate the effects of varying laser power in our instrument and see if a similar effect existed for an oa-TOF.

4.4.1 Procedures Used in Assessing Effects of Digitiser and Laser Power

TPP was used as an analyte for these experiments, with DHB matrix. C60 was used in laser power experiments only. Samples were prepared by the dried droplet method for both series of experiments and accelerating potentials used were set to the levels given in chapter 2, except where specified otherwise in the results below. The mole ratio of DHB to analyte was approximately 1500:1. In the first series of experiments spectra of the M+/[M+H]+ cluster were obtained with the 4 GS/s LeCroy 9384 oscilloscope. These spectra were compared with earlier results obtained with the 400 MS/s oscilloscope (LeCroy 9450).

In the second series of experiments laser power was varied with the attenuator, set from 0.2 (highest power) to 1.0 (lowest power). The number corresponds to the absorbance setting of the attenuator. The percentage of incident radiation transmitted at each of these settings can be read from figure 4.6. The specifications for the laser give a nominal pulse energy of 300ÊµJ and allow the incident radiation to vary by 3.5% (standard deviation) in intensity. Using these values and the size of the laser spot on the target (approximately 0.1ÊxÊ0.1Êmm, determined from figure 4.5) the laser power per unit area and fluence

58 (energy per unit area on target) can be calculated for each attenuator setting. These estimated powers and fluences are given in table 4.3. In experiments with C60 the resolution obtained when laser power was varied from high power to threshold levels, attenuator settings of 0.2 to 1.0, was measured. In an experiment with TPP the attenuator was set to two levels, 0.2 and 0.4, with five spectra acquired at each level. Spectra were acquired in a random order to reduce effects of systematic instrument variations (for instance drifts in power supply potentials). Peak width and time of flight were recorded.

4.4.2 Measured Effect of Digitiser Rate A sample spectrum obtained with the 4ÊGS/s digitiser is given in figure 4.7. The spectra obtained with optimal instrument conditions had resolutions (FWHM definition) of 3,800 to 4,400. Spectra of TPP obtained earlier with the 400ÊMS/s digitiser under similar conditions had resolutions of 2,800 to 3,400. The improvement in resolution due to increased digitiser rate is over 20%, taken from the means, clearly a significant effect. Almost all of this was due to a decrease in the peak width and not an increase in time between peaks.

Increased sampling rate is able to have a significant effect on resolution because the slower digitiser undersamples the signal. For instance, in figure 4.2, where TPP was obtained with the 400ÊMS/s oscilloscope, a 1ÊDa wide peak only contains 8 data points. This would have been approximately 80 if the 4ÊGS/s digitiser had been used instead. The effects of this difference in sampling rate can be observed using an ideal (theoretical) example. In figure 4.8 a gaussian curve has been modelled, first with eighty points across the peak and then with only eight points, to generate results expected with both oscilloscopes. A gaussian curve was used since this type of oa-TOF mass spectrometer generates signals that approximate gaussian peaks. The 80 point peak gives a fairly accurate representation of the original curve, while the 8 point example provides a poorer representation. The apparent peak width at half height is over 10% larger at the lower sampling rate, since the poorer representation misses the top of the peak and provides a less even peak shape. In this example, all the data points fitted the gaussian curve perfectly. The presence of noise could make this effect worse. Much of the noise in real data is electronic, due to the high frequency digitiser, with a higher frequency than the desired signal. When the number of sampling points is barely sufficient to characterise the signal the effect of higher frequency noise is almost impossible to ignore. A higher sampling rate allows high frequency noise to be filtered out or viewed as spikes and largely ignored when determining the width of mass peaks. In addition to the effects of sampling rate and noise, oscilloscope bandwidth also contributed to a larger peak width with the slower oscilloscope. The 400ÊMS/s digitiser had a bandwidth of 350ÊMHz,

59 Table 4.3: Estimated laser power and fluence incident on the sample for a 300µJ, 5ns pulse at the given attenuator settings.

attenuator % energy power (µJ.mm-2 ns-1) fluence (µJ.mm-2) setting transmitted 0.2 63.10 3786 18929 0.4 39.81 2389 11943 0.6 25.12 1507 7536 0.8 15.85 951 4755 1.0 10.00 600 3000

Table 4.4: Effects of varying laser power on C60 samples

attenuator setting signal resolution (±100) 0.2 very strong 500 0.6 strong 900 0.8 strong 1100 1.0 weak 3100

[M+H] +

[M]+

m/z 614 m/z 615 m/z 616

Figure 4.7: Molecular ion region of a spectrum of TPP in DHB acquired with a 4GS/s digitiser. The resolution is approximately 4400.

Number of points on peak 80 points 8 points

0 2 4 6 8 10 12

Figure 4.8: An example of the effects of undersampling. Both traces attempt to represent the same ideal gaussian peak. The trace with 8 points on ∆ ∆ the peak has a xfwhm of 1.63. The 80 point trace has a xfwhm of 1.45. The undersampled value is approximately 10% larger. causing broadening of peaks with rise times approaching the bandwidth, while the 4ÊGS/s digitiser had a bandwidth of 1ÊGHz, resulting in less broadening of peaks in the 100s of MHz range. Thus, the combined effects of bandwidth limits, undersampling and noise were likely to be responsible for introducing an additional~20% to the peak width at 615ÊDa.

4.4.3 Effect of Laser Power

+ The results for the C60 M cluster (m/z 720 Ð 721ÊDa) are presented in table 4.4. They indicate that, for C60 at least, there was a marked improvement in resolution, from 500 to over 3,000, when laser fluence was decreased from approximately 19,000ÊµJ.mm-2 (0.2 attenuator) to 3,000ÊµJmm-2 (1.0 attenuator). Unfortunately, the increase in resolution was accompanied by a large drop in signal. This, coupled with the sensitivity limits of the instrument, dictated that it was not practical to routinely acquire spectra at the irradiance threshold. In response to this a more detailed study was conducted with the TPP [M+H]+ cluster (m/z 615 Ð 616ÊDa) over a smaller laser power range, to determine whether smaller changes in laser power, well above the threshold, had a significant effect.

The changes in times-of-flight and resolution obtained in that experiment for masses of 615 and 616ÊDa are summarised in table 4.5. Resolution data was also pooled, since there was no theoretical reason for an observable difference in resolution between adjacent peaks. Resolution results were a little disappointing, since while there was a clear increase in resolution of 21% for pooled data when laser power was decreased, this was less than the sum of the standard deviations and thus not significant. Similar results were obtained for the individual masses when analysed. The results overlapped at the 68% confidence level (one standard deviation) and thus the difference in these results was not statistically significant. For time-of-flight data there appeared to be a trend to slightly longer time of flight values. However, similarly to the resolution data, the time-of-flight results overlapped within the sum of the standard deviations of the low and high laser power data sets, so the results were not statistically significant or conclusive.

While the TPP results were not significant on their own, they can be combined with data from other sources to indicate trends. The result that showed increasing resolution with decreasing laser power were consistent with those obtained with C60 over a larger range of laser powers. It may be concluded that this increase in resolution was most likely real, but not large enough to appear significant over the small laser power range used. The difference in flight times due to laser power can be compared to the mass accuracy specifications of the instrument, ±90Êppm, given by Mlynski and Guilhaus [166]. The

60 Table 4.5: Effects of Laser power on TPP flight time and resolution. Grid 2 potential was set to Ð108 V, with all other potentials set as per chapter 2. The relative differences of values from the means are given in parentheses.

(a) m/z 615

time-of-flight (µs) resolution attenuator mean standard mean standard deviation deviation 0.2 53.6969 0.0041 (76ppm) 1880 330 (18%) 0.4 53.7001 0.0012 (22ppm) 2460 350 (14%)

difference 0.0032 (60ppm) 580 (27%) ( 0.4 Ð 0.2)

(b) m/z 616

time-of-flight (µs) resolution attenuator mean standard mean standard deviation deviation 0.2 53.7388 0.0040 (74ppm) 2230 350 (16%) 0.4 53.7458 0.0039 (73ppm) 2600 150 (5.8%)

difference 0.0070 (130ppm) 370 (15%) ( 0.4 Ð 0.2)

resolution attenuator mean standard deviation 0.2 2050 358 (17%) 0.4 2530 251 (10%)

difference 480 (21%) ( 0.4 Ð 0.2) increase in flight time due to the lower laser power, of 60 and 130Êppm, was essentially within mass accuracy specifications. Thus no conclusion can be drawn from these results and the explanation offered above must remain speculative.

4 . 5 Summary of Early Experiments and Sample Preparation Many of the results presented here were only qualitative or semi-quantitative, but they provided information crucial for later experiments with the MALDI-oa-TOF instruments. The improvements in sample preparation developed in this chapter allowed deposition of samples with much better spot to spot reproducibility. Enhanced results obtained with the higher frequency digitiser clearly indicated another way to improve the instrument’s performance. Better understanding of the effects of laser power also had a positive effect on future experiments. In combination, these discoveries allowed the fundamental experiments discussed in the next chapter to be carried out.

61 Chapter 5: Fundamental Studies Ð Desorption

Velocity/Energy Measurements on the Linear

Instrument

5 . 1 Importance of Desorption Velocity / Energy A thorough understanding of the MALDI process requires knowledge of what happens to the excess energy given to ions and neutral species formed following absorption (predominately by the matrix) of 337Ênm photons from the laser. According to quantum mechanics, the photon’s energy is absorbed by the species and can result in translational motion, rotational motion and vibrational motion of molecules, electronic excitation of species and / or fragmentation of chemical bonds. Investigation of most of these states is beyond the scope of these experiments. Translational motion is the exception, since the motion of the ionized molecules and fragments (after collisions) can be measured within the confines of TOF mass spectrometers from times of flight, by determining the desorption velocity of the ions. Such velocity measurements have been performed by a number of research groups [105, 121-127, 198-200].

In addition to giving information about the MALDI process, knowledge of the desorption velocity of analyte ions is important in MALDI oa-TOFMS. This is because for oa- TOFMS all ions require a limited range of energies in the desorption axis to reach the detector. This translates to a range of allowed velocities at each mass. The initial velocity component in the TOF direction is also important, since this can affect the focusing and calibration characteristics of the instrument, as will be explained below. Reported velocity/energy results have tended to depend on the instrument used. The situation concerning desorption energy for oa-TOFMS instrument geometry is given graphically in figure 5.1. Assuming ions reach their final desorption velocity close to the target, the oa- TOF method allows for calculation of the permitted range of source axis desorption velocities (usa) for detected ions, based upon distance from the target to the fill up region and the delay time between triggering of the laser and push out pulse. Accordingly, it was decided to measure desorption velocity with the linear MALDI-oa-TOFMS.

desorption plume slit push-out pulse

target u sa

u TOF

fill up region u D expanded

Limits on ion packet given by slit

} source axis

Ions, showing velocity in the ToF axis to detector (assumes correlated case)

Figure 5.1: A diagram of the source to fill up regions of an oa-TOF instrument, illustrating the path taken by ions from the MALDI plume to the fill up region. A general vector diagram of ion velocities after acceleration is given in the top right. 5 . 2 Velocity Distribution in the Desorption Axis 5.2.1 Introduction: Published MALDI Desorption Velocities and Relevance to oa-TOF Instrument Geometry Results from other researchers have indicated that desorption velocity depends upon sample preparation, analyte and matrix, amongst other factors [121]. The method used to determine the velocity of species created in the MALDI process also influenced the results, as detailed below. Initial ion velocity values were obtained with delayed extraction instruments, utilising the principle that the rate of increase in ion flight time with delay time is (approximately) proportional to the initial velocity [122]. The velocities found with the delayed extraction method were approximately 300 to 600 ms-1 for a variety of species and matrices [121, 122]. Experiments performed with a postionisation instrument, where the velocity of desorbed neutrals was determined by the intensity of signal as a function of the delay between the desorbing and ionising laser pulses, gave similar results [123]. Experiments in an instrument that compared the flight times for ions analysed after prompt acceleration with flight times obtained after ions traversed a field free region provided desorption velocities of ~750Êms-1, for ions of over 1,000ÊDa [105]. Similar values for initial velocity were obtained by Chan and colleagues with another instrument that utilised field free desorption, for both positive ions (840±70Êms- 1) and negative ions (750±40Êms-1) of cytochrome C [199]. In oa-TOF instruments higher velocity values were obtained, from approximately 800 to 1,000 ms-1 for analyte ions from 1,000ÊDa to over 10,000ÊDa [124, 125]. In a purpose built instrument Bökelmann and coworkers measured desorption velocities at different angles and found that velocity was approximately 700 to 1,500 ms-1 for substance P in DHB matrix [126].

Importantly, experiments with all geometries showed matrix ions had higher velocity than analyte ions. Above c. 1000ÊDa, velocity was essentially constant for analyte molecules measured by most methods (including oa-TOF), at least within each reported method of measurement. It is believed that energy is directly gained by matrix ions in the MALDI process, and then transferred to analyte molecules via collisions, explaining why matrix ions are faster and why MALDI tends to be a constant velocity, rather than constant energy source.

Interestingly, the oa-TOF instruments and other methods that used fixed field free regions gave consistently higher velocity values than DE-TOF instruments. It has been suggested that this is at least partly due to field penetration [121]. Field penetration may have occurred in some devices, but it is difficult to see how there could be significant penetration of accelerating fields into the source region of oa-TOF instruments, since the shielding required between the source and orthogonal accelerator ensures that field

63 penetration from the accelerator is only likely to have a significant effect in the fill up region, if indeed there is any field penetration. Thus the higher velocities measured in oa- TOF instruments are due to other effects, most likely a combination of (i) a long field free delay time after desorption and (ii) restrictions in the acceptance of ion divergence in the TOF axis. Orthogonal acceleration instruments tended to have longer delays than delayed extraction instruments (µs rather than ns) prior to acceleration with electric fields, allowing more collisions between matrix and analyte. Restrictions in permitted ion divergence from the source axis in typical oa-TOF instruments (see figure 5.1) ensure that only ions expressing most of their kinetic energy in the source axis enter the fill up region of the oa. In contrast, conventional instruments sample almost all the ions. Thus conventional instruments could be expected to give lower values for the average velocity, since many ions would express a significant proportion of their energy in other directions. Imaging experiments have confirmed that matrix molecules (3- hydroxypicolinic acid, 139ÊDa) do indeed have a wide distribution of radial velocities, but that high mass analyte species (~30,000ÊDa dye tagged protein) propagate in narrow jets [127]. This suggests that divergence restrictions only have a significant effect on velocity measurements for matrix species and (possibly) low mass analytes, so the generally higher velocities in oa-TOF measurements are probably largely the result of additional collisions between species prior to extraction.

In summary, published results indicate that the desorption velocity depends on the instrument and sample preparation conditions used. Thus knowledge of desorption velocity on a particular instrument can only be obtained with certainty by direct measurement. Knowledge of the desorption velocity is very important in oa-TOF mass spectrometry, since the range of source axis desorption velocities allowed in an oa-TOF mass spectrometer are governed by instrument geometry and the energy given to ions during orthogonal acceleration. The required velocity is primarily related to the angle of the drift trajectory relative to the source axis, given by equation 1.14, reproduced below for convenience:

−  u  θ = tan 1 TOF    usa θ where is the angle, while utof and usa are the final velocity components in TOF and source axes respectively, illustrated on the vector diagram in figure 5.1. The range of allowed values for θ can be determined from the size of the oa aperture, orientation and width of the drift tube and size of the detector. In other words a range for θ is fixed by the geometry of the instrument. utof is directly linked to the fields acting upon the ion during orthogonal acceleration and thus utof can be calculated for any m/z value from the 64 energy given to ions during acceleration. This ‘energy’ value is quite useful, since it represents the fraction of kinetic energy expressed by one of the velocity vectors and is constant for all masses provided accelerating potential is not changed. It is not strictly the true kinetic energy, since kinetic energy is not a vector quantity that can be split into components with both direction and magnitude, so it is defined to be the partial kinetic energy for the TOF axis velocity component, PKEtof: 1 PKE = mu2 (5.1) tof2 tof

PKEsa is defined in a similar fashion from velocity in the source axis. Combining equations 1.14 and 5.1 we obtain equation 5.2a, which when rearranged to 5.2b allows calculation of desorption energy:

− PKE θ = tan 1 tof (5.2a) PKEsa PKE PKE = tof (5.2b) sa tan2 θ From this and the mass to charge ratio, the ranges of desorption velocities and energies allowed by the instrument were determined. These results are summarised in table 5.1 for the masses investigated in this series of experiments.

5.2.2 Experimental: Methods used to Measure Source Axis Velocity Distributions Samples of TPP, gramicidin S and substance P were prepared. Each was deposited separately with DHB matrix by electrospray deposition. The TPP and gramicidin S solutions described in chapter 4, table 4.2 were used. Solution (I) was used for TPP and solution (III) for gramicidin S. Substance P was made in ethanol at 0.5Êmg/mL, with 10Êmg/mL DHB. The distance between the spray needle and target was varied from 4.5 to 7Êmm and spray potential from 2.8 to 3.5ÊkV, to obtain stable spraying and allow the target to remain dry during spraying. Each solution was deposited on a separate target at a rate of 4ÊµL/min for 25Êminutes.

Individual spectra were obtained in the following manner: (a) the laser attenuator was set to 0.2, to ensure maximum ion yield; (b) the desired delay time was set; (c) an appropriate number of laser shots were fired, in sets of 10, until signal to noise was reasonable in the averaged spectrum or a total of 100 shots had been averaged. (d) the TOF spectrum was then saved and peak area recorded if the m/z of interest had given a clear peak.

65 Table 5.1: Calculated range of θ and desorption energy allowed in the linear oa-TOF instrument by geometry and accelerating potential. The allowed range of usa for masses analysed in this section are also shown. PKEtof was calculated to be 2966ÊeV, using values θ from table 2.1 (chapter 2) and assuming zero average initial utof. PKEtof and were then used to find PKEsa.

for minimum velocity for maximum velocity θ 87.2¡ 84.5¡

PKEsa 6.9 eV 27.5 eV

usa for m/z 145 Da 3030 ms-1 6050 ms-1 615 Da 1470 ms-1 2940 ms-1 1164 Da 1070 ms-1 2130 ms-1 1348 Da 990 ms-1 1980 ms-1 After recording a spectrum, a fresh sample spot was exposed to the laser for the next analysis.

The drift tube was connected to the accelerator chamber with a fixed length bellows. This θ meant that it was possible to change the angle of the drift tube to the source axis, ins. It θ θ should be noted that for each ins there was a small range of that provided ion trajectories that could reach the detector, based upon the width of the drift tube. For each θ mass of interest the full range of delay times that gave a signal at a set angle, ins, between the source axis and drift tube was scanned. The detector was then moved approximately 5 θ mm to change ins and delay times scanned again. This process was repeated for each of the masses investigated, 145 Da (DHB), 615.3 Da (TPP), 1164 Da (gramicidin S) and 1348 Da (substance P).

The data was analysed in the following manner for each mass. Peak area was plotted against delay time for each detector position that gave a discernible signal for at least six delay times. A gaussian curve was fitted to the data. The delay time and area corresponding to the top of each of these curves was obtained. Combining these results with mass of interest (in kg) and distance between the sample probe and centre of the oa allowed calculation of the derivatives dN/dusa and dN/dPKEsa. From these results the plume density as a function of both energy and velocity was plotted, with a Boltzmann curve fitted to the data for each mass.

5.2.3 Results and Discussion: Source Axis Velocity and its Effect on Instrument Performance Peak area against delay time for TPP is shown in figure 5.2. Each detector position had one delay time for which ions in the middle of the fill up region (source axis direction) had the correct source direction velocity to reach the middle of the detector. This corresponded to the delay time for the maximum area at each detector position in figure 5.2. Ion trajectories were parallel to the drift tube and almost all ions entering the drift tube reached the detector, utilising its entire active surface. At other delay times, ion trajectories were not parallel to the drift tube, resulting in clipping of the ion packet, giving fewer ions, which hit only a fraction of the detector’s surface. This resulted in the areas, as a function of delay time, being approximately normally distributed within the limits of reproducibility.

The maxima of the fitted curves reached their highest values for the detector positions of 85 to 95 mm. Hence a large proportion of TPP ions had velocities within the range 1450 66

4500

Detector Position 4000 60mm 65mm 70mm

3500 75mm 80mm 85mm 90mm 3000 95mm 100mm 105mm

2500

2000 Area (arbitrary units)

1500

1000

500

30 35 40 45 50 55 60 65 70

Delay time (µs)

Figure 5.2: Peak area plotted against delay time for TPP, m/z 615 at a range of detector positions. A separate gaussian trace was fitted to each detector position, with solid curves for positions 60, 70, 80, 90 and 100 mm and dashes for positions 65, 75, 85, 95 and 105 mm. ms-1 to 1850 ms-1, corresponding to these detector positions. A similar picture emerged for gramicidin S, substance P and the matrix, DHB, with the highest values appearing at different detector positions and delay times. Since the graphs of these results were similar to that of TPP, they were not included here. Instead, the areas and delay times corresponding to the maxima at each position have been used to generate the data summarised in table 5.2. These values were used to generate figure 5.3, for all compounds analysed. Figure 5.3 plots the number density as a function of both velocity

(dN/dusa against velocity) and energy (dN/dE against PKEsa).

Table 5.3 summarises the overall results for maximal density at each mass. The results obtained lie within the range previously calculated for our instrument, based on θ knowledge of and PKEtof (see table 5.1). The maximum for the DHB derived ion (m/z 145ÊDa) lies near the middle of its allowed range, while all the analytes (m/z 615, 1164 and 1348ÊDa) tended to give the strongest signals near the lower end of their allowed velocity ranges. This suggests that many analyte ions formed were unable to be analysed in our instrument, since their PKEsa values were too low, an unfortunate result giving reduced sensitivity. In fact, these results suggested that PKEsa required for our instrument favoured some matrix ions over the analytes. One solution to this was to bias the sample probe, with a potential of 5 to 25ÊV, to accelerate ions during and after desorption, shifting the maximal analyte ion densities closer to the middle of their range. The results of this experiment were disappointing, as even lower signals were obtained. This was most likely due to the probe bias having the effect of a defocusing lens at the sampling orifice, located 1 mm from the probe. This unintended lens would disperse the ions in the TOF axis, causing very few to pass through the sample slit to the oa, 26Êmm later. Thus, there was no simple way to adjust an ion’s PKEsa without actually decreasing sensitivity, so signals in this instrument were reliant upon ions receiving the required desorption velocity from the MALDI process. While the limitations caused by the restricted range of

PKEsa were not pursued further in the linear instrument, measures were taken to overcome this issue with the reflecting instrument, as will be discussed in chapters 6 and 7.

5 . 3 Correlation of Velocity and Position in the TOF Axis 5.3.1 Relevance of Correlation or Lack of Correlation of Position and Velocity to Instrument Performance Minimising initial spatial and velocity spreads in the TOF axis were important reasons for creating the MALDI-oa-TOF instrument, as explained in chapter 1. The small spreads that remain in an oa-TOFMS are due to excess energy from the desorption / ionization process, as restricted by the geometry of the instrument. The resulting non-zero initial 67 Table 5.2: Data calculated for the fitted maxima for signals at each detector position for (a) TPP at 615.3 Da, (b) gramicidin S at 1164, (c) substance P at 1348 Da and (d) DHB at 145 Da

(a) m/z 615.3 Da -1 ‡ ‡ velocity (ms ) dN/dusa PKEsa (eV) dN/dPKEsa 1510 0.796 6.98 0.864 1640 1.00 8.22 1.00 1760 0.923 9.45 0.860 1940 0.828 11.5 0.700 2210 0.564 15.0 0.417 2520 0.283 19.4 0.184 2820 0.0794 24.5 0.0460

(b) m/z 1164 Da -1 ‡ ‡ velocity (ms ) dN/dusa PKEsa (eV) dN/dPKEsa 1160 0.536 7.74 0.562 1220 1.00 8.51 1.00 1470 0.799 12.4 0.663 1620 0.234 15.1 0.176 1640 0.188 15.4 0.139 1910 0.136 21.0 0.0867 2190 0.0516 27.5 0.0287

(c) m/z 1348 Da -1 ‡ ‡ velocity (ms ) dN/dusa PKEsa (eV) dN/dPKEsa 1060 0.650 7.63 0.732 1200 1.00 9.70 1.00 1316 0.693 11.7 0.632 1480 0.412 14.8 0.333 1670 0.277 18.8 0.199

(d) m/z 145 Da -1 ‡ ‡ velocity (ms ) dN/dusa PKEsa (eV) dN/dPKEsa 3330 0.244 8.41 0.297 3450 0.364 9.00 0.428 3600 0.563 9.81 0.633 4090 1.01 12.7 1.00 4860 0.632 17.9 0.527 5390 0.604 22.0 0.455 5550 0.494 23.3 0.361 ‡ units are not given for dN/dusa and dN/dPKEsa since N is proportional to the population density and is in arbitrary units.

gramicidin S DHB (m/z 145) A TPP (m/z 615) GS (m/z 1164) 1.2 substance P SP (m/z 1348)

1.0

0.8 TPP

(arbtrary units) 0.6 sa

0.4 dN/du DHB 0.2

0.0

2000 3000 4000 5000 -1 Velocity (ms )

TPP gramicidin S B DHB (m/z 145) 1.2 TPP (m/z 615) GS (m/z 1164) substance P SP (m/z 1348) 1.0

0.8 DHB 0.6 (arbitrary units) sa 0.4

dN/dPKE 0.2

0.0

10 15 20 25

PKE sa (eV)

Figure 5.3:Plume density as a function of ion (a) velocity and (b) PKE for m/z 145, 615.3, 1164 and 1348 Da. The Maxwell- Boltzmann distribution was fitted to each data set. Table 5.3: Region of greatest ion density for DHB matrix and TPP, gramicidin S and substance P analytes, as a function of both ion desorption velocity and partial kinetic energy. Uncertainty was taken as ±1/2 the values between adjacent data points near the maxima.

Species Velocity (ms-1) Energy (eV) DHB (145 Da) 4390 ± 390 14.3 ± 2.6 TPP (615.3 Da) 1680 ± 120 7.5 ± 1.2 gramicidin S (1164 Da) 1320 ± 150 9.4 ± 2.3 substance P (1348 Da) 1207 ± 130 9.9 ± 2.0

Table 5.4: Grid 2 potential for optimal resolution at m/z 615.3ÊDa (TPP) and 1163.7ÊDa (gramicidin S), obtained from experiments and theoretical models.

mass experimental modelled correlated uncorrelated 615.3 Da -109 V -85 V -110 V 1163.7 Da -100 V -82 V -110 V

Table 5.5: Sequence and molecular weight information for the synthetic peptides. peptide sequence N-C (1 letter code‡) composition MW

A DTAGD AAAAA ALGAA C128H215O48N43 3124.4 NAKAA AELGA ANAAA AAAAT AR

B DTASD AAAAA ALSAA C131H221O51N43 3214.5 NAKAA AELSA ANAAA AAAAT AR ‡ found in Mathews and van Holde [201] velocity and spatial spread that ions have in the TOF axis, just prior to application of the push-out-pulse, affects the tuning potentials and resolution of the mass spectrometer.

There are two main models that can explain the relationship between the initial spatial and velocity spreads in the TOF axis: correlated or uncorrelated. In the first, the velocity and starting position are correlated, as illustrated in the expanded region of figure 5.1. This would occur if the MALDI process acted as an effective point source, with collisions between desorbed and ionized species occurring only within the first fraction of a millimetre of the sample probe’s surface. In the second, there is no correlation between position and velocity in the TOF axis, or in other words, ions have the full range of velocities at each position in the TOF axis in the fill up region of the oa. Understanding which model applies has ramifications for how the instrument is tuned to obtain optimal results. Tuning for the correlated situation reduces to a special case of Wiley-McLaren time lag focusing [22], with small spatial and velocity spreads, while for the uncorrelated case tuning reduces to spatial focusing. Experimental results and simulations can be compared, in principle, to determine which model applies. Additionally, the conclusions reached based upon the results of these experiments can be combined with the desorption velocity distributions determined in section 5.2 to provide a more complete understanding of the MALDI process.

5.3.2 Experiments and Simulations used to Determine the Degree of Correlation Samples of TPP and gramicidin S in DHB matrix were prepared by electrospray deposition, using solutions (I) and (III) described in chapter 4, section 4.3.1. The distance between the spray needle and target was 5Êmm for both depositions. Each solution was deposited on a separate target at a rate of 4ÊµL/min for 25 minutes.

Spectra were obtained in the following manner: (i) the laser attenuator was set to 0.2 (high power), to ensure maximum ion yield; (ii) the delay time was set to 50Êµs for TPP and 60Êµs for gramicidin S; (iii) grid 2 potential was set to a value in the range Ð65 to Ð155ÊV for TPP or Ð75 to Ð125ÊV for gramicidin S; (iv) the push out pulse and accelerating potential were set as listed in chapter 2 (table 2.1); and (v) an appropriate number of laser shots were fired, in sets of 10, until signal to noise was reasonable in the averaged spectrum or a total of 100 shots had been averaged. The TOF spectrum was then saved if the m/z of interest had given a clear peak. Peak width at half height and time of flight were recorded. After this, a fresh sample spot was exposed to the laser for the next analysis.

68 The correlated model was simulated using a combination of SimTOF and a spreadsheet (figure 5.4) created in Microsoft Excel, since neither method of computer simulation was able to accurately simulate all contributions to peak width by itself. The spreadsheet was used to estimate the contribution of the spatial and velocity spreads to peak width, while SimTOF was used to calculate the peak width contribution of the other factors, such as those resulting from grid effects, high voltage power supply ripple or the detector. The two peak widths were added in quadrature to determine the overall peak width for each grid 2 potential.

The spreadsheet required the user to enter data for several parameters, as can be seen in figure 5.4. Important parameters included distances (in column B) and accelerating voltages for the instrument (in column C), together with the mass of interest (in column G). The other parameters that had to be entered were initial position and initial velocity, both in column G of the spreadsheet. The initial position of the ions in the oa was set within a maximum range of ±1Êmm (TOF axis) from the location where the laser hit the target. The maximum range of velocities allowed was determined from the geometry and delay times to be ±20Êms-1 (TPP) or ±17Êms-1 (gramicidin S) and entered in the spreadsheet. The spreadsheet was able to calculate the initial velocity (uTOF) at each of ten points across the full range of initial positions in the TOF axis, assuming direct correlation between position and velocity. TOF data were obtained for each position and the width of the peak calculated from the range of time values. Grid 2 potential was then changed and the new peak width and time-of-flight were recorded. This was repeated across the full range of grid 2 potentials, -50 to Ð160ÊV (TPP) or Ð60 to Ð150ÊV (gramicidin S), in 10ÊV increments. The Solver algorithm installed with Excel was used to obtain the grid 2 potential that gave the narrowest TOF peak width. For each of these grid 2 potentials, SimTOF was used to calculate the peak width for ions of the appropriate mass, starting on the beam axis (no spatial spread) with zero initial velocity. The peak width result obtained by combining SimTOF and spreadsheet values (added in quadrature) were converted to resolution values by dividing the calculated TOF by twice the peak width.

Calculations for the uncorrelated case were obtained using SimTOF alone, since SimTOF was able to accurately simulate peaks obtained for species where initial spatial and velocity spreads were uncorrelated. After the correct distances and potentials had been entered for the instrument, resolution was obtained directly from SimTOF, using the spatial and velocity spreads given above for each of the masses. Grid 2 potential was then changed by 10ÊV and the new resolution recorded for the appropriate mass. This was repeated until the full range of potentials, -50 to Ð160ÊV (TPP) or Ð60 to Ð150ÊV (gramicidin S), was covered. 69

ABC DEF G H

TOFMS Calculator distance/mm potential*/V Entered data**: ONLY ENTER DATA IN BOXED AREA initial position/mm = -1 -0.8 -2 8 8 initial velocity/ms-1 = -16 -12.8 ion beam (1) 0 m/z "Da" = 615 615 3.36 0 Calculated Data#: 1st accel. (2) 8.36 -66.36 m/z "kg" 1.0212E-24 1.0212E-24 2nd accel. (3) 20.36 -2858 KE gain in (1)/J 1.1469E-17 1.0943E-17 KE gain in (2)/J 1.0632E-17 1.0632E-17 field free (4) KE gain in (3)/J 4.4727E-16 4.4727E-16

v at end of (1)/ms-1 4.7393E+03 4.6293E+03 detector 1525.1 -2858 v at end of (2)/ms-1 6.5790E+03 6.5002E+03 v at end of (3)/ms-1 3.0319E+04 3.0302E+04 v in (4)/ms-1 3.0319E+04 3.0302E+04 Width of time peak (ns) = 1.47 time in (1) 1.8462E-06 1.8022E-06 *Potential at end of region time in (2) 8.8352E-07 8.9851E-07 time in (3) 6.5044E-07 6.5214E-07 **Place a minus sign in front of m/z (Da) value in time in (4) 4.9630E-05 4.9658E-05 negative ion mode. Total time/s 5.30E-05 5.30E-05

Figure 5.4: The Excel spreadsheet used for the correlated model. Columns I to Q (not shown) are similar to column H (on the right edge of the figure), with initial position and velocity linearly scaled up to +1 mm and +16 m/s. The width of the time peak was calculated from the range of TOF values. Grid 2 potential was chosen as the tuning potential in these experiments for a number of reasons. Resolution was fairly sensitive to changes in grid 2 potential and changing grid 2 did not affect total ion energy gained during acceleration. It merely changed the relative field strength between the second and third accelerating regions and had a small effect on time of flight. Tuning with the push-out-pulse or main accelerating potential would have adjusted PKEtof and have had a larger effect on the time of flight. Additionally, and of lesser importance, the mass spectrometer was designed so that it was easy to adjust and accurately measure grid 2 potential to within 0.1ÊV.

5.3.3 Comparison of Experimental Results with Simulations The resolution as a function of grid 2 potential obtained for TPP from experiments and simulations are given in figure 5.5, with those for gramicidin S presented in figure 5.6. The error bars for experimental resolution were calculated from the sampling limits of the Le Croy 9450 oscilloscope, which introduced an uncertainty of 1Êns on peak width (FWHM) calculations. In these experiments, the most important result is the grid 2 voltage giving optimal resolution for each species, both in practice and with the two models. The exact value of the resolution found with the simulations was not important, since resolution was influenced by parameters such as detector contributions and grid effects that had been estimated, which would not influence the grid 2 potential required for optimal resolution. Any confusion that differences in resolution would have caused has been removed, by scaling the results from the simulations to fit on the same vertical scale as experimental measurements in figures 5.5 and 5.6. The grid 2 values that provided optimal resolution are provided in tableÊ5.4 for all experimental and simulated data for both species.

On first inspection, the experimental results for TPP favour the uncorrelated model, while the results obtained for gramicidin S gives an optimal potential between that predicted by the correlated and uncorrelated simulations, although with a tendency towards the uncorrelated value. However, before any conclusions are made, uncertainty in experimental results must be determined. Uncertainty was taken to be 2.5Êns in FWHM values, based upon the sampling rate of the LeCroy 9450 oscilloscope. This generated error bars over 30% of the resolution for data points near the maxima in experimental results. As a result of this uncertainty maxima were taken from gaussian fits to the data, since scatter ensured there was no trend to a clear maximum in the raw data. A gaussian fit was used, since while not a perfect match the gaussian curve approximated the curves predicted by the simulations. Based upon the scatter, a subjective estimate of the uncertainty in the maxima of experimental data would be that it was correct to within

6000 uncorrelated

correlated

5000

4000 resolution

3000

experimental 2000

1000

-160 -140 -120 -100 -80 -60 grid 2 (V)

Figure 5.5: Resolution as a function of grid 2 potential for TPP m/z 615.3 Da. Experimental results are plotted with a gaussian fitted to the data. Results from the correlated model are given by the dashed curve, while results from the uncorrelated model are given by the dotted curve. The values on the resolution axis correspond to experimental results, with the curves from simulations scaled to fit on the graph.

6000 uncorrelated

correlated

5000

4000 resolution

experimental 3000

2000

-140 -120 -100 -80 -60 grid 2 (V)

Figure 5.6:Resolution as a function of grid 2 potential for gramicidin S m/z 1164 Da. Experimental results are plotted with a gaussian fitted to the data. Results from the correlated model are given by the dashed curve, while results from the uncorrelated model are given by the dotted curve. The values on the resolution axis correspond to experimental results, with the curves from simulations scaled to fit on the graph. ±10ÊV of the grid 2 value for optimal resolution in each case. This would seem to account for the discrepancy between the TPP and gramicidin S results. Additionally, this favours the uncorrelated model, since both sets of data only overlap with the uncorrelated predictions, within the limits of uncertainty.

This, however, represents only half of the analysis. Predictions from the models also had an uncertainty associated with them, related to the entered data defining potentials and distances. The voltage values were very accurate with a combined effect of less than 200Êppm, as evidenced by mass accuracy experiments performed previously [166]. However, uncertainties in distances between crucial elements could have a more dramatic effect. The engineering tolerances were ±0.1Êmm for each part. The effect on optimal voltage of varying each of these components by ±0.1Êmm was calculated and the square root of the sum of the squares of these uncertainties was taken, since each engineering tolerance had an independent effect. The overall uncertainty in grid 2 for optimal resolution in both simulations was calculated to be ±20ÊV. Unfortunately, all experimental and simulated values overlapped within this uncertainty and thus the results of these grid 2 experiments favoured neither simulation.

It is difficult to draw a conclusion from these experiments owing to overlap of results. However, the fundamental differences between theoretical models allows some further analysis. The correlated model, a special case of time lag focussing, requires grid 2 potential to be varied to obtain optimal resolution for different masses, dependent on both mass and delay time. If velocity and position in the TOF axis are uncorrelated, allowing only spatial focussing, a single grid 2 potential would give optimal resolution for all masses. The only way to establish which of these two models hold is to extend these experiments over a much larger mass range, with many more compounds. Then it can be determined whether grid 2 potential for optimal resolution is dependent or independent of mass, since with many data points the effects of engineering tolerances reduces to a (hopefully) clear systematic error.

5 . 4 Extension to Other Samples: Higher m/z and Matrix Effects Additional experiments investigating the energy given to ions during the desorption/ionization process required the use of new samples. Initial experiments were aimed at extending the mass range, utilising slightly larger peptides which had given results in other mass spectrometers. Insulin chain A (2,532ÊDa), insulin chain B (3,492ÊDa) and two synthetic peptides (3,124 and 3,214ÊDa) were chosen. It was hoped that these molecules would allow extension of the work conducted in section 5.3 and provide a conclusive answer on whether the initial TOF axis position and velocity of ions

71 were correlated or uncorrelated in the fill up region of the oa. Another experiment involved investigating the effects of matrix on desorption velocity. Gramicidin S was analysed with α-cyano-4-hydroxycinnamic acid (HCCA) matrix, to see if it gave a different MALDI desorption velocity/energy profile to earlier results obtained with DHB matrix.

5.4.1 Methods of Sample Preparation and Analysis For the first set of experiments all samples were dissolved in 1:1 (v/v) acetonitrile and 0.5% aqueous trifluoroacetic acid and deposited by the dried droplet method, as explained in section 2.5.3. The dried droplet technique was used, instead of ESD, to conserve the peptides. DHB matrix was made to 20Êmg/mL and Insulin chains A and B were each made to 1Êmg/mL. Sequence information and molecular weight of the two synthetic peptides used are listed in table 5.5. These peptides were only sparingly soluble in the solvents used and could only be made to 0.5Êmg/mL, so 4ÊµL, rather than 2ÊµL, of each was deposited in the dried droplet sample preparation.

Gramicidin S, deposited with HCCA matrix, was investigated on both the linear oa-TOF instrument and a commercial MALDI instrument, the Voyager DE-STR. The solution used consisted of 1 part 10Êmg/mL gramicidin S in a 1:1 acetonitrile to water solvent mixture, mixed with 99 parts saturated solution of HCCA. A large supply of this solution was available, so samples were prepared by ESD. ESD occurred at a needle potential of 2.9 Ð 3.2ÊkV, with distance between the sample and needle of 8.3 Ð 8.7Êmm. A flow rate of two microlitres per minute was maintained for a period of 60 minutes for samples prepared for the linear oa-TOF, with the sample slide on the turntable. Samples prepared for analysis with the Voyager instrument were sprayed onto single spots under the same conditions for a period of 30 minutes.

Experiments were performed on the linear oa-TOF instrument in the manner described for section 5.2.2, except the detector was moved in 15Êmm rather than 5Êmm increments, with up to 60 rather than 100 laser shots averaged before changing parameters. Any reasonable spectra obtained were saved. Spectra were obtained in both the linear and reflectron mode with the Voyager, using the settings given in chapter 2, section 2.4. Individual spectra were averaged until the signal to noise was good, typically at 40 to 70 shots.

72 5.4.2 Analysis of High m/z and HCCA Matrix Results No spectra were obtained with the three larger peptides, other than signals for matrix ions. All of these peptides had earlier given positive ion mass spectra in conventional MALDI-TOF systems with DHB matrix, so it was reasonable to assume ions were also formed in the experiments performed with the linear oa-TOF instrument. The lack of signal was most probably due either to the ions formed not having the correct source axis velocities to reach the detector, or that the ions reached the detector, but did not generate a signal. The source axis velocities required for detection of these species ranged from 730 Ð 1,400Êms-1 at 2,500ÊDa to 620 Ð 1,200Êms-1 at 3,500ÊDa. This is within the velocity range expected from desorption/ionization for at least some ions with this mass, based upon experiments performed in other oa-TOF instruments [124, 125], so at least some ions would have been expected to have the correct velocities for analysis with this instrument. Regarding generation of a signal, semiconductor surfaces similar to those in MCP detectors are known to have a low conversion efficiency for ions with velocities of less than 104Êms-1 [110, 111, 112]. The accelerating field used in this instrument imparted ~3ÊkeV to the ions, providing TOF axis drift velocities of 1.3 Ð 1.4Ê×Ê104Êms-1 for ions of m/z 2,500 to 3,500ÊDa. This is above the expected threshold velocity required for detection, so at least some of these heavier peptide ions would be expected to generate a signal.

The result of the analysis in the preceding paragraph is quite perplexing, since it suggests that the peptides should have been detected with this instrument, while clearly they were not. Two factors that the above analysis have not taken into account, however, are that (i) while ions can be analysed with a range of source axis velocities, this instrument can only analyse a small portion of this velocity range for any given detector position and (ii) the transmission efficiency of the instrument is quite low, with a narrow sampling slit and relatively low transmission grids, ensuring that most ions with appropriate source axis velocities would not reach the detector in any event. Further, the MCP multiplier used in this study was several years old and may no longer have been sensitive to 3ÊkeV ions in this mass range. Ion energy and TOF axis velocity could only be increased by setting a higher analyser potential, but unfortunately this analyser could not be operated at a higher potential, owing to the danger of discharges, so ion velocities could not be increased with this instrument. Thus, it is likely that the failure to detect any signal for these peptides resulted from a combination of: (i) relatively few ions having the appropriate velocity for analysis at any given detector position, (ii) low transmission efficiency for any ions with the appropriate velocity, and (iii) low detector efficiency ensuring that almost no ions were detected, with resulting signals indistinguishable from background noise.

73 Experiments with gramicidin S in HCCA on the linear oa-TOF mass spectrometer were also not successful. Only the occasional single ion was detected around 1,140 to 1,170ÊDa, the molecular ion region, regardless of the detector position and delay time. Below 200ÊDa ion peaks were observed for every m/z, which had not been observed with DHB matrix, suggesting that the HCCA matrix generated significant fragmentation, at least for matrix species. Combining this with our knowledge of the instrument, this suggested three possible explanations for the lack of signal for gramicidin S molecular ions: (i) no gramicidin ions formed, (ii) ions formed but virtually all fragmented prior to acceleration, or (iii) the ions formed had source axis velocities outside the range that gave appropriate θ values for the instrument.

It is unlikely that the samples generated no quasimolecular gramicidin S ions, since HCCA matrix was known to generate spectra for proteins and peptides [192]. It is possible that ions formed and then decayed prior to acceleration, since HCCA is known to be a ‘hotter’ matrix than DHB, giving more excess energy to analyte molecules during the ion formation process. The delay before orthogonal acceleration (>Ê50Êµs) is much greater than that prior to acceleration in conventional delayed-extraction time-of-flight instruments, allowing more ions to decay. However, it is highly unlikely that all ions would decay during this period, for this would have been observed on conventional reflecting geometry TOF instruments as a large metastable peak in post source decay experiments for gramicidin S. Finally, all of the gramicidin S ions may have been generated with the wrong desorption velocities to be detected in our instrument.

It was decided to investigate which of the three explanations applied by obtaining spectra on the Voyager delayed extraction MALDI-TOF instrument (“the Voyager”). The Voyager utilised a similar laser to the linear MALDI-oa-TOF, so ions were likely to be formed by the same type of MALDI process. The Voyager also incorporated an ion mirror and allowed detection of ions in both linear and reflectron modes. Since it was a conventional TOF instrument, desorption energy would affect its mass accuracy rather than sensitivity. If no ions were observed in either mode, it meant that none were formed. If gramicidin S ions were observed in the linear mode, but largely replaced by a lower mass metastable peak in the reflectron spectrum, it would suggest that ions were forming but probably decayed before orthogonal acceleration in our instrument. If both modes gave a good spectrum, with no large metastable peaks, it would indicate that molecular ions of gramicidin S were formed, but they do not have the correct desorption energy to reach the detector in the linear oa-TOF instrument.

+ [M+H] [M+Na] +

1040 1080 1120 1160 1200 1240 1280 1320 1360 1400 m/z (Da)

Figure 5.7: A spectrum of gramicidin S in HCCA matrix obtained on the Voyager DE-MALDI-TOF, in reflectron mode. A blow-up of the molecular ion cluster (m/z c.1144) is inset. Spectra obtained with the Voyager gave good results in both linear and reflectron mode. No large metastable peaks were noted. The main difference between spectra was that resolution was better for the reflectron mode, giving isotopic resolution, as indicated in figure 5.7. Thus this sample, made from the same solutions as those used for the oa-TOF instrument, gave good [M+H]+ and [M+Na]+ signals on the Voyager in both operating modes, while no such ions reached the detector of the linear oa-TOF instrument. This indicated that the gramicidin S ions formed were not detected, most probably because there was not a sufficient number of ions with the correct PKEsa to generate a signal that could be distinguished from the background noise in the linear oa-TOF instrument.

5.4.3 Conclusions Ð Limits of the Instrument

The linear MALDI oa-TOF instrument was only able to analyse ions over a limited PKEsa range, which translated to a limited range of source axis velocities, dependant upon the mass of the analyte ions. This was because the analyser used for this instrument was originally designed for use with an electron impact (EI) ion source, which gave ions with a constant PKEsa. Unfortunately for sensitivity in this instrument, MALDI generates ions with a much larger range of PKEsa than an EI source. Additionally, the accelerating potential used limited detector utility to low masses, suitable for the analyser’s original EI source, but once again unsuitable for MALDI. Thus many ions could not be analysed and even those ions that could be analysed had low sensitivity when compared to conventional MALDI-TOF mass spectrometers, due to the combined effects of PKEsa restrictions and detector mass limits. There were no simple modifications that could resolve either of these issues for the prototype linear oa-TOF. Instead, a purpose built reflecting geometry MALDI-oa-TOF instrument was designed, utilising the knowledge gained from experiments described in this and the proceeding chapter. Importantly, this purpose built instrument was designed to permit the analysis of ions with a much larger range of PKEsa (without requiring movement of the detector) and operated at a higher accelerating potential, to give ions higher velocities. Experiments relating to characterisation of the reflecting geometry MALDI-oa-TOF mass spectrometer are described in the remaining chapters of this thesis.

75 Chapter 6: Testing and Improving MALDI-oa-TOF

with an Ion Mirror

6 . 1 Introduction A new instrument, the compact MALDI-oa-TOF with an ion mirror, was constructed. The basic design and operation of this instrument was discussed in chapter 2, section 2.3, and is not repeated in this chapter. A number of important improvements were present in this instrument, compared with the linear oa-TOF instrument: (a) It was designed to operate at a higher accelerating potential, -20ÊkV, with ions receiving (approximately) 800ÊV of additional accelerating potential from the push out pulse, giving a final energy of ~20.8ÊkeV to assist in the detection of higher mass ions. (b) A wide orthogonal accelerator was combined with an open drift region and large detector, as opposed to the narrow drift region and smaller detector of the linear

instrument. This increased the allowed range of PKEsa, improving instrument sensitivity and permitting the analysis of a larger range of substances. (c) Custom designed grids were installed, with a more open rectangular mesh, to improve sensitivity without compromising resolution. (d) A single stage ion mirror was incorporated, to allow improved resolution, without increasing the overall size of the instrument significantly.

The instrument was designed to operate at an analyser potential of Ð20ÊkV, as mentioned above. It was discovered, however, that the detector had to be conditioned with lower potentials for a period of time (approximately 2-3 months) before its gain had reduced to a level where it was possible to use it at Ð20 kV, without causing damaging discharges. During this conditioning phase, the mass spectrometer was operated at lower potentials (- 3, -15 and Ð17ÊkV), generating useful results, if with somewhat lower resolution than that expected at Ð20ÊkV. This chapter will discuss and explain those experiments, together with a number of modifications made to the instrument throughout the project. Characterisation of the analyser at Ð20 kV and other important experiments relating to detector performance and grid effects will be discussed in subsequent chapters.

All experiments discussed in this chapter were performed with an old Brandenburg +15ÊkV Alpha Series power supply (Thornton Heath, Surry, UK) connected to the mirror backplate, instead of the Glassman supply mentioned in chapter 2. The

76 Brandenburg supply did not allow accurate measurement of its potential during experiments, so the values for mirror potential given below are only approximate. The Glassman supply, which allowed accurate measurement of the voltage and current, was obtained at a later date and used for all experiments in subsequent chapters.

6 . 2 Initial Experiments with Fullerene Standard, Gramicidin S and Insulin in DHB Matrix In the very first experiments no spectra were obtained. This was found to be due to field penetration into the target region. This electric field penetration was stopped, by placing a metal shield with a small aperture over the opening. Experiments related to field penetration are explained in section 6.3.1. After installation of the shield, some spectra were obtained with an accelerating potential of Ð3ÊkV. This potential was used, since it had given results in the prototype instrument. The original detector failed when the magnitude of the accelerating potential was increased, due to a manufacturing defect. It was replaced and the new detector was used with a larger accelerating potential of Ð15ÊkV. Some of the spectra were calibrated by fitting data to the mass calibration equation, given in chapter 1 (equation 1.10).

6.2.1 Techniques Used in Initial Experiments for Fullerene Standard, Gramicidin S and Insulin The fullerene standard (1Êmg/mL) was deposited with DHB matrix (10Êmg/mL) and prepared by the dried droplet procedure, explained in chapter 2, for experiments at both potentials. For experiments at Ð15ÊkV samples of insulin (1Êmg/mL) gramicidin S (1Êmg/mL) and DHB (10Êmg/mL) were also prepared by the dried droplet technique.

For experiments at Ð3ÊkV the mirror backplate was set to approximately +800ÊV and the push out pulse adjusted until optimal resolution was obtained. Only single shot spectra were acquired. When experiments were performed with an analyser potential of Ð15ÊkV the mirror back-plate was set to approximately +1,000ÊV and not altered. The push out pulse was adjusted across the range 600 to 800ÊV until the optimal resolution was obtained. SimTOF simulations had indicated that the approximate potential required was in this range, assuming the mirror back-plate was within 10% of 1,000ÊV.

Mass accuracy within a single spectrum (internal standard) was also investigated at an accelerating potential of Ð15ÊkV for carbon clusters obtained from the fullerene sample. Carbon clusters were used because it was easy to assign the correct mass to every peak, since all major peaks of 720ÊDa or heavier only contained an even number multiple of carbon atoms. Additionally, clusters of many different masses were formed, providing

77 enough peaks of known mass for statistical calculations. TOF data was obtained for all masses that gave reasonable signal-to-noise ratios. The true mass for each of these peaks was determined from the known composition. The square root of the mass to charge ratio ( mz/ ) was calculated, and then plotted against the measured times and a line of best fit was fitted to the data. Basing the calibration on all of the data points, rather than two or a small number, ensured the calibration obtained reflected the true accuracy of the instrument, with a mean error close to zero. The mass error in the calibration was determined from the residuals, representing the difference between the true masses and the masses calculated from the corresponding times of flight with the calibration function. This mass error was quoted as both an absolute (mDa) and relative (ppm) error, with means and standard deviations, indicating accuracy and precision of the measurements.

6.2.2 Signals and Mass Accuracy for Fullerene Standard, Gramicidin S and Insulin at - 3 and -15ÊkV Figure 6.1 shows the first spectrum acquired with the new instrument, a single shot + acquisition of the molecular ion region for the C60 [M] cluster, obtained with an analyser potential of Ð3.0ÊkV. Sensitivity was much higher than that obtained with the linear instrument, since with the linear oa-TOF spectra were always averaged to give reasonable signal to noise. In fact, the number of ions obtained from a single shot saturated the detection system, evidenced by the flat topped peak for m/z 720. Resolution was calculated to be approximately 1,000, using the peaks for m/z 721 and 722. Other similar spectra were obtained, including some containing a few lower intensity higher mass carbon clusters. The accelerating potential was then increased to Ð 20ÊkV. The bias voltage across the MSP detector was adjusted to somewhat less than 3ÊkV, well within the manufacturer’s allowed range. Unfortunately, owing to a flaw in MSP construction, an electrical discharge occurred across the detector 2 hours after application of the higher potential. This destroyed the detector. A new improved MSP detector was obtained as a replacement and experiments were continued at Ð15ÊkV, with a detector bias of 2.2ÊkV, the lowest potential that gave a reasonable gain.

At Ð15ÊkV the optimal resolution, 1,600 for the C60 cluster at m/z 720, was obtained with the push out pulse set to 705ÊV. A spectrum of carbon clusters, obtained from the fullerene sample is shown in figure 6.2. The spectrum was mass calibrated with the all carbon-12 C60 and C70 peaks, at m/z 720.000 and 840.000 Da. Laser attenuator was set to 0.2 (high power) and delay time to 30Êµs. No spectra were obtained for insulin or gramicidin S, regardless of the delay time used. It must be noted that the resolution in this spectrum (1,600) was 60% higher than that obtained when the accelerator was set to Ð3.0ÊkV. This improved resolution results from improved analyser focusing at the higher

-6 66.5 67.0 67.5 68.0 68.5x10 time (s)

Figure 6.1: The first recorded signal from the MALDI-oa-TOF with an ion mirror, a spectrum of C 60 , obtained with an accelerating potential of -3 kV. Only a single laser shot was fired.

R = 1600

1000 2000 3000 4000 5000 m/z (Da) Figure 6.2: A spectrum of carbon clusters, from 720 to over 5000, obtained with an accelerating potential of -15 kV. A close up of the C60 molecular ion region (m/z 720) is inset, with its resolution. potential, largely due to a reduction in turn around time resulting from the higher push out pulse voltage. Performance should be even better at an analyser potential of Ð20.0ÊkV.

The most likely reason for the lack of spectra with the two peptides is that their PKEsa are outside the range required for the geometry/PKETOF at Ð15ÊkV. For an accelerating potential of Ð15ÊkV, the full range of PKEsa allowed in this instrument was 20 Ð 130ÊeV, or approximately 50 Ð 100 eV for good sensitivity. According to experiments performed with the linear oa-TOF, PKEsa for gramicidin S was only approximately 7 Ð 12 eV, lower than the values required. In experiments performed by Dworschak et al [125] on a MALDI oa-TOF, the velocity of insulin in a 2,5 DHB rich matrix was found to be approximately 900 ms-1. The results of Dworschak et al for substance P were similar to those obtained in the linear instrument (c. 1,100 ms-1, compared to 1,200 ms-1 in the linear instrument), so it is reasonable to assume that the usa for insulin in our instrument -1 would be approximately 900 Ð 1,000 ms . From this, the likely PKEsa for insulin can be calculated to be 48 to 59ÊeV, within the lower end of the values required for detection.

Thus the majority of ions for gramicidin S probably did not have sufficient PKEsa and hence would be reflected back either towards the orthogonal accelerator or to strike analyser liner between the orthogonal accelerator and detector. Insulin ions were expected to have PKE sa within the required range, but it is possible that many of the ions had a lower than predicted velocity. Biasing the probe with a positive potential may have given positive ions formed with these samples kinetic energy components in the source axis that would have enabled the majority of the ions to reach the detector, an approach described in section 6.4.

Interestingly, the results obtained with carbon clusters at Ð15 kV indicated that these species had unexpectedly high desorption velocities, ranging from 900 Ð 2200 ms-1 for -1 the large clusters at m/z 5,000, to between 2,400 Ð 5,900Êms for the C60 ions observed at m/z 720. In experiments conducted at Ð3 kV, the required PKEsa was calculated from experimental times of flight and geometry restrictions to be 5 Ð 28ÊeV, requiring -1 desorption velocity of between 1,100 and 2,700Êms for C60. Thus the carbon clusters generated in a MALDI source appear to have a larger range of desorption velocities than the other species investigated, suggesting that the carbon cluster ions may be formed in a different manner to other MALDI analytes. Additional evidence for this was provided by the observation that the fullerene standard provided similar spectra without any matrix. The only significant effect of the matrix was that when matrix was present it took more laser shots to deplete the sample. Furthermore, high mass clusters only appeared at higher laser powers, suggesting that at least some energy may be absorbed by the fullerenes, with numerous reactions and collisions involved in creating the larger ions. In

79 combination, these factors suggest that carbon cluster ions formed by a direct laser desorption/ionization mechanism, rather than conventional MALDI.

The spectrum shown in figure 6.2 was used for the mass accuracy determination. Signals for clusters ranging from C60 to C308, with m/z 720 to 3,700ÊDa, were sufficiently strong to be used for this purpose. Above this mass range signals were still clear, but noise made it difficult to determine the centroids of the peaks with enough precision for this experiment. Figure 6.3 shows the calibration plot obtained from this spectrum, with the relative errors for each mass given above it. Errors over most of the mass range were normally distributed within approximately 100 ppm of their correct values, except for ions with the four shortest flight times, which are outliers, trending to large positive values, with error decreasing with decreasing flight time. These fast ions are those from the C60 and C70 clusters, with m/z of 720, 721, 840 and 841. The positive ppm errors indicate that these ions have a higher apparent mass than true mass, with a longer than expected flight time. The most plausible explanation for this is that these lower mass ions had slightly less energy than expected, giving a longer flight time. One reason for this could be the rise time of the push out pulse generator. This will be discussed in more detail in chapter 7, section 7.2.1, where the push out pulse generator was characterised. Regardless of the reason, it is clear that the low mass values were outliers with respect to the calibration function, so a second set of calibration coefficients was obtained, excluding the four lowest masses. These results are summarised in table 6.1, with column (a) showing results for all the data and (b) results after excluding the outliers. The most important of these results is the standard deviation for the relative mass error in (b), 30 ppm, which indicates the mass accuracy of the instrument for internal calibration, when operated at Ð15ÊkV.

After completing these experiments the sensitivity of the detector decreased, requiring the accelerating potential to be increased to Ð17.5ÊkV, which increased the detector bias to 2.57ÊkV. The effects of applying a small accelerating potential to the probe were investigated at this new potential, after all field penetration issues had been resolved.

6 . 3 Solving Field Leakage Problems in the Target and Fill Up Regions The effects of field penetration were investigated on a Macintosh IIfx computer (Apple Computer Inc., Cupertino, CA, USA) running the I-Opt program,6 ion optics software described in chapter 1, section 1.6.2. Field penetration from one region to another could have its most significant effects when ions with little translational energy were in a supposedly field free region. There were two such areas, both occurring before the ions

6© J.H.J. Dawson and M. Guilhaus, 1988. 80

300

200

100 ppm error

-100

1/2 45 m/z 40

35 40 45 50 55 60 65 time (µs)

Figure 6.3: Mass calibration plot from carbon clusters spectrum shown in Figure 6.2. Clusters

from C60 to C 308 (720 to 3700 Da) were used, giving a total of 110 measurements. The mass error in ppm, between calculated and actual m/z for each mass, is plotted against time above the calibration trace. Table 6.1: A summary of mass accuracy results obtained for the carbon cluster spectrum shown in Figure 6.2. Column (a) gives results for the clusters C60, C70 and C98 Ð C308 (720 Ð

3700 Da) and column (b) gives results for calibration and analysis of clusters from C98 Ð C308 (1180 Ð 3700 Da). The calibration constants were used with equation 6.1.

parameter (a) (b) No data points 110 106

calibration constants A (Da1/2/s) 910551 910502 B (Da1/2) -1.66506 -1.66223

absolute mass errors mean error (mDa) 1.51 -1.68 standard deviation 88.41 66.85

relative mass errors mean error ppm 2.35 -0.53 standard deviation 65.04 29.74 were accelerated for analysis: the target region and the fill up area of the orthogonal accelerator. Field penetration in both of these areas could have dramatic consequences.

6.3.1 Target Region When the instrument was originally commissioned, no signals were obtained. The most probable reason for this was deduced to be that the electric field generated between the drift chamber liner (20ÊkV) and the grounded chamber and probe target was penetrating into the target region, which was supposed to be at ground. A simulation was performed with I-Opt, to investigate whether this field was able to deflect the ions. In the scale simulation (planar geometry) of the ion source and surrounding regions, each element of the potential array corresponded to a 1Êµm distance. To take into account the “worst case” scenario the accelerating potential was set to –20 kV, the largest accelerating potential the instrument was designed for, and the sample slide was placed 2Êmm from the source shield, a larger distance than that used in most experiments. The orthogonal accelerator between grids 1 and 3 was approximated with a uniform field with potential ranging from 0 to Ð20ÊkV. This was close enough to the actual accelerator for the simulations, since the difference between this and the true accelerator would have little impact on the field near the target shield. After the correct electric field in the spaces had been calculated by I-Opt, ion trajectories were calculated for ions of m/z 100 and 10,000ÊDa. The lower mass ions were given an initial velocity of 2,000Êms-1 (4ÊeV) and the higher mass ions a velocity of 1,000Êms-1 (10ÊeV).

The results obtained can be observed in figureÊ6.4. A larger than expected field gradient penetrated into the grounded target shield. The top left part of the target cup, labelled ‘field penetrating into target cup’ in figure 6.4(A), was at a potential of -30 to –40 V, while this dropped to Ð6ÊV near the middle of the cup, in the region undeflected ions would have traversed. This deflected the ions, so that they did not enter the orthogonal accelerator, even at m/z 10,000ÊDa. The much lighter 100ÊDa ions were so highly deflected that they actually left the source completely in the simulation. Most of the ions that we wanted to investigate had mass values of hundreds to thousands of dalton. The level of deflection for those ions lay between the high and low mass extremes simulated, thus none would enter the accelerator.

The 30Êmm diameter opening in the target shield was covered and a 1Êmm diameter aperture left instead, to allow the ions to enter. I-Opt was allowed to determine the correct electric potentials once again, and the trajectories for m/z 100 and 10,000ÊDa were obtained. The results are presented in figureÊ6.5. It was immediately obvious that there was essentially no electric field within the shielded region, including the 2Êmm gap

-20 kV

A field penetrating into target cup

00 V V

ions trajectories: 10 kDa 100 Da -20 kV

orthogonal accelerator (even field, from 0 to -20 kV)

B fill-up region (0 V)

sample slide (0 V)

target shield (0 V)

outer chamber (0 V)

Figure 6.4: I-Opt simulation of the ion source region, without the shielding apperture. (A) shows the level of field penetration with an electric field contour plot with the surface inverted and (B) shows the effect on ion trajectories for m/z 100 and 10,000 Da.

-20 kV

0 V

-20 kV ion trajectories for both m/z values

orthogonal accelerator (even field, from 0 to -20 kV)

B fill-up region (0 V)

sample slide (0 V)

target shield (0 V)

outer chamber (0 V)

Figure 6.5: I-Opt simulation of the ion source region, with the shielding apperture. (A) shows the field contour plot (with the surface inverted for clarity) indicating no field penetration into the source region and (B) gives ion trajectories for m/z 100 and 10,000 Da. between the sample slide and source shield. Ions at m/z 100 and 10,000ÊDa were able to enter the fill up region of the orthogonal accelerator, without experiencing any deflection.

The simulations conducted at 20ÊkV were repeated with the analyser at 3ÊkV. While the deflections experienced by ions of 100 and 10,000ÊDa were smaller at 3ÊkV, they were sufficient to prevent ions from entering the fill up region of the orthogonal accelerator. Thus simulations confirmed that ion deflections due to field penetration into the target area was indeed responsible for the lack of signal in the original experiments.

6.3.2 Grid 1 and Field Penetration into the Fill Up Region Leakage of accelerating fields into field-free ionization regions of TOF mass spectrometers was known to decrease resolution in time-lag focusing instruments, although such leakage could be readily corrected with a small bias potential applied to an electric field defining element in the ion source [22]. In an oa-TOF analyser, an equivalent resolution decreasing effect would result if, prior to the application of the push out pulse, the fill up region contained an electric field with a component in the TOF direction.

In this reflecting oa-TOF instrument, it was possible that the strong electric field between grids 1 and 2 (Ð197ÊkV/m for a Ð17.5ÊkV accelerating potential) could penetrate between the wires of grid 1 into the fill up area. A small field gradient from this would slowly accelerate ions toward the mirror, with ions entering the fill up area earliest being most affected. This would result in a larger than expected initial spatial and energy spread for the ion packet, reducing optimal resolution. It was decided to investigate the level of field penetration through grid 1 with I-Opt, to determine whether a non-zero field existed in the fill up region and estimate the grid 1 bias required to eliminate it.

It would be impossible to accurately simulate the entire grid, since there were a large number of grid wires and I-Opt could not manipulate an array with the tens of millions of elements that would be required. Instead, five grid wires were simulated with planar repeat symmetry and the array was made effectively infinite, utilising the ‘wrapping’ technique, where the first column is mapped onto the last, explained in more detail in chapter 1. This allowed an accurate simulation to be performed with a much smaller array. The wires had a cross section of 22 × 5Êµm, with 60Êµm gaps between the wires.

This was approximately simulated in an array where each element corresponded to 5 ×

5Êµm. The grid wires were made 4 × 1 element, with 12 element gaps between the wires. The fill up region was located on the left side of this and the region between grids 1 and 2 on the right. The array extended 100 elements (0.5 mm) to each side of the grid. The grid

82 and the left and right edges were set to their appropriate values and I-Opt was left to calculate the correct potentials for the free space (all elements not corresponding to electrodes or the left or right edges) until the sum of residuals was zero.

In the first simulation grid 1 was set to 0ÊV, the value used in all experiments before this point. The result of this simulation is given in figureÊ6.6(A). It is immediately obvious that field penetration occurs in the supposedly field free region. The potential contour diagram does not give a flat surface. In fact, there is a 1 kV/m field in the fill up region. In an attempt to prevent field penetration, grid 1 potential was biased to a positive value and the correct field obtained with I-Opt. This process was repeated until the potential gradient in the fill up region was zero, the field free state, demonstrated in figure 6.6(B). Grid 1 was biased to a potential of +2.5ÊV to achieve this. Applying this small bias potential to grid 1 would have little effect on the tuning characteristics of the instrument, but would definitely improve resolution if it eliminated the field gradient in the fill up region, by reducing the initial spatial spread.

An experiment to determine this effect was conducted, with an accelerating potential of Ð17.5ÊkV. Insulin chain B was chosen as the analyte, since for its main ion cluster, with m/z of approximately 3,500ÊDa, only the tips of adjacent mass peaks could be resolved. Any improvement in resolution would be immediately obvious. A more detailed study was later conducted at Ð20.0ÊkV, with gramicidin S. Gramicidin S was used since it provided isotopic resolution when grid 1 was grounded, permitting a more accurate quantitative measurement of any change in resolution. For all these experiments, samples were prepared using the dried droplet technique discussed in chapter 2, with a matrix to analyte ratio of 1,000:1. Grid 1 potential was set to 0, 2.5 and 4ÊV and intermediate potentials for experiments at Ð17.5ÊkV or 0, 2.5 and 5.0ÊV at Ð20.0ÊkV and spectra were obtained by averaging the results from 60 laser shots for chain B or 30 shots for gramicidin S. Mirror and push out pulse potentials were set to their optimal levels, specified in section 6.4 (-17.5ÊkV) and table 2.2 (-20.0ÊkV).

The result of the experiment at Ð17.5ÊkV is given in figure 6.7. Visual inspection clearly shows that resolution improved significantly when grid 1 potential was changed from 0.0ÊV (A) to +2.5ÊV (B). FWHM resolution was measured to be approximately 5,600 for (B), and estimated, from comparison with simulated spectra, to be roughly 3,800 in (A). This represents an improvement in resolution of approximately 45% when grid 1 was biased to 2.5ÊV. When bias potential was increased further to 4.0ÊV, resolution obtained was similar to that shown for a bias of 0ÊV. Changing the potential by small amounts, of 0.5 ÊV or less, gave small changes in resolution that were difficult to quantify, owing to the effects of noise on the signals. 83

0 V

1 kV/m in fill up region grid wires A

to grid 2 197 kV/m

grid wires 0 V at +2.5 V 0 V/m in fill up region

to grid 2

197 kV/m

Figure 6.6: Field contour diagrams from I-Opt simulations of field penetration through grid 1into the fill up region when the push out pulse is not applied. In (A) grid 1 is set to 0 V. In (B) grid 1 is set to +2.5 V to give a relatively field free fill up region.

R ≈ 3800

61.4 61.6 61.8 62.0 62.2 62.4 time (µs)

R = 5600

61.4 61.6 61.8 62.0 62.2 62.4 time (µs)

Figure 6.7: The effects of field penetration through grid 1 into the fill up region on resolution of insulin chain B (3496.98 Da for [M+H]+ cluster). (A) is a spectrum obtained with grid 1 at 0 V and (B) is a spectrum obtained with grid 1 at +2.5 V. Table 6.2: Effect of biasing grid 1 on peak width and resolution for gramicidin S [M+H]+ cluster at an analyser potential of Ð20.0ÊkV. Results are quoted with uncertainty at 1 standard deviation.

grid 1 (V) FWHM (ns) FWHM resolution change in (normalised) resolution (%) 0.0 3.92±0.57 1.16±0.17 4400±640 0 2.5 3.37±0.19 1.00±0.06 5030±280 +14 5.0 5.67±1.30 1.68±0.39 3150±720 -28

Table 6.3: Important parameters for the masses used in plotting the required delay time and probe potential at an accelerating voltage of Ð17.5 kV

required for TOF analysis from gained during ionization geometry and potentials (assumed) µ -1 -1 mass (Da) TOF ( s) usa (ms ) PKEsa (eV) udes (ms ) PKEdes (eV) 100 10.158 7,400 - 18,200 28 - 173 4000 8.29 500 22.802 3,310 - 8,170 28 - 173 2000 10.4 1000 32.261 2,340 - 5780 28 - 173 1500 11.7 5000 72.163 1040 - 2,580 28 - 173 1000 25.9 10000 102.057 740 - 1,820 28 - 173 1000 51.8 20000 144.333 520 - 1,290 28 - 173 1000 104 The effects on FWHM for the experiments conducted at Ð20ÊkV are summarised in table 6.2. Ten measurements of individual peak width were obtained at each grid 1 potential for gramicidin S. The resolution of the [M+H]+ cluster improved by 14% when a bias of 2.5ÊV was applied to grid 1. Importantly, signal reproducibility was also higher when grid 1 was biased to 2.5ÊV, reflected by the smaller range of variation in FWHM. Improved reproducibility was probably related to the increased signal to noise resulting from narrower and sharper signals. Both resolution and reproducibility (the spread in FWHM measurements) deteriorated significantly when grid 1 was set to 5.0ÊV, with the broader signals than those obtained with a grounded grid.

Overall, the experiments conducted demonstrated that applying a bias potential of 2.5ÊV to grid 1 gives the best achievable resolution at an analyser potential of Ð17.5 and Ð20ÊkV. This confirms the I-Opt simulation, which indicated that this potential ensured the fill up region was essentially field free. A lower grid 1 potential allowed the electric field from the adjacent accelerating region to penetrate into this area, attracting ions to grid 1 before the push out pulse was applied. A larger positive bias gave a field that deflected ions towards the push out plate before the accelerating pulse was applied. In both these cases a small (offset) velocity spread was introduced, together with a spatial spread, which reduced the instrument’s resolution. Thus grid 1 was set to 2.5 V for all subsequent experiments at Ð17.5 and Ð20ÊkV.

6 . 4 Establishing Resolution at Ð17.5 kV After the new detector had been conditioned for sufficient time, the accelerating potential was increased to Ð17.5 kV. The effects of applying a potential to the probe were then investigated, permitting the analysis of a much larger mass range and enabling spectra to be recorded for sufficient species to determine the resolution characteristics of the analyser.

6.4.1 Methods Used in Experiments Establishing Resolution at Ð17.5 kV Samples of the fullerene standard (1 mg/mL), insulin chain B (1 mg/mL), insulin (2 mg/mL), ubiquitin (2 mg/mL) and ribonuclease (3 mg/mL) were prepared with DHB matrix (10 mg/mL), utilising the dried droplet technique explained in chapter 2, section 2.5.3. The accelerating potential was set to Ð17.5ÊkV and mirror backplate to approximately +1,100ÊV. Grid 1 was biased to 2.5ÊV, reducing field penetration into the fill up region. The push out pulse potential was adjusted from approximately 800 to 960ÊV for each mass, to determine whether the voltage for optimal resolution varied as a function of mass. Laser power was set just above threshold levels, with the attenuator set

84 from 0.4 to 0.8, as required. It was anticipated that this experiment would allow the TOF axis velocity and position results obtained on the linear instrument to be extended over a larger mass range. The sole difference in experimental design was that push out pulse, rather than grid 2, was used as the variable potential for this instrument.

It would have been preferable to adjust grid 2 to optimise resolution, since the grid 2 potential only affects the focusing characteristic of the instrument and does not affect the final TOF axis component of ion energy/velocity, unlike variations in the push out pulse voltage, which affects both focusing and final energy/velocity. It was not, however, possible to adjust grid 2 without changing the accelerating potential in the reflecting geometry oa-TOFMS, since design restrictions ensured there was no separate feedthrough for grid 2. Instead, grid 2 was set to be a fixed proportion of the accelerating potential by means of a resistance installed inside the vacuum chamber. Fortunately, the variation in final TOF axis ion energy from a 160ÊV variation in the push out pulse was less than 1% of the total, so the effects of energy changes were relatively small, although it must be noted that any reduction in the push out pulse, if required to improve overall resolution, would provide a small increase in the turn around time.

A positive potential was applied to the sample probe, to accelerate MALDI ions in the source axis and give them the velocity required to reach the detector. The probe potential would act to accelerate ions to a final source axis velocity in the space between the probe and target region shield. The required velocity was obtained from the following calculations. The flight times for a number of ions accelerated with the potentials indicated above (push out potential was assumed to be 880ÊV) were calculated with an Excel spreadsheet. From this, the desorption velocity and PKEsa required for ions to travel 130Êmm, the distance from the middle of the oa to the middle of the detector, during this flight time was obtained. The same calculation was performed, with the ions travelling 75Êmm or 185Êmm, instead of 130Êmm during this period. These represented desorption axis displacements of the slowest or fastest ions respectively, given by the edges of the oa and active detector surface. This gave the allowed range of delay times for the accelerating potential, plotted for m/z 100 to 20,000ÊDa in figure 6.8. The potential required to give ions the correct final velocity in the desorption axis could be calculated, provided the desorption velocity reached at the end of the MALDI process was known. Expected desorption velocities were estimated from experiments on the linear instrument and previously published values. Table 6.3 gives important parameters and the desorption velocities assumed for the m/z values used in the calculations. The plot of required desorption potentials is given in figure 6.8.

200

150 s)

µ A 100

Delay time ( B

0 0 2000 4000 6000 8000 10000 12000 14000 16000 18000 20000 m/z (Da)

200

150

100

173 eV 50

Desorption potential (V) 0 85 eV

-50

28 eV

-100 0 2000 4000 6000 8000 10000 12000 14000 16000 18000 20000 m/z (Da)

Figure 6.8: Graphs of predicted delay times and desorption potentials required for analysis of ions from m/z 100 to 20,000, when accelerating potential was set to -17.5 kV. The best sensitivity is expected with delay time set between the two solid curves on the top graph, which represent the delay for the slowest ions to reach the begining of the oa (line A) or the fastest ions to go to the end of the oa (line B). The dashed lines on the top graph represent the extremes of delay, beyond which no ions are expected to be detected. In the bottom graph, the solid line is an estimate of the desorption potential required to give ions 85 eV, the potential for ions to travel from the middle of the oa to the middle of the detector. The dashed curves represent the estimated potentials required to give ions the maximum (173 eV) or minimum (28 eV) source axis energy required to reach the detector at this analyser potential. It must be noted that the plots provided in figure 6.8 are only correct if the following assumptions are true: (i) ions reach their final source axis velocity very rapidly, required for the delay time plot; and (ii) for the desorption potentials plot to be correct, both (a) the estimated desorption velocities are correct and (b) ions experience the full potential applied to the probe. Assumption (ii)(b) in particular is unlikely to be (strictly) correct, since it ignores the finite thickness of the sample and the distance desorbed neutrals may travel prior to ion formation, both of which could be significant on the scale of the 1- 2Êmm distance over which the probe potential acts. Thus it was possible that higher probe potentials could be required than those predicted by figure 6.8. Further, the desorption velocities for analyte ions may be reduced by the accelerating potential, because this could reduce the number of collisions between desorbed matrix and the analyte ions.

Owing to the limitations of the assumptions, the two graphs given in figure 6.8 were only used as a guide in determining the probe potential and delay before triggering the push out pulse to be used in this series of experiments. The number of shots averaged for spectra obtained, together with actual probe potentials, will be given with the results. Delay time was always set within the range given by figure 6.8.

6.4.2 Resolution Obtained at Ð17.5 kV Good spectra were obtained for all the species investigated, up to the mass of insulin (m/z 5,733ÊDa). Results at higher mass, for ubiquitin and ribonuclease, required improvements in sensitivity, explained below in section 6.5. Figure 6.9 shows the result for C60, figure 6.10 for low mass background ions and figure 6.11 for insulin. The result for insulin chain B has already been shown in figure 6.7(b). Thus applying a potential to the probe allowed the mass spectrometer to be used over a wide mass range, enabling analysis of samples that did not provide spectra with the linear oa-TOF instrument.

Optimal resolution was obtained for all these species, including ubiquitin and ribonuclease, when push out pulse potential was set to 895±10ÊV. The uncertainty of 10ÊV was based on the accuracy of FWHM peak measurements. The observation that a single potential provided optimal resolution for m/z from less than 100 to over 12,000ÊDa indicated that initial velocity and position in the TOF axis were not correlated in this instrument. This lack of correlation may be due, at least in part, to the application of a potential to the probe, since the potential between the probe and shield acts as a lens, deflecting ions in the TOF axis, as can be seen from figureÊ6.12. Spectra for ubiquitin and ribonuclease were obtained using a focusing lens as well, which had an even larger effect on TOF axis position and velocity. Thus any possible correlation between position

R = 2800

28.8 28.9 29.0 29.1 29.2 time (µs)

Figure 6.9: Spectrum of C60 in DHB, with peaks for m/z 720 - 723 Da, obtained at an accelerating potential of -17.5 kV from 20 averaged shots, listed with the mean resolution. The probe was set to 162 V.

matrix fragments

K+ + Na

6 8 10 12 14 time (µs)

Figure 6.10: Spectrum of low mass background obtained from DHB matrix, with an accelerating potential of -17.5 kV from 20 averaged shots. The probe was set to 162 V.

77 78 79 80 81 time (µs)

Spectrum of insulin in DHB obtained at an accelerating potential of Figure 6.11: + -17.5 kV from 35 averaged shots. The large peak represents the [M+H] cluster (m/z 5734.6 Da). The probe was set to 74 V and a 200 MHz Chebyshev was applied.

ion trajectories

part of the fill up region and oa

probe +100 V

grounded shield 5 mm

Figure 6.12: I-Opt simulation (planar) of the source region with a biased probe, showing the deflection of m/z 500 Da ions at different starting positions. Initial ion velocity (desorption axis) was 2000 m/s and probe was set to 100 V. The ions had an initial velocity of zero in the TOF axis. and velocity due to the desorption/ionisation process was not conserved in these experiments. This is different to experiments performed in the linear instrument, where ions were left to reach the fill up region only with energy from the desorption process. However, these experiments indicate that a single push out pulse potential can be used for all masses, providing mirror and accelerator potential do not change.

Resolution for the C60 molecular ion region was 2,800. This was a 75% increase on results obtained with a Ð15ÊkV accelerating potential, a much larger improvement than that due to grid 1 field penetration effects alone. Since the same methods of sample preparation were used for all accelerating potentials, at least part of this improvement in resolution was due to improved analyser focusing characteristics at the larger accelerating potential. These improved focusing characteristics probably related to (1) decreased turn around time, due to the higher push out pulse required at -17.5ÊkV and (2) reduced initial TOF direction spatial dispersion, since the application of the probe potential had a lens effect that could reduce the population of ions sampled with a significant TOF direction spatial dispersion, since these ions were deflected more than ions closer to the lens axis (see figure 6.12).

A working mass spectrometer requires more than just good resolution. Mass accuracy is a critical requirement and sensitivity is also important. For a well designed oa-TOF instrument, external calibration should give results almost as good as those obtained with an internal standard, owing to the inherent mass axis stability (explained in chapter 1, section 1.3). Mass accuracy when calibrated with an internal standard was previously determined to be approximately 30Êppm (one standard deviation) in earlier experiments performed with the fullerene standard. Unfortunately, mass accuracy found with external calibration in these experiments was substantially poorer than 30Êppm. Measured TOF for a single mass was observed to drift by up to tens of nanoseconds over a period of days, corresponding to several masses or thousands of ppm in the mass range used. Significant drifts in TOF were even observed over a single day. When the time axis is this unstable it is not practical to use external calibration for the identification of unknown species. Thus, all the spectra given in this section were obtained from standards, for which mass assignments could be made from approximate time of flight values predicted with SimTOF, and plotted with a time rather than mass axis. Mass accuracy had to be improved before the instrument could be considered fully functioning. The most likely causes of mass axis instability were power supply drift or timing drift. Detector sensitivity fell before these possible causes of mass calibration drift could be thoroughly investigated, so this important issue was resolved later, at an accelerating potential of Ð20.0ÊkV and the process is described in chapter 7, section 7.3.

87 6 . 5 Improving Sensitivity 6.5.1 S/N and Approaches to Improving Sensitivity This instrument provided very poor sensitivity at high mass. Signals were only obtained with difficulty for insulin (~5,700ÊDa) and no significant signal was detected for species with m/z of over 6,000, such as ubiquitin (~8,600ÊDa) and ribonuclease (~13,700ÊDa). Thus attempts were made to improve the sensitivity of the instrument, to provide an extension to mass spectrometer’s mass range.

The key factor that determines sensitivity in any analytical technique is the signal to noise ratio (S/N). The quality of spectra is directly linked to it, as is the limit of detection. It is known that, in general, S/N increases with the square root of the number of samples [202]: ∝ S/Noverall N =× or S/Noverall S/N acq N (6.1) where ‘S/Noverall’ is the S/N of the averaged signal, ‘N’ is the number of acquisitions and

‘S/Nacq’ is the S/N in (an average) single acquisition.

Detector signals in mass spectrometry are often proportional to the number of incident ions, at least within certain limits, permitting accurate measurements of relative isotope abundances. This suggests that overall signal in a single acquisition can be determined from the multiple of the average signal of a single ion peak (Ssi) by the number of single ions in an acquisition (n). Substituting this into equation 6.1 provides: × = Ssi n × S/Noverall N (6.2) Nacq

It had been observed that the main source of noise in spectra recorded on the reflecting MALDI-oa-TOF was from the high speed digitiser. This contributed a certain level of noise for each time channel occupied by the ion signals. The spread in detector arrival times and contributions of isotope distribution ensure that the signal for individual clusters is spread over a larger number of time channels than single ions, with noise proportional to the number of time channels. Assuming no other significant contribution to noise, this relationship between time channels and noise can be (crudely) accounted for by replacing Nacq with the multiple of the noise per single ion width (Nbin) and number of single ion widths (‘bins’), generating equation 6.3: S × n S/N = si × N (6.3) overall × Nbin bins

88 Owing to the assumptions made, equation 6.3 can only be used to provide a rough estimate of the S/N expected for a given experiment. The main function of this equation is to indicate the parameters that affect the signal and those that affect the noise, allowing an assessment of methods that can be used to improve S/N. According to the equation, improvements in signal can be obtained by (i) increasing N by averaging over a larger number of acquisitions, (ii) increasing Ssi, the signal from single ions, or (iii) increasing n, the number of single ions per acquisition; while noise can be reduced by (i) decreasing

Nbin, the noise on a single ion signal, or (ii) decreasing the width of the signal to decrease the number of bins.

Averaging over more acquisitions is perhaps the most commonly used approach for improving S/N, adopted in most MALDI instruments. Signal averaging involves a simple mathematic function available with many data systems and it does not require modification of the analyser. Averaging acquisitions theoretically allows the attainment of almost any S/N, provided each individual acquisition has a S/N of greater than 1. In practice this is not possible, since there are only a limited number of samples and S/N only improves with the square root of the number of acquisitions, creating diminishing returns as acquisition number increases. Additionally, samples that generated very weak MALDI signals on our instrument, typically higher mass species, did not give a signal for every shot, so averaging many shots gave little improvement in S/N. Thus improvements in sensitivity had to involve adjusting other factors.

Decreasing the level of noise, by decreasing either the noise per single ion or number of single ion peak bins per mass would also provide some improvement in S/N. The width of single ion peaks was, however, essentially fixed by the detection system and the level of noise during this period was also fixed, since it was intrinsic to the oscilloscope used to record spectra. The number of bins per mass, including isotope effects, was determined by the width of the peaks. The mass spectrometer had already been designed to minimise peak width, since this provided the highest resolution. Thus it was not possible to improve S/N further by reducing the magnitude of the noise.

It was possible to try to improve the other two parameters given in equation 6.3, the signal for single ions or number of ions detected from a single acquisition. An amplifier was installed, in an attempt to improve signal for single ions, without providing a similar increase in the background noise. This is discussed in section 6.5.3. The number of ions detected from a single acquisition is dependant upon many factors, including grid transmission efficiencies, the detection yield of the detector and sampling efficiency. Grid transmission and detector efficiencies will be discussed in chapters 8 and 9, but improving the sampling efficiency is within the scope of this section. 89 Sampling efficiency is linked to the proportion of ions generated that are able to reach the detector. Even for the reflecting MALDI-oa-TOF sampling efficiency is quite low, since the narrow sampling slit ensures that only the central section of the desorption plume is able to enter the analyser. This is necessary to ensure high resolution. Improving sensitivity by widening the slit significantly would result in an unacceptable decrease in resolution. A small increase in slit width would be unlikely to increase sensitivity much, since application of a small accelerating potential to the probe has ensured an even larger dispersion of ions in the TOF axis. Another approach to improving sampling efficiency was to use a lens to prevent ions from dispersing, due to ion velocity in the TOF axis, investigated in detail below.

6.5.2 Sensitivity Improvement with an Einzel Lens An einzel lens was constructed for use in experiments at Ð17.5ÊkV. The lens electrode was 10Êmm long, with an internal diameter of 15.5Êmm and an outer diameter of 25.5Êmm. The installed lens was simulated with a scale model drawn in I-Opt, with cylindrical symmetry. A cross sectional view of the model is illustrated in figureÊ6.13. The lens was designed to give ions a parallel trajectory, providing they started from the middle of aperture and appropriate potentials were applied to the lens and sample probe, as illustrated for m/z 2,000ÊDa in figure 6.13(a). This would provide some improvement in sensitivity, by reducing the spatial spread of ions. Resolution would not be sacrificed, since ions that entered the fill up region would not have an increased velocity spread in the TOF axis.

It was experimentally difficult to determine the exact potential that gave parallel ion trajectories, so the lens potential was adjusted instead to give maximum sensitivity. This had the undesirable side effect of decreasing resolution, by overcorrecting ion divergence and introducing a spatial and velocity spread in the fill up region of the orthogonal accelerator, illustrated in figure 6.13(b).

Preliminary qualitative experiments indicated that the einzel lens provided a significant increase in S/N and sensitivity. It allowed the useful mass range of the instrument to be extended up to almost 14,000ÊDa. Good signals were obtained for both ubiquitin (figureÊ6.14) and ribonuclease (figureÊ6.15) from the samples prepared for the Ð17.5ÊkV resolution experiments. Neither of these samples gave reasonable spectra without using the lens.

ion trajectories

to fill up region lens (82 V)

probe (150 V)

ion trajectories

lens (105 V)

probe (150 V)

Figure 6.13: An I-Opt simulation of 2000 Da ions being focused by the einzel lens. Ions start on the desorption axis, with a velocity of 1000 m/s and an angle of 0 - 25°, in 5° increments, to the axis. All unlabelled parts are electrically grounded. In (A) the lens potential is set to create approximately parallel ion trajectories, while in (B) the lens potential was increased to ensure the maximum number of ions entered the accelerator.

8564.9 Da

92 93 94 95 96 97 98 time (µs)

Figure 6.14: Spectrum of ubiquitin in DHB obtained at an accelerating potential of -17.5 kV from 60 averaged shots. The probe was set to 150 V and the lens to 80 V. A 20 MHz Chebyshev was applied to reduce noise.

13683.3 Da

118 119 120 121 122 123 124 time (µs)

Figure 6.15: Spectrum of ribonuclease in DHB obtained at an accelerating potential of -17.5 kV from 80 averaged shots. The probe was set to 135 V and the lens to 80 V. A 5 MHz Chebyshev was applied to reduce noise. A small number of experiments were conducted to quantify the sensitivity gain obtained with the lens, utilising the fresh samples of fullerene standard and insulin, both of which had given results without the lens. The samples were deposited in the same manner and from the same stock solutions as those used in section 6.4. Analyser potentials were set as given in section 6.4. Only one sampling point was used for each sample to obtain all the data, to improve the reproducibility. For the fullerene sample the signal for C60 was monitored and the probe bias was set to 161ÊV. The lens potential was sequentially set to 0 and then 60ÊV. Ten shots were fired at each lens setting and the peak area recorded for m/z 720 and 721ÊDa. This process was performed twice before the sample became depleted. In the experiment with insulin the probe was set to 153ÊV and the delay time to 50Êµs. Lens potential was sequentially set to 80ÊV, 0ÊV and 80ÊV. Thirty shots were averaged for each and the area of the M+H (5734.6ÊDa) cluster recorded. After this the sample became depleted. For both experiments the lens potential was chosen on the basis of I-Opt simulations.

The results from these experiments are given in table 6.4, showing an overall gain of approximately 7 for the C60 molecular ion cluster and 10 for insulin M+H cluster. This suggests that the lens functions more efficiently as mass increases, but this conclusion must be tentative, since only two compounds were investigated. However, it is safe to conclude that these experiments confirmed that the lens does increase sensitivity, as predicted, by ensuring more ions enter the fill up region. Unfortunately, it also decreased resolution, as was expected.

Thus, overall, the performance of the einzel lens was somewhat disappointing, since its modest increase in gain was at the expense of resolution. The most significant limitation of this type of lens design was that it could only generate parallel trajectories for ions with a single PKEsa in a given analysis. Ions with a higher PKEsa would not be corrected, resulting in a loss of sensitivity, while ions with a lower PKEsa would be over corrected, resulting in a lower resolution. Thus it could not efficiently correct ions generated by MALDI, since isobaric ions generated with this technique are known to have a range of desorption energies (PKEsa in the oa-TOF geometry) as has been detailed in chapter 5.

6.5.3 Sensitivity Improvement with a Pulsed Lens The pulsed lens was designed to address a significant limitation of the einzel lens system, by providing a reduction in the spread of PKEsa for isobaric ions. The most important innovation of this design was that the accelerating and focusing potentials were not applied continuously and were instead pulsed on a short period (µs delay) after the ions

91 Table 6.4: Increase in signal due to the einzel lens for (A) C60 and (B) insulin. Results in (A) are the average of two acquisitions, while the results in (B) were obtained in the order 1 Ð 3.

(A) Gain Observed for C60 lens (V) relative areas (normalised) 720 Da 721 Da pooled 0 1.0 1.0 1.0 60 7.1 6.4 6.8

(B) Gain Observed for Insulin [M+H]+ cluster spectrum lens (V) relative area 18011 2 0 1.0 38011

Table 6.5: Increase in signal due to the pulsed lens for insulin chain B and insulin

insulin chain B insulin relative area relative area probe (185 V) 1.0 probe (180 V) 1.0 probe (185 V) 1.6 lens (120 V) 27 lens (130 V) 30 lens (120 V) 17 mean gain 23 mean gain 22 were formed. This permitted the initial spread in ion kinetic energy to translate into a spatial dispersion in the source and TOF axes, which can be corrected with time-lag focusing methods. The idea was that appropriate settings of the lens potential and delay time would permit the spread in PKEsa to be reduced to a level that would permit a large proportion of ions to be analysed with a single delay time.

The most critical element in the pulsed lens was a single cup shaped focusing electrode. This electrode was 5 mm wide (diameter) and 12.3 mm long, with a 1 mm diameter sampling orifice in its base, replacing the original sampling aperture. The lens was simulated with a scale model of the source region drawn in I-Opt, with cylindrical symmetry. A cross sectional view of the model is illustrated in figureÊ6.16. There were important differences between this device and the einzel lens, which suggested that sensitivity should be higher with this device. Firstly, this lens assembly was narrower, preventing the ion plume from expanding as much in the TOF axis. Secondly, the main focusing element in this device was pulsed on at some time after the laser was fired, but before the push out pulse was triggered. This allowed the lens to act as a time-lag velocity focusing device, and if the delay and potential were set appropriately it would ensure that isobaric ions had approximately the same amount of PKEsa, allowing virtually all of them to be analysed with a single delay time. Finally, because the pulsed element was cup shaped, ions could receive all additional energy required from the lens and the probe could be left at ground, preventing the sampling orifice from dispersing the ions in the TOF axis.

FigureÊ6.16 shows the simulated trajectories for ions of 2,000ÊDa, with an initial desorption velocity of 1,000Êms-1, being accelerated and focused with a delay of 7Êµs after formation. The top two trajectories show how the lens is able to focus ions that are that located on the source axis (same as the lens axis) when the lens is triggered, with the result that these ions have trajectories parallel to the source axis and are able to pass through the fill up region entry slit, even where the initial trajectory is 10¡ away from being parallel. This is an improvement on the focusing ability of the first lens, which created a parallel ion beam, but allowed the ions to disperse. Unfortunately, like the first lens, it is unable to focus ions that are not on the axis, shown by the bottom two trajectories. These ions, which start only 0.5Êmm below the axis (a distance corresponding to the geometric limits of the sampling orifice) are unable to pass through to the fill up region, even though they could pass through the first aperture. Note that this failure of transmission is difficult to see in figure 6.16, owing to the limits of resolution in the bitmap image file.

trajectories for ions starting from the axis

to fill up cup shaped lens region element (150 V)

trajectories for ions starting 0.5mm below the axis

Figure 6.16: An I-Opt simulation of 2,000 Da ions with an initial velocity of 1,000 m/s being focused by the pulsed lens.The ions were allowed to desorb for 7 µs before the potential was applied. Ions were either started on the desorption axis, with an angle of 0 or 10°, or started with similar angles to the axis, 0.5mm below it. All parts in the diagram, except the lens, were electrically grounded. An experiment was performed to determine the actual gain due to this lens. The accelerator was set to Ð20.0ÊkV, with the push out pulse at 1,005ÊV and mirror at 1,300ÊV, the potentials used in chapter 7. Samples of insulin chain B and insulin were prepared by the dried droplet method. The area of the [M+H]+ cluster was obtained for each species, after 50 to 100 shots were averaged. All spectra were obtained from a single sample for each species, with three spectra recorded for each species. For each spectrum the lens was grounded and the probe biased or the lens biased and the probe grounded, as specified in the results. The delay between laser firing and application of the lens potential was set to 7.9Êµs with a Data Dynamics 5113 (analog) pulse generator (Englewood, NJ, USA). The Data Dynamics device triggered a high voltage pulse generator, which applied the preset voltage to the lens.

The normalised area results for each analyte are given in table 6.5, clearly indicating that the increase in signal for both these masses is ~20, twice the gain experienced for insulin with the einzel lens. However, resolution was still affected. For instance, resolution for insulin chain B was approximately 7,800 when the lens was not used, but this dropped to approximately 4,500 when the lens was used. Thus, while experiments show that this lens is superior to the cylindrical design, it is still not perfect. It may be possible to improve the result by fine tuning the delay between laser and lens triggering, in the same manner as the push out pulse delay, but this would require a more precisely adjustable set digital delay device.

6.5.4 Signal Preamplifier The background noise generated by the high speed digitiser of the LeCroy 9384 Oscilloscope had a larger relative effect when the digitiser was operated at its highest gain levels of 2 or 5 mV/division, required for very weak signals. S/N was measured to be approximately 5 for single ions acquired with these gain settings. It was believed that most of the measured noise was due to the digitiser, and that the (relative) noise from the digitiser decreased when the gain on oscilloscope was reduced. This means that one way to improve S/N for weak signals was to amplify the detector’s signal before measuring it with the oscilloscope and providing a good amplifier was used, the background noise would increase by a smaller amount than the signal, increasing S/N.

A signal preamplifier was obtained, the EG&G ORTEC Model 9306 1 GHz Preamplifier (Oak Ridge, TN, USA). It provided a nominal gain of 100 to two separate output channels, from a single input. Noise was rated at <Ê100ÊµV rms over the full bandwidth of 100ÊkHz to 1ÊGHz (3Êdb) and the claimed output risetime was 350Êps. When

93 required for experiments, the input of the preamplifier was connected to the detector’s collector plate. Output 1 was then connected to the oscilloscope, which was set to an appropriate vertical axis setting (0.2ÊV/div or greater). The remaining output channel was either terminated (50ÊΩ) or connected to a second measuring device, if required. The effect of the preamplifier on single-ion like signals monitored with our oscilloscope was characterised in a simple experiment, by comparing results obtained with and without the preamplifier attached.

The analyser potential was set to Ð20ÊkV, with a detector bias of 2.94ÊkV, and left to warm up for 30Êminutes. The oscilloscope was set up as shown in table 6.6, to trigger on the background signal, which produced narrow peaks that (on average) did not appear significantly smaller than single ion signals. After the oscilloscope had finished acquiring data (30 acquisitions) the spectrum was saved. Fifteen spectra were obtained with the amplifier attached and another fifteen without it. The entire acquisition period took less than 30 minutes. FigureÊ6.17 shows a sample spectrum (a) obtained without and (b) with the preamplifier attached. After all spectra were acquired the peak heights, areas and FWHM were extracted from the saved data. These means and standard deviations of these results are given in tableÊ6.7, together with the calculated range of gains.

The overall gain on the signal, measured from both the height and area of the peak, are well within the specifications of the device (gain of 50 to 150). Peak area evidenced a higher gain, due to the broadening of the peak. This was to be expected, since the mean peak broadening, measured at half height, was 380Êps, only slightly larger than the manufacturer’s quoted risetime of 350 ps. This is equivalent to an approximately 10% increase in peak width for a signal at a single m/z value. Importantly, there was a significant improvement in the signal to noise ratio, with a relative increase of between 1.1 Ð 2.6, measured at the 66% confidence limit (one standard deviation from the mean). Besides a slight broadening, the only other negative effect of the amplifier was an increase in ringing, evidenced by the presence of a small satellite peak, approximately 20Êns after the main peak, observed in all spectra. This may have been due to an impedance mismatch between the cabling and the amplifier. Thus the preamplifier behaved largely as expected, giving a significant improvement in signal to noise for only a slight increase in peak width and small ghost peak. Use of the preamplifier caused a lower decrease in resolution than the lenses, so the preamplification was the first method used to improve sensitivity in subsequent experiments.

94 Table 6.6: Oscilloscope settings for experiments used to characterize the preamplifier

parameter preamplifier not used preamplifier used volts per division 5 mV 0.5 V trigger level (on signal) -1.0 mV (LFREJ) -0.1 V (LFREJ) sweeps averaged 30 30 time scale 0.2 µs 0.2 µs sampling rate 4 GS/s 4 GS/s

Table 6.7: A summary of results from experiments characterizing the preamplifier, with uncertainties expressed as ± the standard deviation of the results. The range of gains given by the full range of uncertainty is quoted.

preamplifier not used preamplifier used overall gain (multiple) peak height (V) 0.0150 ± 0.0011 1.37 ± 0.06 81 Ð 103 peak area (V.s) (2.51 ± 0.16) x 10-11 (2.65 ± 0.26) x 10-9 90 Ð 124 peak width [FWHM] (ns) 1.51 ± 0.23 1.89 ± 0.27 0.93 Ð 1.7 S/N 31.9 ± 8.0 51.9 ± 9.9 1.1 Ð 2.6

A 14

6 signal (mV)

-40 -20 0 20 40 time (ns)

B 1.2

1.0

0.8

0.6 signal (V)

0.4

0.2

0.0

-40 -20 0 20 40 time (ns)

Figure 6.17: Signal obtained from background counts (A) without and (B) with the 1 GHz preamplifier attached 6 . 6 Overall Performance at Ð3ÊkV, Ð15ÊkV and Ð17.5ÊkV and Initial Performance at -20ÊkV: Conclusions A number of difficulties encountered with the instrument were solved or reduced in the course of the experiments described in this chapter. Problems due to field penetration were countered and sensitivity was improved. These early experiments indicated that the reflecting oa-TOF provided a better platform for MALDI experiments than the linear oa- TOF, since the larger range of energies accepted, together with its higher sensitivity, allowed a much larger range of m/z values to be investigated. Additionally, it was possible to obtain comparable resolution to commercial MALDI-TOF instruments of similar size, with resolution improving each time accelerating potential was increased. Only one more issue had to be solved before it could be called an excellent instrument. It had to demonstrate good mass axis stability, a requirement for any mass spectrometer. This issue was resolved when the instrument was operated at Ð20 kV and will be discussed in the next chapter.

95 Chapter 7: Improving Mass Accuracy and

Characterizing the Analyser of the Reflecting

MALDI-oa-TOF

This chapter describes the detailed experiments performed to assess two important characteristics of the reflecting MALDI-oa-TOF analyser, resolution and mass accuracy, when the analyser was operated at an accelerating potential of Ð20ÊkV. Prior to assessing these characteristics, it was important to determine what caused the drift in mass measurements mentioned in chapter 6, in order to improve the mass accuracy characteristic of the instrument. A section has also been included to illustrate the range of samples that were mass analysed with this mass spectrometer and an estimate has been made of the reflecting MALDI-oa-TOF instrument’s sensitivity. Before commencing this analysis, it is important to understand the theoretical focusing characteristics of this oa- TOF analyser.

7 . 1 Ion Optics at Ð20.0 kV This MALDI-oa-TOFMS was designed to operate at an accelerating potential of Ð20.0ÊkV, with this potential chosen to provide a small turn around time and permit the detection of higher mass ions. A higher potential was not used, owing to the risks of discharging in the vacuum chamber and the upper voltage limits of the feedthroughs used. The performance of the mass analyser, in particular its resolution is, to a large extent, based upon the ion optics. In a TOFMS, the most important parameters determining ∆ ∆ ∆ resolution are the magnitude of the initial dispersions ( t0, s0 and u0) and the ability of the mass spectrometer to minimise the effect of these dispersions on resolution, as ∆ explained in sections 1.2.3 and 1.2.5. Temporal dispersion ( t0) is very small owing to the pulsed nature of the oa. The ability of the mass spectrometer to minimise the effects of ∆ ∆ ∆ spatial ( s0) and energy ( U0, calculated from u0) dispersions is related to the locations of two key focusing points: the spatial focus and the energy focus.

The location of the spatial focus in this instrument could not be accurately calculated with equationÊ1.12, since, in practice, the oa consisted of a three stage accelerator, rather than the two stages required for equation 1.12. The spatial focus could have been found with an equation for a 3 stage accelerator [41], but it was decided to determine its location

96 with SimTOF instead. The mass analyser was simulated in linear mode, with accelerator potentials and dimensions given in chapter 2, table 2.2. The location of spatial focus was determined by adjusting the length of the drift region to give the highest resolution for ∆ ions with a s0 (TOF direction) of ±1Êmm, corresponding to the width of the oa entrance slit. The location of the spatial focus was found to be 610Êmm from gridÊ3, i.e. beyond the entrance grid of the ion mirror. This is different from a typical reflecting geometry TOFMS, which has a spatial focus close to the accelerator, able to act as a virtual source for the mirror, as illustrated in figure 1.6 and explained in section 1.2.5. The reflecting oa-TOF instrument was, however, designed so that it could be easily adapted in the future to switch between linear and reflecting oa-TOFMS operation, if required. Thus the spatial focus was located close to the backplate of the mirror, since this distance represented the longest drift region available within this chamber if the mirror field was turned off (linear mode). The location of the energy focus was determined with SimTOF simulating the instrument in reflecting mode, by determining the effect on resolution of changing the location of the detector, to find the location providing the highest resolution. The simulation indicated that the energy focus was located at the detector plane, for ions with ∆ U0 (TOF direction) of ±35ÊmeV (the maximum energy distribution permitted for a point source from geometry restrictions), within the limits of engineering tolerance (±0.1Êmm for the push out plate, grids and detector).

An attempt was made to use SimTOF to calculate the expected resolution, utilising the geometry of the instrument and the TOF direction energy acquired by ions with 20ÊkV accelerating potential, assuming zero averaged distributions and determining the maximum range for ions travelling from the midpoint of the oa to the midpoint of the ∆ detector in the source axis direction. This provided a s0 (TOF direction) distribution of approximately ±1Êmm, within the radial distribution of ions observed in MALDI plume ∆ imaging experiments conducted by Puretzky and colleagues [127], and a U0 (TOF direction) energy distribution of approximately ±35ÊmeV. The resolutions calculated with this distribution were much lower than those observed in the spectra provided in chapter 6 (or given later in this chapter). This suggests that the initial dispersions in this instrument ∆ were smaller than the geometry limits. A number of arbitrary combinations of s0 and ∆ U0 (for instance ±0.8Êmm and 8ÊmeV, or ±1Êmm and 5ÊmeV) would provide similar resolutions to those observed in the experiments, but it was not possible to independently measure the actual dispersions with the instrument. Thus the initial dispersions could not be determined and simulations could not be used to predict the actual resolution that would be obtained, although the simulations could still be used to investigate the relative

97 effects on instrument performance (resolution or TOF) of changing tuning potentials or ∆ electrode positions, using experimental results to confirm reasonable settings for s0 and ∆ U 0.

7 . 2 Optimising the Source and Analyser Potentials 7.2.1 Push Out Pulse Shape and Potential The push out pulse (POP) generator is a critical component of any oa-TOF mass spectrometer. Ideally, its potential would reach the set POP voltage almost instantaneously when triggered and remain there until after the highest mass ions had entered the next accelerating region. In practice it is difficult to create a device that can perfectly ramp a potential from ground to approximately 1,000ÊV almost instantly and hold it there for about 10 µs, so ideal performance could not be assumed. This could affect instrument performance and the accuracy of mass calibration, as will be explained below. Thus it was important to measure the output from the POP generator of the MALDI-oa-TOF, as well as determine the potential that provides optimal resolution.

All measurements of the POP shape were performed with the LeCroy 9384 oscilloscope with a x10 divider, 1ÊM probe attached. The instrument was set up as if ready to acquire a TOF spectrum, to ensure the POP measured was as close to that of a real experiment as possible. The accelerating potentials were turned on, with the drift region at Ð20ÊkV and mirror at approximately Ð1,250ÊV. Since no spectra were acquired, the laser was switched off. The oscilloscope probe was attached to the back of the POP connector, used to transfer the pulse to the push out plate. The following oscilloscope settings were used: 1ÊM input impedance, 10 V/division vertical scale, 2Êµs/division horizontal scale, 2ÊGS/s sampling rate and trigger on the rising edge at 1ÊV. The POP generator was then turned on and the desired potential set. After allowing 30 minutes for the system to warm up, the POP was triggered 100 times and the averaged signal recorded. Unfortunately, the oscilloscope probe had a maximum operating voltage of 700ÊV, so the full POP potential of 1,000ÊV could not be directly observed. Instead the POP was set to 50ÊV, 400ÊV, 600ÊV and 700ÊV.

The actual 11.2Êµs POP obtained at a potential of 600ÊV is shown in figure 7.1 (inset), with the leading edge given in the main figure. The POP must be maintained at maximum amplitude until after all ions to be analysed have left the region between the push out plate and grid 1 (the “push out region”), to ensure all ions of all m/z values receive the same energy, important for both focusing (resolution) and mass calibration. This instrument 98

600 600 500 400 500 300 200 100 400 0 0 5 10 15 time (µs)

300

Pulse Amplitude (V) 200

100

-200 -100 0 100 200 time (ns)

Figure 7.1: Leading edge of a 600 V push-out-pulse, showing a rise time of 40 ns, with the full push-out pulse inset.

4000

C60

C70 3500 average

3000 resolution 2500

2000

980 990 1000 1010 1020 1030 push out pulse (V)

Figure 7.2: Resolution as a function of push out pulse potential for C 60 (720 - 722 Da) and C 70 (840 and 841 Da). At least three resolution measurements were performed at each potential. Table 7.1: Characterizing the push out pulse Ð risetime (5% Ð 95%) of the leading edge and rise in potential over the 11.2 µs duration of the pulse.

push out pulse rise time (10% - 90%) change in potential over pulse (V) (ns) % total time† (V) % total potential 50 18 0.16 1.5 3 100 22 0.20 2.5 2.5 400 32 0.29 10 2.5 600 40 0.36 15 2.5 700 43 0.38 18 2.6

† It would more informative to express this as % total time ions spend in the first accelerating region, but this value would be mass dependent.

Table 7.2: The effects of POP risetime of (60Êns to 100% for 1000ÊV) on mass calibration, determined from TOF data obtained with SimTOF. Calibration coefficients were calculated from the TOF for 3,000 and 5,000ÊDa ions.

correct m/z (Da) TOF (µs) fitted m/z (Da)‡ difference between fitted and correct m/z (ppm) (mDa)

10 3.051 388 10.004 400 4 50 6.784 360 50.009 180 9 100 9.581 969 100.013 130 13 720 25.659 283 720.024 34 24 840 27.712 701 840.026 30 26 1,000 30.234 296 1,000.027 27 27 2,000 42.744 140 2,000.020 10 20 3,000 52.343 131 3,000.000 0 0 5,000 67.564 930 5,000.000 0 0 10,000 95.534 463 10,000.000 0 0 ‡ Fitted m/z were calculated from TOF with the calibration equation: m/z1/2=a.TOF+b, with constants a = 1.04752931ÊDa1/2µs-1 b = -0.032627603ÊDa1/2 was intended for analysis and detection of ions of up to 50,000ÊDa and calculations indicate ions of up to this m/z leave the push out region in under 5Êµs. This means that all detectable ions will have left the push out region while the POP is applied, as required, and thus the falling edge of the POP is of little interest. The pulse shows little ringing and a rise time of 40Êns (from 5% to 95% of maximum height). After reaching its maximum the potential dropped by 15ÊV (2.5%) and then comparatively slowly increased to its maximum over for the remainder of the pulse, before rapidly falling. The rise time and potential change over the duration of the pulse for all experiments is summarised in table 7.1. Extrapolating these results up to a push out potential of 1,000ÊV gives a rise time of the order of 60 Ð 70Êns and a rise in potential across the peak of approximately 25ÊV.

Instability in the push out pulse is known to affect the mass calibration curve, causing it to deviate from linearity, with rise time also having a small effect on the calibration [203]. It was suspected that the rise time of the pulse and its potential change over the full duration may have an effect on mass accuracy, due to mass related energy differences. Low mass ions leave the first accelerating region more rapidly than heavier ions and could be expected to have an energy deficit, since the push out plate potential does not stop rising. It was hoped that this would explain the mass accuracy results for carbon clusters, given in figure 6.3. The effects of rise time on mass accuracy were estimated with SimTOF, by calculating times of flight for ions for a POP of 1,000ÊV with a risetime of 60Êns. The small potential change over the remainder of the pulse was not able to be simulated, but it was likely that the 25ÊV increase over the remainder of the peak would only increase any low mass energy deficit in a similar manner to risetime, expected (qualitatively) to increase any mass error due to risetime. The other potentials used in the simulation were grid 1 at ground, grid 2 at -1.335ÊkV, -20ÊkV for the analyser and 1.3ÊkV for the mirror backplate. Simulated times of flight were obtained for ions with m/z ranging from 10 to 10,000ÊDa. The TOF values obtained for ions of 3,000 and 5,000ÊDa were used to calculate the constants for the calibration equation (equation 1.10) and the calibration was used to calculate a fitted m/z for each simulated flight time. The results are presented in table 7.2 and they reveal a positive relative mass error (ppm) for masses lower than the calibration, as expected, with larger ppm errors at lower mass. The higher apparent masses at low m/z are due to the energy deficit. Calibration was performed with m/z of 3,000 and 5,000ÊDa to allow comparison with the carbon cluster results (figure 6.3) for 720 and 840ÊDa. The mass errors calculated here for these two masses, 30 - 34Êppm, is much less than the errors of over 200Êppm observed in the experiment. This suggests that instability in the potential over the remainder of the POP has a much larger effect on TOF than the risetime, as suggested by Hoyes et al [203].

99 Thus a more stable push out pulse should provide a more accurate calibration relationship when analysing a large m/z range.

An experiment was performed to determine the POP potential that gave optimal resolution. Experiments at Ð17.5ÊkV indicated that a single value should give optimal resolution for all masses, so only two peak ion clusters were investigated, C60 (720 Ð

722ÊDa) and C70 (840 Ð 842ÊDa). A sample of the fullerene standard (1Êmg/mL) with DHB matrix (20Êmg/mL) was prepared by the dried droplet technique. The acceleration potential was set to Ð20.0ÊkV and the mirror back plate held at approximately 1200ÊV. The probe potential was set to 100ÊV to ensure that the analyte ions had a high enough desorption velocity to reach the detector. The POP was adjusted from 975 to 1,035ÊV in 10ÊV increments and two spectra were obtained at each potential. The delay between laser firing and triggering of the POP was 20 µs. FWHM resolution was calculated for each isotopic mass.

The results for each cluster were averaged separately and then together, allowing figure 7.2 to be plotted for resolution as a function of POP potential. Optimal resolution appeared over the potential range 995 Ð 1,005 V for C60. The trend was less clear for C70, but the optimal potential for the average result was 1,005 V, so this potential was applied to the plate in subsequent experiments.

7.2.2 Mirror Backplate Potential It was very difficult to determine the exact effect of varying mirror back plate potential with the Brandenburg power supply, since it could not be adjusted or measured precisely during experiments. The Brandenburg power supply was thus replaced with a Glassman power supply, to allow for precise control and measurement of the backplate potential. This Glassman supply was the same model as the power supply used for the POP generator. Before connecting the Glassman supply to the back plate, the voltage from the Brandenburg supply was measured to be 1,239.2±0.1ÊV. The Brandenburg supply had not been adjusted since the POP potential had been tuned. The Glassman supply’s potential was adjusted until it was at 1,240.2 V, according to the same measuring device. According to the Set Potentials VI, running on the computer, the potential on the Glassman supply was 1,255.2ÊV. The Glassman supply was then connected to the mirror back plate and monitored only with the computer.

An experiment, similar to that used to determine the effect of adjusting the POP potential, was performed to investigate the effect of adjusting the mirror back plate potential on resolution. The fullerene sample and accelerator potential given in section 7.2.1 were

100 used again in this experiment. The push out plate was set to 1,005ÊV and mirror back plate adjusted from 1,100 to 1,280ÊV. It was important to ensure the back plate potential was higher than the POP potential, to ensure that all ions would be reflected. Two spectra were obtained at each potential. Resolution results were pooled, since there was no significant difference in resolution for ions in the C60 and C70 clusters.

The results are given in figure 7.3, with error bars on resolution corresponding to one standard deviation around the average values. It was clear that adjusting the back plate potential only had a small effect on resolution, since a change of 180ÊV, almost 15% of the mirror back plate potential, did not appear to have a clear effect on resolution, within the limits of uncertainty in resolution measurements. This can be contrasted with the result for POP, where a much smaller change in potential of 60 V, around 6% of the total POP potential, had a significant effect on resolution. There was however a slight trend to higher resolution as mirror back plate potential was increased, evidenced by the line of best fit. Simulations confirmed this slight trend, so it was decided to set the mirror back plate to 1,300ÊV in future experiments.

7.2.3 Explaining Focusing Criteria: Push Out Pulse and Mirror Potentials Small adjustments to the POP were observed to have a much larger effect on resolution than similar changes in mirror back plate potential due to the ion optics, indicating that the focusing characteristic of the instrument is more sensitive to small changes in the POP than mirror backplate. This is principally due to (i) the different roles (ion optically speaking) of the mirror and first accelerating regions and (ii) more importantly in this instrument, the relative contribution of the POP potential and mirror backplate potential to the total potential change experienced by ions in their respective regions.

The POP represents the potential used to accelerate ions in the first accelerating region. Experiments described in chapter 6 suggested that, when probe potentials were applied or lens elements utilised, the initial energy and spatial spreads were not correlated. Assuming this, the main effects of the POP are (i) to establish the turn around time for ions with TOF direction velocities away from the analyser; and (ii) to provide a spread in the TOF direction energy of ions based upon the initial spatial distribution. A higher POP potential reduces the turn around time, while a single setting of the POP will result in optimal spatial focusing (allowing the ions initially further from the detector to catch up with the initially closer ions at the detector plane), as explained in section 1.2.5. The role of the mirror, detailed in section 1.2.5, is to correct for final TOF direction energy spreads, since for ions of the same m/z the ions with higher energy will penetrate further into and spend longer in the ion mirror than the ions with less energy. An appropriate

101

3800

3600

3400

3200

resolution 3000

2800

2600

2400 1100 1150 1200 1250 1300 mirror backplate (V)

Figure 7.3: Resolution for C60 and C70 as a function of mirror backplate potential. The backplate potential was adjusted from 1100 - 1285 V. A line has been fitted to the data to indicate the overall trend. mirror potential ensures that the difference in time in the ion mirror offsets the longer time that the slower ions spend in the other regions of the TOF analyser. In practice, the entire instrument provides combined spatial and energy focusing, since the energy spread provided by the initial spatial distribution in the fill up region is corrected by the mirror in a similar manner to any other energy spread for ions of the same m/z.

The ability of the reflecting MALDI-oa-TOFMS to focus ions is, amongst other things, related to the velocity ions have in each region; and the lengths of the various regions and electric fields, with the fields defined by the potential differences between elements. In the fill up region, the potential difference experienced by ions is approximately 0.77ÊkV (the exact value depends upon s0 of the individual ions) with all of this energy provided by the 1ÊkV POP. In the ion mirror the ions experience approximately 20.77ÊkV of potential difference in deceleration to zero followed by reacceleration through the same potential, with 6.1Ê% of this potential provided by the mirror backplate supply and the remainder by the accelerator supply. Experimental results confirm that it is more appropriate to compare the effects of changes in field strength, rather than element potential, in affecting resolution. For instance, a 1% change to the field in the fill up region of the accelerator (POP ±10ÊV) near conditions for optimal resolution caused resolution to change by (on average) 400 (see figure 7.2). This change in resolution is, within the limits of uncertainty, roughly equivalent to the loss of 300 in resolution demonstrated by the trend line in figure 7.3, which is due to a 0.9% change in the mirror field (185ÊV change in the backplate).

7.2.4 Delay Time and Probe Potential Ranges Most experiments were performed with a biased probe. After the correct POP and mirror back plate setting had been found, it was necessary to estimate the delay times and probe potentials require for the analysis of ions at an accelerating potential of Ð20.0 kV. The method used to obtain these values was the same as that described in section 6.4.1. The only difference here was the potentials used: 1,005ÊV for the POP, -20.0ÊkV for the analyser and 1,300ÊV for the mirror. Hereinafter these will be referred to as the standard operating potentials, which includes setting grid 1 to 2.5 V. For these settings ions needed a PKEsa of 32 Ð 197ÊeV to be successfully analysed. The delay times and probe potentials required for m/z 100 to 20,000ÊDa are given in figure 7.4. This figure was used to determine the approximate range of delay times and probe potentials for all subsequent experiments where the probe was biased.

102

200

180

160

140

120 s) µ A 100

Delay time ( 80 B 60

0 0 2000 4000 6000 8000 10000 12000 14000 16000 18000 20000 m/z

250

200

150

100 197 eV

50 Desorption potential (V) 0 97 eV

-50 32 eV

-100 0 2000 4000 6000 8000 10000 12000 14000 16000 18000 20000 m/z

Figure 7.4: Graphs of predicted delay times and desorption potentials required for analysis of ions from 100 to 20,000 Da, when accelerating potential was set to -20 kV. The best sensitivity is expected with delay time set between the two solid curves on the top graph, which represent the delay for the slowest ions to reach the begining of the oa (line A) or the fastest ions to go to the end of the oa (line B). The dashed lines on the top graph represent the extremes of delay, beyond which no ions are expected to be detected. In the bottom graph, the solid line is an estimate of the desorption potential required to give ions 97 eV, the potential required for ions to travel from the middle of the oa to the middle of the detector. The dashed curves represent the estimated potentials required to give ions the maximum (197 eV) or minimum (32 eV) source axis energy required to reach the detector at this analyser potential. 7 . 3 Mass Accuracy Instability (External Calibration) 7.3.1 Contributions to Mass Axis Instability The mass spectrometer had suffered from poor mass axis stability in experiments performed with an analyser potential of below Ð20ÊkV, despite its good resolution, making external mass calibration almost useless. This was a significant problem that had to be resolved before the instrument could be considered a success, since good mass axis stability was an expected key advantage of MALDI-oa-TOFMS. Potential sources of this problem may be discerned from the simple calibration relationship, relating the square root of the mass to charge ratio to time of flight (equation 1.10) reproduced here for convenience: mz/.=a TOF + b Any change that affects the TOF or its measurement can alter the coefficients, a and b, introducing discrepancies in external calibration. From the equations used to derive the calibration relationship (equations 1.1 to 1.8) it is clear that anything that affects: (d) the distance ions travel in the TOF axis; (e) the energy (and hence velocity) that ions receive in the TOF axis; or (f) the accuracy and precision of time measurement will change the TOF for ions of the same m/z. If these parameters were varying on a time scale from tens of seconds to two minutes it would have shifted the calibrated m/z of ions within a single spectrum, since spectra were acquired by averaging signals obtained over periods of up to approximately two minutes. This would have caused peak broadening and reduced resolution. Since resolution was still good for spectra acquired over two minutes, whatever affected mass accuracy must have varied significantly over a longer period, either tens of minutes or hours.

The distance ions had to travel was unlikely to have altered significantly for this instrument. Distances between the push out plate, grids, mirror backplate and detector could have varied, albeit very slightly, due to thermal expansion and contraction of the instrument. The instrument was, however, located in an air conditioned laboratory, with the critical components largely insulated by the vacuum chamber, so any changes in temperature after the warm up period were likely to be very small. Further evidence that the laboratory temperature was unlikely to affect distance and hence TOF was provided by the linear MALDI-oa-TOF instrument, which was located within a 2-3Êm distance of the reflecting instrument in the same laboratory. The linear instrument had shown no clear drift in its calibration, so laboratory temperature fluctuations did not result in measurable changes to the distance ions travelled in that instrument. Thus it was reasonable to assume that the distance ions travelled in the TOF axis in reflecting MALDI-oa-TOF were also not noticeably affected by temperature fluctuations.

103 Energy given to ions during acceleration was governed by the power supplies used for the analyser. It was reasonable to assume that there was some variation in these potentials over time and hence it was important to determine their stability.

Actual time-of-flight was measured with a LeCroy 9384 oscilloscope. It automatically checked its calibration every time it was switched on and had been recently checked and serviced by the manufacturer, so it was unlikely that the actual measuring device was inaccurate. Additionally, it was not imprecise, with a sampling rate of 4 GS/s. However, there could be a problem with the reproducibility of timing triggers for the start time of the experiment, which could also result in errors in the measurement of time of flight. The critical start events were the rising edge of the POP, which accelerated ions into the analyser, and initiation of spectral acquisition on the oscilloscope. Any drift in the time difference between these events would change the measured time of flight and hence apparent mass of ions. Thus experiments were performed to determine the contribution of power supply drift and timing triggers to mass errors.

In addition to measuring the drift in accelerating potentials and timing triggers, it was important to quantify the long term mass measurement drift of the instrument. This measured mass drift would represent the overall effect and the combined effects of the various contributions could be compared to it, to ensure that the experiments reported in this section of the thesis did indeed include all significant contributions to the long term mass measurement drift.

7.3.2 Determining Drift on the Power Supplies Drift on the power supplies was measured with a new computer program, the Plot Potentials VI. It allowed the voltage output of the POP, analyser and mirror back plate power supplies to be monitored by the computer over an extended period. This data could then be saved for later analysis. A more detailed description of this program is given in section 3.3.2. The power supplies were started cold for this experiment, so the results would also indicate the warm up time.

The power supplies were switched on to standard operating potentials, with the POP (1,005ÊV) and accelerator (-20.0ÊkV) set from the computer, while the back plate supply (1,300ÊV) was set with its local control, a ten turn potentiometer. The Plot Potentials VI monitored all three power supplies, together with an additional background channel, used to monitor any noise or drift that originated in the computer measurement system, rather than the power supply potentials. The signal and reference inputs for this channel were

104 linked with a 56Ê resistor. Ten measurements were performed for each channel every minute. The experiment was left to run for over 63 hours. Readings from the background channel were scaled in an identical manner to readings from the POP and mirror supplies, where a 1.5ÊkV output from the high voltage supply corresponds to a monitor level of 10ÊV. This value would have to be multiplied by Ð16.67 to be scaled to the same level as the accelerator supply, where a monitor level of 10ÊV corresponds to a high voltage output of Ð25,000ÊV.

The results obtained from monitoring the analyser’s power supplies over an extended period are given in figure 7.5. The top trace, AccÊ(kV), was obtained from the main acceleration supply; BP (V) corresponded to the mirror back plate’s supply; and POPÊ(V) gives the measured potentials for the POP power supply. The trace labelled RefÊ(mV) corresponded to the zero reference potential. A 25 point Savitsky-Golay smoothing function was applied to the data, reducing the effects of changes occurring on a time scale of less than 2 minutes. The actual potentials did not drift significantly on such a short time scale, since resolution did not noticeably decrease in individual averaged spectra, typically obtained over time intervals of up to 2 minutes. Thus, most of this higher frequency variation was due to noise in the measurement system and not actual voltage variation, allowing the smoothing function to be applied.

Variation in potential over the entire experiment was evidenced by the range in potential values in plots smoothed with a 601 point box average (hour timescale) and any overall trend is indicated by the change in voltage of the lines of best fit for each supply. Both the range and trendline were obtained after taking account of the increase potential for all three supplies during the warm up period (first 2 hours). Any variation on a shorter time scale, of several to tens of minutes, was best determined by the standard deviation of applied potentials. The standard deviations and ranges (excluding the first 2 hours) are given in table 7.3, together with the average potential of each power supply. As can be seen, all variations were within the manufacturer’s specified stabilities, with the accelerator and push out pulse supplies well within specifications.

The difference between measured and set potential for the accelerator (-2.41ÊV or 0.012%) and POP (+0.314ÊV or 0.031%) indicates the accuracy (offset error) of this computer based method of setting and monitoring power supplies. The relative error for both of these is close to the 1 bit error expected from the 12-bit DAC used to set these power supplies, 0.024Ê%. In any event, a small offset between the set and monitored value of analyser potentials does not affect mass accuracy, providing it occurs

105

∆V line of best fit

10 ppm -20.0020 3 180 mV -20.0024

-20.0028 Acc (V) x10 -20.0032 1300.10

1300.05 40 ppm

1300.00

BP (V) 1299.95 -75 mV

1299.90

1005.34 10 ppm

1005.32 -1.3 mV

1005.30 POP (V)

1005.28

10% 80 75 70 -0.90 mV 65

Ref (mV) 60 55

0 10 20 30 40 50 60 time (hours)

Figure 7.5: Measurement of the drift on MALDI-oaTOF high voltage power supplies (started cold) from Fri, 15 Oct 1999, 6:19 PM to Mon, 18 Oct 1999, 9:35 AM. A 25 point Savitsky-Golay smoothing function was applied to smooth out variations of under 2 minutes. Two other plots are also included for each power supply: (1) a plot of the data after smoothing with a 601 point box average, to indicate trends in drift on the hour timescale and (2) a line of best fit has been included for each power supply, with change in potential (∆V) as listed. Table 7.3: Drift in analyser power supplies’ potentials, measured with the standard deviation of the measurements (short to medium term drift) and overall change in potential (range). Corrected values were obtained by subtracting the zero reference value from the supplies’ values and the relative part per million (ppm) drifts were calculated from the corrected results only.

accelerator mirror backplate push out pulse zero reference mean potential -20002.41 V 1299.964 V 1005.314 V 0.0675 V

(a) short-medium term drift standard deviation 580 mV 69 mV 27 mV 13 mV corrected standard 370 mV 56 mV 14 mV 0 mV deviation‡ RSD (ppm) 18 43 14 0 RSD total potential (ppm) 17 2.5 0.64 0

(b) Range for box smoothed results after warm up ∆V 238 mV 106 mV 15 mV 4.2 mV corrected ∆V¤ 234 mV 102 mV 11 mV 0 mV ppm 12 78 11 0 ppm total V 10 4.6 0.49 0

(c) stability specifications 1 hour ± 2500 mV ± 150 mV ± 150 mV Ð 8 hours ± 12500 mV ± 750 mV ± 750 mV Ð ‡ standard deviation less standard deviation of the zero reference, scaled as appropriate (x16.7 for accelerator, x1 for mirror backplate and push out pulse). ¤ ∆V less ∆V of the zero reference, scaled as appropriate (x16.7 for accelerator, x1 for mirror backplate and push out pulse). consistently7, since this represents the systematic error in the method of measurement, while mass accuracy is only related to the stability of the output potential of each supply.

The relative stability of the different supplies, with respect to their own potential, is given in ppm units in table 7.3. The mirror back plate supply was clearly the least stable on both the medium and long term. It was also the power supply that showed the largest change in potential after the two hour warm up period in figure 7.5. This was probably due to the method used to set its potential. The accelerator and POP supplies were set from the computer, with 12 Ð bit D/A converters, while the mirror back plate supply was controlled and set with a 10ÊV supply connected to a 10 turn potentiometer, since no further D/A converters were available. It should be possible to improve the mirror supply’s stability by controlling it from the computer. This was tested in a simple experiment, where the mirror supply was set by the computer and the push out potential by a ten turn potentiometer, together with a reference channel. This would determine whether the difference was due to method of control alone. The potentials were monitored for 16 hours. Corrected standard deviation for the mirror was 16ÊmV (12Êppm) and 58ÊmV (58Êppm) for the POP, almost the exact reverse of results obtained in the initial experiment, quoted in table 7.3. Thus obtaining additional hardware to enable computer control of all three power supplies would improve power supply stability further.

The ppm total potential values were determined by dividing the corrected standard deviation or corrected ∆V value by the sum of the absolute values of the three voltages, then multiplying by 106. This was an important measurement, since it could be used to indicate the approximate effect of power supplies on mass stability for an ion of unit charge since: mz/∝ TOF (expression 1.9) and 1 2 =∆ PKEtof =2 mutof Vq rearranged to ∆ 2 V= PKEtof ∝ m/q = 2 with m/q m/z utof q so ∝∝2 mz/ TOF PKEtof (7.1) ∆ The relationship applies as effectively all ion PKEtof was due to the power supplies ( V in the above equation). Ions spend most of their time in field free drift regions, so the sum

7 The offset between set and measured supply potentials had been observed to be fairly consistent over all

106 of total potential errors gives the approximate mass error due to power supply drift.8 Since the various power supply effects were not observed to be correlated (see figure 7.5) the different effects were added in quadrature. This gives the total m/z error due to power supplies in the short to medium term as approximately 17 ppm (1 standard deviation). The effective long term error due to power supply drift is calculated from the sum of short to medium (standard deviation) and long term (range) effects and was 20 ppm, to one standard deviation (68% confidence level).

7.3.3 Long Term Mass Error It was impossible to record time of flight spectra while monitoring power supplies, owing to the limits of the instrumentation, so the mass accuracy was measured in a separate experiment. A sample of melittin (2 mg/mL) with DHB matrix (10 mg/mL) was prepared by the dried droplet technique. Sixteen spectra were obtained at irregular intervals over a period of two days. Analyser power supplies were set to their standard potentials. The sample probe was biased to 185 V and the POP delay to 25 µs, with 20 sweeps averaged for each spectrum. Time of flight measurements were obtained for the three largest ion peaks in the [M+H]+ cluster. The time measurements were obtained for each peak and the pooled results used to calibrate the mass scale. The mass errors found with this experiment were compared to those expected from measurements of the power supplies.

The times of flight obtained for melittin [M+H]+ with 0 Ð 3 carbon-13 atoms are given in figure 7.6. The total range of time values drifted by 70 ns, i.e. over 1,000 ppm, and the change in times of flight over the second day indicated that significant drift occurred on a time scale of hours, suggesting that the variation was not random. Qualitatively, this appears to be much larger than the measured drift of the power supplies. A quantitative comparison requires calculation of the mass errors, given in table 7.4, which shows mass error to have a standard deviation of 980 ppm, 49 times the error expected from power supply drift alone. Even when the results considered were restricted to those from the second day, time drift was still large, 33 ns, over 600 ppm of the total time-of-flight. The standard deviation of the mass error in this smaller set of results was 440 ppm, still more than 20 times larger than the results expected from power supply drifts alone. Some of this error was due to uncertainty in determining the centre of the peak with manual integration, since noise made it difficult to accurately determine the baseline in many of these spectra. The effect of this on the centroid value was estimated by adjusting the baseline for integration in the noisiest spectra, and it was found to shift the centre of mass experiments performed with this instrument.

107

5.215E-05

5.214E-05

5.213E-05

5.212E-05 2 C-13 5.211E-05

5.210E-05

5.209E-05

5.208E-05 TOF (s) 1 C-13 5.207E-05

5.206E-05 no C-13 5.205E-05

5.204E-05 16:48 19:12 21:36 0:00 2:24 4:48 7:12 9:36 12:00 14:24 16:48

Time (hours:minutes)

Figure 7.6: Measured times of flight for ions from the melittin [M+H]+ cluster, obtained over two separate days. Ion time of flight is given on the left axis and the time of day is given on the bottom axis. Adjacent points for isomass ions are joined, to illustrate any systematic drift in time of flight during the experiment. Table 7.4: Mass errors obtained with the melittin [M+H]+ cluster, for results obtained over two working days.

time M+H M+H (1 C-13) M+H (2 C-13) acquired m/z (Da)… error (ppm) m/z (Da) … error (ppm) m/z (Da) … error (ppm) (hh:mm) 18:20 2840.517 -1843.085 2841.566 -1826.424 2842.522 -1842.436 18:25 2840.632 -1802.597 2841.612 -1810.231 2842.649 -1797.915 18:55 2840.644 -1798.548 2841.623 -1806.183 2842.626 -1806.010 11:05 2844.851 -320.170 2845.843 -324.020 2846.811 -336.259 11:10 2845.070 -243.183 2846.119 -226.791 2847.134 -222.845 11:20 2846.050 101.267 2847.054 101.394 2848.057 101.231 11:45 2846.315 194.481 2847.319 194.592 2848.311 190.361 12:00 2847.031 445.777 2848.069 458.001 2849.026 441.567 12:12 2847.838 729.535 2848.957 770.084 2849.927 757.646 12:20 2848.323 899.809 2849.280 883.581 2850.342 903.545 12:38 2848.265 879.538 2849.280 883.581 2850.227 863.016 12:50 2846.477 251.223 2847.457 243.218 2848.438 234.928 13:03 2846.327 198.534 2847.365 210.800 2848.288 182.258 13:13 2846.904 401.190 2847.850 381.001 2848.900 396.996 14:00 2848.265 879.538 2849.315 895.741 2850.388 919.756 14:13 2848.634 1009.279 2849.546 976.815 2850.584 988.657

true m/z: 2845.762 2846.765 2847.769

mean error: -1.088 0.322 -1.594 standard deviation: 981.859 981.466 983.108

… Mass was calculated with equation 7.1. The coefficients, a (1.08092×106) and b (-2.96943) were determined from the pooled data. by up to 0.1 Ð 0.2ÊDa. This corresponded to a worst case error of 33 Ð 67 ppm for melittin [M+H]+, far less than the discrepancy between the measured mass error for melittin and that predicted by power supply stability.

Another factor must have been responsible for the larger than predicted error. Having eliminated power supply drift, the next most likely cause was considered to be the drift between the commencement of ion flight in the TOF axis and recording of the time spectrum.

7.3.4 Drift in Start Triggers: ∆t Between the Application of the Push Out Pulse and Oscilloscope Zero Time In the reflecting MALDI-oa-TOFMS, acceleration of ions in the TOF direction commences when the POP is triggered by a signal sent by the computer, while the zero time for a recorded spectrum is based upon a trigger signal sent by the POP generator to the oscilloscope. The overall triggering sequence used in this instrument was the same as that which was used in the linear oa-TOFMS, illustrated in figure 2.3 (chapter 2). Provided the time interval between the application of the POP and the zero time for the oscilloscope remains constant, this ∆t will be included in the zero-time offset for m/z in the mass calibration equation (equation 1.10). If this ∆t shows a significant drift over time, it will be impractical to use the mass calibration equation with external standards.

An experiment was performed to determine whether a significant mass error could result from a drift in ∆t between the POP and oscilloscope triggers. This ∆t was measured by simultaneously monitoring the POP and signal used to trigger the oscilloscope. The POP output was connected to the LeCroy 9384 with a x10 divider, 1ÊM probe attached, as described in section 7.2.1, with the same oscilloscope settings. The accelerator and mirror supplies were set to their standard settings. The POP supply was set to 750ÊV, the maximum input potential of the oscilloscope probe. The TTL trigger output of the POP generator was used to trigger the oscilloscope. The POP generator was triggered 10 times, at 2ÊHz. The averaged signal for the TTL trigger and POP were simultaneously recorded with different channels of the oscilloscope. The difference in time between the two pulses at half height was measured and the display then cleared. This procedure was repeated numerous times over a 7 hour period. The time interval between the TTL trigger and actual POP were plotted as a function of time. Unfortunately, it was impossible to

8 This linear relationship does not strictly hold for the accelerating regions, but since ions spend most time in the drift regions it is a fair approximation. Additionally, ions received only 770ÊeV per unit

108 obtain time of flight spectra and monitor this time difference contemporaneously, so these results could not be compared directly with shifts in isobaric ion times of flight.

The time difference between the two start events, ∆t, was plotted in figure 7.7. The full range of drift, 16.5Êns, was still only half the 33Êns time drift obtained for melittin experiments in a single day, but it was possible that changes in ∆t had been larger on the day when experiments were performed with melittin. Unlike drift in the power supplies, which had a constant relative effect on mass accuracy, the effect of changing ∆t had a larger relative effect at lower mass, where a 16Êns shift in flight time represented a greater proportion of an ion’s total flight time. The absolute mass error still increased with mass, since mz/ ∝TOF2 (from expression 1.9 in chapter 1) and thus peaks always arrive closer together in time as mass increases in a TOFMS instrument, ensuring that a 16.5Êns drift translates to a larger range of m/z at higher mass values. For instance, with the calibration obtained in the proceeding section, the error for an ion of almost 3,000ÊDa would be 650Êppm (2.0ÊDa), while the error for a heavier ion of 30,000ÊDa would be 210Êppm (6.2ÊDa). An ion of 300ÊDa would have an even larger relative error, of 2,100Êppm and an absolute error of 0.6ÊDa. All these errors were quite large, so it was important to reduce the level of drift between timing triggers, since this was at least a major, if not the main, cause of medium term mass axis instability.

The measured ∆t between triggers was able to change so much, because the POP and oscilloscope TTL trigger were separate pulses, generated in parallel by separate circuits in the POP generator. The simplest way to resolve this problem would be to initiate oscilloscope acquisition with the same pulse as ion acceleration, in theory reducing ∆t to zero. It was impossible to interface the POP directly to the oscilloscope, but it was possible to use the POP test output, a ×100 attenuated 2Êk output, to initiate spectral acquisition. Before doing this the test output had to be monitored to determine whether it was suitable for this purpose.

The test output was connected to the oscilloscope with 1ÊM DC coupling, 2V/division input, via a second x10 divider, 1ÊM probe. The oscilloscope trigger level was set to 1.00ÊV, rising edge, low frequency rejection mode. The leading edge of the POP test output is shown in figure 7.8, with the entire pulse inset. There is a significant level of ringing in the output, which can be contrasted with the much better signal for an actual

charge from the 1,005ÊV POP, due the starting location of ions in the TOF axis. 109

1.808

1.806

1.804

1.802 s)

µ 1.800 t ( ∆ 1.798

1.796

1.794

1.792

12:00 PM 2:00 PM 4:00 PM 6:00 PM time

Figure 7.7: Time difference between triggering of the push out pulse and oscilloscope, with measurements conducted over 7.25 hours. It drifted by 16.5 ns.

6 10 signal (V) 4 5 signal (V) 2 0

0 0 4 8 12 time (µs) -2

-0.1 0.0 0.1 0.2 0.3 time (µs)

Figure 7.8: The leading edge of the signal from the push out pulse test output, averaged over 100 acquisitions. The push out pulse was set to 1005 V for this experiment. The entire trace is also inset. The arrow indicates a steeply rising portion of the rising edge, suitable for triggering the oscilloscope. POP (figure 7.1). This was due to the quality of the ×100 divider used in the test output and impedance mismatch between the 2Êk output impedance and 1ÊM oscilloscope input. In spite of the ringing, the signal was still very reproducible, so it was decided to monitor the drift in ∆t between the actual POP and the test output.

The oscilloscope was set up as it had been for monitoring the POP (see section 7.2.1), except with different trigger settings. The POP was set to 700ÊV, close to the maximum input allowed for the oscilloscope probe. The trigger was set to external channel/10, 1ÊM input impedance, low frequency rejection and trigger at a leading edge level of 0.42ÊV. A trigger level of 0.42ÊV would cause the oscilloscope to trigger at a steep part of the rising edge, corresponding to the location of the arrow in figure 7.8, scaled for a 700ÊV POP. It was preferable to trigger the oscilloscope at a steep part of the rising edge, since this would minimise any drift in ∆t due to small changes in the risetime of the test output. The drift in ∆t was observed qualitatively by looking at the thickness of the POP rising edge with the oscilloscope screen set to persistence mode for 360 sweeps, acquired over regular intervals within 3 minutes. The screen was cleared and the process repeated, for 360 sweeps acquired over regular intervals within 60 minutes. This experiment revealed residual timing drift, since the previous triggering method had shown a much larger timing drift over an hour than a few minutes.

The drift in ∆t over 3 minutes was 1.2Êns and the drift over 60 minutes was 1.4Êns, very close to the best results expected for the oscilloscope, since triggering was specified to be accurate to 1ÊGHz, suggesting a triggering accuracy of approximately 1Êns. After these experiments it was decided to trigger the oscilloscope directly from the test output, with trigger level set to 0.6ÊV for a POP of 1,005ÊV.

A 1.4Êns timing drift corresponds to a mass error of approximately 30Êppm for melittin, a vast improvement on the earlier error. Combining the results obtained in this section with mass errors due to signal to noise and power supply drift, the expected mass error (external calibration) for a species like melittin, with an m/z of approximately 3,000ÊDa can be estimated, assuming all effects are uncorrelated: =++222 ppmtot ppm S/N ppm power ppm timing =++(33 to 6700)222 2 3 ≈−50 80ppm

This is not much worse than the expected mass error with internal calibration, ppmS/N in the above equation, so it was concluded that most significant contribution to mass axis 110 instability had been largely resolved, allowing a thorough mass accuracy experiment to be performed.

7 . 4 Characterizing the Analyser After adjusting tuning potentials until optimal resolution was obtained with the main accelerating potential, at Ð20.0ÊkV, and solving the main cause of mass axis instability, the mass spectrometer was ready to be characterized. Mass Accuracy and resolution are important characteristics of any mass spectrometer, as explained in sections 1.2.3 and 1.2.4. Accordingly, characterization of the reflecting MALDI-oa-TOF instrument required knowledge of both mass accuracy and limiting analyser resolution, to indicate the effectiveness of this particular ionization method - analyser combination.

7.4.1 Mass Accuracy Experiments Determining mass accuracy was relatively straightforward, requiring (1) spectra to be acquired at a number of masses and over a period of time, (2) the use of the mass calibration equation (equation 1.10) and (3) calculation of the resulting mass errors with internal and external calibration. This was performed over a number of days, in the experiments described below.

Gramicidin S, melittin, insulin chain B, insulin and ubiquitin were used as analytes, with DHB matrix, allowing spectra to be acquired for m/z ranging from 1,140 to 8,590 Da. Fresh samples were deposited by the dried droplet method for each day of experiments, as described in section 2.5.3, and inserted into the mass spectrometer. Power supplies were set to their standard potentials, 1,005ÊV (POP), -20.0ÊkV (analyser) and 1,300ÊV (mirror backplate) and left to warm up for 5 hours before acquiring data, to ensure maximum stability had been reached. Power supplies were turned off at the end of each day.

The oscilloscope was set to trigger on external channel/10, 1ÊM input impedance, at a level of 0.6ÊV (rising edge) off the POP test output. A ×10 divider was used, in addition to the ×100 divider of the test output, which would cause it to trigger at a steep part of the rising edge, indicated by the arrow in figure 7.8. The time scale was set to 4 GSample/s, with an offset to ensure that ions with the time of flight expected for the analyte of interest would be recorded. The vertical scale was set from 10ÊmV/div to 0.1ÊV/div, depending on the size of analyte signals. Most spectra were acquired with a positive potential applied to the probe. The probe potential and the delay times used were selected to be within the ranges given in figure 7.4 for the mass of the analyte under investigation. For some spectra with insulin chain B, insulin and ubiquitin the pulsed lens was used to increase

111 Table 7.5: Times when spectra were acquired in MA Set B.

sample and spectrum number time of acquisition gramicidin S 1 15:32 gramicidin S 2 16:55 gramicidin S 3 18:30 melittin 1 15:37 melittin 2 17:01 insulin chain B 1 17:03 insulin chain B 2 17:17 insulin chain B 3 17:23 insulin 1 16:11 insulin 2 16:24 insulin 3 18:10 ubiquitin 1 16:45 ubiquitin 2 18:25 sensitivity, since it did not affect mass accuracy, although it would reduce resolution, preventing resolution of multiplets in some cases. In these instances the probe was grounded and a potential of 130ÊV (insulin chain B), 120ÊV (insulin) or 110ÊV (ubiquitin) was applied to the lens instead, after a delay of 7.9Êµs. POP delay time was increased by 8Êµs over that used with the biased probe, to take account of the delay in application of the electric field used to increase desorption velocity. Spectra were acquired after averaging 20 to 150 acquisitions, as required to give at least moderate signal to noise. Most spectra of gramicidin and melittin required 50 or fewer sweeps and the compounds of higher m/z usually required 50 to 100 sweeps. Typical spectra were acquired, similar to those obtained in routine use, rather than spectra with better than average signal to noise, to give a realistic indication of the mass accuracy that could be expected in routine experiments, since the effects of signal to noise on peak shape limited mass accuracy.

Measurements were performed on three different occasions, spanning a full week, to indicate the long term (day to day) mass axis stability. For ease of analysis, the results from each of these three occasions have been labelled ‘MA Set A’, ‘MA Set B’ and ‘MA Set C’. The results presented in MA Set A were obtained with all samples except ubiquitin. The results presented in MA Set B were taken from experiments performed two days later, with all of the analytes. The experiments providing results for MA Set C, which only involved measurements of gramicidin S, were performed four days after MA Set B. Mass calibration was performed using the data in MA Set B, since this set of data provided the largest mass range. In an attempt to reduce the effects of systematic (with time) sources of error skewing the calibration, spectra were not acquired in order of mass. Instead, the order in which spectra were acquired in MA Set B, for different samples, was interlaced, as indicated in table 7.5. This calibration was then used to determine mass errors with external calibration for data from all three sets of experiments.

The results used for mass calibration (MA Set B) are presented in figure 7.9. The fitted line had a slope of 1.05652×106ÊDa1/2µs-1 and an intercept of 5.05207×10-3ÊDa1/2, with a correlation coefficient >Ê0.9999995. The relative error is plotted above the fitted data as a ppm mass error, calculated from the difference between m/z obtained from the calibration equation and the true m/z value. The relative error has also been plotted on a frequency histogram (figure 7.10), to demonstrate that ppm errors were (approximately) normally distributed, justifying statistical treatment. The full set of results is also given in table 7.6, together with absolute (mDa) errors, average errors and the spread in error, measured with the standard deviation. The relative mass error was fairly constant over the entire

112

300

200

100

0 mass error (ppm) -100

-200

insulin

80 ubiquitin

70 1/2 melittin m/z 60 insulin chain B

40 gramicidin S

30 30 40 50 60 70 80 90 time (µs)

Figure 7.9:Mass calibration curve obtained from data collected on a single day (MA Set B). Slope of the fitted line is 1.05652x106 and the y- intecept is 5.05207x10 -3 . The relative error, in ppm, between fitted and actual mass is plotted above the calibration curve.

6 count

0 -100 0 100 200 300 mass error (ppm)

Figure 7.10: Frequency histogram for relative mass error results given in figure 7.9. Table 7.6: Mass accuracy results for MA Set B. These results were used to calibrate the mass spectrometer.

file TOF (µs) Assignment true m/z (Da) fitted m/z (Da) ∆m (ppm) ∆m (mDa) GS1 31.9766 M+H 1141.71376 1141.69446 -16.91 -19.30 31.9905 M+H, 1*C13 1142.71712 1142.68710 -26.27 -30.02 32.0040 M+H, 2*C13 1143.72047 1143.65159 -60.23 -68.88 GS2 31.9779 M+H 1141.71376 1141.78728 64.39 73.52 31.9913 M+H, 1*C13 1142.71712 1142.74424 23.74 27.13 32.0049 M+H, 2*C13 1143.72047 1143.71590 -4.00 -4.57 32.2835 M+Na 1163.69571 1163.71147 13.55 15.76 32.2973 M+Na, 1*C13 1164.69906 1164.70642 6.32 7.36 32.3110 M+Na, 2*C13 1165.70242 1165.69459 -6.72 -7.83 GS3 31.9779 M+H 1141.71376 1141.78728 64.39 73.52 31.9915 M+H, 1*C13 1142.71712 1142.75853 36.24 41.41 32.0055 M+H, 2*C13 1143.72047 1143.75878 33.49 38.31 32.2844 M+Na 1163.69571 1163.77635 69.30 80.64 32.2982 M+Na, 1*C13 1164.69906 1164.77133 62.05 72.26 32.3118 M+Na, 2*C13 1165.70242 1165.75230 42.79 49.89 Mel1 50.4858 M+H 2845.76200 2845.61535 -51.53 -146.64 50.4951 M+H, 1*C13 2846.76535 2846.66373 -35.70 -101.62 50.5035 M+H, 2*C13 2847.76871 2847.61082 -55.44 -157.88 50.5119 M+H, 3*C13 2848.77206 2848.55807 -75.12 -213.99 50.5206 M+H, 4*C13 2849.77542 2849.53932 -82.85 -236.10 Mel2 50.4864 M+H 2845.76200 2845.68299 -27.76 -79.01 50.4957 M+H, 1*C13 2846.76535 2846.73138 -11.93 -33.97 50.5040 M+H, 2*C13 2847.76871 2847.66720 -35.64 -101.50 50.5130 M+H, 3*C13 2848.77206 2848.68213 -31.57 -89.93 50.5215 M+H, 4*C13 2849.77542 2849.64083 -47.23 -134.58 Ins ChB1 55.9487 M+H 3494.65132 3494.69822 13.42 46.89 55.9564 M+H, 1*C13 3495.65468 3495.66012 1.56 5.44 55.9651 M+H, 2*C13 3496.65803 3496.74711 25.48 89.08 55.9729 M+H, 3*C13 3497.66139 3497.72180 17.27 60.41 55.9799 M+H, 4*C13 3498.66474 3498.59663 -19.47 -68.11 55.9684 M+H (clust) 3496.97536 3497.15946 52.65 184.10 Ins ChB2 55.9483 M+H 3494.65132 3494.64825 -0.88 -3.07 55.9558 M+H, 1*C13 3495.65468 3495.58516 -19.89 -69.51 55.9635 M+H, 2*C13 3496.65803 3496.54719 -31.70 -110.84 55.9721 M+H, 3*C13 3497.66139 3497.62182 -11.31 -39.56 55.9804 M+H, 4*C13 3498.66474 3498.65912 -1.61 -5.62 55.9757 M+H (clust) 3496.97536 3498.07172 313.52 1096.36 Ins ChB3 55.9749 M+H (clust) 3496.97536 3497.97174 284.93 996.38 56.1438 M+Na (clust) 3518.96536 3519.11151 41.53 146.15 Ins1 71.6730 M+H (clust) 5734.64083 5734.88296 42.22 242.13 Ins2 71.6650 M+H (clust) 5734.64083 5733.60288 -181.00 -1037.95 Ins3 71.6683 M+H (clust) 5734.64083 5734.13090 -88.92 -509.93 Ubi1 87.5936 M+H (clust) 8565.90 8565.39928 -58.46 -500.72 87.7120 M+Na (clust) 8587.89 8588.56931 79.10 679.31 Ubi2 87.5970 M+H (clust) 8565.90 8566.06420 19.17 164.20 87.7125 M+Na (clust) 8587.89 8588.66722 90.50 777.22 mean error 9.03 standard deviation 81.04 mass range, as can be seen by examination of the ∆m (ppm) column of table 7.6 or ppm mass error in figure 7.9, with most errors less than 100Êppm. The two largest errors of 314 and 285Êppm, are clear outliers, more than 3 standard deviations from the mean error. They were both obtained from insulin chain B, for results obtained over the [M+H]+ cluster. These results were probably due to the effects of noise distorting the peak, since both of the clusters had worse than average signal to noise, indicating the importance of signal quality on mass accuracy results. Almost all other results were within or close to 81Êppm of the mean error, so it was concluded that this result gave the precision of the mass accuracy measurements obtained from the instrument in this set of experiments, for data collected in a single afternoon. The absolute mass error cannot be used to determine the overall error at all masses, since it increases dramatically with mass, so the mean and standard deviation were not taken for these results. It does however indicate the actual mass errors observed at each mass, experimentally useful to know when allocating accurate masses, and so has been included for completeness.

Table 7.7 presents the results obtained in MA Set A. Fitted m/z was calculated with the coefficients given in the proceeding paragraph. The standard deviation of relative mass errors, 77Êppm, is similar to that obtained for MA Set B. In MA Set C only one compound was investigated, gramicidin S, with the errors in mass measurements summarised in table 7.8. The standard deviation of mass errors in MA Set C was 24Êppm, less than that obtained in the first two sets of experiments. This was hardly surprising, since when compared to the other two data sets MA Set C involved (i) the smallest mass range and (ii) spectra were obtained in the smallest time window, both factors expected to decrease the range of mass errors. Pooling all three sets of data, it can be assumed that 80Êppm represents the approximate standard deviation of relative mass error for a single day, with masses from 1,100 to 8,600ÊDa. The true mass error (external calibration) of the mass spectrometer, to 1 standard deviation (68% confidence level), was twice this value, 160Êppm, representing data distributed on both sides of the mean.

The overall day to day variation in mass calibration stability can be obtained from the differences between the mean errors for each set of data. The largest difference was between the MA Set A and MA Set B, with an overall difference of approximately 110Êppm. This was not significantly larger than variations within a single day, so it was decided to calculate the overall spread of mass, using all 80 measurements. The overall mean error found to be -21Êppm with a standard deviation of 82Êppm, only slightly larger than the 80Êppm standard deviation obtained for a single day. Thus it has been

113 Table 7.7: Mass accuracy results for MA Set A, calibrated with fitting coefficients found from results in table 7.6.

file t (µs) assignment true m/z (Da) fitted m/z (Da) ∆m (ppm) ∆m (mDa) GS 32.2819 M+Na 1163.69571 1163.59614 -85.56 -99.57 32.2964 M+Na, 1*C13 1164.69906 1164.64152 -49.40 -57.54 32.3100 M+Na, 2*C13 1165.70242 1165.62244 -68.61 -79.97 Mel 50.4832 M+H 2845.76200 2845.32229 -154.51 -439.70 50.4922 M+H, 1*C13 2846.76535 2846.33680 -150.54 -428.55 50.5011 M+H, 2*C13 2847.76871 2847.34021 -150.47 -428.50 Ins ChB 55.9636 M+H (clust) 3496.97536 3496.55969 -118.87 -415.67 55.9440 M+H 3494.65132 3494.11114 -154.57 -540.18 55.9520 M+H, 1*C13 3495.65468 3495.11045 -155.69 -544.23 Ins ChB 2 55.9668 M+H (clust) 3496.97536 3496.95953 -4.53 -15.83 55.9441 M+H 3494.65132 3494.12363 -151.00 -527.69 55.9522 M+H, 1*C13 3495.65468 3495.13543 -148.54 -519.25 55.9599 M+H, 2*C13 3496.65803 3496.09740 -160.33 -560.64 55.9681 M+H, 3*C13 3497.66139 3497.12198 -154.22 -539.41 Ins 71.6742 M+H (clust) 5734.64083 5735.07498 75.71 434.15 Ins 2 71.6728 M+H (clust) 5734.64083 5734.85096 36.64 210.13 mean error -99.66 standard deviation 76.79 Table 7.8: Mass accuracy for gramicidin S in MA Set C, using the calibration from data listed in table 7.6.

[M+H]+ [M+Na]+ true m/z (Da): 1141.71376 true m/z (Da): 1163.69571 File & time apparent m/z ∆m (ppm) ∆m (mDa) apparent m/z ∆m (ppm) ∆m (mDa) GS1, 3:03 PM 1141.697 -14.68 -16.76 1163.708 10.56 12.29 GS2, 3:05 PM 1141.684 -26.07 -29.76 1163.701 4.55 5.29 GS3, 3:35 PM 1141.688 -22.56 -25.76 1163.657 -33.26 -38.71

true m/z (Da): 1142.71712 true m/z (Da): 1164.69906 File & time apparent m/z ∆m (ppm) ∆m (mDa) apparent m/z ∆m (ppm) ∆m (mDa) GS1 3:03 PM 1142.677 -35.11 -40.12 1164.676 -19.80 -23.06 GS2 3:05 PM 1142.684 -28.98 -33.12 1164.721 18.84 21.94 GS3 3:35 PM 1142.706 -9.73 -11.12 1164.675 -20.66 -24.06

true m/z (Da): 1143.72047 true m/z (Da): 1165.70242 File & time apparent m/z ∆m (ppm) ∆m (mDa) apparent m/z ∆m (ppm) ∆m (mDa) GS1 3:03 PM 1143.667 -46.75 -53.47 1165.647 -47.54 -55.42 GS2 3:05 PM 1143.676 -38.88 -44.47 1165.664 -32.96 -38.42 GS3 3:35 PM 1143.658 -54.62 -62.47 1165.611 -78.42 -91.42 mean error (all species): -26.45 standard deviation (all species) error: 23.73 Table 7.9: Mass accuracy for internal calibration from all three sets of mass accuracy experiments

spectrum m/z range (Da) largest error (ppm) MA Set A (see Table 7.7) GS 1163.7 Ð 1165.7 36.16 Mel 2845.8 Ð 2847.8 4.04 Ins ChB 3494.7 Ð 3495.7 1.12 Ins ChB 2 3494.7 Ð 3497.7 11.79

MA Set B (see Table 7.6) GS 1 1141.7 Ð1143.7 43.32 GS 2 1141.7 Ð 1165.7 81.35 GS 3 1141.7 Ð 1165.7 30.90 Mel 1 2845.8 Ð 2849.8 47.15 Mel 2 2845.8 Ð 2849.8 35.30 Ins Ch B 1 3494.7 Ð 3498.7 44.95 Ins Ch B 2 3494.7 Ð 3498.7 30.82 Ubi 1 8565.90 Ð 137.56 8587.89 Ubi 2 8565.90 Ð 71.33 8587.89

MA Set C (see Table 7.8) GS 1 1141.7 Ð 1165.7 58.10 GS 2 1141.7 Ð 1165.7 57.72 GS 3 1141.7 Ð 1165.7 68.69 mean error 47.52 standard deviation 32.27 shown that external calibration mass accuracy was truly 160Êppm, providing the analyser power supplies were given a sufficient warm-up time.

Mass accuracy with internal calibration on this instrument was largely limited by instrument S/N, as could be seen in figure 6.3. Data from the three sets of experiments used for external calibration could also be used to determine accuracy with internal calibration, from individual spectra containing more than 1 peak. Since mass accuracy was quite good with external calibration, the spectra did not need to be recalibrated to determine internal calibration errors. Instead, the range of external calibration ppm errors (largest Ð smallest error) in single spectra indicated the worst ppm error by internal calibration in each spectrum containing more than 1 peak. Where both cluster and individual mass results were available for one mass region in a single spectrum, only the results for individual masses were used, reflecting the calibration procedure that would have been used if it was an internal calibration experiment. Internal calibration results are summarised in table 7.9, showing a mean error of 47.5Êppm with standard deviation of 32.3Êppm. This gives the error in internal calibration to be 15 Ð 80Êppm (68% confidence limit). The effects of S/N also impact on external calibration and a contribution of 47Êppm represents a significant proportion of the 160Êppm external calibration error.

The overall errors obtained, 160Êppm for external calibration, appeared larger than the 50 to 80Êppm relative error predicted for melittin in section 7.3.3, based on the measured power supply and timing trigger drifts, including an estimate of internal calibration errors. However, if the melittin results are only compared to other spectra obtained on the same day, by comparing the mass errors for melittin in tables 7.6 and 7.7 with the mean error on each table, the results obtained for melittin are in approximate agreement with the prediction, with relative errors of 20-90Êppm (table 7.6) and 50-55Êppm (table 7.7). This is a more realistic comparison of the observed and predicted errors, since the drift between timing triggers and drift in power supplies used to estimate the predicted errors were monitored over periods of one to two days and these drifts were probably larger during the week long period over which the spectra were recorded. In particular, it is likely that the drift in timing triggers contributed significant additional drift in ∆t between days, suggested by the large overnight drift observed in figure 7.6. It can be concluded from this that the overall mass accuracy of the instrument with external calibration was 160Êppm, although many results obtained on the same day as the calibration had an error of less than 80Êppm, while internal calibration gave errors of 80Êppm or less.

114 7.4.2 Determining the Limiting Resolution and Quantifying Contributions to Peak Width Resolution in a time of flight mass spectrometer increases in a non linear manner with mass [42], with the effect easily observed as long as ∆t at half height for each peak is small enough to resolve peaks due to individual isotopes in observed spectra. The overall instrument resolution can be determined by simply finding m/∆m at individual masses. Unfortunately, this does not give the true analyser or limiting resolution, since a number of factors, only some of which are related to the analyser, contribute to peak width. The factors include: (a) detector pulse width for a single ion; (b) timing drift (jitter) between the start time for ions (POP) and data recording (oscilloscope trigger), when averaging a number of acquisitions; (c) electric field inhomogeneity, both spatial (imperfect fields near the grids) and temporal (ripple and short term drift in power supplies); and (d) initial ion energy (and spatial) dispersions. The first two factors are essentially independent of analyser performance, instead depending upon the detection and recording system. They generate a constant ∆t for all masses. The effect of the other two factors, field inhomogeneities and initial dispersions, gives an essentially constant energy spread to ions, regardless of mass, and causes an arrival time spread for ions of the same m/z. This arrival time spread is directly linked to analyser performance, with a well designed analyser minimising the size of the spread and hence its contribution to peak width. It is these factors which give rise to the limiting analyser resolution.

A mathematical expression can be derived which allows separate determination of both detection system and analyser contributions to peak width. This method has been used by Coles and Guilhaus to determine the resolution limitations in the original linear oa-TOF analyser [42]. The following analysis is based upon the approach taken in that paper, extended to apply to an analyser incorporating an ion mirror. The starting point is the kinetic energy TOF equation, relating the mass ‘m’ (in kg) of an ion to its final velocity -1 ‘u’ and initial velocity ‘u0’ (both in ms ), charge ‘q’ (in C) and accelerating potential difference ‘V’ (volts): 1 −=2 2 mqV()uu0 giving a velocity of: 2qV uu=+ (7.2) m 0

115 The time an ion spends in the drift region of length ‘D’ (in metres) may be calculated with a different equation: D t = u The initial TOF axis velocity in an oa-TOF instrument is essentially zero, so when the above equation is substituted into equation 7.2 the following is obtained: m t = D (7.3) 2qV Total time of flight is of greater interest than time spent in the drift region, so ‘D’ is replaced with ‘S´’, the effective drift length. This is the theoretical displacement an ion would undergo, travelling at the final velocity ‘u’, to require the same total flight time ‘t’, that an identical ion would incur in the mass spectrometer when accelerated from rest into the analyser, including time spent in the ion mirror. This value is independent of mass, depending upon electric fields and dimensions, allowing equation 7.3 to be modified to: m t = S© (7.4) 2qV The factors that give rise to the limiting analyser resolution, the energy effects, are independent of mass and effectively reduce to a finite spread in potential, ∆V, in equation ∆ 7.4. This causes an analyser linked spread in ion arrival time, ta, found by differentiating equation 7.4 with respect to potential:

© Sm−3 ∆∆t=− VV2 a22q or SV©∆ m ∆t =− (7.5) a 22V qV Dividing equation 7.4 by equation 7.5 generates equation 7.6: t V =− (7.6) ∆∆ 2 ta V Differentiating equation 7.4 with respect to mass and converting to a finite mass interval leads to: Sm© ∆m ∆t = (7.7) 22qV m

116 Dividing equation 7.7 by equation 7.4 and rearranging to make resolution (m/∆m) the subject gives the resolution relationship (equation 1.11) reproduced below with mass in kg rather than Da: m =t ∆∆m 2t

Combining this with equation 7.6 shows how the limiting analyser resolution, R0, is linked to the magnitude of the mass independent energy spread, tV R == (7.8) 0 ∆∆ 2 ta V

The temporal spread introduced by the detection and recording system gives a fixed value ∆ for all masses, td. Both forms of peak broadening create effects that are approximately Gaussian, so the total peak width at any mass can be determined by adding the two separate effects in quadrature: ∆∆∆222=+ tttmad (7.9) ∆ Substituting equation 7.5 into this expression, with V/V replaced with 1/R0 gives:

©2 ∆∆2 = S()+2 ttm2 m/q d (7.10) 8R0 V or, converting m/q (mass in kg per C) to m/z (mass in Da per elementary charge, e) gives: S©2 ∆∆tmzt2= ()/+2 (7.11) m× 82 d 710.RV.71876 0 =+ ∆ 2 This is an equation of the form yxab, where ‘y’ is tm , ‘a’ is S©2 , ‘x’ is m/z and ‘b’ is ∆t 2 . Since all parameters contained in ‘a’ are × 82 d 710.RV.71876 0 ∆ 2 known constants, except R0, plotting tm against m/z for a range of masses allows both ∆ 2 td and R0 to be determined from the line of best fit.

Spectra obtained for mass accuracy experiments were also used to determine the limiting analyser resolution obtained in routine experiments. Peak width at half height was measured for all masses that gave unit resolution to this level, including matrix ions. This was possible for all spectra obtained with biased probe, up to the mass of insulin, giving an overall mass range of approximately 38 to 5,700ÊDa. The mass at each peak was determined from the mass calibration function. Peak widths for individual masses across each cluster (up to 5 adjacent masses in the case of insulin) were not expected to vary significantly, allowing data to be pooled for each mass and between different spectra. This was necessary to prevent masses where more measurements had been performed 117 having a larger weighting when curve fitting was preformed. The square of the peak ∆ 2 width ( tm ) was plotted against m/z, according to equation 7.11, with error bars on ∆ 2 tm calculated from the standard deviation in widths and m/z from the mean of the pooled results. ‘S´’ was calculated from the times of flight of masses used in this experiment and ‘V’ from the accelerating potentials and instrument geometry. A line was then fitted to the plotted data, allowing accurate determination of limiting resolution and the contribution of the detection system and timing electronics to peak width.

∆ 2 Figure 7.11 contains the plot of tm against m/z. The theoretical displacement, ‘S´’, was calculated to be 1.8967Êm from time of flight values and the overall accelerating potential, ‘V’, to be 20,815 V. The contribution of the detection system and timing jitter to peak ∆ 2 width, td , was found to be 1.21Êns from the intercept and the limiting analyser resolution, R0, was calculated to be 7,990.

Unfortunately, the fit obtained was not perfect, with a correlation coefficient of 0.952, due to the large scatter in width at data points. This scatter was partly due to the accuracy of time measurements, based on the digitiser rate (4ÊGS/s). Typical peak width at half height was from 1.25 to 4.0Êns, with measurement to the nearest 0.25Êns, giving an uncertainty of between 20% to 6% of the measurement of width, with a greater impact for the narrower peaks obtained at low mass. Another reason for the large variation could be ∆ 2 that the overall timing jitter (included in td ) was linked to the number of averaged sweeps in a spectrum, with larger numbers of sweeps allowing small increases in this ∆ 2 value. A single sweep would have no timing jitter, with td reducing to detector width effects, while several hundred sweeps over several minutes would give a timing jitter of ∆ 2 approximately 1.2Êns, found in section 7.3.3, and td would result from combination of this with detector effects. This experiment was designed to find limiting resolution obtained under standard operating conditions, so spectra contained varying numbers of sweeps, from 20 to 100, as required to obtain reasonable signal to noise, with more sweeps used at higher masses. This could explain why the insulin result (m/z c. 5,730ÊDa), which required the largest sweeps per spectrum, lay the furthest above the plot. Melittin results are the furthest below the line, with a small error bar, indicating better than predicted (by the plot) resolution. The melittin spectra were particularly good, with excellent signal to noise and good reproducibility, so it was assumed that the excellent resolution obtained for melittin was due to it being a model compound.

Notwithstanding the above comments, plotting ∆t2 against m/z provided the best way to estimate the limiting analyser resolution and other contributions to width, averaging

118

20 ) 18

2 15 x10 2 m t (s ∆

0 1000 2000 3000 4000 5000 6000 m/z (Da)

Figure 7.11: A plot of ∆t2 against mass, obtained from data acquired in mass accuracy experiments. Errors on ∆t2 correspond to the standard deviation of ∆ t for multiple vaues, or 30% of ∆t2 for a single point. Peak broadening due to the detection system can be calculated from the intercept. The limiting analyser resolution can be found from the slope. variations at different masses to find an overall result. However, scatter in the plotted data does suggest that results should only be quoted to 2 significant figures, giving limiting resolution of 8,000 and the contribution of detection system and timing jitter to peak width of 1.2Êns. It must be remembered that these values represent the results for typical spectra, obtained in routine operation. Better resolution has been obtained from above average spectra, as demonstrated below. However, it was decided that it was more realistic to determine limiting resolution for routine experiments, rather than calculating it from the best possible results.

7.4.3 Spectra of Species Used to Find Mass Accuracy and Limiting Resolution

Some examples of the spectra obtained during the mass accuracy experiments are presented here, with average resolution for the ion peaks displayed indicated on the figures. Gramicidin S, melittin and insulin results are shown, in figures 7.12, 7.13 and 7.14 respectively. Insulin chain B and ubiquitin gave similar spectra to those shown in chapter 6, with an accelerating potential of Ð17.5ÊkV, so additional spectra have not been included here.

The predicted resolution, according to the plot in figure 7.11, for gramicidin S is 7,000 (2 significant figures), less than the 8,000 resolution obtained. Similarly, melittin was expected to have a resolution of 7,400, not 9,100, and insulin resolution should have been approximately 7,800, rather than 8,500, as observed. Not only have these three spectra given better than expected results, they all had resolution above the calculated limiting analyser resolution! This result was not as strange as it appears, when the methods used in these experiments were taken into account. Mass accuracy and resolution experiments were intentionally performed with most spectra of only average quality, to give a realistic estimate of mass accuracy and resolution results expected during routine performance. Some spectra were of above average quality, including the spectra presented here, with better than expected (average) resolution and signal to noise. Limiting resolution calculated from the best results was approximately 9,000 to 10,000, for samples of melittin and insulin chain B.

7 . 5 Extension of the Instrument to other Samples: TPP, Myoglobin, DNA and Polyethylene Glycol A number of other samples were also analysed with the Ð20ÊkV MALDI-oa-TOF mass spectrometer, to illustrate the range of compounds that could be successfully analysed with this instrument. These samples included a larger protein (myoglobin), an oligonucleotide (DNA decamer) and a synthetic polymer (polyethylene glycol, PEG). Spectra were also obtained from TPP, for comparison with results obtained from the 119

+ [M+Na]

R ≈ 8000

[M+H] +

1120 1140 1160 1180 1200 m/z (Da)

Figure 7.12:A Gramicidin S spectrum obtained during mass accuracy experiments. 20 sweeps were averaged. A 100 MHz Chebyshev filter was applied to reduce noise.

R ≈ 9100 [M+H] +

2820 2830 2840 2850 2860 2870 m/z (Da)

Figure 7.13:Melittin spectrum obtained during mass accuracy experiments. 30 sweeps were averaged.

[M+H] +

R ≈ 8500

5700 5710 5720 5730 5740 5750 m/z (Da) Figure 7.14: Insulin spectrum obtained during mass accuracy experiments. 150 sweeps were averaged. A 50 MHz Chebyshev filter was applied to reduce noise. prototype linear MALDI-oa-TOF. All samples were prepared by the dried droplet method, with DHB matrix in a 1,000 fold molar excess for all samples except the DNA decamer, which used a 1,000 fold molar excess of 3-hydroxypicolinic acid in 0.5ÊM aqueous ammonium citrate solution instead. Further details of the dried droplet method were given in chapter 2. The mirror, POP and accelerator potential were all set to standard values for experiments at -20ÊkV. The number of sweeps and other instrument conditions are listed in the figure captions or following text.

The TPP result is shown in figure 7.15, mass calibrated with the fitting coefficients found during mass accuracy experiments. Resolution was 3,900 and signal to noise was excellent. Predicted resolution was 6,200, significantly higher than that obtained. This unexpectedly low resolution was probably due to a combination of (i) space charge effects, since this species provided far more ions per shot than any other analyte; and (ii) a different ion formation process providing a larger initial spatial and energy distributions, since TPP gives similar spectra without matrix compound.

Figures 7.16 and 7.17 show spectra obtained with myoglobin, the highest mass molecule analysed with this mass spectrometer. The spectra have not been mass calibrated, since both results were obtained before mass axis instability problems had been solved. Figure 7.16, with poor signal to noise, was obtained with the einzel lens and figure 7.17, with very good signal to noise, when the pulsed lens was installed. No clear signal was obtained without using one of the lenses. This provided qualitative evidence of the signal enhancing ability of the pulsed lens, since it was required to obtain a myoglobin spectrum with good signal to noise. The quasimolecular ion peak in figure 7.17 has a width at half height of 28ÊDa, which represents width of the isotope envelope for the cluster. Thus the presence of isotopes prevents the determination of the analyser resolution in this spectrum.

A spectrum of a synthetic DNA decamer, GAC-CGC-TTG-T (5’ - 3’), is given in figure 7.18, mass calibrated with results from mass accuracy experiments. Signal to noise was not particularly good, due to the large number of alkali metal adduct ions, so the pulsed lens was used to increase the signal in the spectrum. Peak width at half height is approximately 4ÊDa across each cluster, including the effects of the isotopes, since resolution is not sufficient for unit mass resolution. The instrument resolution can not be calculated from this and all that can be concluded is that the resolution is Ê3,000, since resolution above this value would resolve at least the tips of isotope peaks. Thus the use of the pulsed lens has caused a significant drop in resolution, from the value of ~7,400 expected without the lens, as would be expected from the results discussed in section 6.5.3, since the lens is introducing an additional spatial and energy spread to the ions 120

[M+H] +

R ≈ 3900

[M] +

605 610 615 620 625 m/z (Da)

Figure 7.15: TPP spectrum obtained after the instrument had been mass calibrated. The probe was biased to 188 V and a 10 µs delay was used. 20 sweeps were averaged.

16952 Da

120 122 124 126 128 130 time (µs)

Figure 7.16: Best spectrum obtained for myoglobin with the einzel lens installed. 70 sweeps were averaged, with the lens set to 81 V and the probe to 152 V. Smoothing was performed with a 5 MHz Chebyshev filter.

16952 Da

28 Da

120 122 124 126 128 130 time (µs) Figure 7.17: Best spectrum obtained for myoglobin with the pulsed lens installed. 200 sweeps were averaged, with the lens set to 92 V and a 7.9µs delay before the lens was set to that potential. Smoothing was performed with a 5 MHz Chebyshev filter.

+ [M+H]

R ≤ 3000 + [M+2Na-H]

+ [M+K] + + [M+Na+K-H] [M+2Na+K-2H]

+ + [M+Na] [M+2K-H]

+ [M+Na+2K-2H]

3000 3050 3100 3150 3200 m/z (Da) Figure 7.18: Spectrum of a DNA oligomer, showing a number of adduct clusters. 160 sweeps were averaged, with the lens set to 130 V and a 7.9 µs delay before the lens was set to that potential. A 10 MHz Chebyshev filter was applied to reduce noise.

27 26 24 25

28 23 H(OCH CH ) OH 2 2 n 29

30 R ≈ 5300 21

31 32

600 800 1000 1200 1400 1600 1800 m/z (Da) Figure 7.19: Spectrum of PEG-1000, with sodiated adducts the main species above n = 19, with the value of n listed near the adduct clusters. 100 sweeps were averaged, with the probe set to 182 V. A 50 MHz Chebyshev filter was applied to reduce the effects of high frequency noise. Table 7.10: Mass accuracy data for the DNA oligonucleotide, 5’-GAC-CGG-TTG -T-3’, using results shown in Figure 7.18

measured m/z (Da) assignment true m/z ∆m (ppm) ∆m (mDa) 3020.20 M+H 3020.15 16.6 50.0 3042.07 M+Na 3042.13 -21.0 -63.8 3058.04 M+K 3058.24 -67.0 -205.0 - M+2Na-H† 3064.11 - - 3080.05 M+Na+K-H 3080.23 -58.7 -180.9 3096.29 M+2K-H 3096.34 -15.8 -49.1 3101.69 M+2Na+K-2H 3102.21 -167.8 -520.7 3117.86 M+Na+2K-2H 3118.32 -147.5 -459.9

† This species has been mentioned in the table for completeness, since it was assigned in figure 7.18, but measured mass could not be obtained, because the peak was insufficiently resolved from the larger [M+K]+ peak. entering the fill up region of the orthogonal accelerator. The results could, however, still be used for a mass accuracy determination. This spectrum was obtained 14 days after mass calibration and could be used to assess mass accuracy for different types of samples, since the calibration was obtained with peptides, not DNA, using DHB and not 3-HPA with ammonium citrate as a matrix. The mass accuracy results are presented in table 7.10. Most results are within 70Êppm, very close to the results expected for internal standards and better than had been expected with a two week old calibration, considering the fairly poor signal to noise. The largest peak, due to [M+H]+, which would have been used to give the mass had this been a sample presented for analysis, had an error of less than 20Êppm. The worst results were for the smallest peaks, which also had the lowest signal to noise. Thus results from this sample indicated that the instrument had a very stable mass scale. This can be contrasted with DE-TOF systems, where the effects of initial velocity ensure that high levels of mass accuracy with external calibration depend upon sample preparation and the initial velocity, which is different for different type of analyte and matrices [121]. MALDI-DE-TOF systems do, however, provide better mass accuracy with appropriate calibrations, reported at ~15Êppm with external calibration and ~5Êppm internal calibration [141], so the stability of the mass scale in this reflecting MALDI-oa-TOF instrument will not prove that significant an advantage until the average mass error is decreased.

The polyethylene glycol sample also had fairly poor signal to noise when acquired under conventional operating conditions. Using the pulsed lens to improve this, by forcing more ions into the analyser, would cause a large reduction in resolution. This was not desired, so sensitivity was improved with the signal preamplifier (see section 6.5.4) instead, which improved signal to noise for each ion, with only a small drop in resolution, allowing the spectrum shown in figure 7.19 to be obtained. Resolution was approximately 5,300, as can be seen from the inset cluster for species with 27 ethylene glycol units. This is lower than the resolution of 6,900 expected from limiting analyser resolution and the contribution of the detection system found in section 7.4.2, because the contribution of the detection system to peak width has been increased by the preamplifier, by approximately 0.4Êns (based upon measurements given in section 6.5.4) without ∆ 2 affecting the limiting analyser resolution. Increasing td by 0.4Êns gives a predicted resolution of 6,200, closer to the value obtained.

7 . 6 Estimating Instrument Sensitivity Instrument sensitivity is dependent upon a number of factors, including sample preparation, transmission efficiency of the analyser and detector efficiency. It was beyond the scope of the series of experiments presented in this chapter, designed to characterize

121 the analyser, to determine the true sensitivity of the instrument, since this would require sensitivity optimised sample preparation and characterization of the detector. Instead, it was decided to estimate the moles of ions (Nion) required for a spectrum with fair signal to noise, using equation 7.12 and extrapolate from this to estimate the detection limit. =×× NnpionN tot (7.12) a p = As× Parameters are defined as follows: ‘n ’ is the number of laser shots averaged to acquire a single spectrum, ‘p ’ is the proportion of the sample spot desorbed per shot and ‘Ntot’ is the total number of moles of analyte in the sample spot. When the area of the sample spot (‘A’) is larger than the area of the target illuminated by the laser (‘a’), the factor ‘p ’ can be calculated as given above, provided the average number shots required to remove all the sample from the target at a single location (‘s’) is known. Note that this method assumes the sample is evenly distributed over the entire spot, which is not strictly correct for samples prepared by the dried droplet method (illustrated in figure 4.1).

The spot size obtained with dried droplet sample preparation was approximately 15Êmm2, the area of target illuminated by a laser shot was estimated to be 0.01Êmm2, from the desorption craters left on the target (see chapter 4, figure 4.5) and approximately 50 shots were required to deplete the sample at a single position for most samples, giving a ‘p ’ of 1.3 x 10-5. This data was utilised to estimate the amount of sample required to obtain a single spectrum with fair signal to noise for gramicidin S, melittin, insulin and myoglobin, given in table 7.11. The S/N in the spectra used for this determination ranged from 100 to 40 for the compounds assessed, with signal to noise decreasing with mass. The detection limit of the instrument was estimated from these results by extrapolation to the defined limit of detection (S/NÊ=Ê3) using the assumption that S/N is approximately proportional to moles of analyte desorbed. The detection limits estimated in this manner ranged from the low tens of fmol for gramicidin S and melittin, to approximately 90Êfmol for myoglobin, with the predicted insulin detection limit in between, as indicated in table 7.11. These estimated detection limits are higher than the low fmol to amol detection limits of other MALDI instruments [87] and several times higher than the low fmol detection limits reported for MALDI-oa-TOFMS with a collisional cooling interface [167]. The higher detection limits on this instrument are not surprising, since many of the ions are not sampled by the analyser, owing to sampling restrictions in the source direction, the effect of probe potentials and limits on PKEsa for ions to be detectable. Thus the detection limits in this instrument could be improved by designing a more efficient means of producing a narrow parallel packet from the ions formed in MALDI, with a small energy range, to ensure a larger proportion of the ions enter the fill up region of the analyser, without reducing resolution. 122 Table 7.11: An estimation of the detection limit, based upon the number of moles of analyte used to acquire a single spectrum with good signal to noise, extrapolated to the limit of detection (S/N=3).

analyte amount of sample No shots per approximate moles desorbed approximate estimated per spot (pmol) spectrum per shot per spectrum S/N detection limit gramicidin S 1750 20 - 40 23 fmol 460 - 920 fmol 100 14 - 28 fmol melittin 700 30 - 60 9 fmol 270 - 540 fmol 85 10 - 20 fmol insulin 700 60 - 150 9 fmol 540 - 1400 fmol 50 30 - 85 fmol myoglobin 480 c. 200 6.2 fmol 1200 fmol 40 90 fmol 7 . 7 Conclusions on Analyser Performance The large drifts in mass calibration were found to be due to drift in the ∆t between triggering of the POP and the oscilloscope. This drift was minimised by triggering the oscilloscope with the POP test output, which provided a mass axis that was stable enough to permit external calibration and allow assessment of the analyser performance. Performance was assessed with species ranging from m/z 615-8,600. Mass accuracy was found to be within 15-80Êppm of the true value with internal calibration, with mass errors of up to 160Êppm within a single day when external calibration was used. Mass accuracy between measurements on different days was approximately 100Êppm. Importantly, the calibration with external standards was very robust, with this level of stability observed in samples analysed over a period of two weeks, with different analytes and matrices utilising a single calibration. The spectra used to assess mass accuracy also enabled calculation of the limiting analyser resolution, which was ~8,000 for average spectra, with the limiting analyser resolution increasing to ~10,000 in very good spectra. Sensitivity was determined by extrapolation from the results obtained with average spectra, and the estimated detection limit was in the region of 10s of fmol.

Resolution would be difficult to improve without altering the dimensions of the analyser, but an analysis of the results suggests a number of more straightforward modifications that could be made to the analyser, to improve its mass accuracy and sensitivity. For instance, the main limit on mass accuracy is still the long term timing jitter, since the oscilloscope is currently triggered by a pulse with extensive ringing. Timing jitter could be reduced, by replacing the existing POP generator with another that provides a better trigger pulse for the oscilloscope with negligible drift in ∆t. A smaller, but still significant, source of mass error is provided by the power supply drift. This could be readily improved by (i) installing a more stable Ð20ÊkV supply, since this supply contributes much of the total drift due to the power supplies and (ii) extending computing capabilities to permit computer setting of the mirror supply, since the computer controlled power supplies were observed to be more stable. These modifications, if successful, have the potential to reduce mass errors by external calibration to approximately the same level as those observed with internal calibration. Finally, mass accuracy with both internal and external calibration could be improved by increasing the signal to noise, since this reduce the effects of noise on the measurement of a peak’s centroid. Signal to noise can be increased by improving the sampling efficiency of the analyser, by designing a more efficient means of generating parallel packet of ions with a small range of PKEsa than that obtained with the pulsed lens.

123 Thus the resolution of the analyser is quite good for a reflecting MALDI-oa-TOF providing a total flight path of ~1.5Êm; the mass axis calibration is robust when compared to conventional MALDI-TOF mass spectrometers, with relatively simple modifications to the instrument expected to reduce errors to a similar level to those observed in delayed extraction instruments; and the estimated sensitivity is fair, although 1 or more orders of magnitude worse than those observed in commercial MALDI -TOF instruments.

124 Chapter 8: Evaluation of the Detector

8 . 1 Why Characterize the Detector? 8.1.1 Importance of the Detection System The detector is a critical component in any mass spectrometer, since without an effective ion detector it is impossible to record a spectrum. The detector in any TOFMS system should have a rapid response time, a high gain and a good conversion efficiency, factors that allow the detection of (preferably) down to single ions with excellent time resolution. Ideally, the detector should have a time resolution much better than the analyser, so it will have a negligible effect on the instrument’s final resolution. [42]

A MALDI ion source has the ability to generate ions of over 1 x 106 Da [113] and many instruments operate at high voltages. Thus, it is desirable for the detector to function at high potentials and give good responses at high mass. Additionally, in quantitative analytical work it is highly desirable that response be proportional to the amount of analyte present.

It is difficult to find a detection system that adequately fulfils all of the above criteria. Most current MALDI-TOFMS instruments, including the linear MALDI-oa-TOF instrument used in the initial stages of this project, utilise microchannel plate (‘MCP’) detectors, with either a time to digital converter or integrating transient recorder monitoring the output [204]. MCP detectors provide time resolution of the order of nano seconds and provide a high gain at low mass. This results in a good overall performance, balancing the various requirements.

While a MCP provides a good compromise for conventional MALDI-TOF instruments, use of an oa-TOF analyser adds a further requirement. MALDI is a pulsed ion source, which, as explained in chapters 1 and 5, produces ions with increasing energy as mass increases, while an oa-TOF analyser works best with ions of constant desorption energy, regardless of mass [18]. This spread in desorption energies over mass due to the MALDI ion formation process requires a long orthogonal accelerator (in the desorption axis) for most of the ions to enter the field free flight region. The spatial distribution of ions has increased even further by the time ions reach the detector plane. Detection of most of these ions requires a large detector, when compared to other TOF instruments. Most of these ions have to be detected to provide a high sensitivity and optimal signal to noise.

125 Large MCPs are expensive, so a detection system was designed incorporating a microsphere plate (‘MSP’), a newer and less expensive alternative. Use of a MSP had not been reported in a TOF mass spectrometer, although a patent had been granted that described a linear TOF instrument with a MSP detector [205]. Since the detector was relatively novel, it was important to evaluate it, to determine how well it satisfied the necessary criteria.

8.1.2 Microsphere Plate Theory

A MSP is formed by sintering glass beads of 20 to 100Êµm diameter into a glass plate (0.7Êmm thick for a single thickness plate, such as the one used in this study). This creates a thin plate with irregularly shaped channels between the planar faces. The surfaces of the beads are covered with an electron emissive material and the faces of the plate are coated to make them conductive. When an appropriate potential is applied accross the plate, it acts as an electron multiplier, functioning in a similar manner to a MCP. An outline of its operating principles is discussed below. More detailed operating principles are given in papers by Tremsin and co-workers [206, 207] and Naaman and Vager [208].

An operational MSP is placed inside a vacuum chamber. A potential difference of between 1.5 and 3.5ÊkV is applied across the plate, with the back face of the MSP at the more positive potential. When a fast moving particle or a photon is incident on the front of the detector, it causes the emission of an integral number of initial secondary electrons. The electric field ensures these electrons are accelerated into the channels of the detector. When one of these electrons hits the surface of a bead inside the channel, it causes the emission of a number of secondary electrons, provided the electric field has accelerated it sufficiently. Each of the secondary electrons is itself accelerated and causes the release of a number of secondary electrons when it collides with the walls. This process is repeated a number of times, as illustrated in figure 8.1, significantly amplifying the signal. When monitored with an integrating transient recorder, the output of the detector is recorded as a pulse.

Earlier investigations of MSP behaviour utilised UV photons [206, 207]. Performance as an ion detector was unlikely to be identical. For instance, electron multiplier performance varies with the mass of the ions, and it is known that they operate less efficiently with higher mass ions [109, 209]. Thus, the detection system of the mass spectrometer was thoroughly evaluated, with particular attention paid to background noise, pulse width, effect of bias potential on gain, and any mass discrimination in gain.

126

primary ion

output electrons

electron multiplication

funneling of surface emitted electrons into the MSP

Figure 8.1: Diagram of a MSP, illustrating the sintered glass construction and secondary electron multiplication cascade. 8 . 2 Design of the Detection System The heart of the detection system was a 70Êmm diameter single thickness MSP electron multiplier, obtained from El-Mul (Yavne, Israel). The manufacturer’s specifications are given in table 8.1. The maximum potential drop allowed across the detector was 3.5ÊkV, while the baseline potential required for the oscilloscope used to monitor the detector’s output was ground. The front of the detector had to be at the acceleration potential, Ð20.0ÊkV, with the back at approximately Ð17ÊkV. This meant that the electrons emerging from the back of the MSP could not be directly monitored by the oscilloscope, since the oscilloscope’s input had to be close to ground. Thus the detection system was designed to allow the electrons to fall through an uniform field, 41Êmm in length, before they were absorbed by the collector plate, with the resulting transient signal monitored by an oscilloscope. The uniform field was set up with five spacer rings and a chain of six resistors, as was illustrated in figure 2.7. Photographs of the detector and the electron optics assembly are given in figure 8.2.

The bias across the MSP was controlled by adjusting the potential applied to the back of the detector, by increasing or decreasing the resistance (‘R’) of a device connected in parallel to the MSP. Two settings of R were available, providing a detector bias of 2.94ÊkV (low setting) or 3.15ÊkV (high setting) when the accelerator supply was set to Ð20.0ÊkV. The resistance of R was also kept significantly lower than the actual resistance of the MSP, to ensure the potential across the MSP was set by it. Additionally, keeping R much lower than the MSP resistance ensured that only a small fraction of total current flowing through the system was drawn through the MSP. This meant that even strong signals from the detector would only have a small impact on total current drawn from the power supply and hence the detector response would not decrease the accelerating potential applied to the instrument, important for mass axis stability and instrument resolution.

Electron multipliers are typically designed to operate under vacuum conditions, where they have high resistance values to avoid electrical discharges. An increase in background pressure decreases the resistance, allowing discharging to occur if operating potentials are applied. Such discharges can damage or destroy the sensitive surfaces of the detector, by drawing too much current through the device. While the MSP used in this instrument was more tolerant of increased background pressure than a typical MCP [208], safely operating at pressures of up to 3×10-2 mbar, it could still be damaged if the instrument’s pressure went higher. The risk of occurrence of damaging discharges was minimised, by limiting the total current from the power supply. This ensured no more than 0.5ÊmA was

127

-20 kV feedthrough

front of MSP

-17 kV feedthrough

direction of ion motion

-20 kV feedthrough PEEK support

} field spacer rings

-17 kV feedthrough

resistors

Figure 8.2: Photograph of the detector assembly from (A) a top view and (B) a side view. Table 8.1: Manufacturer’s specifications for the MSP used in the reflecting geometry MALDI- oaTOF (when new)

Physical Specifications plate diameter 70 mm active input surface diameter 67 mm active output surface diameter 68.6 mm thickness 0.70 mm sphere diameter 60 µm

Operational Specifications maximum bias potential 3.5 kV resistance 23 - 31 M pulse width < 1 ns dark current 12 nA

Table 8.2: Background count rates obtained from the MSP detector with bias set to 2.94 kV and 3.15 kV.

2.94 kV bias 3.15 kV bias overall results number of samples 154 132 286 raw results mean count rate 20.4 ct.s-1 18.5 ct.s-1 19.5 ct.s-1 standard deviation 7.8 ct.s-1 9.3 ct.s-1 8.6 ct.s-1 range 3.5 - 46.9 ct.s-1 3.7 - 59.2 ct.s-1 3.5 - 59.2 ct.s-1 rate per unit area… mean count rate 0.58 ct.s-1cm-2 0.52 ct.s-1cm-2 0.55 ct.s-1cm-2 standard deviation 0.22 ct.s-1cm-2 0.26 ct.s-1cm-2 0.24 ct.s-1cm-2 range 0.10 - 1.33 ct.s-1cm-2 0.10 - 1.68 ct.s-1cm-2 0.10 - 1.68 ct.s-1cm-2 … This was found by dividing the raw results by the active input surface area of the detector (35.3 cm2). allowed to flow through the MSP, significantly less than the 1ÊmA limit given by the manufacturer.

8 . 3 Evaluation of Detection System Performance Evaluation of the detector was performed with the analyser potential set to Ð20.0ÊkV, and the bias (potential difference) across the MSP set to either 2.94 or 3.15ÊkV, referred to as low gain and high gain respectively in the experiments below. The system was left to warm up for a two hour period before any measurements were commenced.

Initial experiments, which determined the background signal and relationship between bias potential and gain, utilised the signals from background counts. Background counts are usually due to field emission, outgassing, or decay of radioactive isotopes within the detector and use of background counts allowed the effect of the detector to be isolated from any effects due to the analyser. Subsequent experiments required single ions to determine the temporal response of the signal and the mass dependence of gain.

8.3.1 Measuring the Background Count Rate and Dark Current The detector’s background count rate and dark current had to be measured, since the combination of these parameters determined its background noise characteristic. It was, however, difficult to monitor the direct output of the detector, due to the small output voltages corresponding to the background signals. This was overcome by using the Ortec preamplifier (see chapter 6, section 6.5.4 for details of this device) to increase the potential of the output signals to a voltage that could be easily monitored, without increasing the number of counts from that outputed from the detector. One of the output channels was connected to the LeCroy 9384 Oscilloscope, to provide direct monitoring of the output signal. The other output was connected to a Hewlett-Packard 5382A 255ÊMHz frequency counter, set to a 10Ês gate (Hz frequency range), with a ×10 attenuator. The counter’s threshold for detection was 2.5 mV and with a net gain/attenuation this corresponded to a pulse amplitude threshold of 0.25ÊmV. After the warm up period the background count rate was monitored for 25 minutes, with the MSP set to the low gain setting. The frequency of background counts during this interval was measured with the frequency counter, giving approximately 150 readings, with the trace on the oscilloscope confirming that background counts and not electronic noise was monitored. MSP bias was adjusted to the high setting, left to warm up for 30 minutes, and the process was repeated.

Histograms of the count rates obtained are presented in figure 8.3, and a summary of the background count rates obtained is given in table 8.2. Both the total count rate and the

128

50 2.94 kV 40

frequency 20

0 0 10 20 30 40 50 60 counts/s

40 3.15 kV 30

20 frequency

0 0 10 20 30 40 50 60 counts/s

75 combined

50 frequency 25

0 0 10 20 30 40 50 60 counts/s

Figure 8.3: Frequency histograms of background count rate, for both detector bias settings and overall results. rate per unit area are given. The total rate is important in determining the contribution of background counts to noise in experiments on this instrument, while the rate per square centimetre allows these count rates to be compared with those obtained for other detectors.

Dark current was monitored in a separate experiment. The instrument was set up as described above for background counts, except that a picoammeter was connected in place of the frequency counter. This allowed the background current, rather than frequency of counts, to be monitored. The device used was a Keithley 485 Autoranging picoammeter (Keithley Instruments Inc., Clevland, Ohio). The MSP was set to the low gain setting and the background current monitored for 25 minutes by inspection of the readings. MSP bias was adjusted to the high gain setting and the process repeated after a further warm up period. The dark current was observed to vary between 14.5 and 15.3 nA at both gain settings. This was only slightly above the manufacturer’s quoted dark current of 12ÊnA.

The background count rate, of 20±10 per second (1 standard deviation), with each signal approximately 2Êns at the baseline, gives background interference for 10 to 30 discrete events over 20 Ð 60Êns in every second. If we assume these events are more or less evenly distributed we can estimate the contribution of this background to spectra. Spectra recorded on this instrument spanned no more than 100Êµs. Each 100Êµs sweep had an (approximately) 0.1 to 0.3% chance of having a background peak, and a vanishingly small probability of recording more than one background peak. The probability of recording zero to five background peaks over 100 averaged 100Êµs sweeps may be calculated with the binomial distribution (equation 8.1). The results of this are given in table 8.3. It is clear from these calculations that the background from the detector has a negligible effect on the spectra recorded on this instrument.

n! nk− ppp= k()1− (8.1) kkn!!()− k

pk is the probability of k events, n is the number of repetitions, and p is the probability for a single event The background count rate and dark current appeared to be related. Dark current was below measurable limits in conditions that gave no background counts. This is because the dark current is due to the flux of charge, which only occurred at potentials where background counts were detected. The relationship between charge and current can be determined from the definition of current [188]:

129 Table 8.3: Probability of zero to five background signals being recorded over 100 x 100 µs sweeps

No bkg signals Chance for a bkg signal per sweep 0.1 % 0.3%

0 90.48 % 74.05 % 1 9.057 % 22.28 % 2 0.449 % 3.319 % 3 0.015 % 0.326 % 4 0.0004 % 0.024 % 5 0.0000 % 0.001 % Total chance 100.0000 % 99.9999 %

Table 8.4: Settings for each sample in mass dependence of gain experiments

Sample m/z range Probe (V) Lens (V) Delay (µs) recorded DHB 130 - 270 50 0 3 - 5 gramicidin S 900 Ð 1,300 178 Ð 179 0 33 insulin 5,300 Ð 6100 178 - 179 74 - 76 18 Ð 19 myoglobin 16,200 Ð 17,600 171 74 - 76 60

Table 8.5: Mean gains at different masses and bias potentials

Detector bias m/z 2.94 kV 3.15 kV Mean Std dev. Mean Std dev. 130 - 270 2.13 1.78 3.50 2.80 900 Ð 1,300 2.12 1.14 2.57 1.65 5,300 Ð 6,100 2.19 1.38 2.88 1.96 16,200 Ð 17,600 2.33 1.30 3.27 2.24 pooled 2.19 1.42 3.06 2.22 All means and standard deviations have been divided by 1×106 prior to tabulation. dq i = (8.2(a)) dt which becomes, for a finite interval of time ∆q I = (8.2(b)) ∆t where I is the current in amps, ∆q is the charge in coulombs and ∆t is the time interval in seconds. Thus, for a dark current of 15ÊnA, the required charge over a period of one second is 15ÊnC. This corresponds to 94ÊxÊ109 electrons. This would give a gain per background peak, based on the measured rate of 20Ês-1, of 5ÊxÊ109, assuming all dark current is due to the background signals. This is over 100 times greater than the gain claimed by the manufacturer, indicating that most of the dark current was due to sub- threshold events that could not be detected with our transient recorder or pulse counter.

8.3.2 The Effect of Bias Potential on Gain It was decided to investigate the effect of bias potential on gain, just prior to using the analyser to record routine spectra at an accelerating potential of Ð20.0ÊkV. In this experiment, the effect of gradually changing the bias potential on background count gain was examined. The detector was set to the low bias potential for all measurements. The accelerating potential was set to Ð17.5ÊkV, the lowest potential for which the detector readily gave a signal, and left to warm up for 30 minutes. After this the area of the averaged signal from 30 background acquisitions was measured. This measurement was repeated. The magnitude of the accelerating potential was then increased by 100ÊV and left to stabilise for 5 minutes. Two area measurements were taken at the new potential. This process was repeated up to an accelerating potential of Ð20.0ÊkV.

The bias potential was calculated from the accelerating potential in the following manner. The bias potential at Ð20.0ÊkV was known to be 2.94ÊkV. The bias potential is proportional to the applied voltage, so the bias at each accelerating potential could be calculated with a simple formula:

Bx = x By y where Bx is the bias at an accelerating potential of x and By is the bias at potential y. Rearranging and substituting in the known bias at a potential of Ð20.0ÊkV gives: =− Bx 0. 147x (8.3) with x measured in kV. The areas obtained were then normalised, with the largest taken as 1.

130

1.00

0.95

0.90

0.85

relative signal 0.80

0.75

0.70

2.60 2.65 2.70 2.75 2.80 2.85 2.90 2.95 bias

Figure 8.4: Relative area of the background signal as a function of the detector bias. A line of best fit (r2 = 0.502) was added to indicate the trend in signal area. The relative area was plotted against bias potential. This is presented in figure 8.4. The data points were quite scattered, but the overall trend appeared to indicate that gain increased with bias, as expected. A line of best fit was plotted, to give a better indication of the overall trend. This demonstrated that gain does increase with bias, although this trend is of a smaller magnitude than the pulse area variability at each setting.

Background counts may originate anywhere in the detector and thus would not experience the same gain as single ions incident on the front plate. It is, however, likely that ion signals would show a similar trend for the change in gain with bias potential, albeit with a different magnitude to that shown by background counts. Thus the results in this section indicate that, in general, quantitative experiments should be performed at a single detector gain, or at least within a narrow range of gain settings, where the difference in gain will be much less than that due to the pulse height distribution of the detector. In a typical mass spectrum the pulse height distribution of the detector will have a smaller effect on the final result, due to the effects of averaging. Mass spectra usually consist of at least hundreds, or more typically thousands or tens of thousands of ions per mass used for quantitation. This allows the effects of pulse height distribution to be averaged out much more effectively than in the current experiment, which consisted of only 30 counts per peak.

8.3.3 Temporal Response The analyser power supply was set to -20ÊkV with the detector set to the low bias setting. After the warm up period the peak width (FWHM) was measured for 120 single ion peaks, obtained from a sample of DHB (m/z 130-270). The results are given in histogram format in figure 8.5. The mean width was 840Êps, with a standard deviation of 130Êps. The narrowest peak was 680Êps and the widest peak was 1,500Êps at half height. Over 75% of signals were less than 1Êns, as indicated by the histogram, which corresponded with the manufacturer’s pulse width specification.

The distribution of pulse width represents the detection system’s contribution to peak broadening in the mass spectrometer, indicating that on average 0.84Êns of each ion signal measured with the mass spectrometer is due to peak broadening by the detection system. Experiments discussed in chapterÊ7 determined that the combined contribution of the detection system and timing jitter was 1.21Êns. Detector broadening and timing jitter are not expected to be correlated, since detector broadening is due to the width of the electron cascade emitted from the MSP, with some contribution from the measuring system, while timing jitter is due to the timing reproducibility of the DSO start trigger,

131

100

frequency 40

0 0.6 0.8 1.0 1.2 1.4 1.6 peak width (ns)

Figure 8.5: Frequency histogram of peak width (FWHM) for 120 single ions of m/z 130-270 at a detector bias of 2.94 kV. relative to the push out pulse. Thus timing jitter can be estimated from these values to be approximately 0.94Êns, based upon quadrature addition.

8.3.4 Mass Dependence of Gain and Detector Conversion Efficiency It was very important to determine the relative gain and conversion efficiency of the detector at different masses, since MALDI is able to generate ions with mass to charge ratios ranging over at least five orders of magnitude. Samples of DHB and the analytes gramicidin S, insulin and myoglobin in DHB matrix were used to obtain signals over the mass range 130 to 17,600ÊDa. The accelerating potential was set to Ð20.0ÊkV and push out pulse to 1,005ÊV, providing velocities ranging from 1.8ÊxÊ105 for ions of 130ÊDa to 1.5ÊxÊ104 for ions of 17,600ÊDa. The mirror and push out pulse potentials were set to their standard values, although the exact value of these potentials was not critical, since this experiment depended upon single ion kinetic energy and over 95% of this energy was imparted to the ions by the accelerating potential. The instrument was left to warm up for at least 30 minutes before commencing experiments. Conditions were adjusted for each sample such that, in the appropriate mass range, only single or no ions were detected. Single ions signals were taken to be narrow peaks (no wider than 2.25Êns at half height) with a S/N of 3 or more. The settings and mass range monitored for each sample are listed in table 8.4. 100 single ion signals were recorded for each sample. The signal obtained in the appropriate mass range from a clean portion of the sample slide was monitored, before commencing these measurements, to ensure that there was no interfering background signal. In all instances there was no such background signal. Once this experiment was completed, the process was repeated with the detector bias set to the high gain setting.

The areas of the single ion peaks were measured, after the data had been rescaled to reflect the voltage setting of the oscilloscope. These areas were then converted to gain measurements. Gain is determined according to equation 8.4: q g= (8.4) ne where ‘g’ is the gain; ‘q’ is the charge measured by the oscilloscope; ‘n’ is the number of unit charges incident on the detector (1 for these experiments); and ‘e’ is charge of an electron. While n and e were known, q had to be calculated from the measured signal. A reference to dq is found in equation 8.2(a), the definition of current. Current can also be calculated from the definition of resistance: V I = R combining equation 8.2 with this relationship, then integrating with respect to time and making q the subject gives: 132 1 qtdt=∫V( ) (8.5) R ∫V(tdt ) corresponds to the area under the single ion peak (‘A’, in volt.sec). Substituting equation 8.5 into equation 8.4 gives: A g = (8.6(a)) Rne The effective impedance of our detection circuit was 25ÊΩ. Substituting in the known constant values for ‘R’, ‘n’ and ‘e’ gives equation 8.6(b): A g = (8.6(b)) 401. × 10−18

The calculated gain values ranged from 0.5×106 to 10×106 for the low gain results and

0.6×106 to 12×106 for the high gain results. The means and standard deviations of the gains for each mass at each bias potential were determined. A pooled result was also calculated over all the masses at each detector bias setting. These results are presented in table 8.5. The mean gain for the pooled values, 2.2×106 at 2.94ÊkV and 3.1×106 at 3.15ÊkV, are very good for a single thickness plate, although lower than the gain of 6×106 claimed by the manufacturer for a new plate. The mean gain increased for all mass ranges at the higher bias setting when compared to the lower bias setting and the pooled mean increased by 39%. This effect is smaller than the relative standard deviation for each set of measurements, which ranged from a low of 53% for m/z 900-1,300 at a bias of 2.94ÊkV, to a high of 84% for m/z 130-270 at the same bias. This result confirmed the conclusion from the experiments conducted over a range of bias settings (see section 8.3.2), that increasing bias does increase gain, although single ion peaks have a wide distribution of signal strengths. Surprisingly, the result did not show the expected trend with mass.

Detector gain for single ions was expected to decrease with mass, in an analogous manner to other electron multipliers [109]. The experiments described in chapter 7 had shown a decrease in sensitivity at higher mass, particularly for masses of over 5,000ÊDa. This decrease can be attributed to (1) the geometry limitations of the instrument, which meant that not all ions at higher mass would be detected; (2) any decrease in the conversion efficiency of the detector as mass increased; and (3) increasing fragmentation of ions with mass prior to the final drift region, between the ion mirror and the detector.

133 The analyser had been especially designed with a wide oa and large detector to minimise losses due to geometry, for known ion desorption velocity spreads. Conventional MALDI instruments are known to have lower sensitivity in reflectron mode for high mass ions, due to fragmentation, a feature utilised in post source decay experiments [210], but this alone could not explain our dramatic fall in sensitivity. Commercial MALDI-TOF instruments, utilising similar lasers, incorporating ion mirrors and providing similar flight times, are able to record spectra of substances such as myoglobin. Thus neither geometry losses nor ion fragmentation in the first drift region were the critical cause of loss of sensitivity. Hence, it was postulated that a very significant contribution to decreasing sensitivity was from the detector performance.

Detector conversion efficiency is likely to decrease with increasing mass, due to the decreasing average number of initial secondary electrons generated when an ion collides with the detector. Therefore, the results were analysed in this context. The global mean and standard deviation of a set of gain data only provide a good measure of the detector’s statistical performance when the raw data are normally distributed. The mean and standard deviation do not, of themselves, indicate whether this is the case, and it is not safe to assume that single ion signals from an electron multiplier are normally distributed. As described in section 8.1.2, electron multipliers work by causing increasing numbers of secondary electrons to be emitted after each collision. Each later collision releases an integral number of electrons and it is likely that, for each initial secondary electron, the actual signal will be normally distributed. Each ion incident on the detector can be expected to dislodge either zero or an integral number of initial secondary electrons (‘ISEs’). If no electrons are emitted the ion will not be detected. If one electron is emitted a signal will be detected. A large number of “one electron” results would be expected to provide normally distributed results around a mean. If two electrons are emitted by an ion, the detected signal would be larger. A large number of “two electron” results would be expected to be normally distributed around a mean value approximately twice that of a “one electron” result. The process is similar where there are a larger number of initial electrons. Thus if different numbers of ISEs are emitted the single ion data would be expected to be multimodal or have one or more high gain shoulders, and not be normally distributed around a single mean. The lowest gain value mode would be expected to indicate the “one electron result”, the next the “two electron” result and so on for the higher gain modes.

In a real detector there are of course other effects, such as sensitivity variations in the channels of the multiplier or gain saturation near the ends of the channels, that may overwhelm the effects on gain due to initial secondary electron emission. However, it was considered important to plot frequency histograms of the data, as this will show the 134 distribution of the results. A shift in the mode or skewed distribution would be expected when gain increases slowly, due to increasing emission of subsequent secondary electrons over part or all of the detector. This may occur where gain increases due to an increase in the detector bias. Changes in a multimodal distribution or changes in shoulders, which do not change the gain value of the primary mode, would be expected for differences in initial secondary electron emission. Such a result may be expected for different momentum and/or structure of the impacting ion.

In these experiments, all data were binned in the same manner, to allow direct comparison of the histograms. A total of 27 bins of equal width were used, spanning the full range of gain values obtained. The lowest value bin contained the gain range of 0.65×106 to

1.08×106. The resulting distribution of gains for the different masses are plotted in figure 8.6 for a detector bias of 2.94ÊkV and figure 8.7 for a detector bias of 3.15ÊkV. In each figure, ‘A’ gives the result for m/z 130-270, ‘B’ for m/z 900-1,300, ‘C’ for m/z 5,300- 6,100 and ‘D’ for m/z 16,200-17,600.

Net gain was expected to decrease with mass because of decreasing numbers of initial secondary electrons (a reduction in the conversion efficiency). Such changes should not affect the location of the modes, unlike varying the bias. Thus results for the different masses were pooled at each detector bias to make it easier to detect multimodal effects, since the larger number of data points should provide a smoother distribution. The resulting histograms are plotted as figure 8.8. A gaussian fit is shown for the two plots in figure 8.8, to provide an indication of whether the data is normally distributed.

All histograms show a sharp low gain cutoff, modal values at a gain of 1×106 - 2.5×106 and a high gain tail. The sharp low gain cutoff was due to the instrument’s limit of detection (3 times S/N). There probably were smaller single ion signals, but these were classed as noise and not recorded or analysed. The modal value does not appear to shift much for the individual masses and there is no clear shift in the distribution to lower gain as mass increases. In fact, the distribution for the highest mass, given at D in the figures, shows more high gain results than those found for lower mass analytes (B and C), the opposite to what would be expected if high mass ions caused less initial secondary electron emission. This suggests that, regardless of mass, most ions do not cause the emission of many initial secondary electrons. Alternatively, other effects, such as detector saturation from even one initial secondary electron, or hot spots on the surface of the MSP, are eliminating or swamping any effect due to initial secondary electrons.

135

20 m/z 200±70 15 A 10

5 Occurences (by bin)

0 1 2 3 4 5 6 7 8 9 10 11 Gainx106

20 m/z 1100±200 15 B 10

5 Occurrences (by bin)

0 1 2 3 4 5 6 7 8 9 10 11 Gainx106

20 m/z 5700±400 15 C 10

5 Occurences (by bin)

0 1 2 3 4 5 6 7 8 9 10 11 Gainx106

20 m/z 16900±700 15 D 10

5 Occurences (by bin)

0 1 2 3 4 5 6 7 8 9 10 11 Gainx106

Figure 8.6: Distribution of gains obtained for single ion peaks when detector bias was set to 2.94 kV. A represents results for matrix ions (m/z 130-270), B for gramicidin S (m/z 900-1,300), C for insulin (m/z 5,300-6,100) and D for myoglobin (m/z 16,200-17,600).

15 m/z 200±70

10 A

5 Occurences (by bin)