Faculty of Performance, Visual Arts and Communications

School of Music


Bach: Sonatas and Partitas for solo violin BWV 1001-1006 - Duncan Druce

The 19th-century revival of interest in music of the baroque period was spearheaded by the work of J. S. Bach; if this is true in general it also applies to the violin music of the period and especially to one piece, the Ciaconna from the Second Partita for solo violin, BWV 1004. As with the revival of the St. Matthew Passion, Mendelssohn again played a crucial role, providing an improvised piano accompaniment when Ferdinand David introduced it at a Gewandhaus concert in Leipzig in 1840. The complete Sonatas and Partitas had been published in 1802 by Simrock [No. 169] but it was David who was responsible for the first annotated edition, brought out by Kistner in 1843, the year in which the Leipzig Conservatoire opened its doors. This may not be coincidence; in close contact with Mendelssohn, David may well have seen these works as crucial studies for the advanced violin student. David’s edition presents the works in their original form, i.e. without keyboard accompaniment, and consists of two lines of text – his edited version and, underneath, what purports to be a transcription of the composer’s manuscript. This demonstrates the seriousness of David’s enterprise, but in fact the “original” text is not taken from Bach’s autograph MS (whose whereabouts was unknown at the time) but probably from the copy made by Anna Magdalena Bach, which was also taken as the basis for the Urtext Bach Gesellschaft edition of 1875. This second line, however, doesn’t give an accurate transcript of the Anna Magdalena text, either.

Following the David/Mendelssohn introduction of the Chaconne into Leipzig’s concert life, it was taken up by David’s pupil, Joseph Joachim. In July 1847, the sixteen-year-old violinist performed it in London, with the Mendelssohn accompaniment played by 'Herr Luders'. The Musical World (22 (1847), p. 137) commented – 'The quaint chaconne which brought the concert to a close, was vigorously performed by Joachim in spite of the obstructions of heat – so fatal to the comfort and success of the violin player.' The Musical World article also quotes a review in the Times:

Bach’s Chaconne is one of the most extraordinary productions of that great master, and the pianoforte accompaniment, added by Mendelssohn, is entirely in accordance with the spirit of the original, and scarcely inferior to it as a work of art and inspiration. The difficulties of modern writers for the violin shrink into insignificance before those exhibited in the Bach's Chaconne, which demands a mastery of double and triple stopping rarely to be met with. In Joseph Joachim, however, this requisite, and indeed every other, both of mechanism and expression, is found in singular perfection, and the recondite works of Bach are as familiar to him as the less elaborate achievements of the present day. [Ibid.]

It is clear from these comments that Bach’s unaccompanied violin music was unfamiliar, but other works had been on concerts programmes for some time. Taking London performances as an example, Ignaz Moscheles frequently played Preludes and Fugues on the pianoforte (presumably from the Well-Tempered Klavier), and other preludes and fugues had been heard on double bass and piano (Dragonetti and Julius Benedict, also James Howell and Cipriani Potter). [MW, 5 (1837) , p. 59.]

Within the next decade, the Chaconne had been taken up by a number of violinists, among them Henry Blagrove, Bernhard Molique, Louis Eller, Henri Vieuxtemps, and Heinrich Wilhelm Ernst. Most of these made use of the Mendelssohn accompaniment, but a performance by Ernst in Geneva on October 27th 1855 is described as for “violin solus” [reported in MW, 33 (1855), p. 728]. Ernst was certainly no stranger to the idea of performing without accompaniment – witness his transcription for solo violin of Schubert’s Erlkönig and his later set of concert etudes. Meanwhile, Molique, settled in London, had been inspired by Mendelssohn’s accompaniment to add piano parts to several further movements from the Sonatas and Partitas – the opening slow movements and fugues from the three sonatas, the Gavotte en Rondeau from the Third Partita, and the Tempo di Borea and Double from the First Partita. He frequently performed these, along with the Chaconne (usually accompanied by one of his daughters) and they were published by Kistner. Molique’s Bach editions take over the violin part of David’s Kistner edition virtually unaltered; the piano parts are discreet and sensitive (though to modern ears quite redundant). Contemporary with Molique’s editions are Schumann’s complete set of piano accompaniments to the Sonatas and Partitas.

David’s 1843 edition proved extremely influential, including his description of the set as Six Sonatas, doing away with the distinction between sonatas and partitas, or suites. (Only with Joachim/Moser and Marteau is this distinction re-instated.) Later 19th-century prints, and some from the early twentieth century, continue to follow many of his suggestions for bowing, fingering, articulation, and dynamic changes. An interesting illustration occurs in the Chaconne, where David has provided interpretations of the two places (at the end of the first minor section, and at the close of the major section) where Bach simply indicates arpeggio. In the first, longer passage , he suggests four different patterns of notes, to provide variety and a cumulative effect. These suggestions are followed by Hellmesberger (1865),Alard (1866), Sitt (1888), Hermann (1894), Wilhelmj (1897), and Auer (1917). For the third and fourth of his arpeggio patterns, David writes staccato dots under the slurs, indicating a springing bow, just like the arpeggios at the end of the Mendelssohn Violin Concerto cadenza.  This suggestion is followed by Alard and Wilhelmj, and partially by Auer. Other editors follow Hellmesberger in deleting the dots.  Even the Joachim edition of 1908, which takes a new look at this (and many other passages), retains some of the aspects of David’s interpretation. (For evidence of Joachim's thoughts about editing this passage, see Clive Brown's article).  Of the editions on the CHASE website, only Marteau  (1922), offers a truly independent version of this passage.

David’s notation of the second arpeggio passage , in the major, is taken over by Hellmesberger, Sitt, Wilhemj, Joachim, Auer, and (with a slight alteration) Marteau. (Alard’s edition suggests no arpeggiation pattern here; only Hermann provides an alternative reading.) Indeed, David’s version of these bars can often still be heard in present-day performances. In the Chaconne, David also elaborated the  bariolage passage in the final minor section, changing Bach’s pattern of slurred pairs of 16th-notes to triplets (after four bars) and then to 32nd-notes. This elaboration, not sanctioned by the composer, proved less influential, but Hellmesberger took it over (minus the triplet stage), as well as Sitt and Wilhelmj. Auer, too, elaborates the passage in a similar way, but suggests a smoother version, without David’s brilliant repeated notes.

A more influential innovation of David’s, though one that is somewhat unfortunate, is his placing of a chord on the last 8th-notes of the Chaconne’s first two bars (and similar places), where Bach has written just a single note. All the 19th-century editors, plus Auer, follow him, and despite the return by Joachim/Moser and Marteau to the original notation, most twentieth-century violinists continued to play the David version.


David’s fingering is comparatively sparse, mostly confined to what he considers essential – changes of position to avoid awkward string crossing, or to prevent a diminished fifth having to be taken by the same finger. In common with all the editors of these works on the CHASE website, however, David operates within a style of violin playing in which the portamento is commonplace. Semitones taken by the same finger are frequent, for instance the successive third fingers from F sharp to G in the Adagio first movement of the First Sonata (p. 3, line 1, from the 3rd to the 4th beat). An interesting comparison of fingering occurs in the Gavotte en Rondeau in the Third Partita at the single point where Bach’s MS includes a fingering, showing how a passage can be conveniently played in the second position. In David (1843) these bars occur on page 6, line 3, bars 3-5, and it will be seen that he avoids the second position, necessitating no fewer than four changes between first and third positions. But David is also at pains to indicate that the bars can be given a striking colour by using, wherever possible, the open E string. It is not surprising, perhaps, that David’s fingering here is not usually followed by later editors; in fact Sitt’s revision of David’s edition is the only one not to put the passage in second position. However Hermann keeps the repeated open Es (specifically excluded by Joachim, Auer, and Marteau) and several editors are happy to follow David’s bowing suggestion of 8th-notes slurred in pairs, with two pairs taken in a single bow. 

The Largo third movement of the Third Sonata, with its elegant, curving melody, offers opportunities for expressive fingerings, keeping a phrase on a single string, and introducing expressive portamenti. These opportunities are limited by the need to return to lower positions for the frequent double stops and chords. David adds only a few expressive fingerings – noted on the scanned copy on the website – and these are taken up by most of the later editors. Hellmesberger follows all David’s suggestions, and adds quite a few of his own (also noted on the scanned copy); these are incorporated into Sitt’s revision of David (plus an extra one of his own), Hermann and Auer. In the first bar, however, by detaching the fourth 16th-note (as in the autograph manuscript), Hermann, though remaining on the A string, obviates the need for portamento. Joachim/Moser, following many of Bach’s bowings, also suggests an interpretation with fewer portamenti, for example in bars 14 and 15 (line 5, bars 2 and 3) where, whilst still reproducing David’s fingerings, the upward position changes can now be accomplished without any slide. Marteau’s edition, however, goes further than any of the others in making use of higher positions on the A string, no doubt to achieve a sweet, expressive quality. Typical examples occur in bars 10 and 11 (line 4, bars 1 and 2) – noted on the scanned copy

Two passages in the Chaconne show the long influence of David’s fingerings. On p. 7 of the Second Partita, on line 5, bars 3 to 5 show a series of portamenti  to the third 8th-note of each bar. David has added slurs, making the slides unavoidable. Virtually identical fingerings are seen in the editions of Hellmesberger (p.4, lines 7-8), Alard (p.6, line 8), Sitt (p.5, lines 6-7), Hermann (p.6, line6), Wilhelmj (p.2, line 8), and Auer (p.5, lines 6-7). The rather more elaborate fingerings shown in the Joachim/Moser and Marteau editions are still inspired by David, retaining the same string for each 4-note group, and, in Joachim/Moser, keeping the harmonic A. A few bars later (line 6, from the end of the third bar), David instituted a fingering that can often still be heard (at least when the Chaconne is played with a modern violin set-up). He has the dramatic idea that the scale passages in this variation can be played on the G string. All the editors featured on the CHASE website are influenced by this fingering, Wilhelmj (a pupil of David) following it precisely and most of the others only deviating in bar 5 of the variation, where they prefer (following Hellmesberger) to stay in third position. Hermann changes to a less showy fingering for the last two bars of the variation, whereas Sitt (and, following him, Marteau) ascend as far as the upper B flat on the G string at this point – David’s G string excursion only takes him as far as the G harmonic. 

Bowing and Articulation 

The autograph MS of the Sonatas and Partitas is bowed with unusual care; there is occasionally some ambiguity about the exact placing of slurs, and in the Anna Magdalena MS the composer’s slurring is not always accurately transcribed, though in general the character of the original has been preserved. The most common slurring pattern found in Bach’s string music is where the first three notes of a group of four or six are slurred; as an eighteenth-century slur suggested a diminuendo from the first note this is clearly an indication that the first note of the group is to be stressed. The design of bow familiar to Bach meant it was comparatively easy to move the bow back towards the heel in time for the next down-bow slur, if several such bowings are shown in succession. In David’s 1843 edition, the second line, purporting to show the original version of the music, often changes these 3-note slurs into 4-note slurred groups. An example occurs during the second movement – Fuga – of the First Sonata (on p. 6, the 2nd and 3rd bars ). A 4-note slur here is quite awkward to manage; David’s typical solution is to separate the first note, adding a staccato dot, and slur the remaining three notes of the bar. This fits in well with a bowing style that concentrates passagework in the upper part of the bow. In the opening Adagio of this Sonata, with its written-out ornamentation in the Italian manner, it is no surprise that David is content to preserve the longer slurs of the original, but there are still places where he extends a slur to make a smoother effect - for example on p. 2, line 5, the 1st bar, where he includes the upward arpeggio on the first beat in the following slur. It’s interesting to observe that all the subsequent editors follow this suggestion (even Joachim/Moser and Marteau, who have the autograph ms for reference, which incidentally gives a different rhythm to that transcribed by David and the other earlier editors). However the bowing change in the fugue is followed by few of David’s followers – Sitt invents a different new bowing, but most of the others follow Hellmesberger, who reverts to the original bowing. 

The Double (variation) of the Sarabande in the First Partita provides an unusual instance of a movement not bowed by the composer. Today this piece can sometimes be heard with the continuous movement in triplet 8ths taken as separate bows, but it is by no means clear that this is what Bach would have intended, and all the CHASE editors provide detailed bowings, with alternations of different lengths of slur and separate bows. Taking the first section of eight bars as example, we find that David alternates two patterns: a four-note slur, followed by two separate notes and a three-note slur (bars 1, 2, 4, 6, 7, 8), and a bar where the first three notes and the last six are slurred (bars 3 and 5). This bowing is followed by Sitt, Joachim/Moser (who, however, omit the staccato dots on the separate notes), and Marteau, who, though providing an independent solution, makes use of David’s bar one pattern. Hellmesberger provides an alternative bowing, with generally shorter slurs and more separate bows. These are distinguished by either dots or lines, to show the relative lightness or smoothness of the required bowing, and expressive nuances and points of stress are shown by accents (> or <>), or else by means of tenuto lines under the first note included in a slur. Hellmesberger's bowing is followed, with minor alterations, by Hermann and Auer. This particular movement poses issues regarding tempo and character - most violinists perform it at a considerably faster speed than the Sarabande. Though none of the editors gives a specific tempo or metronome mark, several suggest an expressive character: "P con espressione" (David); "Dolce, ma con suono" (Joachim/Moser). Auer, whilst not indicating a general character, writes "staccato volante" under a series of up bows in the first bar, which implies an unusually swift tempo. A number of further issues concerning bowing are shown by annotations on the scanned copies: N.B. especially the final movements of the Second and Third Sonatas (BWV 1003 and 1005). 

David’s 1843 edition makes use of only a limited number of articulation marks. Later in his editing career (and especially in his manuscript annotations) he made regular distinction between staccato dots, tenuto lines, and a combination of lines and dots, but here he only uses dots. It seems they are designed not only to show separate bows, but also a lighter, less sustained bow stroke; in the Fuga of BWV 1001, different semi-quaver passages are marked “leggieramente” (with dots) [p. 5, bar 2], and "largamente" (without dots) [p.5, line 3]. A similar alternation occurs with the forte/piano contrasts at the start of BWV 1003’s finale; the forte semi-quavers unmarked, but the piano ones with dots. David also combines dots with slurs. In one place in the Chaconne, [BWV 1004, p. 8, line 3] he writes “saltato” under such a passage; elsewhere it is less certain whether he intends to bow to leave the string [BWV 1002, start of Corrente – on p. 4]. Often, indeed, this notation is used just to show two notes articulated within the same bow, as in the fugue subjects in BWV 1001 and 1003. For dotted rhythms, he is inclined to place a staccato dot only on the short note, as at the start of BWV’s Siciliana. This implies that in such a situation the dotted note is not shortened more than necessary; where he wishes a sharper, more detached character, David uses notational means, omitting the dot after the note as substituting a rest, as in the Allemanda of the First Partita, BWV 1002. 

David also uses notation to show articulation in the Andante of the Sonata BWV 1003, where the steady quavers of Bach’s bass line are shortened to semiquavers, to show separation. David indicates notes to be accented, or brought out by means of an > accent, more rarely by <>, as in the Sonata BWV 1001 [p. 6, lines 2 and 3] where it may perhaps suggest a note prolonged beyond its true value. He is also occasionally more precise about the intended bow stroke, as in the Double of the Tempo di Borea in BWV 1002 [p.9] where he writes “du milieu de l’archet”, or, in BWV 1005, “talon” [p. 10]. His indication, sostenuto, at the major-key section of the Chaconne [BWV 1004; p. 11, line 3] may be intended to suggest an especially smooth bow stroke. For strongly stressed, clearly articulated series of notes, David will suggest a succession of down bows [e.g. in BWV 1001 on p. 5, the bottom two lines]. There are places, too, where David shows one slur inside another. In the Third Partita’s Gavotte [BWV 1006, p. 5] this simply shows two slurred pairs articulated within the same bow. The more elaborate, virtuoso patterns in the Chaconne [BWV 1004; top of page 8] similarly shows a long, travelling up bow, with each pair of notes articulated. But in the Fuga of BWV 1003 [p. 6, line 4], David appears to suggest independent articulation for the two voices, the lower part maintaining the paired slurs of the inverted fugue subject whilst the upper part moves smoothly upwards without articulation. As before mentioned, in this edition David never uses the tenuto line, but he does write ten. over notes he wishes sustained at the start of the Corrente of the Second Partita {BWV 1004; p. 3]. Hellmesberger’s 1868 edition does make copious used of the tenuto line – at this very place in the Second Partita’s Corrente, for instance. Though this is a notable difference of editorial procedure, in many cases it is simply a more explicit way of showing the same variations in bowing. This applies in the semiquaver passages of the First Sonata’s Fuga, as well as at the start of the Second Sonata’s finale. He, too, uses an accent mark [>] to indicate notes that should be made prominent [BWV 1001; p. 4, line 6], as well as repeated down bow signs [BWV 1001; P. 4, line 5]. He sometimes uses tenuto lines on single notes [e.g. the third note of BWV 1001’s Siciliana: p. 6, bar 1]. In the finale of this Sonata he makes it very plain which notes should be played with a broad stroke [e.g. bar 1], and which with a lighter, possibly springing one [e.g. the last four bars on p. 6]. 

Generally, Hellmesberger favours more sober patterns of articulation than David. The First Partita’s Corrente has dots on the upward arpeggios, but not David’s repeated up bows; the long arpeggiando passage in the Chaconne does not use springing bowing across the strings. And though Hellmesberger keeps David’s bowing at the place in the Chaconne where David writes “saltato”, he omits the staccato dots on the quickest notes, suggesting that the staccato semi-quavers here are to be bowed on the string. If Hellmesberger’s edition suggests a less virtuoso manner of performance, he implies an approach where the violinist should make the sense of the music as clear as possible. In the Double of the First Partita’s Sarabande, he indicates within a few bars a considerable variety of articulation and emphasis with the use of staccato dots, tenuto lines, as well as > and <> signs for accentuation. It is perhaps symptomatic of his approach that, whereas in the Andante of the Second Sonata he follows David in shortening the repeated quavers to semi-quavers – for clarity – he doesn’t take over David’s notation in the Allemanda of the First Partita, preferring to keep the original dotted 16th plus 32nd notation and its suggestion of a rather more sustained style of performance.

Alard, in his editions of the Second and Third Partitas (BWV 1004 and 1006), follows David in confining himself to staccato dots as a main indicator of articulation. He also makes use of > accents (for instance, in the Allemanda of BWV 1004, and the Gavotte of BWV 1006). Like David, when dotted rhythms are articulated in the same bow, he generally puts a dot only on the short note, but, in the Corrente of BWV 1004 he suggests a sharper, more detached articulation by making both the dotted 8th and the 16th staccato. Also in David’s tradition, he lights on any opportunity for up-bow staccato. I imagine that most of these are designed as on-the-string staccato (as in BWV 1004, p. 12, bar 1), and Alard makes it clear by including an extra slur that David’s Saltato bowing in the Chaconne is not to be followed (BWV 1004, p. 7, line 5, bars 2 and 3. He does, however, suggest bounced arpeggiando bowings later on (BWV 1004, page 9). These follow David, and, indeed, Alard is quite ready in many places simply to reproduce David’s bowings, for example, in the Bourrée and Gigue in the Third Partita (BWV 1006).

Hans Sitt’s edition of 1889, announced as a revision of David’s, unsurprisingly reproduces many of his bowing and articulation markings. He does, however introduce, to a modest extent, the dash, to show a broad bow stroke (at the start of BWV 1004’s Allemanda, for instance), and occasionally rethinks the bowing of particular passages. In the Fuga of the Second Sonata, BWV 1003, he introduces dashes under a slur (page 4, line 9) to show broad but articulated notes within a single bow, and later in the same movement a series of up-bow staccato notes bears the instruction “saltato” (page 5, line 7). In the corresponding movement in the Third Sonata (BWV 1005), Sitt completely changes David’s bowing in the two passages over a pedal bass; whereas David’s idea is to play in the upper half with each bar starting on an up bow (to make the string crossing more convenient), Sitt indicates a down-bow start, with the additional instruction “talon”.

Hermann’s 1895 edition comes the closest to the original score of the 19th-century annotated prints. In a considerable number of places – the Gigue of the Third Partita (BWV 1006), the opening of the finale of the Third Sonata (BWV 1005) – he is happy to accept Bach’s short slurs beginning on the beat, and he is reluctant to employ the more virtuoso staccato or bounced bowings. This is not to say he presents himself as an editor anxious to follow the original text. In the finale of BWV 1003, for example, he is just as ready as the other editors to add to or change the original bowings. He uses the same markings as the other editors (staccato dot, lines, and accents) but tends to deploy them economically; where he adds lines to show a broad détaché stroke (at the start of BWV 1001’s finale) a single bar is sufficient. He also resorts to lines to show that individual notes are to be played sostenuto – the Siciliana in BWV 1001 makes a good case study of the contexts in which he indicates this. Overall, one gets the impression that Hermann himself, or students following his suggestions, would aim to give a polished account of the music, quite smooth in effect, and avoiding the virtuoso style that lies behind David’s, Alard’s, Sitt’s, and even Hellmesberger’s editorial interventions.

Of the twentieth-century editions, Joachim's and Marteau's are based for the first time on Bach's autograph manuscript, with the original text appearing underneath the annotated version. (Bach's text is not always accurately transcribed, however.) A number of erroneous readings are eliminated, for example Bach's back-dotted rhythm in bar 5 of the First Partita's Allemanda is given correctly by both editors. (The change of time signature from C to crossed C for the following Double, however, is ignored by Marteau.) Both editions show a greater readiness to retain the original slurring patterns, for example in the Allemanda of the Second Partita (BWV 1004). Alongside this, however, is an equal readiness to make changes to bowing and articulation to fit in with an early twentieth-century technical style. In the finale of the Second Sonata (BWV 1003) neither editor retains Bach's separate bowing in the third bar. But in the corresponding movement of the Third Sonata (BWV 1005), Marteau retains many features of the original articulation, whereas Joachim, in the 5th bar, remains faithful to David's style of not slurring from the first note in the bar, but isolating this note before an up-bow slur. These two editions share something in editorial approach, but they present a very different picture on the page. Joachim makes only limited use of dots, dashes and accents, preferring general indications such as "leggieramente" to give the player an idea of the character of a passage. Marteau, on the other hand, is liberal in his use of these signs, so that his edition appears far more prescriptive. A particular feature of his edition is the precision with which he shows the intended bowing. For instance, in the Chaconne (BWV 1004)(see p.19 of the scanned copy) he writes "Zwischen Frosch und Mitte des Bogens" (between the nut and the middle of the bow) over a passage where each note is marked semi-staccato - with lines and dots.

Leopold Auer's 1917 edition shares with Marteau's a concern to indicate in detail variations in articulation and bow stroke, using verbal descriptions as well as the usual dots, lines, and accents. In the Doubles (variations) of the Corrente and Tempo di Borea in the First Partita, he suggests that the repeats be varied - "La Ima volta sempre detaché" in a forte dynamic, and "IIda volta P e sempre spiccato". Unlike Marteau, however, Auer gives no indication of having consulted the autograph manuscript, and has little hesitation about introducing his own bowings in such movements as the Allemanda of the Second Partita (BWV 1004) and the finale of the Third Sonata (BWV 1005). In the Prélude to the Third Partita (BWV 1006) he even provides an alternative, simplified version of the bariolage passages, with slurred pairs of 16ths and the note-order changed. Other suggestions for bowing and articulation are taken over unchanged from earlier editions, especially Hellmesberger's.

Wilhelmj's Edition of the Chaconne

This edition, published in 1897 by Schlesinger, and dedicated to "Meinem Freund Felix Mottl" is for violin with accompaniment for orchestra or piano.  It is clear that Wilhelmj was familiar with his teacher David's edition of 1843; he takes over many bowings and fingerings unchanged.  The violin part is not a slavish transcription, however; where the dotted rhythms begin in bar 9, Wilhelmj does not follow his master in changing Bach's notation by introducing 16th-note rests.  But even where Wilhelmj makes changes, he will often retain aspects of David's editorship.  For example, at the point where David suggests saltato bowing (noted on p. 8 of the scan), Wilhelmj, whilst changing to a slur over the 32nds at the beginning of the bar, retains saltato for the quick notes on the 3rd beat.  The most substantial divergences from David are in the addition of many extra dynamic indications, presumably because of the need to spell out dynamic levels when performing with orchestra.  The relationship between Wilhelmj and David is shown very clearly in the two arpeggiando passages, and the bariolage episode in the final section (discussed above).  Wilhelmj follows David virtually to the letter in the way these sections are elaborated, even to the placing of staccato dots, accents and markings of Fz.  (Wilhelmj's suggestions for articulation in the first bar of the major-key arpeggiando is the most significant divergence.)  Only in the area of dynamic markings does Wilhelmj add to David's text in any substantial way, and even here the aim is more often to make David's idea more explicit than to change it.  For example, during the long crescendo during the passage of bariolage, Wilhelmj adds a half-way forte, followed by sempre crescendo.  In the first, long arpeggiando section, however, he does show a different, more elaborate dynamic structure.